Deep Hooking and Angling Success When Passively and Actively Fishing for Stream-Dwelling Trout with Baited J and Circle Hooks Christopher L

Home , Striped barb, Sun Pass State Forest

This article was downloaded by: [Department Of Fisheries] On: 20 January 2013, At: 19:20 Publisher: Taylor & Francis Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

North American Journal of Fisheries Management Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/ujfm20 Deep Hooking and Angling Success When Passively and Actively Fishing for Stream-Dwelling Trout with Baited J and Circle Hooks Christopher L. Sullivan a , Kevin A. Meyer a & Daniel J. Schill a a Idaho Department of Fish and Game, 1414 East Locust Lane, Nampa, Idaho, 83686, USA Version of record first published: 07 Dec 2012.

To cite this article: Christopher L. Sullivan , Kevin A. Meyer & Daniel J. Schill (2013): Deep Hooking and Angling Success When Passively and Actively Fishing for Stream-Dwelling Trout with Baited J and Circle Hooks, North American Journal of Fisheries Management, 33:1, 1-6 To link to this article: http://dx.doi.org/10.1080/02755947.2012.732670

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material. North American Journal of Fisheries Management 33:1–6, 2013 C American Fisheries Society 2013 ISSN: 0275-5947 print / 1548-8675 online DOI: 10.1080/02755947.2012.732670

ARTICLE

Deep Hooking and Angling Success When Passively and Actively Fishing for Stream-Dwelling Trout with Baited J and Circle Hooks

Christopher L. Sullivan,* Kevin A. Meyer, and Daniel J. Schill Idaho Department of Fish and Game, 1414 East Locust Lane, Nampa, Idaho 83686, USA

Abstract Circle hooks are becoming commonplace in recreational fisheries because they often reduce deep hooking, but there has been little evaluation of their effectiveness in trout fisheries. To compare the occurrence of deep hooking and angling success rates for stream-dwelling trout, we used three baited hook types (i.e., inline circle hooks, inline J hooks, and 4◦-offset J hooks) fished with two angling methods (i.e., active fishing, using a traditional bait fishing hook set; and passive fishing, with no sharp hook set). Of the 583 wild trout caught by anglers, 20% were deep hooked. The deep hooking rate varied by hook type and angling method, but the interaction term hook type × angling method was statistically significant, indicating that the effect of hook type could not be interpreted separately from fishing method. Accordingly, the occurrence of deep hooking was significantly greater for offset J hooks fished passively (28 ± 9% [95% CI about the mean]) and inline J hooks fished actively (27 ± 9%) than for offset J hooks fished actively (9 ± 6%) and inline circle hooks fished actively (10 ± 6%). Fish length affected deep-hooking rates, such that trout smaller than 250 mm were less likely to be deeply hooked than trout 250–350 mm in length. Hooking success (i.e., successful hook-ups divided by strikes) was greatest for actively fished inline J hooks (75 ± 7%), lowest for passively fished inline circle hooks (45 ± 6%) and passively fished offset J hooks (48 ± 8%), and always greater for actively fished hooks than for passively fished hooks of the same type. We found deep hooking was nearly twice as likely for inline circle hooks when fished according to manufacturers’ recommendations (i.e., passively) than when fished actively. These results and those of others suggest that fishing circle hooks actively when bait fishing for stream-dwelling trout will result in less deep hooking than fishing circle hooks passively.

Catch-and-release angling has become more commonplace mortality usually results from fish being hooked in critical loca- in the last several decades due to voluntary release of fish tions, including the gills, esophagus, or other organs, after the Downloaded by [Department Of Fisheries] at 19:20 20 January 2013 by anglers and special-regulation fisheries (Bartholomew and bait is ingested (Mason and Hunt 1967; Schill 1996). Bohnsack 2005). However, catch and release benefits the fish- Because of high mortality rates associated with bait-caught ery only by the rate at which released fish survive to reproduce fish, many special-regulation fisheries that restrict harvest also or are caught again by anglers (Wydoski 1977; Cooke and Suski prohibit bait fishing to maximize survival of fish caught by 2005). Postrelease mortality of trout has been researched exten- anglers (Noble and Jones 1999). However, such restrictions sively, and the type of fishing gear and angling methods used can alienate bait anglers and lead to dissension among angling can affect trout survival (Mongillo 1984; Jenkins 2003). Most groups (Thurow and Schill 1994; Noble and Jones 1999). Alter- research on trout has compared hooking mortality rates of fish native hook types have been developed that often reduce deep- caught when bait is used with those caught with artificial flies hooking rates for bait anglers, thus reducing hooking-related or lures; nearly all have concluded that the use of bait results mortality of bait-caught fish. For example, circle hooks, which in higher mortality (Shetter and Allison 1955; Hunsaker et al. are specifically designed with the hook point oriented perpen- 1970; Mongillo 1984; Meyer and High 2010). This increased dicular to the shank, have been shown to reduce deep-hooking

*Corresponding author: [email protected] Received May 25, 2012; accepted September 17, 2012 1 2 SULLIVAN ET AL.

rates in some species (Aalbers et al. 2004; Cooke and Suski Bull Trout Salvelinus confluentus, Brook Trout S. fontinalis, 2004). Unlike conventional J hooks, which often penetrate sur- Cutthroat Trout O. clarkii, and Rainbow × Cutthroat hybrids. rounding tissue when pressure is applied by the angler, circle Few Brook Trout and Bull Trout were caught and thus were not hooks are designed to slide past critical structures such as the included in the analyses. esophagus and gills and pivot only when exiting the mouth, thus Six experienced trout anglers fished from shore or waded into penetrating the jaw more frequently (Johannes 1981). the stream to cast into trout holding areas, focusing their efforts Circle hooks have been studied intensely in marine envi- in pools and slower runs where trout densities were highest. ronments, but little research has been conducted in freshwater Fishing rods, reels, and lines (Berkley Trilene 6-lb. monofila- systems (Cooke and Suski 2004), especially for stream-dwelling ment) were standardized for all anglers. Anglers used inline cir- trout. A recent study by Meyer and High (2010) found that deep cle hooks (Eagle Claw size 8, model L702G-8), inline J hooks hooking was, on average, almost five times higher and mortality (Eagle Claw size 8, model L214-8), and minor offset (∼4◦)J was almost four times higher for J hooks than for circle hooks hooks (Eagle Claw size 8, model 084-8), all barbed and baited for Rainbow Trout Oncorhynchus mykiss caught in a southern with night crawlers. Size 8 hooks were used because they are Idaho stream. However, because their study involved inline cir- one of the most common sizes used by trout bait anglers in Idaho cle hooks (i.e., nonoffset; point shank is parallel to main hook and elsewhere. Hooks were attached to the end of the line and a shank) and offset J hooks, differences in the effectiveness of removable split shot weight (0.88–2.65 g), was attached to the each hook type may have been confounded by differences in line above the hook. These rigs were drifted through the holding the angle of offset rather than by the point configuration. We water until a strike was detected. are not aware of any studies comparing inline (i.e., zero offset) Anglers fished each hook type both actively and passively. and offset designs of the same hook type on stream-dwelling Active fishing used the traditional hook-setting technique of trout. The few studies conducted on other species have found setting the hook with a sharp, quick lifting of the fishing rod that offset hooks may be more damaging than inline hooks, and when a strike was detected. In contrast, passive fishing, as de- in the case of circle hooks, may reduce or eliminate the benefits fined by Prince et al. (2002), was characterized by lifting the rod of using circle hooks over conventional J hooks (Hand 2001; slightly and gently while slowly reeling up slack line and apply- Prince et al. 2002). ing constant pressure when a strike was detected. It is generally Because limited circle hook research has been conducted on assumed that circle hooks must be fished passively to reduce trout in lotic systems, Meyer and High (2010) suggested that deep hooking (Montrey 1999; ASMFC 2003; Cooke and Suski further research comparing circle hook and J hook performance 2004). Anglers alternated hook type and fishing method (active for trout is needed in other streams before conclusions about or passive) periodically to ensure they caught similar numbers their effectiveness can be reached. To add to this knowledge of fish with each combination of hook type and angling method. base, we angled for trout in several streams in southern Idaho to Landed fish were identified to species, measured to the nearest evaluate (1) deep-hooking rates for actively and passively fished millimeter (mm, total length), and assigned a hook location of inline circle and inline and offset J hooks, and (2) hooking and esophagus, gills, upper jaw or mouth, lower jaw or mouth, or landing success rates and overall capture efficiency for each foul hooked (i.e., head, back, fin, etc.). Deep hooking was char- hook type and angling method combination. If circle hooks acterized by hooks embedded in the esophagus (or deeper) or reduced deep-hooking rates, we hypothesized that it might occur gill arches. Trout that were not landed due to line breakage were at the cost of reduced catchability, in terms of lower hooking not included in these analyses. and landing success rates, and that larger fish might be more The number of strikes and number of successfully hooked susceptible to deep hooking with all hook types. and landed fish were recorded for each hook type and angling Downloaded by [Department Of Fisheries] at 19:20 20 January 2013 method to determine hooking success, landing success, and capture efficiency. Hooking success rate was determined by dividing METHODS the number of successful hook-ups (i.e., the fish was hooked and Rivers selected for this study were the Big Lost, Big Wood, fought for at least 1–2 s) by the number of strikes. Landing suc- Boise, Malad, and South Fork of the Boise rivers in southern cess rate was determined by dividing the number of fish landed Idaho. Bait fishing is legal on all of these rivers except the by the number of successfully hooked fish. Capture efficiency South Fork of the Boise and is the most popular angling method for each hook type and angling method combination was deter- used by anglers statewide (IDFG 2007). These fisheries were mined by multiplying hooking success by landing success. The selected because of their moderate to high trout densities and deep hooking rate was determined by dividing the number of an abundance of fish larger than 200 mm, a size desirable to landed fish that were hooked in the gill arches or esophagus by anglers. Rivers were fished from late June to early October the total number of fish landed. 2010, after high spring flows had receded and angling could be Despite the binary nature of our response variable (i.e., fish effectively conducted. Streams ranged from 10 to 50 m wide, were either deeply hooked or not), we used multiway analysis from 800 to 2,100 m in elevation, and from 0.5% to 2.8% of variance (ANOVA) rather than logistic regression because in stream gradient. Trout present included Rainbow Trout, all independent variables were categorical. Accordingly, we DEEP HOOKING WITH BAITED J AND CIRCLE HOOKS 3

TABLE 1. Mean ﬁsh length and deep-hooking rates for each hook type and angling method.

Mean ﬁsh length (mm) Deep-hooking rate Hook type and angling method n Estimate SE Estimate 95% CI Inline circle (active) 97 282 6.8 0.10 0.05 Inline circle (passive) 100 294 6.4 0.19 0.08 Inline J hook (active) 92 316 7.8 0.28 0.09 Inline J hook (passive) 99 325 7.1 0.24 0.08 Offset J hook (active) 94 313 9.1 0.09 0.05 Offset J hook (passive) 101 326 8.9 0.28 0.08

constructed one global model to test the relationship between fishing (15 ± 4%). However, the interaction term hook type deep hooking and the following five categorical variables: × angling method was also significant (F2 = 4.03, P = 0.018), species (Rainbow Trout, Cutthroat Trout, and Rainbow Trout indicating that the effect of hook type could not be interpreted × Cutthroat Trout hybrids), fish length (binned into sizes of separately from how the hook was fished. Consequently, with <250 mm, 250–350 mm, and >350 mm), hook type (inline hook type and angling method combined, post hoc analysis circle hook, inline J hook, and offset J hook), hook-set method revealed that deep hooking was significantly greater for offset (active or passive), and angler (angler 1–6). Fish length was J hooks fished passively (28 ± 9%) and inline J hooks fished binned rather than considered a continuous variable because actively (27 ± 9%), than for offset J hooks fished actively we anticipated the relationship between fish length and deep (9 ± 6%) and inline circle hooks fished actively (10 ± 6%). hooking might just as likely be parabolic in shape as linear. We These results are summarized in Table 1. also tested for first-order interactions among all combinations of Fish length also affected deep-hooking rates (F2 = 7.76, independent variables. Nonsignificant variables were removed P < 0.001). Post hoc analysis revealed that, in general, trout from the model in a backwards stepwise manner. We used Tukey <250 mm long were less likely to be deeply hooked than trout post hoc tests to assess differences within groups. The ANOVA 250–350 mm long, but this relationship varied greatly between was performed with SAS statistical software (SAS 2009) at hook types and angling method (Table 2). Angler also influenced α = 0.05. deep hooking (F5 = 4.98, P < 0.001), with deep-hooking rates We used contingency tables and chi-square analyses (at α = ranging from 4% to 28% for individual anglers. Deep-hooking 0.05, computed by hand) to test for differences in hooking and rates for anglers differed only between the angler with the lowest landing success rates. For reporting purposes, we also calculated deep-hooking rate and those three with the highest deep-hooking 95% confidence intervals (CIs) around all hooking and landing rates. Species was the only factor in our analysis that did not proportions. influence deep hooking (F2 = 0.39, P = 0.676); this variable was subsequently dropped from the model before the F-value for the global model was reported. There were no other statistically RESULTS significant first-order interaction terms. From June to October 2010, six anglers landed 583 trout. Because of data recording errors by one angler, data for only The majority of trout landed were Rainbow Trout (76%), but 487 of the 583 landed trout were used in analysis of hooking and Downloaded by [Department Of Fisheries] at 19:20 20 January 2013 Cutthroat Trout (18%) and Rainbow × Cutthroat hybrids (6%) landing success and capture efficiency. Hooking and landing were also caught. The size (average ± SE) of landed fish was success rates and capture efficiency varied depending on hook 309 ± 3.3 mm and ranged from 155 to 500 mm. Landed fish type and angling method. Hooking success differed signifi- were hooked in the jaw or mouth most frequently (80%), but cantly among different combinations of hook types and angling 2 deep hooking occurred 20% of the time. Foul hooking was methods (χ 0.05, 5 = 13.0, P = 0.02; Table 3). Hooking success infrequent and accounted for <1% of the landed fish. was greatest for actively fished inline J hooks (75 ± 7%) and Use of ANOVA indicated that a number of factors influenced lowest for both passively fished inline circle hooks (45 ± deep-hooking rates (global model F12 = 5.33, P < 0.001). 6%) and passively fished offset J hooks (48 ± 8%). Once For example, the deep-hooking rate varied between hook types hooked, landing success on average was high (80 ± 3%), varied (F2 = 4.13, P = 0.017), with mean deep-hooking rates (active little between combinations of hook types and angling methods 2 and passive) being greatest for inline J hooks (26 ± 6%), (range 76–86%), and did not differ statistically (χ 0.05, 5 = 0.5, lowest for inline circle hooks (15 ± 5%), and intermediate for P = 0.99; Table 3). However, the significant difference in hook- offset J hooks (18 ± 5%). Also influencing deep-hooking rate ing success carried through to capture efficiency, which differed 2 was angling method (F1 = 7.23, P = 0.007), with deep-hooking between hook types and angling methods (χ 0.05, 5 = 12.2, rates being greater when passive fishing (24 ± 5%) than active P = 0.03; Table 3). Capture efficiency was highest for actively 4 SULLIVAN ET AL.

TABLE 2. Deep-hooking rates (95% CIs about the mean) for different length groups of trout caught with baited inline circle hooks and inline and offset J hooks.

Inline circle hook Inline J hook Offset J hook Active Passive Active Passive Active Passive Fish length (mm) n Rate n Rate n Rate n Rate n Rate n Rate <250 37 0.05 ± 0.07 28 0.03 ± 0.07 23 0.35 ± 0.20 20 0.35 ± 0.21 34 0.06 ± 0.08 23 0.09 ± 0.12 250—350 48 0.04 ± 0.06 51 0.33 ± 0.13 48 0.31 ± 0.13 44 0.25 ± 0.13 27 0.19 ± 0.15 37 0.43 ± 0.16 >350 15 0.40 ± 0.25 21 0.05 ± 0.09 30 0.20 ± 0.15 38 0.18 ± 0.13 38 0.05 ± 0.07 41 0.24 ± 0.13

fished inline J hooks (63 ± 8%), lowest for passively fished upon exiting the mouth. Our results and those of Meyer and inline circle hooks (35 ± 8%), and intermediate (41–50%) High (2010) suggest that fishing passively provides no benefit for the remaining combinations. Hooking success and capture when bait fishing for stream-dwelling trout with circle hooks efficiency was always greater for actively fished hooks than for and actually results in more deep hooking. One explanation for passively fished hooks of the same hook type. this may be that these studies were conducted in flowing water, and baited circle hooks drifting laterally through flowing water may perform differently than in lentic environments, where DISCUSSION bait is usually fished vertically (e.g., longline marine fisheries). Results of the present study indicate that the incidence of Zimmerman and Bochenek (2002) reported that circle hooks deep hooking when bait fishing for stream-dwelling trout may appeared to be less prone to deep-hooking flounder when drift be reduced when fishing with inline circle hooks, especially if speed was highest. Bait (and hook) retention time inside the fish they are fished actively. Considering that angling related mortal- (before a strike is detected) may be different between lentic and ity is predominantly caused by injuries related to deep hooking lotic environments, and actively setting the hook should mini- (Wydoski 1977; Mongillo 1984), we conclude that actively fish- mize hook retention time compared with not setting the hook. ing with inline circle hooks should result in lower postrelease This may influence deep-hooking rates as much as the type of mortality when anglers bait fish for stream-dwelling trout. In- hook set. It is also important to consider that anglers may not deed, these results suggest that the benefit of using inline circle have always detected the intake of the bait prior to the hook set hooks may be greatly reduced if they are fished passively rather when active fishing, thus making those “active” hook sets more than actively. This concurs with the findings of Meyer and High “passive” in nature. This could have biased the deep-hooking (2010), which showed an increase in deep hooking of Rainbow rates for active fishing and actually made them slightly higher. Trout with circle hooks fished passively rather than actively, The high deep-hooking rate for inline J hooks, regardless of although these results contradict conventional wisdom. Accord- whether they were fished actively or passively, may have been ing to circle hook manufacturers (see Montrey 1999; ASMFC due in part to hook configuration and dimensions. The swallow 2003; Cooke and Suski 2004), passively fishing circle hooks diameter of the hooks (i.e., outside diameter) was 5.8 mm for should result in lower deep-hooking rates than active fishing inline J hooks, 6.9 mm for offset J hooks, and 9.5 mm for inline because any deeply ingested hooks can slowly slide past the circle hooks. Thus, the streamlined shape of the inline J hook Downloaded by [Department Of Fisheries] at 19:20 20 January 2013 esophagus and gills before penetrating the corner of the jaw compared with the other hooks may have made it inherently

TABLE 3. Hooking and landing success and capture efﬁciency rates (95% CIs about the mean) for each hook type and angling method.

Hook type and Number of Number Number Hooking Landing Capture angling method strikes hooked landed success success efﬁciency Inline circle (active) 168 103 80 0.61 ± 0.07 0.78 ± 0.08 0.48 ± 0.08 Inline circle (passive) 248 111 86 0.45 ± 0.06 0.77 ± 0.08 0.35 ± 0.08 Inline J hook (active) 149 112 94 0.75 ± 0.07 0.84 ± 0.07 0.63 ± 0.08 Inline J hook (passive) 143 95 72 0.66 ± 0.08 0.76 ± 0.09 0.50 ± 0.08 Offset J hook (active) 186 113 90 0.61 ± 0.07 0.80 ± 0.07 0.49 ± 0.08 Offset J hook (passive) 160 76 65 0.48 ± 0.08 0.86 ± 0.07 0.41 ± 0.08 All hooks combined 1,054 610 487 0.58 ± 0.03 0.80 ± 0.03 0.46 ± 0.08 DEEP HOOKING WITH BAITED J AND CIRCLE HOOKS 5

easier for trout to swallow the hook once it was ingested. Sim- et al. 2005). The smallest trout captured in this study were prob- ilar results have been found for Bluegill Lepomis macrochirus ably not as capable of deeply ingesting bait (and hook) due to and Pumpkinseed L. gibbosus, where deep-hooking rates in- smaller mouth and esophagus openings, and thus had less deep creased with decreasing hook size (Cooke et al. 2005). Both hooking. hook size and gap width can affect deep hooking (reviewed in Six anglers collected data during the study, and deep-hooking Cooke and Suski 2004), and it is not possible to standardize rates differed between the angler with the lowest deep-hooking both with the same hooks. For example, because the hook point rate and the three anglers with the highest deep-hooking rates. bends back toward the shank, the inline circle hook in our study This may be explained in part by the size of fish caught by had the largest swallow diameter but also a smaller gap width those anglers. Angler 1, whose overall deep-hooking rate was (4.1 mm) than the inline J (4.7 mm) and offset J (4.9 mm) hooks. 4%, caught trout that averaged 213 mm in length, whereas the Larger hooks may be harder to swallow than smaller hooks, but other anglers with higher deep-hooking rates caught trout that they may also cause greater damage when they puncture tissue averaged 291, 323, and 349 mm. Because smaller trout were less (Pauley and Thomas 1993; DuBois et al. 1994). The relationship likely to be hooked deeply in our study, this probably explains between hook size, fish size, and hook performance has varied some of the difference in deep hooking between these anglers. widely among studies (Muoneke and Childress 1994), and fur- Thus, the fish size difference between anglers was the result of ther research comparing circle and J hooks with varying gap differences in fish length frequency between streams and the widths and swallow diameters would be useful to better assess number of days these anglers spent fishing different streams, how hook dimensions affect deep-hooking rates. Additionally, not differences in angling skill. comparing circle and J hooks with identical dimensions (i.e., Circle hooks must perform at least nearly as well as con- swallow diameter, gap width, or hook configuration) would be ventional hook types to gain acceptance among anglers (Jordan useful for evaluating circle and J hook performance. 1999). Cooke and Suski (2004) reported that the use of circle Inline and offset J hooks were used to determine whether this hooks generally resulted in less capture efficiency than conven- characteristic would affect deep-hooking rates. Other studies tional hooks. We found that capture efficiency for inline circle have attempted to assess the impacts of inline and offset hook de- hooks was lower than for inline and offset J hooks. However, signs, and have produced equivocal results. Hand (2001) showed these differences were inversely related to swallow diameter, that circle hooks with a minor offset (∼4◦) had slightly higher which for inline circle hooks was larger than for inline and off- deep-hooking rates (13%) than inline circle hooks (6%) for set J hooks. When fished actively, capture efficiency for inline Striped Bass Morone saxatilis, whereas Graves and Horodysky circle hooks was similar to capture efficiency for offset J hooks. (2008), studying White Marlin Kajika albida, showed no differ- Our study was not designed to control for differences in swallow ence in deep hooking or survival between 5◦ offset and inline diameter, so it is difficult to speculate whether hook size may circle hooks. Intuitively, the exposed point on an offset hook have influenced capture efficiency. However, we believe our re- should be more likely to lodge deeply in the fish. However, we sults suggest that anglers could transition to using circle hooks found just the opposite in this study, with offset J hooks resulting without suffering drastically reduced catch rates, regardless of in less deep hooking than inline J hooks. As mentioned above, fishing method. Nevertheless, our results are based on only three this may have been related to swallow diameter, but the offset common hook types, all of similar size. More research is clearly angle may also have further reduced the streamlined nature of needed on a variety of hook shapes and sizes before definitive the offset J hook relative to the inline J hook (beyond swallow conclusions can be drawn relative to capture efficiency with diameter alone), making it more difficult to quickly swallow the circle hooks on stream-dwelling trout. hook deeply. Further research using offset and inline circle and We acknowledge it is nearly impossible for managers to stip- Downloaded by [Department Of Fisheries] at 19:20 20 January 2013 J hooks would help elucidate the effect that hook alignment may ulate angling methods with regulation changes, especially in have on deep-hooking rates for bait anglers. stream trout fisheries. However, it is important for fishery man- We found a relationship between deep-hooking rate and fish agers to understand that active fishing with circle hooks does length, in that smaller trout (i.e., <250 mm) were less likely to be not appear to affect their ability to reduce deep hooking, as has deeply hooked than intermediate-sized trout (i.e., 250–350 mm). been reported in a number of marine fisheries studies. Therefore, The few previous studies investigating the relationship between managers can institute the use of circle hooks in trout fisheries fish length and deep-hooking rates when bait fishing have pro- without concern as to whether anglers fish them actively or pas- duced conflicting results. Grixti et al. (2007) showed a positive sively, and still gain the benefit of reduced deep-hooking rates. correlation between fish length and deep-hooking rates for Black In summary, circle hooks produced deep-hooking rates sim- Bream Acanthopagrus butcheri bait fished with longshank ilar to or lower than both J hook styles, regardless of fishing hooks, but Schill (1996) showed no relationship between Rain- method and therefore appear to be a viable option for fisheries bow Trout length and deep-hooking rates when bait fishing with managers attempting to reduce bait angling deep hooking (and size 8 offset J hooks. Relationships between fish size and deep subsequent related mortality) on stream-dwelling trout. This re- hooking are probably not causative, but rather may simply be a duction in deep hooking will likely be accompanied by slightly function of hook size (or bait size) relative to fish length (Cooke lower capture efficiency. However, further research is needed on 6 SULLIVAN ET AL.

circle hook performance in which a variety of trout bait-angling Hand, G. R. 2001. Evaluation of circle hooks as a means of reducing catch and scenarios, such as lentic and lotic environments, the effect of release mortality of Roanoke River Striped Bass. North Carolina Wildlife hook size, the inclusion of other circle hook designs (i.e., offset Resources Commission, Division of Inland Fisheries, Federal Aid in Fish Restoration, Project F-22, Raleigh. circle hooks), gap width compared to swallow diameter, and Hunsaker, D., II, L. F. Marnell, and F. P. Sharpe. 1970. Hooking mortality of the degree of offset in the hook, are evaluated before definitive Yellowstone Cutthroat Trout. Progressive Fish-Culturist 32:231–235. conclusions can be made about the performance of circle hooks IDFG (Idaho Department of Fish and Game). 2007. Fisheries management plan when bait fishing for stream-dwelling trout. If future researchers 2007–2012. IDFG, Boise. continue to find that stream trout caught with actively fished cir- Jenkins, T. M., Jr. 2003. Evaluating recent innovations in bait fishing tackle and technique for catch and release of Rainbow Trout. North American Journal cle hooks have lower deep-hooking rates than trout caught with of Fisheries Management 23:1098–1107. passively fished circle hooks, then hook manufacturers, man- Johannes, R. E. 1981. Words of the lagoon: fishing and marine lore in the Palau agement agencies, and outdoor media will need to modify their District of Micronesia. University of California Press, Los Angeles. hook set recommendations for circle hooks used to bait fish for Jordan, J. 1999. Going full circle. Big Game Fishing Journal 12(3):52–62. stream-dwelling trout. Mason, J. W., and R. L. Hunt. 1967. Mortality rates of deeply hooked Rainbow Trout. Progressive Fish-Culturist 29:87–91. Meyer, K. A., and B. High. 2010. Wild trout competition studies: project 2— wild trout investigations. Idaho Department of Fish and Game, Report 10-13, ACKNOWLEDGMENTS Boise. We thank the following individuals for their help in this Mongillo, P. E. 1984. A summary of salmonid hooking mortality. Washington project: S. Elle, E. Herrera, A. Dias, T. Lamansky, and A. Department of Game, Fish Management Division, Olympia. Montrey, N. 1999. Circle hooks ready to boom: design pierces fish through jaw, Sievers. Funding was provided by the Federal Sport Fish Aid promotes conservation. American Sportfishing 2(January 1999):6–7. and Restoration Act. Muoneke, M. I., and W. M. Childress. 1994. Hooking mortality: a review for recreational fisheries. Reviews in Fisheries Science 2:123–156. Noble, R. L., and T. W. Jones. 1999. Managing fisheries with regulations. REFERENCES Pages 455–477 in C. C. Kohler and W. A. Hubert, editors. Inland fisheries Aalbers, S. A., G. M. Stutzer, and M. A. Drawbridge. 2004. The effects of catch- management in North America, 2nd edition. American Fisheries Society, and-release angling on the growth and survival of juvenile White Seabass cap- Bethesda, Maryland. tured on offset circle and J-type hooks. North American Journal of Fisheries Pauley, G. B., and G. L. Thomas. 1993. Mortality of anadromous coastal Cut- Management 24:793–800. throat Trout caught with artificial lures and natural bait. North American ASMFC (Atlantic States Marine Fisheries Commission). 2003. Circle hook def- Journal of Fisheries Management 13:337–345. inition and research issues. ASFMC, Special Report 77, Arlington, Virginia. Prince, E. D., M. Ortiz, and A. Venizelos. 2002. A comparison of circle hook and Bartholomew, A., and J. A. Bohnsack. 2005. A review of catch-and-release “J” hook performance in recreational catch-and-release fisheries for Billfish. angling mortality with implications for no-take reserves. Reviews in Fish Pages 66–79 in J. A. Lucy and A. L. Studholme, editors. Catch and release Biology and Fisheries 15:129–154. in marine recreational fisheries. American Fisheries Society, Symposium 30, Cooke, S. J., B. L. Barthel, C. D. Suski, M. J. Siepker, and D. P. Philipp. Bethesda, Maryland. 2005. Influence of circle hook size on hooking efficiency, injury, and size SAS (Statistical Analysis Systems). 2009. SAS/STAT 9.2 user’s guide, 2nd selectivity of Bluegill with comments on circle hook conservation benefits in edition. SAS Institute, Cary, North Carolina. recreational fisheries. North American Journal of Fisheries Management 25: Schill, D. J. 1996. Hooking mortality of bait-caught Rainbow Trout in an Idaho 211–219. trout stream and a hatchery: implications for special-regulation management. Cooke, S. J., and C. D. Suski. 2004. Are circle hooks an effective tool North American Journal of Fisheries Management 16:348–356. for conserving marine and freshwater recreational catch-and-release fish- Shetter, D. S., and L. N. Allison. 1955. Comparison of mortality between fly- eries? Aquatic Conservation: Marine and Freshwater Ecosystems 14: hooked and worm-hooked trout in Michigan streams. Michigan Department 299–326. of Conservation, Institute for Fisheries Research, Publication 9, Ann Arbor. Cooke, S. J., and C. D. Suski. 2005. Do we need species-specific guidelines for Thurow, R. F., and D. J. Schill. 1994. Conflicts in allocation of wild trout

Downloaded by [Department Of Fisheries] at 19:20 20 January 2013 catch-and-release recreational angling to effectively conserve diverse fishery resources: an Idaho case history. Pages 132–140 in R. Barnhart, B. Shake, resources? Biodiversity and Conservation 14:1195–1209. and R. H. Hamre, editors. Wild trout V: wild trout in the 21st century. Trout DuBois, R. B., T. L. Margenau, R. S. Stewart, P. K. Cunningham, and P. W. Unlimited, Arlington, Virginia. Rasmussen. 1994. Hooking mortality of Northern Pike angled through ice. Wydoski, R. S. 1977. Relation of hooking mortality and sublethal hooking stress North American Journal of Fisheries Management 14:769–775. to quantify fishery management. Pages 43–87 in R. A. Barnhart and T. D. Graves, J. E., and A. Z. Horodysky. 2008. Does hook choice matter? effects Roelofs, editors. Catch-and-release fishing as a management tool. California of three circle hook models on postrelease survival of White Marlin. North Cooperative Fishery Research Unit, Arcata. American Journal of Fisheries Management 28:471–480. Zimmerman, S. R., and E. A. Bochenek. 2002. Evaluation of the effectiveness Grixti, D., S. D. Conron, and P. L. Jones. 2007. The effect of hook/bait size of circle hooks in New Jersey’s recreational Summer Flounder fishery. Pages and angling technique on the hooking location and the catch of recreation- 106–109 in J. A. Lucy and A. L. Studholme, editors. Catch and release in ally caught Black Bream Acanthopagrus butcheri. Fisheries Research 84: marine recreational fisheries. American Fisheries Society, Symposium 30, 338–344. Bethesda, Maryland. This article was downloaded by: [Department Of Fisheries] On: 20 January 2013, At: 19:23 Publisher: Taylor & Francis Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

North American Journal of Fisheries Management Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/ujfm20 Comparison of Detection Efficiency among Three Sizes of Half-Duplex Passive Integrated Transponders Using Manual Tracking and Fixed Antenna Arrays Nicholas J. Burnett a , Keith M. Stamplecoskie a , Jason D. Thiem a & Steven J. Cooke b a Fish Ecology and Conservation Physiology Laboratory, Department of Biology, Carleton University, 1125 Colonel By Drive, Ottawa, Ontario, K1S 5B6, Canada b Fish Ecology and Conservation Physiology Laboratory, Department of Biology and Institute of Environmental Science, Carleton University, 1125 Colonel By Drive, Ottawa, Ontario, K1S 5B6, Canada Version of record first published: 27 Dec 2012.

To cite this article: Nicholas J. Burnett , Keith M. Stamplecoskie , Jason D. Thiem & Steven J. Cooke (2013): Comparison of Detection Efficiency among Three Sizes of Half-Duplex Passive Integrated Transponders Using Manual Tracking and Fixed Antenna Arrays, North American Journal of Fisheries Management, 33:1, 7-13 To link to this article: http://dx.doi.org/10.1080/02755947.2012.734895

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material. North American Journal of Fisheries Management 33:7–13, 2013 C American Fisheries Society 2013 ISSN: 0275-5947 print / 1548-8675 online DOI: 10.1080/02755947.2012.734895

MANAGEMENT BRIEF

Comparison of Detection Efﬁciency among Three Sizes of Half-Duplex Passive Integrated Transponders Using Manual Tracking and Fixed Antenna Arrays

Nicholas J. Burnett,* Keith M. Stamplecoskie, and Jason D. Thiem Fish Ecology and Conservation Physiology Laboratory, Department of Biology, Carleton University, 1125 Colonel By Drive, Ottawa, Ontario K1S 5B6, Canada Steven J. Cooke Fish Ecology and Conservation Physiology Laboratory, Department of Biology and Institute of Environmental Science, Carleton University, 1125 Colonel By Drive, Ottawa, Ontario K1S 5B6, Canada

decoding their unique alpha-numeric tag code with a handheld Abstract wand. However, innovations in reader technology have led to the We compared detection efficiency for three lengths (12, 23 and development of systems capable of manually tracking fish using 32 mm) of half-duplex (HDX) passive integrated transponder (PIT) back-pack units (e.g., Roussel et al. 2000; Bubb et al. 2006; Hill tags for both manual tracking and fixed array applications. In a stream we used a wand-type manual tracking antenna and deter- et al. 2006) and fixed arrays capable of detecting tagged animals mined that detection efficiency was considerably influenced by tag as they pass near antennas (e.g., Castro-Santos et al. 1996; Bond size (i.e., 20% for 12 mm, 43% for 23 mm, and 81% for 32 mm) and et al. 2007; Aymes and Rives 2009). Upon detection, tagged fish water depth. Vertical and horizontal read range also varied among can provide valuable insight into life history traits, such as habi- tag sizes (lower for smaller tags) and orientation (12-mm and 23- tat utilization, migratory processes and patterns, behaviors, and mm tags oriented perpendicularly failed to read in the horizontal test). Using a fixed PIT array, we also compared the detection ef- survival (Brann¨ as¨ et al. 1994). Passive integrated transponder ficiency of the same three sizes of PIT tags in Shorthead Redhorse telemetry in fisheries research has developed into a more af- Moxostoma macrolepidotum released into a fishway. Again, detec- fordable method of tracking compared with traditional radio tion efficiency increased with tag size: 12 mm = 55.8 ± 9.2% (mean and acoustic telemetry methods (Cooke et al. 2012). ± = ± = ± SE), 23 mm 91.0 1.8%, and 32 mm 97.0 1.5%. When Manual tracking and use of fixed antenna arrays have in- using PIT telemetry on smaller fish species and/or life stages, we suggest that researchers consider the tag size, as the diminished decreased in popularity and have been subject to continuous testing tection efficiency of 12-mm tags could introduce a bias and impede and improvement. For example, over the past two decades, fish- Downloaded by [Department Of Fisheries] at 19:23 20 January 2013 the ability to address some research questions. eries scientists have been experimenting with several portable PIT detector designs (Morhardt et al. 2000; Cucherousset et al. Since its initial development in the mid-1980s, passive inte- 2005; Linnansaari and Cunjak 2007), aiming to improve the tag- grated transponder (PIT) technology has been applied across a reading performance of the portable unit, with the ultimate goal range of taxa, including fish, invertebrates, reptiles, amphibians, of increasing the detection range of tags. The general premise birds, and mammals (reviewed in Gibbons and Andrews 2004). of these manual tracking devices is that an antenna on the distal However, PIT tags have perhaps been most embraced by the fish- end of a pole is systematically swept over the surface of streams eries science community to mark and track fish in laboratory, (open water and through ice) interrogating tags and recording aquaculture and natural systems (Prentice et al. 1990; Lucas and tag information. Fixed arrays are deployed at strategic sites to Baras 2000; Cooke et al. 2012). Although PIT tags provide a detect and record tagged fish as they pass near (or through) long-term mark often used to identify broodstock (Moore 1992), antennas. From an applied perspective in the wild, PIT tags have early studies relied on physically recapturing tagged fish and been used for monitoring small-scale fish movements through

*Corresponding author: [email protected] Received July 11, 2012; accepted September 20, 2012 7 8 BURNETT ET AL.

stationary streamwide antennas and above flatbed antennas Comparative analyses that examine the influence of HDX (Armstrong et al. 1996; Aarestrup et al. 2003; Bond et al. 2007; tag size on detection efficiency via both manual tracking and Aymes and Rives 2009), for evaluating the passage efficiency of fixed arrays would inform researchers in selecting optimal tag fishways (Castro-Santos et al. 1996; Baumgartner et al. 2010; size for a given application. Therefore, we used three sizes of Thiem et al. 2011; Thiem et al. in press), for estimating the sur- HDX tags (12, 23 and 32 mm) representing the contemporary vival of out-migrating salmonid smolts in hydropower-impacted range of commercially available tags, and two detection strate- systems (Williams et al. 2001; Hockersmith et al. 2003; Axel gies (manual tracking and a fixed array) to evaluate detection et al. 2005; Welch et al. 2008), and for identifying critical efficiency. Specifically, the objectives of this study were to (1) habitats of small stream fish (Gresswell et al. 2006; Enders et al. compare the detection efficiency of a portable backpack reader 2007). and a fixed PIT antenna array and (2) quantify the maximum Despite advances in manual tracking and fixed-array PIT vertical and horizontal read ranges of the three available tag technology, research has demonstrated that detection efficiency sizes and determine how this varies with tag orientation. varies with several major biotic and abiotic factors. For example, physical habitat features such as undercut banks, deep pools, and woody debris accumulations negatively affect the METHODS detection efficiency of manual tracking systems, allowing fish Experiment 1: detection efficiency in stationary PIT tags.— to take shelter beyond the maximum detection distance of the To test whether the detection efficiency of a portable HDX wand-type antenna and impeding the movement of the opera- backpack reader varied with tag size and orientation, an experi- tor (Bubb et al. 2002; Keeler et al. 2007; Cucherousset et al. ment was carried out in Watts Creek (Kanata, Ontario, Canada; 2008). Subsequently, missed detections can result due to low 45◦2042N, 75◦5219W), a small urban tributary of the Ottawa recapture rates and high habitat complexity, resulting in unrep- River. Ten PIT tags of each HDX size (12 × 2.12 mm, 23 × resentative spatial ecology data. When using fixed arrays, fish 3.65 mm, 32 × 3.65 mm; Texas Instruments Inc., Dallas, Texas) traveling at high speeds or in large groups have been associated were fastened to the top of flat limestone rocks (mean weight = with decreased reader efficiencies, where code collision results 168.1 g; SE, 7.2) using nonconductive tape and Plasti Dip (Plasti in missed detections (Castro-Santos et al. 1996; Morhardt et al. Dip International, Blaine, Minnesota), and randomly deployed 2000). Nonetheless, study objectives will dictate whether man- throughout a 100-m study site reach on 4 April 2012. Lime- ual tracking, fixed arrays, or a combination of strategies are most stone rocks were used because they contain only trace amounts appropriate for addressing research questions. of iron and, thus, were unlikely to influence tag performance. Passive integrated transponders are available in a variety of The study site reach (mean depth = 22.4 cm, SE = 6.1; channel sizes, and it is well established that there is a tradeoff between width range = 3.7–8.5 m) was composed of deep pools, shallow tag size and read range (Cucherousset et al. 2010; Barbour et al. and moderately deep riffles, and undercut banks. No artificial 2011). Larger tags tend to have higher detection efficiencies but structures were present within or in close proximity to the study smaller tags enable the tagging of smaller species or life stages, site. The rocks were placed in order to position the PIT tags ei- thus extending the utility of the technology. In addition to tag ther parallel (upstream–downstream) or perpendicular to stream size, there are two types of PIT tags used in fisheries science that flow. Considerable effort was directed to avoid woody debris ac- differ in their ability to transmit and receive data. While half- cumulations, deep pools, and undercut banks when deploying duplex (HDX) PIT tag systems separate data transmission and tags. Additionally, all three tag sizes were spread equally among receiving periods temporally, full duplex (FDX) communication various potential fish habitats and depths: 12 mm at mean depth systems involve simultaneous data transmission and reception. of 21.8 cm (SE, 1.4), 23 mm at 23.5 cm (SE, 2.2), and 32 mm at Downloaded by [Department Of Fisheries] at 19:23 20 January 2013 As a result, HDX communication systems are much simpler than 22.0 cm (SE, 2.3). There was no significant difference in mean FDX systems, ultimately lowering the cost of the equipment deployed water depth among tag sizes (one-way ANOVA: F = required to undertake a PIT telemetry study (Barbour et al. 0.0089, df = 27, P = 0.925). An experienced operator used a 2011). However, until recently, HDX tags were only available portable HDX backpack reader (Oregon RFID, Portland, Ore- in the 23-mm and 32-mm sizes. Therefore, existing studies that gon) with a watertight polyvinyl chloride (PVC) antenna wand have evaluated the role of tag size (12, 23 and 32 mm) have (1.85-m length) and circular antenna (0.5-m diameter). The op- been limited to FDX technology (e.g., Cookingham and Ruetz erator was visually obstructed using semitransparent sunglasses, 2008; Cucherousset et al. 2009, 2010) or larger sizes of HDX unaware of tag detections (disconnected piezoelectric buzzer) tags (e.g., 23 and 32 mm; Morhardt et al. 2000; Linnansaari and na¨ıve with respect to tag location (see Linnansaari and Cun- and Cunjak 2007). In addition, a number of recent studies have jak 2007). The operator conducted five complete pass-through reported that a transponder oriented perpendicular to the antenna scans of the reach for each of the two tag orientations (time: plane possesses a higher range of reading (Bubb et al. 2002; Hill 16.1 min; SE, 1.1) in the downstream direction (maximum of et al. 2006; Bond et al. 2007). However, it is unclear how tag 50 detections per tag size), sweeping the antenna along the orientation influences detection efficiency of the full spectrum stream surface at a consistent sampling effort of 6.2 m/min. In of HDX tag sizes. deeper sections of the creek, the antenna was submerged to a MANAGEMENT BRIEF 9

maximum of 8 cm below the surface in an attempt to interrogate turning basin between antennas 6 and 7. There was no signif- tags on the streambed. Water quality values at the Watts Creek icant difference in TL among released fish (one-way ANOVA, study site (YSI model 85 water quality meter, Yellow Springs, F = 0.578, df = 57, P = 0.450). Of the 60 fish released, 30 were Ohio) were as follows: temperature = 8.3◦C, conductivity = male, 29 were female, and 1 was a juvenile. Passive integrated 1,404 µS/cm, and dissolved oxygen = 5.55 mg/L. transponders were injected into the coelomic cavity (long axis Experiment 2: PIT tag read range tests.—To evaluate the of tag aligned along sagittal plane) of each fish through the use vertical and horizontal read range of the three PIT tag sizes rel- of a 6-gauge hypodermic needle and syringe (see Thiem et al. ative to tag orientation, an experiment was conducted in still, in press). The entire handling process took less than 2 min, 77-cm-deep water in Watts Creek on 4 April 2012. Similar to and care was taken to minimize air exposure. Fish were placed past studies, one operator held the wand and circular antenna ventral side up in a foam-lined v-shaped trough; river water on the stream surface while another slowly moved a wooden flowed through the trough and was directed towards the mouth meter stick (PIT transponder secured to bottom) away from the of the fish. Shorthead Redhorse were used in this study as they front end of the wand in either a vertical or horizontal direc- were locally abundant during the study period. By releasing the tion (Morhardt et al. 2000; Cucherousset et al. 2005). Using tagged fish into the second turning basin (i.e., seven antennas this method, the maximum read range was noted when the upstream and nine antennas downstream), we were able to portable backpack reader failed to detect a tag (piezoelectric guarantee the greatest amount of detection data of fish either buzzer would cease to alarm) for longer than 30 s (portable unit ascending or descending the fishway. Shorthead Redhorse were reads 10 times/s). Horizontal read ranges were performed with released into the fishway in groups of 15 (5 of each tag size) to the tag in parallel and perpendicular orientations. Vertical tests avoid large group-associated detection errors. Fish were given were conducted exclusively with the tag in a parallel orienta- a total of 5 d to freely ascend (or descend) the fishway. We tion, as this was the only possible orientation to mount the tag were able to infer a direction of movement by comparing time- onto the meter stick. A tag in the parallel orientation was de- stamped detection data from multiple antennas. Throughout the fined as being parallel to stream flow (upstream–downstream), study period, the fishway passed 1 m3/s and the predicted max- while a perpendicular tag was oriented at a right angle to stream imum velocity through each vertical slot was 1.72 m/s (Thiem flow. A horizontal read-range test was also conducted with the et al. 2011). Water quality values in the Vianney-Legendre PIT tag in a perpendicular orientation. During both horizontal Fishway (YSI model 556 water quality meter, Yellow Springs, read-range tests, the meter stick was moved away from the front Ohio) throughout the duration of the study were as follows: of the antenna at a depth equal to half of the mean maximum temperature 14.1◦C(SE= 0.7), conductivity 183.8 µS/cm (SE vertical read range of the 12-mm HDX tag. We believed this to = 3.2), and dissolved oxygen = 11.2 mg/L (SE = 0.1). be an adequate depth to assess the horizontal read range because Data analysis.—Detection efficiency of each stationary PIT all PIT tags were detected within this vertical distance. All read- tag size (and orientation) was calculated as the proportion of range tests were conducted eight times to establish reproducible detected tags compared with the total possible number of detec- results. tions (experiment 1). As both 12-mm and 23-mm tags failed to Experiment 3: detection efficiency of Shorthead Redhorse.— read in the perpendicular orientation during the horizontal read To assess how the tag-reading performance of a fixed PIT an- range test (Table 1), no statistical analysis was possible (exper- tenna array varies with tag size, an experiment was conducted at iment 2). Following release into the second turning basin, 92% the Vianney-Legendre Fishway on the Richelieu River (Saint- (55 of 60) of fish descended the Vianney-Legendre fishway, re- Ours, Quebec;´ 45◦5148N, 75◦0900W). The fishway consulting in the greatest amount of potential detection data from sisted of 15 vertical slots (0.6 m wide), two resting–turning the nine antennas downstream (experiment 3). Some tagged Downloaded by [Department Of Fisheries] at 19:23 20 January 2013 basins, and large entrance and exit basins. The PIT array con- Shorthead Redhorse were detected by the first few antennas and sisted of 15 pass-through antennas, all of which were attached then failed to continue to descend (or ascend) the fishway 5 d on the upstream-facing side of each vertical-slot baffle. Anten- nas were constructed from 12-gauge stranded electrical wire ± that connected to a remote tuner box (Oregon RFID); groups of TABLE 1. Mean vertical and horizontal maximum read ranges (cm; SE) of half-duplex PIT tags (n = 8 for each tag size) in parallel and perpendicular four antennas connected to a multiplexor unit (Oregon RFID) via orientations. Numbers with different letters were significantly different based twin-axial cable. All antennas were tuned prior to conducting the on Tukey post hoc analysis. experiment to ensure optimal reading range (about 0.5 m) and tag-reading performance. Further details regarding the Vianney- Parallel orientation Perpendicular Legendre fishway and antenna array can be found in Thiem et al. Tag size orientation (2011). (mm) Vertical Horizontal Horizontal Sixty Shorthead Redhorse Moxostoma macrolepidotum 12 12.4 ± 0.04 x 6.4 ± 0.02 x = = (mean TL 407 mm, SE 3.6) were captured (10–15 May 23 20.2 ± 0.07 y 10.1 ± 0.02 y 2012, 0900–1400 hours) in a trap near the exit of the fishway, 32 29.0 ± 0.13 z 14.8 ± 0.10 z 4.6 ± 0.35 PIT-tagged (20 of each tag size), and released into the second 10 BURNETT ET AL.

postrelease, which limited the detection data to five antennas. tags (Figure 2). Tukey post hoc analyses revealed that the This detection failure may have been due to tag loss, death within read range (vertical and horizontal in the parallel plane) of all the fishway, predation, or simply passing through antennas un- HDX tags differed significantly from each other, with larger tag detected. Because all fish were known to have passed through sizes yielding larger read ranges (Table 1). While both 12-mm antennas 8 through 12, detected or undetected, we used only and 23-mm perpendicularly oriented tags failed to read in the these antennas to quantify detection efficiency among tag sizes. As one 12-mm-tagged fish was only detected by antennas 8 and 9, this individual was subsequently removed from detection 35 a) efficiency analyses. Portable PIT detection efficiency data (experiment 1) were 30 subjected to an analysis of covariance (ANCOVA) to determine if detection efficiency covaried with tag size and depth. Read range (experiment 2) and fishway detection (experiment 3) data 25 for all tag sizes were compared using a one-way ANOVA and Tukey post hoc analysis. All statistical analyses were performed 20 using R version 2.13.0 (R Development Core Team; www.r- project.org) at α = 0.05 for all tests. 15 RESULTS 10 Experiment 1: Detection Efficiency in Stationary PIT Tags The detection efficiency of stationary HDX tags in Watts a b c Creek varied considerably among tag sizes, with a higher pro- 5 portion of larger tags (23 and 32 mm) being detected than smaller tags (Figure 1). After controlling for the effect of tag size, an AN- read range (cm) vertical maximum Mean 0 COVA revealed that water depth also significantly contributed to the detection efficiency results (F = 13.49, df = 26, P = 12 mm 23 mm 32 mm 0.001). 16 b) Experiment 2: PIT Tag Read Range Tests There were significant differences in the mean maxi- 14 mum vertical (one-way ANOVA: F = 1989.6, df = 2, 21, P < 0.001) and horizontal (one-way ANOVA: F = 1074.3, 12 df = 2, 21, P < 0.001) read ranges among all parallel HDX 10

Downloaded by [Department Of Fisheries] at 19:23 20 January 2013 6

4 a b c 2

Mean maximum range (cm) horizontal read Mean maximum 0 12 mm 23 mm 32 mm Tag size

FIGURE 2. Mean maximum (a) vertical and (b) horizontal read ranges of the three half-duplex PIT tag sizes (n = 8 for each tag size and trial). Read ranges correspond to a parallel tag orientation. Error bars = SEs; different lowercase FIGURE 1. Detection efﬁciency of the three half-duplex (HDX) PIT tag sizes letters represent signiﬁcant differences (P < 0.05) from ANOVA and Tukey in Watts Creek using a portable HDX backpack reader. post hoc tests. Note that the y-axes possess different scales. MANAGEMENT BRIEF 11

20 Past research has identified that read range can be reduced by 100 20 the suboptimal orientation of the tag with respect to the antenna plane (Bond et al. 2007; Linnansaari and Cunjak 2007; Aymes 80 and Rives 2009). Despite the apparent difference in detection 19 between parallel and perpendicular tag orientations, the operator swept consistently throughout the study reach for all trials, mov- 60 ing from one stream bank to the other, effectively approaching tags from all possible angles. Although this sweeping technique 40 is a common antenna-maneuvering method (Bubb et al. 2006; a b b Cucherousset et al. 2008, 2010), it interrogates transponders in the same manner, regardless of orientation. As such, we were Detection efficiency (%) efficiency Detection 20 unable to describe a precise, unbiased difference in detection efficiency between parallel and perpendicular stationary PIT 0 tags. Although the current study did not use live fish to test 12 mm 23 mm 32 mm the handheld PIT tracking system, the results do indicate that Tag size tag size is important to consider when performing an assessment of small-bodied fish behavior and migration in freshwater FIGURE 3. Detection efficiency of PIT-tagged Shorthead Redhorse passing systems. through five antennas in the Vianney-Legendre Fishway. The numbers above Tag-reading performance of the portable reader also varied bars are the sample sizes (n). Error bars = SEs; different lowercase letters represent significant differences (P < 0.05) from ANOVA and Tukey post hoc with tag depth. This finding contrasts previous work by Enders tests. et al. (2007) who found that depth did not affect the detection success of finding 23-mm tagged juvenile Atlantic Salmon horizontal test, 32-mm tags had a mean maximum horizontal Salmo salar and Brown Trout Salmo trutta in water less than read range of 4.6 cm (SE, 0.35; Table 1). 1 m. However, handheld PIT detection technology may be limited to shallow water streams (<0.3 m) so that the water depth Experiment 3: Detection Efficiency of Shorthead Redhorse is effectively less than the maximum read range of the portable Detection efficiency among tag sizes (Figure 3) varied sig- unit (Cucherousset et al. 2008). Although all tags we deployed nificantly in the Vianney-Legendre Fishway (one-way ANOVA: were in water less than 0.4 m deep, the results indicate that F = 17.8, df = 12, P = 0.001). Tukey post hoc analyses iden- the water depth at which the stationary PIT tags were deployed tified that the detection success of 12-mm tags (55.8%, SE = exceeded the maximum read range of the tags. Future work us- 9.2%) was significantly lower than for the 23-mm (91.0%, SE = ing HDX tags on smaller species or life stages of fish should 1.8) and 32-mm (97.0%; SE, 1.5) tags, and there was no sig- conduct read-range tests prior to undertaking extensive manual nificant difference in detection efficiency between 23-mm and tracking to ensure that water depth does not exceed the maxi- 32-mm tags. mum read range. Additionally, when conducting a portable PIT telemetry study, it is prudent to ensure that a wand-type antenna is effectively watertight so that the wand can be submerged to DISCUSSION a depth that falls within the maximum reading range of the PIT Using a handheld wand-type antenna, we found a positive tags. relationship between PIT tag size and detection efficiency. As Read range tests yielded results similar to the manual track- Downloaded by [Department Of Fisheries] at 19:23 20 January 2013 we expected, the smaller HDX tags possessed a reduced read ing detection efficiency, with larger PIT tags (23 and 32 mm) range relative to larger tags, a finding most likely attributable possessing greater vertical and horizontal mean maximum read to the size of the internal copper coil within each transponder ranges when in a parallel orientation. Using portable antennas (Cucherousset et al. 2010; Barbour et al. 2011). Detection ef- on Steelhead Oncorhynchus mykiss Hill et al. (2006) identified ficiency data within the current study (14–82%) corresponded an innate flight response during the sweeping process, causing with past manual tracking studies, with differences probably due fish to move deeper into the water column or up to 50 cm up- to the PIT tag type and size, specifics of the portable unit de- stream from the operator. Although the trout did not move out sign, and physical habitat characteristics (Morhardt et al. 2000; of the portable unit’s detection range, several other studies have Roussel et al. 2000; Zydlewski et al. 2001; Cucherousset et al. commented that fish exhibiting escape behavior could swim by 2005; Hill et al. 2006). Although the tag-reading performance the portable backpack reader undetected, potentially causing of a portable PIT detection system can be influenced by highly spatial ecology assessments on small stream fish to be biased varying, extreme environmental conditions (Cucherousset et al. and unrepresentative of the local fish community (Cucherousset 2009), we believe it is unlikely that the surroundings during the et al. 2005, 2008; Cookingham and Ruetz 2008). We believe study had a direct influence on the Watts Creek detection results, this warrants a further investigation in order to determine the and the findings from this study are largely transferable. role fish escape behavior can play on the detection efficiency of 12 BURNETT ET AL.

portable PIT telemetry tools, particularly for fish tagged with possibility of tracking smaller species or life stages, it is not 12-mm HDX tags, for which read range is diminished. without limitations given the reduced read range and detection Only 32-mm transponders were successfully read in a per- efficiencies noted using both manual tracking and fixed PIT pendicular orientation during the horizontal read-range test. In arrays. reality, most fish species would not be tagged with the transponder oriented perpendicular (or nearly perpendicular) to the long- ACKNOWLEDGMENTS axis of the body, simply because only larger animals (i.e., large- We are grateful to C. Hatry, D. Hatin, F. Archambault, and G. bodied fish, turtles, etc.) would have the body depth to allow the Lemieux for providing field assistance at the Vianney-Legendre tag to be oriented in a different manner. We believe this finding Fishway and to S. Maarschalk for help in Watts Creek. Com- is irrelevant for small-bodied fish because there would probably ments by reviewers on an earlier version of the manuscript not be enough space in the body cavity for a larger transponder, are appreciated. N.J.B. was supported by an Ontario Gradu- which could, if used, introduce a tag burden to the fish. However, ate Scholarship from the Ontario Student Assistance Program it is possible that a fish tagged with the long axis aligned with (OSAP) and the Province of Ontario, and S.J.C. was supported the sagittal plane could be oriented perpendicular to stream flow by the Canada Research Chair program. Additional support was during a flight response or sudden change in direction. Indeed, provided by Natural Sciences and Engineering Research Coun- this may decrease the fish’s detectability when using a portable cil HydroNet and Research Tools and Instruments Grants Pro- backpack reader. gram grants to S.J.C. and the National Capital Commission. Using five antennas we were able to quantify the differ- All fish were handled in accordance with the guidelines of the ence in detection efficiency between the three HDX tag sizes Canadian Council on Animal Care administered by the Carleton in live Shorthead Redhorse in the Vianney-Legendre Fishway. University Animal Care Committee (B12-6). The larger tags (i.e., 23 and 32 mm) provided a significantly greater reader efficiency. Past detection efficiency work using REFERENCES 23-mm PIT tags and stationary detection systems have reported Aarestrup, K., M. C. Lucas, and J. A. Hansen. 2003. Efficiency of a nature-like similar efficiency results (Brann¨ as¨ et al. 1994; Armstrong et al. bypass channel for sea trout (Salmo trutta) ascending a small Danish stream 1996; Zydlewski et al. 2001; Axel et al. 2005; Aymes and Rives studied by PIT telemetry. Ecology of Freshwater Fish 12:160–168. 2009; Thiem et al. 2011). As the fishway used in the current Armstrong, J. D., V. A. Braithwaite, and P. Rycroft. 1996. A flat-bed passive study is designed to attract fish to pass through each vertical integrated transponder antenna array for monitoring behaviour of Atlantic Salmon parr and other fish. Journal of Fish Biology 48:539–541. slot directly against flow, we are confident that tagged Short- Axel, G. A., E. F. Prentice, and B. P. Sandford. 2005. PIT-tag detection system head Redhorse were interrogated with the transponder in an for large-diameter juvenile fish bypass pipes at Columbia River basin hydro- optimal, perpendicular orientation, similar to previous research electric dams. North American Journal of Fisheries Management 25:646–651. (Zydlewski et al. 2001; Bubb et al. 2002; Bond et al. 2007). Fur- Aymes, J. C., and J. Rives. 2009. Detection efficiency of multiplexed passive thermore, Aymes and Rives (2009) discovered that the sagittal integrated transponder antennas is influenced by environmental conditions and fish swimming behaviour. Ecology of Freshwater Fish 18:507–513. plane of Brown Trout were oriented parallel to the antenna plane Barbour, A. B., A. J. Adams, D. C. Behringer, T. Yess, and R. K. Wolfe. 2011. in countless detection-error events. As there was no significant PIT tag antennae arrays as fishery monitoring tools in tropical environments. difference in detection efficiency between 23-mm and 32-mm Proceedings of the Gulf and Caribbean Fisheries Institute 63:118–124. tags, we would expect representative movement and habitat use Baumgartner, L. J., C. A. Boys, I. G. Stuart, and B. P. Zampatti. 2010. Evalu- data through the use of either tag size. Although fixed PIT an- ating migratory fish behaviour and fishway performance: testing a combined assessment methodology. Australian Journal of Zoology 58:154–164. tenna studies on age-0 fish are limited to 12-mm tags (Prentice Bond, M. H., C. V. Hanson, R. Baertsch, S. A. Hayes, and R. B. MacFarlane. et al. 1990; Bond et al. 2007), our results suggest that when sam- 2007. A new low-cost instream antenna system for tracking passive integrated Downloaded by [Department Of Fisheries] at 19:23 20 January 2013 ple sizes are small or fish are unlikely to school, larger tags and transponder (PIT)–tagged fish in small streams. Transactions of the American subsequent larger read ranges are advantageous. Alternatively, Fisheries Society 136:562–566. ¨ ¨ ¨ ¨ smaller read ranges may reduce the chance of code collision Brannas, E., H. Lundqvist, E. Prentice, M. Schmitz, K. Brannas, and B. S. Wiklund. 1994. Use of the passive integrated transponder (PIT) in a fish in schooling species. Thus, a tradeoff exists when attempting identification and monitoring system for fish behavioral studies. Transactions to maximize detection efficiency. Researchers must balance the of the American Fisheries Society 123:395–401. probability of missing an individual due to code collision versus Bubb, D. H., M. C. Lucas, T. J. Thom, and P. Rycroft. 2002. The potential the probability of missing an individual due to a reduced read use of PIT telemetry for identifying and tracking crayfish in their natural range. We acknowledge that this issue is both species and study environment. Hydrobiologia 483:225–230. Bubb, D. H., T. J. Thom, and M. C. Lucas. 2006. Movement patterns of the specific. invasive signal crayfish determined by PIT telemetry. Canadian Journal of To conclude, when performing a PIT telemetry study, we Zoology 84:1202–1209. suggest that one should strongly consider the type of communi- Castro-Santos, T., A. Haro, and S. Walk. 1996. A passive integrated transponder cation system (i.e., FDX versus HDX), tag size, and the method (PIT) tag system for monitoring fishways. Fisheries Research 28:253–261. of interrogation (i.e., fixed versus manual). All of those factors Cooke, S. J., S. G. Hinch, M. C. Lucas, and M. Lutcavage. 2012. Biotelemetry and biologging. Pages 819–881 in A. V. Zale, D. L. Parrish, and T. M. could individually or collectively influence the ability to address Sutton, editors. Fisheries techniques, 3rd edition. American Fisheries Society, some research questions. Although 12-mm HDX tags open the Bethesda, Maryland. MANAGEMENT BRIEF 13

Cookingham, M. N., and C. R. Ruetz III. 2008. Evaluating passive integrated Linnansaari, T. P., and R. A. Cunjak. 2007. The performance and efficacy of transponder tags for tracking movements of Round Gobies. Ecology of Fresh- a two-person operated portable PIT-antenna for monitoring spatial distribu- water Fish 17:303–311. tion of stream fish populations. River Research and Applications 23:559– Cucherousset, J., J. R. Britton, W. R. C. Beaumont, M. Nyqvist, K. Sievers, 564. and R. E. Gozlan. 2010. Determining the effects of species, environmental Lucas, M. C., and E. Baras. 2000. Methods for studying spatial behaviour conditions and tracking method on the detection efficiency of portable PIT of freshwater fishes in the natural environment. Fish and Fisheries 1:283– telemetry. Journal of Fish Biology 76:1039–1045. 316. Cucherousset, J., P. Marty, L. Pelozuelo, and J. M. Roussel. 2008. Portable Moore, A. 1992. Passive integrated transponder tagging of Channel Catfish. PIT detector as a new tool for non-disruptively locating individually tagged Progressive Fish-Culturist 54:125–127. amphibians in the field: a case study with Pyrenean Brook Salamanders Morhardt, J. E., D. Bishir, C. I. Handlin, and S. D. Mulder. 2000. A portable (Calotriton asper). Wildlife Research 35:780–787. system for reading large passive integrated transponder tags from wild trout. Cucherousset, J., J. M. Paillisson, A. Cuzol, and J. M. Roussel. 2009. Spatial be- North American Journal of Fisheries Management 20:276–283. haviour of young-of-the-year Northern Pike (Esox lucius L.) in a temporarily Prentice, E. F., T. A. Flagg, and C. S. McCutcheon. 1990. PIT-tag monitoring flooded nursery area. Ecology of Freshwater Fish 18:314–322. systems for hydroelectric dams and fish hatcheries. Pages 323–334 in N. C. Cucherousset, J., J. M. Roussel, R. Keeler, R. A. Cunjak, and R. Stump. 2005. Parker, A. E. Giorgi, R. C. Heidinger, D. B. Jester Jr., E. D. Prince, and G. The use of two new portable 12-mm PIT tag detectors to track small fish in A. Winans, editors. Fish-marking techniques. American Fisheries Society, shallow streams. North American Journal of Fisheries Management 25:270– Symposium 7, Bethesda, Maryland. 274. Roussel, J. M., A. Haro, and R. A. Cunjak. 2000. Field test of a new method Enders, E. C., K. D. Clarke, C. J. Pennell, L. M. N. Ollerhead, and D. A. for tracking small fishes in shallow rivers using passive integrated transpon- Scruton. 2007. Comparison between PIT and radio telemetry to evaluate der (PIT) technology. Canadian Journal of Fisheries and Aquatic Sciences winter habitat use and activity patterns of juvenile Atlantic Salmon and Brown 57:1326–1329. Trout. Hydrobiologia 582:231–242. Thiem, J. D., T. R. Binder, J. W. Dawson, P. Dumont, D. Hatin, C. Katopodis, Gibbons, J. W., and K. M. Andrews. 2004. PIT tagging: simple technology at D. Z. Zhu, and S. J. Cooke. 2011. Behaviour and passage success of upriver- its best. BioScience 54:447–454. migrating Lake Sturgeon Acipenser fulvescens in a vertical slot fishway on Gresswell, R. E., C. E. Torgersen, D. S. Bateman, T. J. Guy, S. R. Hendricks, and the Richelieu River, Quebec, Canada. Endangered Species Research 15:1– J. E. B. Wofford. 2006. A spatially explicit approach for evaluating relation- 11. ships among coastal Cutthroat Trout, habitat, and disturbance in small Oregon Thiem, J. D., T. R. Binder, P. Dumont, D. Hatin, C. Hatry, C. Katopodis, streams. Pages 457–471 in R. M. Hughes, L. Wang, and P. W. Seelbach, ed- K. M. Stamplecoskie, and S. J. Cooke. In press. Multispecies fish passage itors. Landscape influences on stream habitats and biological assemblages. behaviour in a vertical slot fishway on the Richelieu River, Quebec, Canada. American Fisheries Society, Symposium 48, Bethesda, Maryland. River Research and Applications. Hill, M. S., G. B. Zydlewski, J. D. Zydlewski, and J. M. Gasvoda. 2006. Develop- Welch, D. W., E. L. Rechisky, M. C. Melnychuk, A. D. Porter, C. J. Walters, ment and evaluation of portable PIT tag detection units: PITpacks. Fisheries S. Clements, B. J. Clemens, R. S. McKinley, and C. Schreck. 2008. Survival Research 77:102–109. of migrating salmon smolts in large rivers with and without dams. PLoS Hockersmith, E. E., W. D. Muir, S. G. Smith, B. P. Sandford, R. W. Perry, N. S. (Public Library of Science) Biology [online serial] 6(10):e265. DOI: 10.1371/ Adams, and D. W. Rondorf. 2003. Comparison of migration rate and survival journal.pbio.0060265. between radio-tagged and PIT-tagged migrant yearling Chinook Salmon in Williams, J. G., S. G. Smith, and W. D. Muir. 2001. Survival estimates for down- the Snake and Columbia rivers. North American Journal of Fisheries Man- stream migrant yearling juvenile salmonids through the Snake and Columbia agement 23:404–413. rivers hydropower system, 1966–1980 and 1993–1999. North American Jour- Keeler, R. A., A. R. Breton, D. P. Peterson, and R. A. Cunjak. 2007. Apparent nal of Fisheries Management 21:310–317. survival and detection estimates for PIT-tagged Slimy Sculpin in five small Zydlewski, G. B., A. Haro, K. G. Whalen, and S. D. McCormick. 2001. Perfor- New Brunswick streams. Transactions of the American Fisheries Society mance of stationary and portable passive transponder detection systems for 136:281–292. monitoring of fish movements. Journal of Fish Biology 58:1471–1475. Downloaded by [Department Of Fisheries] at 19:23 20 January 2013 This article was downloaded by: [Department Of Fisheries] On: 20 January 2013, At: 19:24 Publisher: Taylor & Francis Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

North American Journal of Fisheries Management Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/ujfm20 Growth and Population Size of Grass Carp Incrementally Stocked for Hydrilla Control Daniel S. Stich a c , Victor Dicenzo b , Emmanuel A. Frimpong a , Yan Jiao a & Brian R. Murphy a

a Department of Fisheries and Wildlife Sciences, Virginia Polytechnic Institute and State University, 106 Cheatham Hall, Blacksburg, Virginia, 24061, USA b Virginia Department of Game and Inland Fisheries, 1132 Thomas Jefferson Road, Forest, Virginia, 24551, USA c Department of Wildlife Ecology, University of Maine, 5755 Nutting Hall, Orono, Maine, 04469, USA Version of record first published: 09 Jan 2013.

To cite this article: Daniel S. Stich , Victor Dicenzo , Emmanuel A. Frimpong , Yan Jiao & Brian R. Murphy (2013): Growth and Population Size of Grass Carp Incrementally Stocked for Hydrilla Control, North American Journal of Fisheries Management, 33:1, 14-25 To link to this article: http://dx.doi.org/10.1080/02755947.2012.739983

PLEASE SCROLL DOWN FOR ARTICLE

The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material. North American Journal of Fisheries Management 33:14–25, 2013 C American Fisheries Society 2013 ISSN: 0275-5947 print / 1548-8675 online DOI: 10.1080/02755947.2012.739983

ARTICLE

Growth and Population Size of Grass Carp Incrementally Stocked for Hydrilla Control

Daniel S. Stich*1 Department of Fisheries and Wildlife Sciences, Virginia Polytechnic Institute and State University, 106 Cheatham Hall, Blacksburg, Virginia 24061, USA Victor Dicenzo Virginia Department of Game and Inland Fisheries, 1132 Thomas Jefferson Road, Forest, Virginia 24551, USA Emmanuel A. Frimpong, Yan Jiao, and Brian R. Murphy Department of Fisheries and Wildlife Sciences, Virginia Polytechnic Institute and State University, 106 Cheatham Hall, Blacksburg, Virginia 24061, USA

Abstract In weed control plans that use Grass Carp Ctenopharyngodon idella for intermediate control of hydrilla Hydrilla verticillata, the knowledge of population dynamics improves efficacy of management. Our objective was to characterize growth, mortality, and associated population metrics of long-lived (up to 16 years) triploid Grass Carp that were incrementally stocked into Lake Gaston, Virginia–North Carolina, starting in 1995. Grass Carp (ages 1–16) were collected by bowfishers during 2006–2010. Growth of Grass Carp was described by the von Bertalanffy growth model −0.1352(t + 1.52) as Lt = 1,297[1 − e ], where Lt is TL at age t. We used three methods to estimate Grass Carp mortality, and annual abundance and biomass of Grass Carp were estimated from each mortality estimate. Estimated annual mortality ranged from 0.20 to 0.25 depending on the method used. The use of constant mortality rates versus age- specific mortality rates produced divergent models of Grass Carp biomass and represented a different approach for tracking the progress of weed control. Grass Carp biomass (but not abundance) was related to hydrilla coverage in Lake Gaston based on several scenarios that described time lags between Grass Carp stocking in year i and decreases in hydrilla coverage (in years i, i + 1, . . . , i + 5). Regardless of the mortality estimate used to derive Grass Carp biomass, the strongest biomass–hydrilla coverage relationship was observed for a time lag of 4 years. Fish older than age 10 constituted nearly 50% of the total Grass Carp biomass in Lake Gaston during some years, and the relationship between Grass Carp biomass and hydrilla coverage was strongest when fish up to age 16 were included in models. Downloaded by [Department Of Fisheries] at 19:24 20 January 2013 These results indicate that Grass Carp up to at least age 16 are important for weed control, thus highlighting the need for stocking models and bioenergetics models that include contributions of older fish when assessing long-lived Grass Carp populations.

Grass Carp Ctenopharyngodon idella have been widely verticillata, which is a preferred food source for Grass Carp stocked for biological control of aquatic vegetation in the USA (Allen and Wattendorf 1987; Chilton and Muoneke 1992; Kirk since the species’ introduction in 1963 (Mitchell and Kelly et al. 2000). However, variable success has resulted in Grass 2006). The Grass Carp has been proven as an effective control Carp stocking rates that range from 2 to 500 ﬁsh/vegetated agent for invasive aquatic weeds, including hydrilla Hydrilla hectare (Kilgen and Smitherman 1971; Allen and Wattendorf

*Corresponding author: [email protected] 1Present address: Department of Wildlife Ecology, University of Maine, 5755 Nutting Hall, Orono, Maine 04469, USA. Received January 23, 2012; accepted October 3, 2012

14 GROWTH OF GRASS CARP STOCKED FOR HYDRILLA CONTROL 15

1987; Bonar et al. 1993, 2002; Kirk et al. 2000). Traditional (in mass) of Grass Carp can be approximately linear with age approaches to using Grass Carp for biological control of hy- (Gasaway 1978; Morrow et al. 1997). Although some life his- drilla often involve large, isolated stocking events. In many tory studies have been conducted on triploid (sterile) Grass Carp situations, Grass Carp either provide inadequate control of veg- (e.g., Morrow et al. 1997; Kirk et al. 2000), little information etation (Baker et al. 1974; Kirk 1992; Killgore et al. 1998) or exists regarding the characteristics of established populations completely eradicate vegetation (including nontarget species) that have not eradicated all of the vegetation in the stocked wa- from aquatic systems (Stott and Robson 1970; Bettoli et al. ter body. As integrated pest management (IPM) becomes more 1993; Killgore et al. 1998; Schramm and Brice 2000). common as an approach to aquatic weed control (Chilton and In some cases, weed management goals target some interme- Magnelia 2008; Richardson 2008), knowledge of the population diate level of noxious weed infestation that is specific to stake- dynamics of long-lived, incrementally stocked Grass Carp pop- holder views and generally is based on some predetermined sur- ulations will become increasingly important because the IPM face coverage of hydrilla (Bonar et al. 2002). This management approach relies more on long-term, low-level Grass Carp stock- goal is highly controversial but generally is the result of conflicting than on traditional stocking strategies, which are designed ing stakeholder views (Chilton and Magnelia 2008; Richardson to eradicate vegetation in the short term. 2008). One approach to the intermediate control of hydrilla is the Hydrilla was first identified in Lake Gaston, Virginia–North incremental stocking of Grass Carp in combination with low- Carolina, in 1992 (Ryan et al. 1995). Since then, millions level herbicide application (Chilton and Magnelia 2008; Chilton of dollars have been spent on hydrilla control in the reser- et al. 2008). This approach theoretically allows fisheries man- voir. Coverage of hydrilla was initially about 10 ha and later agers to make adjustments to Grass Carp stocking rates so that peaked at 1,364 ha in 2003 (Dodd-Williams et al. 2008). Since fish density can be maintained at a desired level based on knowl- 1995, incremental Grass Carp stocking has been integrated edge of the population’s growth, mortality, and longevity. How- with annual fluridone applications to control hydrilla in Lake ever, information regarding population characteristics of Grass Gaston (Lake Gaston Weed Control Council [LGWCC], unpub- Carp is often lacking; therefore, stocking rates are commonly de- lished; www.lgwcc.org). The hydrilla leaf-mining flies Hydrel- termined on the basis of maintaining a desired number of Grass lia pakistanae and H. balciunasi (Diptera: Ephydridae) were Carp per total surface area of the lake or per unit of surface introduced into the lake in 2004, but they failed to establish weed coverage (e.g., Kirk et al. 2000; Bonar et al. 2002; Chilton viable populations and are considered to have been ineffective and Magnelia 2008; Chilton et al. 2008) rather than based on (Grodowitz et al. 2010). Due to the highly controversial nature the biology of Grass Carp in the system of interest. For exam- of aquatic weed control (Kirk and Henderson 2006) and the va- ple, this management approach assumes that for the purpose of riety of conflicting views among Lake Gaston stakeholders (see making management decisions, the number of Grass Carp in a Richardson 2008), the goal of weed control at the lake is not the system is more important than Grass Carp biomass; however, complete eradication of hydrilla. Instead, the management goal this assumption has not been validated for large reservoirs. for hydrilla control, as established by the LGWCC, is “to de- Mortality and growth rates of Grass Carp vary by geography, velop and maintain a healthy lake ecosystem based on a diverse climate, availability of food, and fish age (Chilton and Muoneke plant community dominated by native species” (LGSB 2005:8). 1992). State and federal agencies have developed software pro- To achieve this goal, one stated objective of management is grams that predict, based on a host of factors, the potential to reduce hydrilla coverage to 120 ha by 2012. The remaining effects of a cohort for up to 10 years after stocking (e.g., Stewart hydrilla coverage of 120 ha is designed to serve as a buffer for ex- and Boyd 1999). Limiting the analyzed effects to 10 years post- pected Grass Carp grazing and to allow for the re-establishment stocking is likely due to the contention in the published literature of desirable aquatic vegetation (LGSB 2005). The target den- Downloaded by [Department Of Fisheries] at 19:24 20 January 2013 that triploid Grass Carp older than age 10 make up a negligible sity for Grass Carp standing stock in 2011 was 37 fish/vegetated proportion of population size and biomass in most systems (Kirk hectare (LGWCC, unpublished; www.lgwcc.org). By 2010, hy- and Socha 2003). However, we suspect that this belief results drilla coverage in Lake Gaston was reduced to approximately from (1) high stocking rates and subsequently high mortality 666 ha (ReMetrix 2011), but this level of coverage was unsatis- rates of Grass Carp due to the elimination of aquatic vegetation factory in relation to management objectives for the lake. in the system that received the stocked fish (e.g., Morrow et al. The present study is the result of research that began in 2006 1997; Kirk et al. 2000; Kirk and Socha 2003); (2) the applica- to assess the current status of Grass Carp with regard to the tion of assumed mortality rates in lieu of either indirect or direct efficacy of weed control in Lake Gaston. One objective was to estimation of Grass Carp mortality (e.g., Chilton and Magnelia characterize the growth and mortality of the long-lived Grass 2008); and (3) the fact that Grass Carp are thought to consume Carp population in Lake Gaston in order to estimate the current less hydrilla and grow more slowly in proportion to body mass standing stock of Grass Carp in the lake. The second objective as they increase in body size and age (Gorbach 1961; Osborne was to use the standing stock estimates to characterize relation- and Sassic 1981). ships between hydrilla coverage and Grass Carp numbers and Research has shown that Grass Carp may live up to 21 years in biomass in Lake Gaston. Our third objective was to explore the systems where food is plentiful (Gorbach 1961) and that growth importance of Grass Carp up to age 16 for weed control and 16 STICH ET AL.

to rank the relative contributions of various age-groups to the where Lt is the mean length of fish at age t, L∞ is the theoret- efficacy of weed control. The information and relationships de- ical maximum length of fish in the population, K is the Brody fined in this study will be useful for assessing the current status growth coefficient, and t0 is the arbitrary origin of the equation of hydrilla control in Lake Gaston and should provide a basis (i.e., theoretical age at zero length; von Bertalanffy 1938). We for improved management through a better understanding of estimated 95% confidence intervals (CIs) for growth parameters the Grass Carp’s contribution to this long-term integrated weed by using bootstrap methods iterated 20,000 times. We predicted management effort. mean length at age, and we constructed 95% CIs for lengths at each age by using the bootstrapped results for VBGM parameter estimates in R. The length–weight relationship for Grass Carp collected from METHODS Lake Gaston was estimated by use of the function W = aTLb, Study site.—Lake Gaston is an impoundment of the Roanoke where a and b are constants, W is fish weight in grams, and River and spans five counties in Virginia and North Carolina. TL is total length in millimeters (Ricker 1975). We used this The total surface area of the reservoir is 8,423 ha, the total equation to predict the weight of Grass Carp at each age from volume is about 5.6 × 1011 L flowing at 1,245 m3/s, and back-calculated lengths at age (Anderson and Neumann 1996). the retention time is 29 d at the full-pond elevation of 61 m The predicted weight at each age was used in combination with (Richardson 2008; Dominion Power 2010). The reservoir is bor- mortality estimates (as described below) to estimate the biomass dered upstream by Kerr Reservoir and downstream by Roanoke of Grass Carp in Lake Gaston. Rapids Lake. Lake Gaston is operated to regulate discharges Mortality.—Morrow et al. (1997) used bowfishing catch data from Kerr Reservoir, the primary flood control station for the to estimate Grass Carp mortality in the Santee–Cooper Reser- Lower Roanoke River; therefore, lake elevation fluctuates little voir system, South Carolina, by using catch curves. However, in Lake Gaston, although flow is variable. The primary purpose our data did not meet the assumptions of catch-curve analysis; of Lake Gaston is hydropower production, but it also supports therefore, we used three alternatives based on VBGM param- popular sport fisheries and is a center of residential develop- eters. We used multiple methods to estimate mortality because ment in the region and therefore is used for a number of non- (1) it is the only estimated demographic parameter used in pre- consumptive recreational activities (Richardson 2008). Human dicting Grass Carp population size in Lake Gaston and (2) there population density is highest at the lower end of the reservoir, is a high degree of uncertainty associated with using indirect whereas the upper portion of the reservoir is sparsely populated methods of mortality estimation (Hewitt and Hoenig 2005). and includes designated wildlife management areas. The reser- The first method we used was based solely on life history the- voir also acts as a major source of drinking water for the City ory (Jensen 1996), the second method was based on empirical of Virginia Beach (Cox 2007). equations derived from meta-analysis of fish life history char- Age and growth.—Specially permitted volunteer bowfishers acteristics (Pauly 1980), and the third method was one that collected 243 Grass Carp from Lake Gaston during 2006–2010. allowed age-specific mortality to be estimated based on fish We measured TL (mm) and mass (g) of individual Grass Carp. life history characteristics (Chen and Watanabe 1989). We esti- Lapillar otoliths were removed from the fish and were prepared mated 95% CIs for each of the three mortality estimates by using and aged by using methods that were documented in the lit- bootstrapped 95% confidence limits (CLs) for the VBGM pa- erature but adapted based on technological advances (Morrow rameters. To assess the importance of using growth parameters and Kirk 1995; Morrow et al. 1997). A consensus was reached from an established population when using indirect estimates between multiple readers in order to assign an age to each Grass of mortality, we also estimated mortality from the VBGM pa- Downloaded by [Department Of Fisheries] at 19:24 20 January 2013 Carp based on annular ring formation in otoliths. rameters reported by Morrow et al. (1997) for Grass Carp in The Fraser–Lee method was used to back-calculate lengths at the Santee–Cooper Reservoir system. Because harvest of Grass each age (DeVries and Frie 1996). We estimated von Bertalanffy Carp is not permitted in Lake Gaston, estimated annual natu- growth parameters (von Bertalanffy 1938) simultaneously from ral mortality (M) represents the total annual mortality rate, and raw age–length data and plotted the von Bertalanffy growth changes in the size of the Grass Carp population depend entirely curve in R software (R Development Core Team 2011). To on mortality (i.e., reproduction is zero due to triploidy). optimize the fit of the von Bertalanffy growth model (VBGM), Jensen (1996) demonstrated that the relationship between M preliminary values of parameter estimates (obtained by using and the growth coefficient K could be expressed as mean length at age in FAST; Slipke and Maceina 2001) were employed as starting values for the final parameter estimation in R. Parameter estimates for the growth curve were reported in Mˆ = 1.50K. (2) the following form:

Mortality rates that we estimated by using the method of −K (t−t0) Lt = L∞[1 − e ], (1) Jensen (1996) will be referred to as Mˆ j . GROWTH OF GRASS CARP STOCKED FOR HYDRILLA CONTROL 17

Pauly (1980) estimated fish mortality based on relationships We estimated the 95% CI for population size based on the between L∞, K, temperature (T), and M for 175 fish stocks as 95% CLs for each mortality estimate; the 95% CI for biomass was estimated by propagating errors around estimated popula- ˆ = . · − . · + . · . loge M 0 654 loge K 0 28 loge L∞ 0 463 logeT (3) tion size and weight at each age (Frishman 1975). For comparison with biomass estimated by using mortality derived from Temperature used for estimating mortality with this method catch-curve analyses in other systems, we used mortality esti- can be based on average annual water temperature or on air mates reported by Kirk and Socha (2003; Santee–Cooper Reser- temperature. We used average annual water temperature at Lake voir system) to project biomass of the Lake Gaston Grass Carp Gaston as measured in Grass Carp telemetry studies (D. S. Stich, population, and we developed 95% CIs for biomass based on unpublished data) that were conducted during the same years as weight at age in Lake Gaston. the present study. Mortality estimates resulting from the Pauly Hydrilla coverage estimates used in this study were privately (1980) method will be referred to as Mˆ p. contracted by the LGWCC on an annual basis during 1995– Finally, we used a method developed by Chen and Watanabe 2010. Because a time lag was expected to occur between Grass (1989) to estimate age-specific mortality rates (Mcw)ofGrass Carp stocking and subsequent effects on hydrilla coverage, we Carp in Lake Gaston based on maximum observed age (tmax), tested the relationship between model-estimated numbers and K, and t0 as described mathematically by standing biomass of Grass Carp in year i and hydrilla coverage + + + + + ⎧ in years i, i 1, i 2, i 3, i 4, and i 5byusing ⎪ K , < simple linear regression (Montgomery et al. 2006). Models of ⎨⎪ − − t tmax 1 − e K (t t0) each time lag were ranked by using an information-theoretic Mcw = , (4) approach based on Akaike’s information criterion adjusted for ⎪ K ⎩⎪ , ≥ small sample sizes (AICc; Burnham and Anderson 2002) in 2 t tmax a0 + a1 (t − tmax) + a2 (t − tmax) SYSTAT version 12 (SYSTAT 2007). When each model was fitted individually with the complete data set, all lag scenarios where a0, a1, and a2 are constants pertaining to senescence. (including a lag of zero) indicated a strong empirical relation- ˆ ˆ Population size and biomass.—We used the Mj and Mp val- ship between total Grass Carp biomass and hydrilla coverage 2 ues to estimate the number of Grass Carp at each age remaining (R > 0.90). However, use of AICc to rank competing models in Lake Gaston at the start of each year (Nt,i) based on the requires that all models be drawn from the same set of observa- number of fish stocked (Ri) at time i: tions (Burnham and Anderson 2002). To meet this requirement, only 1998–2005 data could be used for ranking the relative (−Mˆ ) Nt,i = Nt−1,i−1e + Rt,i . (5) plausibility of different lags because (1) hydrilla data for 1996 and 1997 were not available and (2) the longest lag scenario We also used age-specific estimates of mortality (Mcw)to (5 years) prevented us from including years after 2005. estimate the number of Grass Carp in Lake Gaston at each age We used the zero-lag model of the Grass Carp–hydrilla rela- from 1995 to 2010 (Nt,i): tionship to determine the importance of Grass Carp up to age 16 for controlling hydrilla in Lake Gaston. To accomplish this, we (−Mcw ) Nt,i = Nt−1,i−1e + Rt,i . (6) developed four additive models within a hierarchical framework to test the relative contributions of four Grass Carp age-groups We estimated population size at the start of each year (Nˆ i ) (ages 1–4, 5–8, 9–12, and 13–16) to the efficacy of hydrilla and the biomass of Grass Carp in each age-class at the start of control in Lake Gaston (Table 1). The use of these age-groups Downloaded by [Department Of Fisheries] at 19:24 20 January 2013 each year (Bt,i) by using each method of mortality estimation: was intended to avoid overparameterization of models for age- classes in which data were sparse; however, the age-groups also coincide with previously documented changes in Grass Carp Nˆ = N , (7) i t i growth and mortality. For example, Grass Carp growth has been reported to be most rapid during ages 1–4 and declines to ap- and proximately 2.5 cm/year by age 8 (Gorbach 1961). The maximum age of Grass Carp in the southeastern USA was reported as 12 years (Kirk and Socha 2003), and the growth dynamics Bt,i = Nt,i Wt . (8) of Grass Carp older than age 12 are relatively undocumented in the USA. Because the existence of subsequent age-classes is Finally, we estimated standing biomass (Bˆ )ofGrassCarp i dependent upon the existence of preceding age-classes, we did in Lake Gaston at the start of each year: not construct reduced-parameter models that considered only the effects of the three oldest age-groups. The zero-lag scenario Bˆ i = Bt,i . (9) was used to maximize the data available for model subsets and 18 STICH ET AL.

TABLE 1. Model development and description for additive models used to test the relative contributions of four age-groups of Grass Carp to the efﬁcacy of hydrilla control in Lake Gaston, Virginia–North Carolina.

Model Description

a Hydrilla (B4 ) Response of hydrilla to biomass of Grass Carp ages 1–4 b Hydrilla (B4 + B8 ) Response of hydrilla to biomass of Grass Carp ages 1–8 c Hydrilla (B4 + B8 + B12 ) Response of hydrilla to biomass of Grass Carp ages 1–12 Hydrilla (B4 + B8 + B12 Response of hydrilla to biomass d + B16 ) of Grass Carp ages 1–16 aBiomass of Grass Carp ages 1–4. bBiomass of Grass Carp ages 5–8. cBiomass of Grass Carp ages 9–12. dBiomass of Grass Carp ages 13–16.

because it allowed us to compare the relative weights of models that included the oldest ﬁsh present in the system, whereas long (e.g., 4- or 5-year) time lags would not. The information- theoretic approach based on AICc (in SYSTAT version 12) was used for model selection, and the relative plausibility of models was ranked based on AICc weights.

RESULTS Growth and Mortality Grass Carp ranged in age from 1 to 16 years, and all stocked cohorts except 2010 were represented in the sample (Table 2). The TL of Grass Carp ranged from 417 to 1,350 mm, and mass ranged from 0.95 to 34.0 kg. Grass Carp growth in Lake Gaston was highly variable within age-classes. The relationship between TL (mm) and weight (W;g)ofGrassCarpwasW = (3.25 × 10−5) × TL2.87 (Figure 1b), suggesting that Grass Carp became less rotund as TL increased (see Anderson and Neumann 1996). The relationship between weight (Wt; g) and

FIGURE 1. Growth characteristics of Grass Carp in Lake Gaston, including TABLE 2. Stocking years, number of ﬁsh stocked (N), and annual catch of (a) the von Bertalanffy growth model describing length at age, (b) the length– Downloaded by [Department Of Fisheries] at 19:24 20 January 2013 Grass Carp by bowﬁshers in Lake Gaston, 2006–2010. weight relationship, and (c) the age–weight relationship. Number caught per year

2 Cohort N 2006 2007 2008 2009 2010 Total age t was nearly linear: Wt = 1,448 + 1,623t (r = 0.99, P < 0.001; Figure 1c). Predicted Grass Carp TL (mm) at each age 1995 20,000 12 25 9 17 22 85 was described by the VBGM as L = 1,297[1 − e−0.135(t + 1.52)], 1997 680 2 1 5 2 5 15 t where L is length at age t (Figure 1a; Table 3). 1999 5,000 0 2 1 3 12 18 t TheestimateofMˆ (mean = 0.20; 95% CI = 0.18–0.22) was 2003 25,392 22 14 4 6 7 53 j signiﬁcantly different from the estimate of Mˆ (mean = 0.25; 2006 7,000 1 7 7 16 9 40 p 95% CI = 0.23–0.28) within years (after 1996) based on the 2007 7,720 0 0 0 7 14 21 lack of overlap between 95% CIs. Age-speciﬁc mortality (M w) 2008 100 0 0 0 0 9 9 c declined rapidly between ages 1 and 5 and declined by less than 2009 6,520 0 0 0 0 2 2 2% between subsequent ages after age 5 and by less than 1% 2010 7,347 0 0 0 0 0 0 between ages after age 8 (Table 4). Although this pattern is ex- Total 79,759 37 49 26 51 80 243 pected based on the formulation by Chen and Watanabe (1989), GROWTH OF GRASS CARP STOCKED FOR HYDRILLA CONTROL 19

TABLE 3. Means, SEs, 95% conﬁdence limits (CLs), and test statistics for parameterization of the von Bertalanffy growth function used to describe growth of Grass Carp collected from Lake Gaston (L∞ = asymptotic length [TL, mm]; K = Brody growth coefﬁcient; t0 = theoretical age at zero length). Parameter Estimate SE Lower 95% CL Upper 95% CL tP> t

L∞ 1,297 24.160 1,253 1,348 53.66 <0.0001 K 0.135 0.007 0.122 0.149 19.67 <0.0001 t0 −1.52 0.112 −1.75 −1.31 13.54 <0.0001

the rate of decline in mortality between ages is a function of the To assess the precision of indirect methods in comparison specific growth characteristics of Grass Carp in Lake Gaston. with direct estimation of mortality, we estimated annual biomass When VBGM parameters from the Santee–Cooper Reservoir of Grass Carp in Lake Gaston by using five mortality rates population (Morrow et al. 1997) were used to estimate mortal- (Kirk and Socha 2003) that were based on catch-curve analyses ity via the indirect methods applied in our study, we found that (Figure 4). Biomass estimated from four of the five mortality Mˆ j was equal to 0.93 and that Mcw decreased from 2.76 at age rates fell within the 95% CI for biomass estimated with Mcw in 1 to 0.64 at age 6. the present study; biomass derived from the fifth mortality rate fell within the 95% CIs for biomass values that we estimated by Population Size and Biomass using Mˆ j and Mˆ p. Estimated population sizes (Nˆ i ) and biomass (Bˆ i )ofGrass We did not detect a significant relationship between Grass Carp varied widely dependent upon the method used to estimate Carp population size in year i and hydrilla coverage in any of + + mortality. In all years, population sizes estimated based on Mˆ p the time lag scenarios (i.e., i, i 1, . . . , i 5). A significant were intermediate to and significantly different from those esti- inverse relationship existed between Grass Carp biomass at time mated based on Mˆ j and Mcw (Figure 2). Biomass predicted based i and hydrilla coverage in all time lag scenarios (including the on Mˆ p was significantly different from biomass estimated with zero-lag scenario) using each of the three indirect mortality Mˆ j only for 2010 (Figure 3). Population sizes and biomasses estimates. The best model of hydrilla coverage in Lake Gaston estimated using Mˆ j and Mcw also differed significantly in all was consistently achieved with a 4-year time lag between Grass years. Annual population sizes and biomass predicted with Mcw Carp biomass and hydrilla coverage, regardless of the mortality were consistently smaller than those predicted by using Mˆ p or estimate used (Table 5). Mˆ j . For years of greatest disparity between estimates, popula- Grass Carp greater than age 10 accounted for 1–20% of the tion size and biomass derived from Mˆ j were more than double annual total population size, and they contributed 22% to nearly those derived from Mcw. 50% of the total annual biomass from 2005 to 2010 depending on the mortality estimate used (Figure 5). The oldest age-classes of Grass Carp also appeared to have a significant effect on TABLE 4. Age-specific mortality rates (Mcw; derived by the method of Chen and Watanabe 1989) and associated 95% confidence limits (CLs) for Grass Carp hydrilla coverage in Lake Gaston relative to other age-classes. in Lake Gaston. The best model of the effects of Grass Carp age on the efficacy of weed control included Grass Carp of all ages up to age 16, Age Mcw Lower 95% CL Upper 95% CL regardless of the method used to estimate mortality (Table 6). ˆ ˆ 1 0.47 0.43 0.51 When Mp or Mj was used to estimate biomass, the model that included ages 13–16 was over 200 times more plausible than the Downloaded by [Department Of Fisheries] at 19:24 20 January 2013 2 0.36 0.33 0.38 3 0.30 0.28 0.31 next-best model, which included only ages 1–8. When Mcw was 4 0.26 0.24 0.27 used to estimate biomass, the model including ages 13–16 was 5 0.23 0.22 0.24 45 times more plausible than the model that included only ages 6 0.21 0.20 0.22 1–8. Based on AICc difference ( i) values, models that did not 7 0.20 0.19 0.21 include the oldest age-group of fish had virtually no support in 8 0.19 0.18 0.20 the data. 9 0.18 0.17 0.19 10 0.17 0.16 0.18 DISCUSSION 11 0.17 0.15 0.18 The collection of Grass Carp in Lake Gaston by bowfishers 12 0.16 0.15 0.17 provided an effective means of sampling for age and growth 13 0.16 0.15 0.17 analyses of the population, as has been previously documented 14 0.15 0.14 0.17 (Morrow et al. 1997). However, because we lacked informa- 15 0.15 0.14 0.16 tion on sampling effort and because there was no standardized 16 0.15 0.14 0.16 regime for sampling Grass Carp in the lake, the Grass Carp catch 20 STICH ET AL.

FIGURE 2. Estimated Grass Carp population size (Nˆ i ; ± 95% conﬁdence interval) in Lake Gaston during 1995–2010, presented in relation to hydrilla coverage. Population size was predicted based on (1) mean mortality (across all ages) derived by the method of Pauly (1980; Mˆ p), (2) mean mortality derived by the method of Jensen (1996; Mˆ j ), or (3) age-speciﬁc mortality derived by the method of Chen and Watanabe (1989; Mcw).

data did not meet some of the underlying assumptions for use obtain a large number of individuals of various ages for use in of direct mortality estimation methods (such as catch curves), estimating VBGM parameters for the Lake Gaston population, which have been widely applied in fisheries monitoring for the resulting in well-informed parameter estimates based on data last 50 years (Thorson and Prager 2011). The mortality rates re- that were representative of fish approaching the maximum age ported by Kirk and Socha (2003) for the years 1998 (M = 0.33), in their native system (Gorbach 1961). In contrast, when we used 1999 (0.39), 2000 (0.35), and 2002 (0.38) produced biomass VBGM parameters from the Santee–Cooper Reservoir popula- estimates for the Lake Gaston Grass Carp population that fell tion (Morrow et al. 1997) to estimate mortality with the indirect within the 95% CI of biomass predicted by using Mcw in our methods applied in our study, we found that estimated mortality study. The low mortality rate of 0.22 (for 2001) reported by Kirk was not within the range of estimates that were derived by using and Socha (2003) resulted in a predicted biomass that was within catch-curve analyses in the same system, even within the same Downloaded by [Department Of Fisheries] at 19:24 20 January 2013 the 95% CI of biomass estimated with Mˆ p and Mˆ j in the present years (Morrow et al. 1997; Kirk and Socha 2003). These results study. These results suggest that (1) the indirect methods we suggest that the use of VBGM parameters from populations that used to estimate mortality have precision comparable to that of are short lived (or are expected to be short lived) is not an ap- direct estimation based on catch curves for other systems and (2) propriate approach to mortality estimation, despite large sample age-specific mortality may present more comparable estimates sizes such as those obtained from the Santee–Cooper Reser- over the life span of Grass Carp. Although the reliability of the voir population (Morrow et al. 1997; Kirk et al. 2000; Kirk and indirect methods we used cannot be measured against that of Socha 2003). Because our estimates are based on a long-lived the estimates derived by Kirk and Socha (2003), the agreement population of Grass Carp that were stocked incrementally, they between results of the two studies suggests that the precision should be useful for projecting Grass Carp population sizes in achieved in the present study should be satisfactory for use in other southeastern U.S. systems, especially where the goal is the management. intermediate control of weeds rather than eradication and where The expected longevity of the population should be consid- Grass Carp are stocked incrementally in sufficient numbers to ered when using indirect methods of estimating mortality based persist for more than 10 years (the age accommodated by other on growth of triploid Grass Carp. In this study, we were able to stocking models; e.g., Stewart and Boyd 1999). GROWTH OF GRASS CARP STOCKED FOR HYDRILLA CONTROL 21

FIGURE 3. Annual hydrilla coverage in Lake Gaston during 1995–2010, presented in comparison with estimates of Grass Carp standing biomass that were derived from mortality estimated by the method of (a) Jensen (1996; Mˆ j ), (b) Pauly (1980; Mˆ p), or (c) Chen and Watanabe (1989; Mcw). Biomass was estimated by using the mean and 95% conﬁdence limits (CLs) for mortality at each age, and the 95% conﬁdence interval around each biomass estimate is based on the upper and lower 95% CLs for weight at each age.

The mortality estimates and resulting biomass estimates in population and therefore resulted in biomass estimates that were our study represent two somewhat divergent models of Grass higher than those derived from age-specific mortality estimates Carp population dynamics in Lake Gaston. The constant in- (Mcw). As a result of the differences in biomass estimated from stantaneous mortality estimates (Mˆ p and Mˆ j ) we obtained gen- these methods, we suggest that constant and age-specific mor- erally represented lower overall mortality in the Grass Carp tality estimates represent two potential approaches to assessing Grass Carp in Lake Gaston, depending on how stocking rates are determined in the future. Application of age-specific mortality estimates to the Lake Gaston Grass Carp population results in a lower estimated number of fish per vegetated hectare and a smaller estimated biomass than does the use of constant mor-

Downloaded by [Department Of Fisheries] at 19:24 20 January 2013 tality rates. Because the biomass estimated from age-specific mortality rates is lower, the absolute difference between current biomass and some target level of biomass would be smaller— and thus the estimated addition of biomass required for stocking would be lower—when age-specific mortality rates are used instead of a constant rate to estimate biomass. When the risk of overshooting the target hydrilla coverage is considered, the use of age-specific mortality in stocking models that are based on biomass therefore represents a more conservative approach to Grass Carp stock assessment than does the use of a constant mortality rate. Relative to the use of a constant mortality rate, the use of age-specific mortality in stock assessment and stocking mod- FIGURE 4. Grass Carp biomass in Lake Gaston, as projected by using five mortality estimates (1998–2002) developed for Grass Carp populations in the els is less likely to result in overshooting the target coverage of Santee–Cooper Reservoir system, South Carolina, via catch-curve analysis by hydrilla but is more likely to result in failure to achieve adequate Kirk and Socha (2003). control of hydrilla. Our results demonstrate that the approach to 22 STICH ET AL.

TABLE 5. Model selection statistics for linear regressions characterizing the relationship between Grass Carp biomass in year i and hydrilla coverage in Lake Gaston based on various time lag scenarios (i.e., coverage in years i, i + 1, . . . , i + 5). Biomass was estimated from (1) constant mortality (across all ages) derived by the method of Pauly (1980; Mˆ p), (2) constant mortality derived by the method of Jensen (1996; Mˆ j ), or (3) age-speciﬁc mortality derived by the method of Chen and Watanabe (1989; Mcw). Model selection statistics include the number of parameters estimated (k), Akaike’s information criterion corrected for small sample sizes (AICc), the difference between the AICc value for the given model i and the best model (i), and Akaike weight (wi ), which describes the relative probability that the given model is the best among the models considered. Models were ranked separately for each mortality estimate.

Mortality Lag time estimate (years) k AICc i wi

Mcw 0 2 186.8 19.1 0.00 1 2 187.8 20.1 0.00 2 2 185.3 17.6 0.00 3 2 181.2 13.5 0.00 4 2 167.7 0.0 1.00 5 2 180.8 13.1 0.00

Mp 0 2 193.2 13.8 0.00 1 2 193.2 13.8 0.00 2 2 191.6 12.2 0.00 3 2 188.1 8.7 0.01 4 2 179.4 0.0 0.96 5 2 186.8 7.4 0.02

Mj 0 2 198.7 19.3 0.00 1 2 198.7 19.3 0.00 2 2 197.1 17.7 0.00 3 2 193.7 14.3 0.02 4 2 186.0 6.6 0.92 5 2 191.7 12.3 0.05 FIGURE 5. Proportion of the Grass Carp population size accounted for by individuals older than age 10 (upper panel) and the proportion of Grass Carp population biomass contributed by individuals older than age 10 (lower panel) in Lake Gaston. Population size and biomass were estimated by using mortality Grass Carp stock assessment (i.e., the mortality estimator used) derived from the methods of Pauly (1980; Mˆ p) and Jensen (1996; Mˆ j ) across is dependent upon the specific weed control objectives. all ages or by using age-specific mortality derived from the method of Chen and Our results indicate that Grass Carp biomass is a more ap- Watanabe (1989; Mcw). propriate index of fish density than Grass Carp abundance. We failed to detect a relationship between Grass Carp abundance Downloaded by [Department Of Fisheries] at 19:24 20 January 2013 and annual hydrilla coverage in Lake Gaston under any of the to mass, whereas the standard metabolic rate and energy per time lag scenarios. We did, however, observe strong negative re- gram of wet weight are positively related to mass (Wiley and lationships between Grass Carp biomass and hydrilla coverage Wike 1986). These factors cause an increase in the energy re- under all of the time lag scenarios. The biomass–hydrilla cover- quired per unit of mass gained by Grass Carp as they grow age relationship was strongest when we considered a 4-year lag larger; greater energetic requirements necessitate increased hy- between Grass Carp stocking and observed effects on hydrilla drilla consumption by the fish in Lake Gaston. Growth of Grass coverage. Because the 4-year lag scenario was the best model Carp has been observed to decrease after age 4 in native systems regardless of whether biomass was estimated from age-specific (Gorbach 1961). However, growth (in mass) of Grass Carp in mortality or from a single mortality rate, we suspect that the lag Lake Gaston remained approximately linear with age after age is not a result of age-specific changes in mortality. As all of the 4, thus amplifying the increase in energy needed for growth after lag scenarios provided a good fit to the data, we speculate that that age. Therefore, the 4-year lag between Grass Carp stocking the importance of a 4-year lag is not based simply on the Grass and observable effects on hydrilla coverage likely reflects an Carp population attaining a threshold biomass at which hydrilla increase in hydrilla consumption due to the greater energetic reduction is achieved. Bioenergetics studies of Grass Carp have demands of Grass Carp for maintenance of linear growth in demonstrated that the feed assimilation rate is inversely related mass after they reach age 4. GROWTH OF GRASS CARP STOCKED FOR HYDRILLA CONTROL 23

TABLE 6. Model selection statistics for multiple linear regression models used to determine the relative effects of Grass Carp biomass at different ages on the hydrilla coverage in Lake Gaston. Biomass was estimated from (1) constant mortality (across all ages) derived by the method of Pauly (1980; Mˆ p), (2) constant mortality derived by the method of Jensen (1996; Mˆ j ), or (3) age-specific mortality derived by the method of Chen and Watanabe (1989; Mcw). Models are defined in Table 1; model selection statistics are defined in Table 5. Models were ranked separately for each mortality estimate.

2 Mortality estimate Model kR AICc i wi

Mcw Hydrilla (B4) 2 0.00 209.4 18.9 0.000 Hydrilla (B4 + B8) 3 0.67 198.1 7.6 0.022 Hydrilla (B4 + B8 + B12) 4 0.67 202.9 12.4 0.002 Hydrilla (B4 + B8 + B12 + B16) 5 0.92 190.5 0.0 0.976

Mp Hydrilla (B4) 2 0.01 209.4 19.2 0.000 Hydrilla (B4 + B8) 3 0.63 199.7 10.7 0.005 Hydrilla (B4 + B8 + B12) 4 0.63 204.7 15.7 0.000 Hydrilla (B4 + B8 + B12 + B16) 5 0.92 189.9 0.0 0.995

Mj Hydrilla (B4) 2 0.01 209.4 19.2 0.000 Hydrilla (B4 + B8) 3 0.59 200.9 10.7 0.005 Hydrilla (B4 + B8 + B12) 4 0.59 205.9 15.7 0.000 Hydrilla (B4 + B8 + B12 + B16) 5 0.92 190.2 0.0 0.995

The lag between Grass Carp stocking and observed effects subject to confounding factors like herbicide application, we on hydrilla coverage is likely different than the lags present cannot speculate whether the relationship observed in our study in other systems, such as Lake Austin and Lake Conroe in is necessarily transferable to other water bodies. Texas (Chilton and Magnelia 2008; Chilton et al. 2008) and Although Grass Carp in Lake Gaston appear to have reduced the Santee–Cooper Reservoir system (Kirk et al. 2000; Kirk and the amount of hydrilla to below peak coverage, the management Socha 2003), because stocking densities are much lower in Lake objective of 120 ha by 2012 was not met. The problem of reach- Gaston than in these other systems. For example, Lake Conroe ing a target level of hydrilla without overshooting it suggests that (8,347 ha; Chilton et al. 2008) is similar in size to Lake Gaston there are specific situations in which an aggressive approach to (8,423 ha) but received 371,766 diploid and triploid Grass Carp weed control at Lake Gaston may give way to a more conser- from 1982 to 2007 (Chilton et al. 2008), whereas only 92,959 vative model and vice versa. It is difficult to predict the time triploid Grass Carp were stocked in Lake Gaston from 1995 period (if any) over which the target level of hydrilla coverage to 2011. The Santee–Cooper Reservoir system (70,000 ha) is will be reached at the current stocking rate for Grass Carp in considerably larger than Lake Gaston, and it received the single Lake Gaston; it is possible that the stocking rate will have to be largest Grass Carp stocking (786,500 fish) in history (Kirk et al. increased in order to reach the target level of coverage. 2000). Stocking models that are more refined than the monitoring We recognize that the Grass Carp biomass–hydrilla coverage tools we present here are available; such models are useful relationship reported here is singular and correlative in nature for estimating the numbers of Grass Carp needed to control a and that it cannot be used to infer a cause-and-effect relation given coverage of hydrilla over time (e.g., Stewart and Boyd (i.e., because this is an observational study). However, given the 1999), but these models apply only to the first 10 years after Downloaded by [Department Of Fisheries] at 19:24 20 January 2013 dependency of Grass Carp on hydrilla as their primary energy stocking. Based on the bioenergetics of Grass Carp (Wiley and source, the strength of the observed relationship (R2 > 0.90), and Wike 1986) and the linear patterns of weight gain observed in the expansive time series used, we believe that this relationship Lake Gaston and other systems (e.g., Morrow et al. 1997), it is is a useful means of monitoring the progress of weed control likely that Grass Carp older than age 10 also make important efforts in Lake Gaston. In fact, the regression relationship based contributions to weed control. In Lake Gaston, fish exceeding on Grass Carp standing biomass was useful for predicting the age 10 made substantial contributions to the total biomass—but Lake Gaston hydrilla coverage within 40 ha of the coverage not the number—of Grass Carp in the system. The first year that was estimated with sonar surveys conducted in 2010 and in which Grass Carp older than age 10 were present in Lake 2011 (D. S. Stich, unpublished data). The unexplained portion Gaston was 2005. After 2005, Grass Carp that were older than of variation in the biomass–hydrilla coverage relationship is age 10 accounted for more than 22% of total estimated standing likely due to environmental variation during the 16 years repre- biomass in each year, regardless of the mortality estimate used; sented. We also recognize the potentially confounding effects of in some years and under some mortality scenarios, fish older the low-level herbicide application that occurred in conjunction than age 10 accounted for nearly 50% of the total estimated with Grass Carp stocking in Lake Gaston. Because the rela- population biomass. In most years, the older fish contributed tionship between Grass Carp biomass and hydrilla coverage is less than 10% and as little as 1% (depending on the model used) 24 STICH ET AL.

to the total number of Grass Carp in the system. Since weight Chilton, E. W., II, and M. I. Muoneke. 1992. Biology and management of gain of Grass Carp in Lake Gaston is approximately linear with Grass Carp (Ctenopharyngodon idella, Cyprinidae) for vegetation control: a age, there is reason to believe that fish up to at least age 16 North American perspective. Reviews in Fish Biology and Fisheries 2:283– 320. continue to provide some control of aquatic weeds and should Chilton, E. W., II, M. A. Webb, and R. A. Ott Jr. 2008. Hydrilla management therefore be considered in stocking models. Other research has in Lake Conroe, Texas: a case history. Pages 247–257 in M. S. Allen, S. speculated that large Grass Carp (up to 10 kg) may be just as Sammons, and M. J. Maceina, editors. Balancing fisheries management and effective for weed control as small fish (Osborne and Riddle water uses for impounded river systems. American Fisheries Society, Sym- 1999). Our regression models of hydrilla coverage response to posium 62, Bethesda, Maryland. Cox, W. E. 2007. North Carolina–Virginia conflict: the Lake Gaston wa- various age-groups confirm this for fish up to at least age 16; the ter transfer. Journal of Water Resources Planning and Management 133: strongest model of hydrilla coverage was based on Grass Carp 456–461. biomass estimates that included fish up to 16 years of age. The DeVries, D. R., and R. V. Frie. 1996. Determination of age and growth. Pages contribution of Grass Carp to weed control through age 16 in 483–512 in B. R. Murphy and D. W. Willis, editors. Fisheries techniques, 2nd Lake Gaston highlights the need for all ages to be included in edition. American Fisheries Society, Bethesda, Maryland. Dodd-Williams, L., G. O. Dick, R. M. Smart, and C. S. Owens. 2008. Point stocking models and bioenergetics models, especially as IPM intercept and surface observation GPS (SOG): a comparison of survey becomes increasingly prevalent and as management of long- methods—Lake Gaston, NC/VA.U.S. Army Engineer Research and Develop- lived Grass Carp becomes more commonplace. ment Center, ERDC/TN APCRP-EA-19, Vicksburg, Mississippi. Available: el.erdc.usace.army.mil/elpubs/pdf/apcea-19.pdf. (July 2009). Dominion Power. 2010. Shoreline management plan for the Roanoke Rapids and Gaston hydropower project. Dominion Virginia Power and ACKNOWLEDGMENTS Dominion North Carolina Power, Federal Energy Regulatory Commis- We thank the volunteer bowfishers for Grass Carp collection, sion Project 2009, Richmond, Virginia. Available: www.dom.com/about/ and we are grateful to Cory Kovacs and Catherine Lim (Virginia stations/hydro/pdf/shoreline plan.pdf. (July 2012). Department of Game and Inland Fisheries) for otolith prepara- Frishman, F. 1975. On the arithmetic means and variances of products and ratios of random variables. Pages 401–406 in G. P. Patil, S. Kotz, and J. K. tion and age determination during 2006–2008. Phil Kirk, two Ord, editors. Statistical distributions in scientific work, volume I: models and anonymous reviewers, and the editors provided comments that structures. D. Reidel, Dordrecht, The Netherlands. greatly improved this manuscript. Funding for the project was Gasaway, R. D. 1978. Growth, survival and harvest of Grass Carp in Florida provided by the LGWCC and the Acorn Alcinda Foundation. lakes. Pages 167–188 in R. O. Smitherman, W. L. Shelton, and J. H. Grover, editors. Symposium on culture of exotic fishes. American Fisheries Society, Fish Culture Section, Auburn University, Auburn, Alabama. Gorbach, E. I. 1961. Age composition, growth, and age of onset of sexual REFERENCES maturity of the White Amur Ctenopharyngodon idella (Val.) and the Black Allen, S. K., Jr., and R. J. Wattendorf. 1987. Triploid Grass Carp: status and Amur Mylophryngodon piceus (Rich.) in the Amur River basin. Voprosy management implications. Fisheries 12(4):20–24. Ikhtiologii 1:119–126. Anderson, R. O., and R. M. Neumann. 1996. Length, weight, and associated Grodowitz, M., J. Nachtrieb, N. Harms, and J. Freedman. 2010. Suitability structural indices. Pages 447–482 in B. R. Murphy and D. W. Willis, editors. of using introduced Hydrellia spp. for management of monoecious Hy- Fisheries techniques, 2nd edition. American Fisheries Society, Bethesda, drilla verticillata (L.f.) Royle. U.S. Army Corps of Engineers, Aquatic Maryland. Plant Control Technical Note BC-17, Vicksburg, Mississippi. Available: Baker, G. E., D. L. Sutton, and R. D. Blackburn. 1974. Feeding habits of the el.erdc.usace.army.mil/elpubs/pdf/apcbc-17.pdf. (May 2012). White Amur on waterhyacinth. Hyacinth Control Journal 12:58–62. Hewitt, D. A., and J. M. Hoenig. 2005. Comparison of two approaches for esti- Bettoli, P. W., M. J. Maceina, R. L. Noble, and R. K. Betsill. 1993. Response mating natural mortality based on longevity. U.S. National Marine Fisheries of a reservoir fish community to aquatic vegetation removal. North American Service Fishery Bulletin 103:433–437. Journal of Fisheries Management 13:110–124. Jensen, A. L. 1996. Beverton and Holt life history invariants result from optimal

Downloaded by [Department Of Fisheries] at 19:24 20 January 2013 Bonar, S. A., B. Bolding, and M. Divens. 2002. Effects of triploid Grass Carp trade-off of reproduction and survival. Canadian Journal of Fisheries and on aquatic plants, water quality, and public satisfaction in Washington State. Aquatic Sciences 53:820–822. North American Journal of Fisheries Management 22:96–105. Kilgen, R. H., and R. O. Smitherman. 1971. Food habits of the White Amur Bonar, S. A., G. L. Thomas, S. L. Thiesfeld, G. B. Pauley, and T. B. Stables. stocked in ponds alone and in combination with other species. Progressive 1993. Effect of triploid Grass Carp on the aquatic macrophyte community Fish-Culturist 33:123–127. of Devils Lake, Oregon. North American Journal of Fisheries Management Killgore, K. J., J. P. Kirk, and J. W. Foltz. 1998. Response of littoral fishes in 13:757–765. upper Lake Marion, South Carolina, following hydrilla control by triploid Burnham, K. P., and D. R. Anderson. 2002. Model selection and multimodel Grass Carp. Journal of Aquatic Plant Management 36:82–87. inference: a practical information-theoretic approach, 2nd edition. Springer, Kirk, J. P. 1992. Efficacy of triploid Grass Carp in controlling nuisance aquatic New York. vegetation in South Carolina farm ponds. North American Journal of Fisheries Chen, S., and S. Watanabe. 1989. Age dependence of natural mortality co- Management 12:581–584. efficient in fish population dynamics. Bulletin of the Japanese Society of Kirk, J. P., and J. E. Henderson. 2006. Management of hydrilla in the Santee– Scientific Fisheries 55:205–208. Cooper reservoirs, South Carolina: experiences from 1982 to 2004. Journal Chilton, E. W., II, and S. J. Magnelia. 2008. Use of an incremental triploid Grass of Aquatic Plant Management 44:98–103. Carp stocking strategy for maintaining vegetation coverage in a riverine Texas Kirk, J. P.,J. V.Morrow Jr., K. J. Killgore, S. J. De Kozlowski, and J. W. Preacher. reservoir. Pages 543–555 in M. S. Allen, S. Sammons, and M. J. Maceina, 2000. Population response of triploid Grass Carp to declining levels of hydrilla editors. Balancing fisheries management and water uses for impounded river in the Santee–Cooper reservoirs, South Carolina. Journal of Aquatic Plant systems. American Fisheries Society, Symposium 62, Bethesda, Maryland. Management 38:14–17. GROWTH OF GRASS CARP STOCKED FOR HYDRILLA CONTROL 25

Kirk, J. P., and R. C. Socha. 2003. Longevity and persistence of triploid Grass Richardson, R. J. 2008. Aquatic plant management and the impact of emerging Carp stocked into the Santee–Cooper reservoirs of South Carolina. Journal herbicide resistance issues. Weed Technology 22:8–15. of Aquatic Plant Management 41:90–92. Ricker, W. E. 1975. Computation and interpretation of biological statis- LGSB (Lake Gaston Stakeholder’s Board). 2005. Managing aquatic plants in tics of fish populations. Fisheries Research Board of Canada Bulletin Lake Gaston: a long-term action plan. LGSB, Littleton, North Carolina. Avail- 191. able: www.lgaston.org/Stakeholders/Plan.pdf. (January 2012). Ryan, F. J., C. R. Coley, and S. H. Kay. 1995. Coexistence of monoecious and Mitchell, A. J., and A. M. Kelly. 2006. The public sector role in the establishment dioecious hydrilla in Lake Gaston, North Carolina and Virginia. Journal of of Grass Carp in the United States. Fisheries 31:113–121. Aquatic Plant Management 33:8–12. Montgomery, D. C., E. A. Peck, and G. G. Vining. 2006. Introduction to linear Schramm, H. L., Jr., and M. W. Brice. 2000. Use of triploid Grass Carp to reduce regression analysis, 4th edition. Wiley, Hoboken, New Jersey. aquatic macrophyte abundance in recreational fishing ponds. Proceedings Morrow, J. V., Jr., and J. P. Kirk. 1997. Age and growth of Grass Carp in Lake of the Annual Conference Southeastern Association of Fish and Wildlife Guntersville, Alabama. Proceedings of the Annual Conference Southeastern Agencies 52(1998):93–103. Association of Fish and Wildlife Agencies 49(1995):187–194. Slipke, J. W., and M. J. Maceina. 2001. Fisheries analysis and simulation tools Morrow, J. V., Jr., J. P. Kirk, and K. J. Killgore. 1997. Collection, age, growth, (FAST), version 2.0. Auburn University, Auburn, Alabama. and population attributes of triploid Grass Carp stocked into the Santee– Stewart, R. M., and W. A. Boyd. 1999. The Grass Carp stocking Cooper reservoirs, South Carolina. North American Journal of Fisheries rate model (AMUR/STOCK). U.S. Army Corps of Engineers, Aquatic Management 17:38–43. Plant Control Technical Note MI-03, Vicksburg, Mississippi. Available: Osborne, J. A., and R. D. Riddle. 1999. Feeding and growth rates for triploid el.erdc.usace.army.mil/elpubs/pdf/apcmi-03.pdf. (April 2010). Grass Carp as influenced by size and water temperature. Journal of Freshwater Stott, B., and T. O. Robson. 1970. Efficiency of Grass Carp (Ctenopharyn- Ecology 14:41–45. godon idella Val.) in controlling submerged water weeds. Nature 266: Osborne, J. A., and N. M. Sassic. 1981. The size of Grass Carp as a factor in 870. the control of hydrilla. Aquatic Botany 11:129–136. SYSTAT. 2007. SYSTAT: version 12. SYSTAT Software, Chicago. Pauly, D. 1980. On the interrelationships between natural mortality, growth Thorson, J. T., and M. H. Prager. 2011. Better catch curves: incorporating parameters, and mean environmental temperature in 175 fish stocks. Journal age-specific natural mortality and logistic selectivity. Transactions of the du Conseil, Conseil International pour l’Exploration de la Mer 39:175–192. American Fisheries Society 140:356–366. R Development Core Team. 2011. R: a language and environment for statistical von Bertalanffy, L. 1938. A quantitative theory of organic growth. Human computing. R Foundation for Statistical Computing, Vienna. Biology 10:181–213. ReMetrix. 2011. Lake Gaston 2010 submerged vegetation mapping summary re- Wiley, M. J., and L. D. Wike. 1986. Energy balances of diploid, triploid, and port: hydroacoustic sampling with species point sampling. ReMetrix, Carmel, hybrid Grass Carp. Transactions of the American Fisheries Society 115: Indiana. 853–863. Downloaded by [Department Of Fisheries] at 19:24 20 January 2013 This article was downloaded by: [Department Of Fisheries] On: 20 January 2013, At: 19:25 Publisher: Taylor & Francis Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

North American Journal of Fisheries Management Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/ujfm20 Retention of Large-Format, Soft Visible Implant Alphanumeric Tags in Walleye Jonathan R. Meerbeek a , Joseph G. Larscheid a , Michael J. Hawkins a & George D. Scholten a

a Iowa Department of Natural Resources, 122 252nd Avenue, Spirit Lake, Iowa, 51360, USA Version of record first published: 09 Jan 2013.

To cite this article: Jonathan R. Meerbeek , Joseph G. Larscheid , Michael J. Hawkins & George D. Scholten (2013): Retention of Large-Format, Soft Visible Implant Alphanumeric Tags in Walleye, North American Journal of Fisheries Management, 33:1, 26-31 To link to this article: http://dx.doi.org/10.1080/02755947.2012.739984

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material. North American Journal of Fisheries Management 33:26–31, 2013 C American Fisheries Society 2013 ISSN: 0275-5947 print / 1548-8675 online DOI: 10.1080/02755947.2012.739984

MANAGEMENT BRIEF

Retention of Large-Format, Soft Visible Implant Alphanumeric Tags in Walleye

Jonathan R. Meerbeek,* Joseph G. Larscheid, Michael J. Hawkins, and George D. Scholten Iowa Department of Natural Resources, 122 252nd Avenue, Spirit Lake, Iowa 51360, USA

is dependent on tag type, attachment location, the species being Abstract tagged, and fish size (Guy et al. 1996). High tag retention is We evaluated the effects of fish length, fish sex, and number of a fundamental assumption of most tagging studies; therefore, days posttagging on retention of large-format, soft visible implant it is important to estimate tag retention so that estimates can (VI) alphanumeric tags that were injected underneath the clear tissue on the lower mandible of Walleyes Sander vitreus. We also be adjusted for potential bias resulting from unaccounted tag evaluated whether the direction of insertion or the application of loss (Kallemeyn 1989). If tag retention is poor, then fisheries surgical-grade tissue adhesive to the tag incision site would affect managers cannot accurately develop the necessary data to effec- tag retention. Adult Walleyes were collected with gill nets from tively manage a fishery (Ricker 1975), such as fish abundance natural lakes in Iowa during spring and then were transported or stocking densities. to a hatchery, where they were measured, sexed, and tagged. One worker injected 752 Walleyes (mean TL = 21.8 in; SE = 0.16) The visible implant (VI) tag (Northwest Marine Technology with two identical VI tags; each side (left and right) of the lower [NMT], Inc., Shaw Island, Washington) is an alphanumerically mandible received one tag. Incisions were dried with a cloth, and coded tag that can be inserted beneath clear tissue and can be tissue adhesive was applied to one of the two tag injection sites. read when fish are recaptured (Haw et al. 1990). Visible implant Walleyes were released back into the lake and were recaptured tags are easy to apply, are easily visible in transparent tissue, with gill nets and by anglers. Of the 129 Walleyes recaptured up to $ 5 years posttagging, 80 fish (62%) had retained both tags and the are relatively inexpensive (US 0.92 per tag; average cost per remaining 49 fish had retained one of the tags. Retention adjusted tag during project duration), do not require the sacrifice of the for fish that lost both tags (n = 8; probability = 0.09) was 58% (80 of fish to retrieve tag codes, provide individual fish data, and have 137). Tag retention was significantly related to fish size at the time of little impact on fish growth, condition, or survival (Haw et al. tagging, as smaller fish lost more tags. Consequently, males (mean 1990; Bryan and Ney 1994; Zerrenner et al. 1997). Three types TL = 20.5 in; SE = 0.39) were more likely to lose tags than females (mean TL = 24.3 in; SE = 0.26). Insertion direction, adhesive of VI tags have been developed by NMT for fish identification. application, or the number of days posttagging at recapture did not The small-format (0.098 × 0.039 × 0.004 in), rigid VI tag influence VI tag retention. We recommend that in studies requiring was initially introduced in the late 1980s and was made of Downloaded by [Department Of Fisheries] at 19:25 20 January 2013 high tag retention in Walleyes, the injection of large-format, soft polyester film (Haw et al. 1990). By 1998, NMT had stopped VI tags into the clear tissue underneath the mandible should not production of the small-format, rigid VI tag. In the early 1990s, be considered. NMT developed large-format (0.138 × 0.059 × 0.004 in) and small-format (0.098 × 0.039 × 0.004 in), soft VI tags Fisheries researchers and managers commonly mark fish with consisting of fluorescent elastomer that made the tags easier tags to obtain estimates of population abundance, exploitation to locate and read. In 2010, NMT introduced a new rigid VI rates, population dynamics, and fish movements and migrations alpha tag (0.106 × 0.047 × 0.004 in) that was easier to inject (Hilborn et al. 1990; Guy et al. 1996). Basic considerations for and load than the large- or small-format, soft VI tags and was the use of tags to mark fish include tagging costs, ease of ap- available in four fluorescent colors (Geraldine Vander Haegen, plication, tag retention, tag readability, and tag effects on fish NMT, personal communication). growth, survival, and behavior (Nielsen 1992). The most recog- Several studies have examined retention rates for VI tags nized and well-studied problem is tag retention. Tag retention injected into the clear adipose postorbital tissue of salmonids

*Corresponding author: [email protected] Received February 16, 2012; accepted October 3, 2012

26 MANAGEMENT BRIEF 27

(Bryan and Ney 1994; McMahon et al. 1996; Shepard et al. tissue and the transparent tissue between fin rays). We felt that 1996) or into the transparent tissue between fin rays in Cut- the lower mandible of Walleyes offered an alternative location throat Trout Oncorhynchus clarkii (Wenburg and George 1995) with the potential to provide optimal tag readability and ease and Channel Catfish Ictalurus punctatus (Buckmeier and Irwin of application in this species. Many studies using various tag 2000). Retention of small-format, rigid VI tags after 60 d post- types have been conducted to estimate important Walleye pop- tagging has been highly variable in both laboratory and field ulation statistics (abundance, exploitation, and survival); some evaluations. Mourning et al. (1994) and Zerrenner et al. (1997) have experienced high tag loss rates (24–50%), which would observed retention rates of 64% and 75%, respectively, for cause substantial under- or overestimation of population statis- small-format, rigid VI tags that were injected into postorbital tics if models are not adjusted for tag loss (Kallemeyn 1989; tissue of Rainbow Trout O. mykiss and Brook Trout Salveli- Isermann and Knight 2005). Therefore, for population statistics nus fontinalis, which were held in the laboratory for 120–251 to be meaningful, tag retention must be high (>90%) or the rates d. Field observations have also resulted in similar retention of tag loss must be well known so that adjustments can be made rates (48–65%) for small-format, rigid VI tags injected at the to population models. same location in salmonids that were recaptured 61–365 d later The Iowa Great Lakes, a chain of lakes in the northwest- (Bryan and Ney 1994; Farooqi et al. 1995; McMahon et al. 1996; ern portion of the state, support an important Walleye fishery Shepard et al. 1996). Blankenship and Tipping (1993) recorded that serves as a broodstock source for Walleye propagation and the highest retention rate after 60 d posttagging (94%) for small- stocking across Iowa. An understanding of Walleye population format, rigid VI tags that were inserted into the postorbital tissue dynamics and factors affecting survival and recruitment in these of Cutthroat Trout. Wenburg and George (1995) also found im- populations is vital to the success of Iowa’s Walleye stocking proved retention (97–98%; 4–85 d) of small-format, rigid tags program. Fisheries researchers began tagging adult Walleyes that were injected into clear tissue between anal rays of Cut- with large-format VI tags in 1990 to estimate population statis- throat Trout in both laboratory and field experiments; however, tics. By 1996, researchers had noted that nearly 20% of Walleyes Buckmeier and Irwin (2000) found 0% retention for tags in- had not retained the large-format VI tags, as evidenced by a sec- jected into the clear tissue between dorsal fin rays of Channel ondary fin clip. In subsequent years, the rate of tag loss increased Catfish. Small-format, soft VI tags have had similar retention to 50%. The objectives of this study were to evaluate retention results for salmonids in both laboratory (67%) and field (67– of large-format, soft VI alphanumeric tags injected underneath 96%) studies (Rikardsen 2000; Olsen et al. 2004). In addition the clear tissue on the lower mandible of Walleyes and to de- to variable tag retention, another problem with small-format VI termine the effects of fish size, fish sex, and number of days tags is that the tags can be difficult to read in long-term studies posttagging (i.e., at recapture) on tag retention. We also evalu- (Zerrenner et al. 1997; Hughes et al. 2000). Large-format, soft ated whether the direction of insertion would affect tag retention VI tags were developed to improve tag readability. Few stud- and whether the application of surgical-grade tissue adhesive to ies have evaluated the retention of large-format VI tags since the tag incision site would improve tag retention. their introduction, but the existing studies have indicated that tag retention (80–95%) and readability after 100 d posttagging were improved (Kincaid and Calkins 1992; Frenette and Bryant METHODS 1996; Hughes et al. 2000). Hughes et al. (2000) compared the Walleyes were collected with gill nets (6 × 320 ft; 2.5-in retention rates of small-format (rigid) and large-format VI tags bar mesh) from three natural lakes in northwest Iowa during and found that large-format VI tags had much better reten- annual broodstock collections in April 2005 and April 2006. tion among larger size-groups of Brook Trout (≥15.7 in) than After capture, Walleyes were transported to the Spirit Lake Fish Downloaded by [Department Of Fisheries] at 19:25 20 January 2013 among smaller size-groups (<15.7 in). Hughes et al. (2000) rec- Hatchery and were held in indoor raceways. At the hatchery, ommended that both tag types be used in studies requiring the Walleyes were measured (TL, in) and sexed (determined by tagging of multiple size-classes. Other researchers have found extrusion of gametes), and two identical large-format, soft VI that retention is generally poor (<60%) for small fish (<8.0 in), alphanumeric tags (0.138 × 0.059 × 0.004 in; NMT) were regardless of the VI tag type used (e.g., small format or large injected just below the soft, clear tissue on the underside of the format; rigid or soft; Kincaid and Calkins 1992; Bryan and Ney lower mandible by using a needle and a single-shot syringe (Haw 1994; Mourning et al. 1994; McMahon et al. 1996; Shepard et al. 1990; Bergman et al. 1992; McMahon et al. 1996). Each et al. 1996; Rikardsen 2000). side (left and right) of the lower mandible received one tag. Both To our knowledge, no study has examined VI tag retention tags were inserted in the same direction for a given fish (i.e., (large format or small format) in Walleyes Sander vitreus or anterior to posterior [front insertion]; or posterior to anterior evaluated the retention of VI tags when injected underneath the [rear insertion]). Nexaband surgical-grade tissue adhesive was clear tissue of the fish’s lower mandible. We examined the lower applied to the dried needle incision area on one of the two mandible as a VI tag injection site in Walleyes because we were inserted tags (Figure 1). We alternated tagging methods (front concerned about tag readability in other, more traditionally used insertion or rear insertion; adhesive on the left or right side) to VI tag locations with darkly pigmented tissue (e.g., postorbital reduce experimental error. All tagging was performed by one 28 MEERBEEK ET AL.

and

Rˆ = c(RA + RB + RAB),

where k and c represent the number of recaptured Walleyes, RA is the number of recaptured Walleyes with only a left VI tag, RB is the number of recaptured Walleyes with only a right VI tag, RAB is the number of recaptured Walleyes with both tags, and Rˆ is the estimated total number of recaptures. The probability of losing a VI tag between tagging and recapture events was deﬁned as

RX , (RX + RAB)

where RX is the number of Walleyes with only the left VI tag or the right VI tag (i.e., the number of Walleyes estimated to have lost one tag). Insertion direction, application of tissue adhesive, number of days posttagging, fish sex, and fish TL at tagging (independent variables) were evaluated for their effects on VI tag retention (dependent variable) by using logistic regression in the Statis- tical Analysis System (α = 0.05; SAS 1999). We used t-tests (α = 0.05; SAS 1999) to compare mean TL at tagging (1) between male and female Walleyes; and (2) between Walleyes that retained both VI tags and those that retained only one tag. Ap- proximate t-tests and Satterthwaite’s approximation for df (SAS 1999) were used if the folded form of the F-statistic indicated heterogeneous variances (P < 0.05). FIGURE 1. Illustration of the technique used to insert large-format, soft visible implant alphanumeric tags into Walleyes collected from natural lakes in Iowa during April 2005 and April 2006. Note that the insertion direction (rear or front) RESULTS was consistent for both tags used in a given fish, and tissue adhesive (Nexaband) Visible implant tags were applied to 752 Walleyes (mean was applied to one of the two tag incision areas. Tagging methods (i.e., rear = = insertion or front insertion; left- or right-side application of tissue adhesive) TL 21.8 in; SE 0.16) during April 2005 and April 2006. were alternated to reduce experimental error. [Figure available in color online.] Females (N = 504; mean TL = 24.0 in; SE = 0.13) were significantly larger than males (N = 248; mean TL = 17.3 in; SE = 0.23; t =−25.14, df = 419, P < 0.01). Adhesive was applied to 413 (50%) of the front-inserted VI tags and 331 (49%) of individual who was experienced in VI tagging. Walleyes were the rear-inserted VI tags (Table 1). We recaptured 129 Walleyes Downloaded by [Department Of Fisheries] at 19:25 20 January 2013 released back to the lake of capture within 3 h of tagging. up to 5 years after tagging (2–1,834 d posttagging). Twenty- Walleyes were recaptured during annual gillnetting surveys two Walleyes were recaptured within the first year (2–153 d) and occasionally by angler creel returns. Recaptured Walleyes posttagging, 84 fish were recaptured 1–2 years (352–744 d) were examined for both tags, measured, and released (unless posttagging, and 23 fish (6–10 fish/year) were recaptured 3– harvested by an angler). For instances in which a tag was un- 5 years (1,091–1,834 d) posttagging. Collectively, 80 fish (62%) readable, a flashlight with an 8-W Xenon lamp was used to had retained both VI tags, and the remaining 49 fish had retained illuminate the tags or to interpret alphanumeric codes (Crook either the left tag or the right tag (Table 2). All tags were readable and White 1995). by the naked eye or by using an 8-W Xenon lamp. Ninety-one The total number of recaptured Walleyes (the number es- fish were recaptured only once, whereas 38 fish were recaptured timated to have lost both tags plus the number of observed two or more times. Of the 38 fish that were recaptured more recaptures) was estimated via the method of Krebs (1999): than once, 21 had both tags at the first recapture event and RA RB 19 (90%) retained both tags until their second recapture event k = , (RA + RAB)(RB + RAB) (218–1,831 d between recapture events). 1 The probability that a Walleye would lose the left VI tag c = , 1 − k (0.24) was similar to the probability of losing the right VI MANAGEMENT BRIEF 29

TABLE 1. Number (N) of large-format, soft visible implant alphanumeric tags injected into the lower mandible of Walleyes collected from natural lakes in Iowa during April 2005 and April 2006. Tags were injected by using a front or rear insertion technique with or without the application of surgical-grade adhesive to the incision site. Mean TL at the time of tagging (SE in parentheses) and the number of VI tags that were retained by or lost from the 129 Walleyes recaptured upto 5 years posttagging are also provided for each method.

Insertion Incision N tags N tags method closure method N tags Mean TL retained lost Front No adhesive 414 21.7 (0.22) 49 16 Adhesive 413 21.7 (0.22) 58 11 Rear No adhesive 346 21.9 (0.25) 52 11 Adhesive 331 21.9 (0.25) 49 12

tag (0.23). Tag loss rates were similar between tags to which tag (ploss)was adhesive was applied (27 of 128) and tags that did not receive adhesive (23 of 130), regardless of tag insertion method e(−3.1519+0.1575·TL) (Table 1). Visible implant tag retention adjusted for fish that lost ploss = 1 − . e(−3.1519+0.1575·TL) + 1 both tags (n = 8; probability = 0.09) was 58% (80 of 137). Tag retention was significantly related to fish TL at the time of tagging (logistic regression: df = 1, P = 0.02), as smaller fish lost more tags than larger fish (t =−2.52, pooled df = 127, DISCUSSION P = 0.01). Consequently, due to the size difference between To our knowledge, this is the first study that has evaluated genders, males (mean TL = 20.5 in; SE = 0.39) were more VI tag retention in the clear tissue under the lower mandible likely to lose tags than females (mean TL = 24.3 in; SE = 0.26; of Walleyes. We observed poor VI tag retention (58%) on the logistic regression: df = 1, P < 0.001). The retention of VI tags lower mandible for this species. The most common application was not influenced by insertion direction, application of adhe- of VI tags has involved tag injection into the postocular tissue sive, or the number of days posttagging at recapture (logistic of salmonids; however, retention has been variable for both the regression: df = 1, P ≥ 0.39). Therefore, the best model for small- and large-format VI tags (range of tag retention = 48– estimating the probability that an adult Walleye would lose a VI 98%; Farooqi et al. 1995; McMahon et al. 1996; Shepard et al.

TABLE 2. Number (N) of male and female Walleyes recaptured up to 5 years posttagging that retained one or both large-format, soft visible implant alphanumeric tags that were injected (via the front or rear insertion method) into the left and right sides of the lower mandible. Mean TL, SE, minimum (Min) TL, and maximum (Max) TL at the time of tagging are presented for each group.

Insertion method Location tag retained N Mean TL SE Min TL Max TL Males Front 17 19.9 0.70 13.2 23.4 Both 5 19.6 1.52 14.6 23.0 Left 5 19.8 1.09 16.7 23.4 Downloaded by [Department Of Fisheries] at 19:25 20 January 2013 Right 7 20.2 1.21 13.2 23.1 Rear 19 21.1 0.37 17.8 24.2 Both 8 21.3 0.55 18.8 23.4 Left 6 21.7 0.65 19.8 24.2 Right 5 20.3 0.80 17.8 22.6 Females Front 50 24.2 0.37 18.9 29.4 Both 35 24.4 0.44 19.3 29.4 Left 8 23.1 0.83 18.9 25.4 Right 7 24.5 1.07 19.3 27.5 Rear 43 24.4 0.36 19.2 28.3 Both 32 24.4 0.44 19.2 28.3 Left 6 25.7 0.55 24.0 27.6 Right 5 23.6 0.93 20.1 25.7 30 MEERBEEK ET AL.

1996; Hughes et al. 2000; Isely et al. 2004). Other studies eval- the applicator is left or right handed and the characteristics of uating alternative insertion locations have also reported variable needle insertion, can influence VI tag retention (Hughes et al. retention rates (0–98%; Wenburg and George 1995; Buckmeier 2000). Nevertheless, tagging experience and technique are po- and Irwin 2000). tential sources of bias in any tagging project. Therefore, it is We used this relatively inexpensive technique to efficiently important for tagging studies to maintain consistency in tagging tag numerous Walleyes and to maintain high tag readability. operations so that tag retention rates do not substantially dif- However, another goal of the study was to achieve high tag fer among years. It may also be necessary to perform periodic retention so that we could accurately estimate Walleye popula- “checks” of tag retention so that population models can be ad- tion statistics. We became aware of poor tag retention when a justed accordingly. All of the Walleyes in this study were VI high percentage of recaptured fish were observed to have lost tagged by one individual who was experienced in VI tagging. their tags, as evidenced by a secondary mark. With an unknown This may not be the case in most long-term tag retention studies, rate of tag loss, our estimates of recruitment and abundance thus potentially introducing more uncertainties around the tag were assumed to be biased upward. Results of the present study loss estimate. indicate that our estimates were strongly biased. Managers that The advantage of large-format VI tags over small-format use models unadjusted for tag loss must be cautious when in- VI tags is improved readability with limited additional cost interpreting model results to reduce the risk of making erroneous vestment. Tag readability is essential in long-term studies or in management decisions (Kallemeyn 1989; Isermann and Knight studies of long-lived species. Based on our results, we believe 2005). For results to be meaningful, the tag retention rates must that the number of unreadable large-format VI tags in most stud- be estimated or managers should use conservative estimates that ies will be low, thus producing only small effects on population have been reported from previous studies. parameter estimates. However, the ability to read the tags is of Most studies of VI tag retention have found substantially de- decreased value if a high percentage of individuals lose their creased tag retention for smaller fish. Studies evaluating small- tags. In these circumstances, the cost effectiveness of the tag format VI tags have reported that to obtain high tag retention, type used should be revisited. fish length generally should be at least 8.0 in (McMahon et al. We VI tagged 19,677 Walleyes at a cost of $25,693, but with 1996; Shepard et al. 1996). Hughes et al. (2000) observed a a long-term tag loss rate of 42%, we effectively tagged only similar trend of increased tag retention as a function of fish 11,413 fish. To tag the same number of Walleyes with anchor size for large-format, soft VI tags, but the fish size required tags ($1.00 per tag; Floy Tag Company) and assuming a conser- for high tag retention (97%) was nearly doubled (≥15.5 in). vative tag retention rate of 70% (Cowen and Schwarz 2006), the We observed poor tag retention (58%) of large-format VI tags cost would be similar ($25,580) to that of VI tags, but models injected beneath the lower mandible of adult Walleyes, most would still have to be adjusted to account for high tag loss. The of which were large in size. Over 83% of the 752 Walleyes injection of PIT tags into 19,677 fish would incur nearly double we tagged with large-format, soft VI tags were larger than the the cost ($2.38 per tag; $48,211 per 19,667 fish; Oregon Radio recommended minimum size of 15.5 in for trout, yet poor tag Frequency Identification [RFID], LLC), but high PIT tag reten- retention and size limitations were observed. tion rates (>97%; Wagner et al. 2007; Isermann and Carlson The probability that a VI tag will be retained is dependent 2008; Daugherty and Buckmeier 2009) mean that the models upon many factors. Prior to this study, we hypothesized that would not require adjustment. In our situation, the cost to PIT the larger injector needle used for large-format VI tags, tissue tag the 11,413 Walleyes ($28,003; i.e., assuming 97% retention) erosion caused by movement of the tag, and the direction of that were effectively tagged with VI tags would be similar to tag insertion might be factors affecting the loss of large-format the cost of VI tagging ($25,567), and the use of PIT tags would Downloaded by [Department Of Fisheries] at 19:25 20 January 2013 VI tags. We found no evidence to support the insertion direc- allow for more accurate estimation of population statistics. Prior tion hypothesis. However, based on multiple tag returns, we did to initiating any tagging project, the study objectives, including find that tag retention is high (>90%) once the tag has been in cost effectiveness of the tag, must be thoroughly considered to the fish for at least 1 year. The large injection site and move- obtain a full understanding of the pros and cons of each tagging ment of the tag may be responsible for immediate tag loss, but method. we did not specifically test these hypotheses; our attempt to The “no tag loss” assumption in mark–recapture models must reduce tag loss immediately after injection by applying a surgi- be addressed prior to study design to obtain meaningful esti- cal tissue adhesive did not improve the retention of tags. Since mates of fish population statistics. We suggest that in studies for the majority of our recaptures occurred at least 1 year after which tag retention is a critical component, the injection of large- initial tagging, we could not accurately assess short-term tag format, soft VI tags underneath the lower mandible of Walleyes loss. should not be considered. In addition, since the completion of Shepard et al. (1996) demonstrated the need for experienced this project, NMT has substantially reduced its production of personnel to perform the VI tagging; however, Hughes et al. large- and small-format, soft VI tags (due to inefficiencies in (2000) found that applicator experience did not always result in the tag loading process and production), and currently these tag increased tag retention. Simple factors, such as whether or not types are only available via special order purchases (Geraldine MANAGEMENT BRIEF 31

Vander Haegen, NMT, personal communication). A new, editors. Fish-marking techniques. American Fisheries Society, Symposium smaller rigid VI alpha tag was introduced in 2010; it has an 7, Bethesda, Maryland. improved tag loading process, is available in four fluorescent Hilborn, R., C. J. Walters, and D. B. Jester Jr. 1990. Value of fish marking in fisheries management. Pages 5–7 in N. C. Parker, A. E. Giorgi, R. C. colors (each with 2,500 unique codes), and costs less than the Heidinger, D. B. Jester Jr., E. D. Prince, and G. A. Winans, editors. Fish- soft type ($0.75 per VI alpha tag; NMT). The Iowa Depart- marking techniques. American Fisheries Society, Symposium 7, Bethesda, ment of Natural Resources has used the new small-format, rigid Maryland. VI alpha tags injected into the lower mandible of Walleyes for Hughes, T. C., D. C. Josephson, C. C. Krueger, and P. J. Sullivan. 2000. Com- long-term tagging studies since 2010. Retention rates for this parison of large and small visible implant tags: retention and readability in hatchery Brook Trout. North American Journal of Aquaculture 62:273– new tag type are unknown and are the focus of ongoing in- 278. vestigation in Iowa. If retention, readability, or retrieval rates Isely, J. J., D. G. Trested, and T. B. Grabowski. 2004. Tag retention and survivor- are poor, an alternative tag type, such as PIT tags, will be ship of hatchery Rainbow Trout marked with large-format visible implant considered. alphanumeric tags. North American Journal of Aquaculture 66:73–74. Isermann, D. A., and A. J. Carlson. 2008. Initial mortality and retention associated with using passive integrated transponders in Black Crappies. North ACKNOWLEDGMENTS American Journal of Fisheries Management 28:1157–1159. The Federal Aid in Sport Fish Restoration Program provided Isermann, D. A., and C. T. Knight. 2005. Potential effects of jaw tag loss on funding for the project. We thank Ed Thelen for his contributions exploitation estimates for Lake Erie Walleyes. North American Journal of Fisheries Management 25:557–562. to the project, and we are grateful to the crew at the Spirit Lake Kallemeyn, L. W. 1989. Loss of Carlin tags from Walleyes. North American Fish Hatchery for collecting Walleyes. Northwest Marine Tech- Journal of Fisheries Management 9:112–115. nology donated the VI tags that were used for double tagging of Kincaid, H. L., and G. T. Calkins. 1992. Retention of visible implant tags in the Walleyes. Lake Trout and Atlantic Salmon. Progressive Fish-Culturist 54:163–170. Krebs, C. J. 1999. Ecological methodology, 2nd edition. Benjamin/Cummings, Menlo Park, California. REFERENCES McMahon, T. E., S. R. Dalbey, S. C. Ireland, J. P. Magee, and P. A. Byorth. Bergman, P. K., F. Haw, H. L. Blankenship, and R. M. Buckley. 1992. Perspec- 1996. Field evaluation of visible implant tag retention by Brook Trout, Cut- tives on design, use, and misuse of fish tags. Fisheries 17(4):20–25. throat Trout, Rainbow Trout, and Arctic Grayling. North American Journal Blankenship, H. L., and J. M. Tipping. 1993. Evaluation of visible implant of Fisheries Management 16:921–925. and sequentially coded wire tags in sea-run Cutthroat Trout. North American Mourning, T. E., K. D. Fausch, and C. Gowan. 1994. Comparison of visible im- Journal of Fisheries Management 13:391–394. plant tags and Floy anchor tags on hatchery Rainbow Trout. North American Bryan, R. D., and J. J. Ney. 1994. Visible implant tag retention by and effects Journal of Fisheries Management 14:636–642. on condition of a stream population of Brook Trout. North American Journal Nielsen, L. A. 1992. Methods of marking fish and shellfish. American Fisheries of Fisheries Management 14:216–219. Society, Special Publication 23, Bethesda, Maryland. Buckmeier, D. L., and E. R. Irwin. 2000. An evaluation of soft visual implant Olsen, E. M., J. Gjøsæter, and N. C. Stenseth. 2004. Evaluation of the use tag retention compared with anchor tag retention in Channel Catfish. North of visible implant tags in age-0 Atlantic Cod. North American Journal of American Journal of Fisheries Management 20:296–298. Fisheries Management 24:282–286. Cowen, L., and C. J. Schwarz. 2006. The Jolly–Seber model with tag loss. Ricker, W. E. 1975. Computation and interpretation of biological statistics of Biometrics 62:699–705. fish populations. Fisheries Research Board of Canada Bulletin 191. Crook, D. A., and R. W.G. White. 1995. Evaluation of subcutaneously implanted Rikardsen, A. H. 2000. Effects of Floy and soft VIalpha tags on growth and visual implant tags and coded wire tags for marking and benign recovery in survival of juvenile Arctic Char. North American Journal of Fisheries Man- a small scaleless fish, Galaxias truttaceus (Pisces: Galaxiidae). Marine and agement 20:720–729. Freshwater Research 46:943–946. SAS (Statistical Analysis Systems). 1999. The SAS system for windows, version Daugherty, D. J., and D. L. Buckmeier. 2009. Retention of passive integrated 8.02. SAS Institute, Cary, North Carolina. transponder tags in Flathead Catfish. North American Journal of Fisheries Shepard, B. B., J. Robison-Cox, S. C. Ireland, and R. G. White. 1996. Factors

Downloaded by [Department Of Fisheries] at 19:25 20 January 2013 Management 29:343–345. influencing retention of visible implant tags by Westslope Cutthroat Trout in- Farooqi, M. A., S. A. Nicholson, and M. W. Aprahamian. 1995. Visible im- habiting headwater streams of Montana. North American Journal of Fisheries plant (VI) tag retention in Arctic Charr, Salvelinus alpinus (L.). Fisheries Management 16:913–920. Management and Ecology 2:243–245. Wagner, C. P., M. J. Jennings, J. M. Kampa, and D. H. Wahl. 2007. Survival, Frenette, B. J., and M. D. Bryant. 1996. Evaluation of visible implant tags applied growth, and tag retention in age-0 Muskellunge implanted with passive in- to wild coastal Cutthroat Trout and Dolly Varden in Margaret Lake, southeast tegrated transponders. North American Journal of Fisheries Management Alaska. North American Journal of Fisheries Management 16:926–930. 27:873–877. Guy, C. S., H. L. Blankenship, and L. A. Nielsen. 1996. Tagging and mark- Wenburg, J. K., and G. W. George. 1995. Placement of visible implant tags ing. Pages 353–383 in B. R. Murphy and D. W. Willis, editors. Fisheries in the anal fin of wild coastal Cutthroat Trout. North American Journal of techniques, 2nd edition. American Fisheries Society, Bethesda, Maryland. Fisheries Management 15:874–877. Haw, F., P. K. Bergman, R. D. Fralick, R. M. Buckley, and H. L. Blankenship. Zerrenner, A., D. C. Josephson, and C. C. Krueger. 1997. Growth, mortality, 1990. Visible implanted fish tag. Pages 311–315 in N. C. Parker, A. E. and mark retention of hatchery Brook Trout marked with visible implant tags, Giorgi, R. C. Heidinger, D. B. Jester Jr., E. D. Prince, and G. A. Winans, jaw tags, and adipose fin clips. Progressive Fish-Culturist 59:241–245. This article was downloaded by: [Department Of Fisheries] On: 20 January 2013, At: 19:26 Publisher: Taylor & Francis Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

North American Journal of Fisheries Management Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/ujfm20 Performance of Four Boat Electrofishers with Measured Electrode Resistances for Electrofishing Boats and Rafts Patrick J. Martinez a & A. Lawrence Kolz b c a U.S. Fish and Wildlife Service, 764 Horizon Drive, Grand Junction, Colorado, 81506, USA b 447 Whitetail Lane, Grand Junction, Colorado, 81507, USA c Retired Version of record first published: 09 Jan 2013.

To cite this article: Patrick J. Martinez & A. Lawrence Kolz (2013): Performance of Four Boat Electrofishers with Measured Electrode Resistances for Electrofishing Boats and Rafts, North American Journal of Fisheries Management, 33:1, 32-43 To link to this article: http://dx.doi.org/10.1080/02755947.2012.739985

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material. North American Journal of Fisheries Management 33:32–43, 2013 C American Fisheries Society 2013 ISSN: 0275-5947 print / 1548-8675 online DOI: 10.1080/02755947.2012.739985

ARTICLE

Performance of Four Boat Electroﬁshers with Measured Electrode Resistances for Electroﬁshing Boats and Rafts

Patrick J. Martinez* U.S. Fish and Wildlife Service, 764 Horizon Drive, Grand Junction, Colorado 81506, USA A. Lawrence Kolz1 447 Whitetail Lane, Grand Junction, Colorado 81507, USA

Abstract A misconception in the description of boat electrofishers has resulted from specifying their operational capabilities over ranges of water conductivity without any reference to the electrode configuration with which they will be used. Water conductivity alone is not the valid or most informative electrical variable for defining the operational characteristics of a boat electrofisher and its power supply, but its persistent usage has displaced the correct terminology: electrode resistance measured in ohms. Consequently, inherent design limitations and actual operating characteristics for specific boat electrofishers and their power sources as dictated by electrode resistance have remained undefined. Recognizing this dilemma, total electrode resistance (anodes plus cathodes) was measured for two electrode configurations, metal-hulled boats and inflatable whitewater rafts. Knowing the electrode resistances for these two electrofishing crafts (conductive versus nonconductive hulls) facilitated complete systems’ analyses, provided an analytical understanding of equipment settings, and identified the power capabilities for four boat electrofishers manufactured in the USA. In accordance with the equivalent electrode resistances of these two craft types, electrical load measurements were performed for four boat electrofishers (Smith-Root VVP 15B and GPP 5.0, the ETS Electrofishing MBS 1D-72A, and the Midwest Lake Electrofishing Systems MLES Infinity) to examine their pulsed-DC operational capacities over a range of simulated water conductivities. Graphical plots for each craft type provide an approximation of the range of water conductivity over which each boat electrofisher would be expected to sustain a power level that meets or exceeds power goals for successful electrofishing. This information should improve the understanding of the role of electrode resistance in dictating power demand from boat electrofishers. It should also aid in the selection of a boat electrofisher based on its capacity to sustain standardized power levels in the electrofishing craft in which they will be used and over the range of water conductivities encountered. Downloaded by [Department Of Fisheries] at 19:26 20 January 2013

Boat electrofishers are special purpose, high-voltage power operators must assume the responsibility for understanding the supplies manufactured for the singular purpose of facilitating the electrical features of inadequately specified boat electrofishers capture of fish. Electrical specifications for boat electrofishers (Kolz 2008). marketed in the USA are generally insufficient for adequately Lacking specific criteria regarding what information is war- judging the actual operational or performance characteristics of ranted in the marketing of boat electrofishers, the degree of risk various models (Kolz 2008). Perhaps this explains why elec- and user satisfaction is obviously at the purchaser’s discretion. trofishing personnel often purchase equipment based upon the Further complicating the decision about which boat electrofisher advice, experience, success, or frustrations of colleagues rather to purchase is the misconception specifying that boat electrofish- than the advertised specifications available from manufacturers. ers operate over ranges of water conductivity that lack any ref- Without technical support to verify boat electrofisher outputs, erence to the electrode configuration with which they will be

*Corresponding author: patrick [email protected] 1Retired. Received June 14, 2012; accepted October 3, 2012

32 EVALUATION OF FOUR BOAT ELECTROFISHERS 33

used. Unfortunately, water conductivity alone is not the valid or dardized power output in accordance with the Power Transfer most informative electrical variable for defining the operational Theorem (Kolz 1989) across the range of water conductivi- characteristics of a boat electrofisher and its power supply, but its ties in which the fleet operates. Further, the output of a boat persistent usage has displaced the correct terminology: electrode electrofisher must exhibit an acceptable and stable waveform resistance measured in ohms. Electrofishers are not designed to (i.e., pulse shape, duty, and frequency) that promotes effec- dissipate power in water conductivity; rather they must supply tive fish capture and does not contribute to fish injury, even at power in accordance with the electrical resistance created by the extremes of water conductivity encountered. By measur- the immersed electrodes (Beaumont et al. 2005). Novotny and ing electrode resistance it becomes feasible to perform circuit Priegel (1974) presented graphical methods for determining the analyses, test electrofisher settings, and identify electrofishing resistance of several electrode configurations, and Kolz (1993) system power limitations for standardizing the in-water power described measurement techniques for electrode arrays of any output for each of these two craft types. We tested four boat size and shape. Unfortunately, the significance of these funda- electrofishers manufactured in the USA to determine their op- mental concepts is often underestimated (Reynolds and Kolz erational capacities via static electrical loads simulating the op- 2013), and resistance measurements have often been ignored, eration of electrofishing boats and rafts over a range of water seldom applied, or assumed to be irrelevant. conductivities. Electrode resistance is perhaps the most misunderstood and ignored variable in the electrofishing literature. As a consequence, the operating limits for boat electrofishers and their METHODS power source, as dictated by electrode resistance, have remained Electrofishing system resistance.—The electrofishing john undefined and have probably contributed to a “black box” (i.e., boat used in this study (about 5.3 m long, flat bottom, aluminum turn the knobs and see what happens) approach in selecting and hull) served as the cathode and had two booms, each supporting setting the adjustments. The reality is that few electrode configu- a 23-cm-diameter anodic spheres. The booms extended 2.3 m rations (spheres and cylinders) can be solved with exact theoreti- from the bow and positioned the two anodes 2 m apart. The cal equations to determine their resistance (Novotny and Priegel whitewater rafts (4.3–4.9 m in length) operated with a single, 1974; Beaumont et al. 2005). However, the empirical alternative 23-cm diameter sphere that extended 1.4 m beyond the handrail. is to measure the applied peak voltage and peak current (amps) The two cathodes (each composed of three, 0.6-cm diameter for any submerged electrode configuration and thereby calculate stainless steel cables, 1.2 m long), one dropped along each side its in-water resistance (ohms) with Ohm’s Law (total electrical of the raft, were about 2 m apart and positioned 5 m aft of the resistance = peak voltage divided by peak current; Kolz 1993, anodic sphere. 2008; Beaumont et al. 2005). Examples of electrofishing sys- The measurement of electrode resistance was performed with tems that may differ in terms of their electrical resistance include the spherical anodes half-submerged about 12 cm below the bank electrofishers or barge set-ups having single or multiple surface of the water. Weather and water conditions for these hand-held anodes, aluminum-hulled boats with multiple boom- measurements were ideal, the lake being closed to boating and mounted anodes, or inflatable rafts with anodes of varying size little if any wind and wave action. The initial electrode resistance µ and configuration (e.g., droppers, rings, or spheres). measurements for the boat and raft (R1) were taken at 745 S/cm δ The research performed for this study was based on two elec- ( 1) ambient water conductivity. To facilitate comparison with trode configurations, one for aluminum-hulled boats and one for other electrode designs (R2; in Kolz 1993), measured resistances Ω inflatable whitewater rafts. This is not to imply that boats and ( ) of the boat and raft were normalized to a water conductivity µ δ rafts using identical electrofishers are or can be designed to have of 100 S/cm ( 2; Dean and Temple 2011): Downloaded by [Department Of Fisheries] at 19:26 20 January 2013 the same catchability (conductive versus nonconductive hulls), as that is not possible (Miranda 2005). But, selecting a boat elec- R2 = (R1 × δ1)/δ2. (1) trofisher brand and model that could be used interchangeably between these two styles of electrofishing craft in a multistate The power source for measuring the equivalent electrode resis- or multiagency electrofishing fleet would be operationally and tances of the rafts and boats was a Smith Root (S-R) GPP 5.0, fiscally desirable. Assembling an electrofishing system in a boat powered by its proprietary generator, operating in the pulsed- or raft can be expensive, the boat electrofisher and its genera- DC (PDC) mode. Peak volts and peak amps were measured tor potentially accounting for much of this cost. The ability for with a Fluke Model 99B digital oscilloscope equipped with a a boat electrofisher to be switched between crafts would also model 80i (1,000s) current clamp. The total electrode resistance facilitate troubleshooting and the training of personnel in the (anodes plus cathodes, collectively) was calculated by the ratio selection of proper control settings for safely and effectively of peak voltage to peak amps. These tests required wiring the capturing fish. electrofisher’s output directly to the input of the oscilloscope via For a boat electrofisher to be fully interchangeable between insulated, high-voltage test leads (caution: do not try this with- two different crafts having different electrode configurations out technical assistance). A current clamp provides a convenient of differing system resistance, it must be able to sustain stan- method to monitor the current amplitude in a circuit without 34 MARTINEZ AND KOLZ

exposing and splicing into the system’s wiring. However, available in commercial boat electrofishers, (2) it is recom- current clamps can distort the PDC waveform and may produce mended for PDC electrofishing to monitor a variety of non- erroneous readings if the instructions provided in the operating salmonid, warmwater fishes (Miranda 2005; Reynolds and manual are ignored and the transformer windings in the clamp Kolz 2013), (3) it may induce taxis of fish toward the anode are allowed to saturate. (Burkhardt and Gutreuter 1995), and (4) its effects may be less Boat electrofisher measurements.—We measured, via the harmful to nontarget life stages or species of fish than higher fre- Fluke 99B, peak output voltage, current, and power for four quencies (Holliman et al. 2003; Bohl et al. 2009; Miranda and electrofishers designed for use in boats: (1) the VVP 15B and Kidwell 2010). The independent waveform controls for duty (2) the GPP 5.0, both made by Smith-Root, Inc.,Vancouver, cycle offered with the VVP 15B, MBS-1D-72A, and Infinity Washington, (3) the MBS 1D equipped with a 72-A high-output electrofishers were adjusted between 10% and 30% to provide current option (ETS Electrofishing, LLC, Verona, Wisconsin), additional performance information. and (4) the MLES Infinity (Midwest Lake Electrofishing Sys- The output of each boat electrofisher was measured when tems, Inc., Polo, Missouri). With the exception of the MBS 1D connected to the static electrical loads whose resistance was pro- (which was new), the boat electrofishers were borrowed from gressively decreased, thereby increasing the amount of power active field programs. Three GPP 5.0 and three VVP 15B elec- required from the electrofisher and its power source. For each trofishers were measured, and the data were averaged for each load, the voltage control on the boat electrofisher was incre- model. Although only one ETS MBS 1D-72A and one MLES mentally increased to its maximum, and the peak voltage and Infinity were available, each was considered representative be- current recorded for calculating the corresponding peak power. cause they had recently passed their manufacturer’s inspections. As testing progressed, an electrofisher would eventually become The output of the GPP 5.0, VVP 15B, and ETS 1D-72A could be power-limited, as evidenced by a reduction in its maximum out- switched to either high voltage for fishing in low-conductivity put voltage or the audible power strain from the generator. The waters or to low voltage for higher conductivities; the high volt- testing of an electrofisher was terminated with a particular load age was twice that of the low voltage. Load tests were performed when the unit malfunctioned either by its failure to generate for both voltage ranges to establish the differences in their per- a stable waveform or by the activation of its protective circuit formance characteristics. The Infinity adjusted its voltage range breakers. Although each electrofisher was intentionally adjusted automatically in response to the electrical load in three distinct to produce maximum power, there were no catastrophic failures steps, the switch between each step momentarily and impercep- of any components. tibly ceasing power output. Power transfer into fish.—Electrical theory predicts that, to Electrical measurements were made in the absence of water produce a consistent threshold of immobilization in fish, it is using static loads consisting of parallel and series combinations necessary to adjust the applied electrode power from the elec- of 1,000-W, 57- electric baseboard heaters (CADET model trofisher in a manner that is determined by the effective conduc- 4F1000W) wired to the output of the electrofishers (Martinez tivity of the fish (δf) and the ambient conductivity of the water and Kolz 2009). A minimum resistance (maximum electrical (δw). Based on this principle (Kolz 1989; Miranda and Dolan load) of about 5.7 was achieved with 10 heaters connected in 2003), the minimum threshold of power (Pmin) required to catch parallel. There is an inverse relationship between the electrical fish occurs when the water conductivity is equal to that of the load and the power required from the electrofisher; thus, as the fish. Empirical tests have measured the effective conductivity load resistance is decreased the power from the electrofisher of fish to be 90–150 µS/cm (Kolz and Reynolds 1989; Miranda must increase. By using static loads, we ensured that the electri- and Dolan 2003; Miranda 2009), but it is not known if fish cal measurements were stable, repeatable, and totally removed conductivity is the same for all species of freshwater fish (Kolz Downloaded by [Department Of Fisheries] at 19:26 20 January 2013 from uncontrollable wave and boat motions normally encoun- 2006). While this potential deviation in the conductivity of fish tered when attempting in-water measurements (Kolz 2008; Mar- functionally makes little difference in the calculation of applied tinez and Kolz 2009). The GPP 5.0 units and the S-R generator electrode power levels adjusted for differing ambient water con- remained mounted in the electrofishing boat, which was parked ductivities (described as “target power” by Miranda 2009), the on its trailer. As a safety precaution, the boat hull and trailer analyses herein utilized 115 µS/cm (δf) as the nominal value for were electrically grounded to the soil. The other three units fish conductivity (Miranda and Dolan 2003). The applied elec- were bench-tested, the VVP 15B connected to a 220-A welder trode power that is required when the conductivities of the fish outlet and the MBS 1D-72A and MLES Infinity powered by a and water are not equal (target power = Pt) can be expressed as 6.5-kW generator. While working with the S-R VVP 15B, we a multiplier for sustaining constant power (Mcp) transferred in made measurements on several VVP 15 units manufactured by vivo to the fish: Coffelt Electronics. The major differences observed between the 2 two models were the upgraded metering and indexed controls Mcp = Pt /Pmin = (δw + δ f ) /4δw · δ f . (2) offered with the VVP 15B. All measurements were performed at a pulse rate of 60 Hz. Based on standardized electrofishing boat designs using a ca- This frequency was tested because (1) the setting is commonly thodic hull and two anodes (dropper arrays or spheres), the EVALUATION OF FOUR BOAT ELECTROFISHERS 35

minimum threshold of peak power for water conductivities of 100–150 µS/cm ranged from 2.7 to 3.25 kW (Burkhardt and Gutreuter 1995; Martinez and Kolz 2009; Miranda 2009). In our investigation, for the two spherical anodes, we used the minimum threshold power level of 3 kW for an ambient water conductivity of 115 µS/cm, as recommended by Miranda (2009). This 3-kW recommendation would be excessive for the electrode configuration of the electrofishing rafts, because only a single spherical anode is powered. Therefore, a power analysis was performed to ensure that the single anode on the raft operated with the same electrical current as the individual anodes on the boat at an ambient water conductivity of 115 µS/cm. This correction reduced the minimum threshold power level for the rafts to 1.84 kW. Operators of boat electrofishers are advised that these minimum threshold power levels for boats (3 kW) or rafts (1.84 kW) may not initially result in successful electrofishing. FIGURE 1. Relationship between measured electrode resistances for two The determination of power goals should be performed empir- types of electrofishing crafts, their respective electrode configurations, and am- ically for individual boats under local habitat conditions. This bient water conductivity. Values in parentheses are electrodes resistances (Ω) µ need to identify the capture threshold for the fish community for each craft measured at 745 S/cm ambient water conductivity. The two types of craft include aluminum-hulled john boats (cathodic hulls with dual, being sampled is discussed further below. half-submerged, anodic spheres) and inflatable whitewater rafts (single, half- Graphical presentation of data.—It is convenient to use log- submerged anodic sphere and dual cathodic cables). The Reference Line is the arithmic graphs to plot and analyze the operating characteristics logarithmic plot of the relationship between electrical resistance of the elec- of electrofishers (Kolz 2008). In doing so, the normalized func- trodes and ambient water conductivity, which has a negative linear slope (1). tion for the multiplier for constant power (Mcp) generates a symmetric “U” shaped curve depicting the need for increased power (as the mismatch between water and fish conductivity raft. Equation (1) was rearranged and converted to a logarithmic increases) when superimposed on power graphs depicting its relationship for the convenience of plotting the relationship of δ relationship to water conductivity and output power (Kolz and 2 versus R2 as a straight line having a negative linear slope ◦ Reynolds 1989; Reynolds 1996; Miranda and Dolan 2003). The (−1or−45 ) as demonstrated by the reference line in Figure 1. position of the “U” curve on logarithmic power graphs (Kolz and Reynolds 1989; Kolz 2008) can be identified by measure- log δ2 =−log R2 + log(δ1 × R1)(3) ments of the power, voltage, or current associated with empirical observations of fish immobilization (subjectively under- This logarithmic relationship of electrode resistance (R2)for stood by experienced electrofishing personnel working in situ; the boat and raft for ambient water conductivities (δ2) from Reynolds 1996; Miranda and Dolan 2003; Reynolds and Kolz 10 to 1,000 µS/cm is displayed in Figure 1,which was created 2013). While we have recommended minimum threshold power without calculations by simply drawing straight lines parallel to levels (Pmin) of 3 kW for boats and 1.84 kW for rafts when fish the reference line through the coordinates (R1, δ1) as measured and ambient water conductivity are matched at 115 µS/cm, it is individually for the boat and raft. Figure 1 also identifies the the operators of the electrofishing equipment who provide the differences in electrode resistance values between the boat and Downloaded by [Department Of Fisheries] at 19:26 20 January 2013 data for refining these power levels by identifying the required raft for any water conductivity. Note that portraying ambient power for the desired fish response (i.e., stun, electrotaxis, nar- water conductivity on a decreasing scale facilitates extrapolating cosis, etc.) under field conditions. Refining these initial power resistances at conductivities exceeding 1,000 µS/cm. The utility levels helps to identify optimum electrofisher settings and may of this logarithmic portrayal becomes apparent when placed in result in a slightly adjusted threshold power (Pmin) and reposi- juxtaposition with the performance data acquired during this tioning of the Mcp curve on the power graphs. study and depicted in power graphs.

Boat Electroﬁsher Performance Versus Threshold Power RESULTS Appropriately positioning the resistance-conductivity rela- Water Conductivity versus Electrode Resistance tionship for boat or rafts from Figure 1 below the corresponding The equivalent electrode resistances (R1) measured at boat electroﬁsher power charts in Figure 2 (boats) and Figure 3 745 µS/cm (δ1) were 10.2 for the boat and 25.0 for the (rafts) illustrates how equivalent electrode resistance dictates raft. Normalized to 100 µS/cm (δ2) using equation (1), the cor- the horizontal position of Pmin. Aligning the resistance axis of responding resistances (R2)were76 for boat and 186 for Figure 1 with that of the power chart in a combined graphic 36 MARTINEZ AND KOLZ Downloaded by [Department Of Fisheries] at 19:26 20 January 2013

FIGURE 2. Measured and predicted operational characteristics for four boat electrofishers in relation to the target power level indicated by the multiplier for constant power (Mcp) curve. The electrofishing system was cathodic aluminum-hulled john boats with dual 23-cm diameter hemispheric anodes. Each boat electrofishers was operated at 60 Hz and at different duty cycles (%) in variable levels of ambient water conductivity simulated with static loads: (a) Smith-Root GPP 5.0 = 18%, (b) Smith-Root VVP 15B = 10%, (c) ETS MBS 1D with 72-A high-output current option = 20%, and (d) MLES Infinity = 30%. EVALUATION OF FOUR BOAT ELECTROFISHERS 37 Downloaded by [Department Of Fisheries] at 19:26 20 January 2013

FIGURE 3. Measured and predicted operational characteristics for four boat electrofishers in relation to the target power level indicated by the multiplier for constant power (Mcp) curve. The electrofishing system used was whitewater electrofishing rafts with a single 23-cm diameter hemispheric anode and dual trailing cable cathodes. Each electrofisher operated at 60 Hz and at different duty cycles (%) in variable levels of ambient water conductivity simulated with static loads: (a) Smith-Root GPP 5.0 = 18%, (b) Smith-Root VVP 15B = 10%, (c) ETS MBS 1D with 72-A high-output current option = 20%, and (d) MLES Infinity = 30%. 38 MARTINEZ AND KOLZ

allows incorporation of water conductivity as a fifth interactive Electrofisher Operating Characteristics parameter. At 115 µS/cm (i.e., equal fish and water conductivi- The high and low voltage selections of the VVP, GPP, ties), corresponding electrode resistances are 66 for the boat and MBS electrofishers produced two traces plotted from the and 162 for the raft. Thus, the coordinates for Pmin become measured values of peak voltage in each of the four-parameter 66 and 3 kW (boat; Figure 2), and 162 , 1.8 kW (raft; Fig- power charts for boats (Figure 2a–c) and rafts (Figure 3a–c). ure 3), respectively. The applied voltage and current required The Infinity produced a single trace for its uniquely designed, to meet the minimum threshold power (Pmin) for a water con- automatic voltage range shifts for three ranges of voltage (Fig- ductivity of 115 µS/cm can be estimated at the points where ures 2d, 3d). Although only measurements of peak voltage were they intersect the Mcp curves, as described in Figure 4: boats = required to produce these traces, the power charts are designed 445 V and 6.7 A (Figure 2), rafts = 540 V and 3.3 A (Figure 3). to provide the corresponding values of peak power and peak cur- If desired, the applied volts (AV ) and amps (AA) necessary to rent without further calculations (Kolz 2008). These operating produce any level of power (P) can be calculated by the two traces are fixed by the electrofisher’s circuitry and cannot be al- forms of the power equation: tered by changing the configuration (shape, size, or number) of the anodes and cathodes of an electrofishing boat or raft. Thus, P = AV 2/R, these traces represent the inherent operating characteristics of or (4) each electrofisher model. P = AA2 × R. The GPP 5.0 has selectable outputs of 500 and 1,000 V, but the measured voltages for both settings were slightly higher (Figures 2a, 3a). The 1,000-V setting is necessary for elec- The power and resistance/conductivity charts in Figures 2 and 3 trofishing in low-conductivity water (Martinez and Kolz 2009) can be used with any value of electrode resistance and its corre- at power levels of less than 10 kW and electrode resistances sponding water conductivity to create an Mcp curve for estimat- greater than about 50 (values extrapolated from Figures 2a, ing applied power, voltage, and current necessary to conform 3a). With electrical loads of less than 50 , the GPP 5.0 contin- to the power transfer theory. It should be noted that, without ues to operate at the 1,000-V setting, but there can be significant electrode resistance measurements, it would not be possible to voltage degradation, indicated by excessive generator loading or determine the horizontal coordinate for Pmin at 115 µS/cm, and instability of the waveform near maximum load (oscilloscope the nadir and resulting position of the “U” curve would only display) that may not be detected by the equipment operator be known to be somewhere on the 3-kW ordinate for boat or because the GPP lacks a voltmeter (Pope et al. 2001; Martinez 1.8-kW ordinate for raft. This implies that calculated power and Kolz 2009). Consequently, we recommend that use of the tables, per se, are incomplete without identifying a correspond- 1,000-V setting be avoided when operating in water where the ing electrode resistance. electrode resistance may be less than 50 (Table 1). This is

TABLE 1. Approximate maximum power output (AMPO) measured for four commercially manufactured boat electrofishers using half-submerged 23-cm spherical anodes (two for the boat and one for the raft). High and low approximations of electrode resistances correspond to the ranges of ambient water conductivity over which each electrofisher would be projected to provide satisfactory electrofishing performance at both of its voltage settings (note that the MLES Infinity automatically adjusts voltage in three increments). Approximate operational conductivity ranges were extrapolated from the power output measurements plotted in Figures 2 and 3 for each electrofisher, according to the position of the multiplier for constant power (Mcp)curve.

Boat (cathodic hull) Raft (two cable-array cathodes) Downloaded by [Department Of Fisheries] at 19:26 20 January 2013 Operational Operational Electrode conductivity Electrode conductivity Voltage resistance (Ω) range (µS/cm) resistance (Ω) range (µS/cm) Electroﬁsher AMPO selector model (kW) setting High Low Low High High Low Low High GPP 5.0 19 500 100 6 70 1,200 180 6 100 3,000 1,000 260 >50 a 28 150 500 >50a 38 370 VVP 15B 6.5 300 19 14 380 550 None None None None 600 105 50 70 140 200 50 90 370 MBS 1D (72 amp) 20 300 21 3 360 2,400 20 6 900 3,000 600 110 12 70 700 200 12 90 1,700 MLES Inﬁnity 11 300, 600, 270 7 25 1,000 500 7 36 2,500 and 1,100 EVALUATION OF FOUR BOAT ELECTROFISHERS 39 Downloaded by [Department Of Fisheries] at 19:26 20 January 2013

FIGURE 4. Method for positioning minimum threshold power levels in Figures 2 and 3 for known electrical resistances of electrofishing boats (about 66 Ω)or rafts (about 162 Ω) at 115 µS/cm (conductivity of fish) and identifying and refining target power levels (kW) for electrofishing under local conditions of ambient water conductivity (µS/cm) and selecting boat electrofisher settings using multiplier for constant power curves (Mcp). 40 MARTINEZ AND KOLZ

not viewed as a serious limitation because the characteristics of for the MLES Infinity were about 25–1,000 µS/cm for boats and the 500-V setting are adequate for electrofishing with electrode 36–2,500 µS/cm for rafts (Table 1). resistances between 6 and 50 (i.e., higher ambient water conductivities). The GPP 5.0 sustained a maximum duty cycle of DISCUSSION about 18% at 60 Hz when adjusted to its maximum power output (Figures 2a, 3a). Because the GPP 5.0 is not designed with Electrode Resistance and Configuration independent voltage and duty cycle controls (Pope et al. 2001; This study validates the importance of electrode resistance, Miranda and Spencer 2005; Martinez and Kolz 2009), further which ultimately dictates the operational performance of elec- experimentation with these variables was not performed. The trofishers (Dean and Temple 2011). In fact, the effectiveness of maximum power available with the GPP 5.0 was about 19 kW, an electrofisher can be enhanced or degraded within a particular and an extrapolated operating range for ambient conductivities range of water conductivity by the configuration of the selected was about 28–1,200 µS/cm for boats and 38–3,000 µS/cm for electrodes. Operators and manufacturers of electrofishing equip- rafts (Table 1). ment have experimented with a variety of electrode designs, but The output voltages for the VVP 15B and MBS 1D-72A elec- electrode resistance has seldom been acknowledged as a sig- trofishers (Figures 2b–c, 3b–c) did not deviate appreciably from nificant variable (Reynolds and Kolz 2013). It is only when the their selectable voltage options (300 or 600 V) as the electrical measured characteristics of the electrofishers are complemented load was increased. The measured power characteristics of the by the effects of electrode resistance that this relationship be- Coffelt VVP 15 compared favorably with those of the S-R VVP comes apparent. 15B, and functionally they are essentially identical. The plots The optimum design of an electrode array is not limited to for the VVP 15B and MBS 1D-72A voltage selections terminate measurements of its resistance; the characteristics of the in- abruptly at that the power level where the electrofisher’s protec- water electrical field must also be considered. This combination tive circuit breakers are activated for shutdown. The maximum of variables needs to be explored for possible improvements in peak power for the VVP 15B was about 6.5 kW and the peak electrode configuration, and the measurement techniques pro- power of the MBS 1D-72A approached 20 kW. These power vided by Kolz (1993), Beaumont et al. (2006), and Miranda and limits were the same for either voltage selection: 300 or 600 V Kratochv´ıl (2008) are considered to be applicable. The paired (Figures 2b–c, 3b–c). Due to its lower peak power output, the spherical anodes (half-submerged) used with the boat in this VVP 15B was unable to support a stable PDC waveform that study differed from the ring and cylindrical dropper anode ar- exceeded a 10% duty cycle (Figures 2b, 3b). The higher power rays described for the standard boat proposed by Miranda and output of the MBS 1D-72A sustained a stable PDC waveform Boxrucker (2009). While the electrode resistance for the boat in at 20% duty cycle (Figures 2c, 3c). The projected 20-kW rating Miranda and Boxrucker (2009) was not specified, a resistance for the MBS 1D-72A was extrapolated from the 300-V data of about 60 (at 115 µS/cm) was estimated from its tabu- (dashed line on graph) because we were unable to overload this lated values of power and current (Miranda 2009). This 6- boat electrofisher with our lowest load resistance of 5.7 ohms. difference (60 versus 66 ) between these two standardized The manufacturer advised that the MBS 1D equipped with the electrode configurations is graphically inconsequential. Thus, 72-A option is capable of powering a 3- load (personal com- Figure 2 can be considered to represent the performance of the munications, B. O’Neal, ETS, LLC). The extrapolated oper- electrofishers tested in this study with either of these electrode ating ranges for ambient conductivity for the VVP 15B were configurations. about 70–550 µS/cm for boats and 90–370 µS/cm for rafts This similarity does not imply that the two boats would fish (Table 1). For the MBS 1D-72A, these operating ranges were the same even with the same applied power coming from iden- Downloaded by [Department Of Fisheries] at 19:26 20 January 2013 about 70–2,400 µS/cm for boats and 90–3,000 µS/cm for rafts tical electrofishers and using the same electrical waveform set- (Table 1). tings (Miranda 2005). The spatial and heterogeneous, in-water The power traces for the MLES Infinity (Figures 2d, 3d) electrical variables (voltage gradient, current density, and power did not conform to fixed voltage settings as the load resistance density) are determined by the size, shape, and configuration of was decreased because the electrofisher is designed with control the electrodes, and the cylindrical droppers will create a more in- circuitry that automatically limits its maximum output to about tense voltage gradient field than that of the 23-cm hemispheres. 11 kW at discrete, maximum output voltages of about 300, We predict that, if the anode spacing is the same for these two 600, and just over 1,100 V (Figures 2d, 3d). Compared with the boats, the recommended power threshold of 3 kW at 115 µS/cm other three electrofishers, the MLES Infinity demonstrated three would probably require at least some correction for one of these cycles of maximum output power occurring at those values of boats. Field personnel should anticipate that it will require mul- electrical resistance where the unit automatically switched its tiple fish-response threshold observations in different values voltage range in response to decreasing resistance and increasing of water conductivity before determining the most acceptable power demand. The MLES Infinity sustained stable PDC power power correction (Mcp) for their particular electrode array. While output for both boats and rafts at a duty cycle of 30% (Figures 2d, experimental laboratory observations have contributed to iden- 3d). The extrapolated operating ranges for ambient conductivity tifying power thresholds (Kolz and Reynolds 1989; Dolan and EVALUATION OF FOUR BOAT ELECTROFISHERS 41

Miranda 2003; Miranda and Dolan 2003; Miranda 2005), there 23-cm hemispherical anodes. In this case, the electrofisher sim- are no theoretical methods currently available for avoiding this ply lacks the output power necessary to support the electrical subjective approach. load. These examples verify the significance of electrode resistance in determining the success or failure of an electrofisher’s Performance Evaluations for the Four Boat Electrofishers operation. Our evaluations of the four boat electrofishers were con- Operators of electrofishing equipment are reminded that the ducted at a frequency of 60 Hz. It would require a substantial horizontal position of the power transfer function (Mcp), and the amount of additional time to record data for several frequen- electrode resistance can be altered at their discretion by chang- cies and duty cycles. However, the development of meaningful ing the configuration of the electrode array. As illustrated by this specifications for electrofishers is the responsibility of the man- study, the applied power and electrode resistance differences be- ufacturers, and this task should not be imposed on the limited tween the boat and raft resulted in electrofishing systems with resources of electrofishing operators. Further, the equipment divergent operational capabilities. Analyses of these systems specifications presently offered by manufacturers of electrofish- demonstrate the unique association of electrofishing principles ers typically do not provide adequate explanations regarding the by displaying the power transfer function (Mcp), the electrode performance capabilities of these power supplies when operat- resistance measurements (δw versus R), and the measured vari- ing with different electrical loads. ables for the electrofishers (P versus R) on a single graphic. For years, field personnel have been frustrated by the uncertainty in adjusting settings on their electrofishers. Equip- ment operators desire to know the range of water conductivi- Variations Due to Operational and Conductivity Factors ties in which they can expect to successfully electrofish. This The electrode and boat electrofisher measurements con- would be preferable to determining this limitation by trial-and- ducted during this study were performed under ideal, stable con- failure after having invested in a particular electrofisher that ditions, but field personnel cannot always expect to encounter may not adequately power their electrode configuration under these near-laboratory situations. In practice, normal operational the prevailing conditions in the local habitats they sample. The variability associated with lateral and vertical movements as peak voltage measurements recorded by this study along with electrofishing boats and rafts are maneuvered, fluctuations in known values of static loads resolves this enigma by creating water depth, distance between electrodes, surface wave action, graphic presentations that serve as handy references for field and relative electrode submergence are uncontrolled, electrode personnel. resistance variables. These resistance variations will, in turn, Our findings for boat electrofisher performance, when used affect the output of the electrofisher, suggesting that the applied with the electrode configurations described in this study, are electrode power will be a dynamic range around the target power readily applicable for users of similar electrofishing craft by level that must fluctuate to account for these changes in the active interpreting the power outputs for boats (Figure 2) or rafts (Fig- electrical field (Kolz 2008; Miranda 2009). Thus, the precision ure 3). The position of the Mcp function on each power chart of the electrode resistance lines in Figure 1 must be considered indicates the power, voltage, and current required for any given an oversimplification, and a more realistic presentation should electrical load or water conductivity based upon the initial min- be viewed as a statistical variance relative to the mean. imum threshold power levels of 3 kW for boats or 1.8 kW for The inverse relationship between electrode resistance and rafts at matched ambient water and effective fish conductivities water conductivity obviously contributes to the limitation of (115 µS/cm). When the output measured for an electrofisher electrofisher performance based on its design parameters and (voltage, current, or power) exceeds the theoretical Mcp require- power source. While Table 1 indicates operational capacity for Downloaded by [Department Of Fisheries] at 19:26 20 January 2013 ment, the electrofisher can be successfully operated; otherwise, some boat electrofishers above 2,000 µS/cm, electrofishing may the electrofisher performance will probably prove inadequate. become less efficient at extremely high conductivities (1,000– Accordingly, surplus output from an electrofisher can be re- 3,000 µS/cm; Speas et al. 2004). As power demand at higher duced to the desired level, closer to the Mcp function, but it water conductivity climbs as electrode resistance decreases, is impossible to artificially generate additional output with an electrical waveforms may not conform to control settings or inadequate electrofisher or power source. may display instability (Van Zee et al. 1996; Martinez and Kolz Table 1 provides predicted operational capabilities for each 2009). Under the latter conditions, the power transfer theory and boat electrofisher based upon those segments of their operating Mcp curve are no longer valid because the theory requires con- traces in Figures 2 and 3 that exceed the power requirements sistent waveforms at all power levels. However, some studies of the Mcp function for the boat or raft. The electrode resis- report that there may be no significant effects of conductiv- tance and water conductivity values were extrapolated from the ity less than about 1,500 µS/cm on catchability or electrofishing graphs and did not require any calculations. This simplicity in efficiency (Hill and Willis 1994; Dumont and Dennis 1997; Bay- interpretation can be observed using Figures 2b and 3b to reveal ley and Austen 2002). Further, increasing electrode resistance that the VVP 15B probably offers inadequate operational capa- by using single anodes may allow electrofishing at high conduc- bility with the 300-V setting for electrofishing operations using tivities, as indicated for rafts in Table 1, even for power-limited 42 MARTINEZ AND KOLZ

electrofishers such as the VVP 15 (Hill and Willis 1994; Van loscopes (Miranda and Spencer 2005); scopemeters, i.e., Fluke Zee et al. 1996). 123 with 80i–110 s current clamp (Miranda 2009) and Fluke 97 (Dean and Temple 2011); peak-reading digital multimeters, i.e., Significance of Duty Cycle Fluke 87V with i200 AC clamp (L. Kolz) and Fluke 87V (Dean Miranda and Dolan (2004) reported that duty cycles of 10– and Temple 2011); and computer software (i.e., S-R Power Stan- 50% require the least peak power to immobilize fish, provide the dardization System). Measurement of peak current offers the ad- highest margin between the power levels that narcotized versus vantage of higher resolution and a linear relationship with water tetanized fishes, and allowed a larger radius of immobilization conductivity (Miranda 2009). For water conductivities greater away from the anode. This range of intermediate duty cycles is than the fish conductivity (e.g., >115 µS/cm), it is suggested considered to be less injurious to warmwater fishes (Dolan and that peak current should be considered the preferred measure- Miranda 2004; Miranda and Dolan 2004). Further, duty cycles ment to assess electrofishing performance and conformance to of 20–30% are associated with electrotaxis of fishes (Burnet target power levels for maintaining constant transfer of in vivo 1959; Habera et al. 2010), which tends to facilitate fish cap- power. ture (Novotny and Priegel 1974; Burkhardt and Gutreuter 1995; Maret et al. 2007); however, conclusive laboratory or field tests have not been performed. SUMMARY COMMENTS The GPP 5.0 can be adjusted to increase applied power within Independent measurements of electrode resistance and boat the first half of its voltage control (Percent of Range knob), but electrofisher power output characteristics were combined in this it lacks an independent control for duty cycle (Miranda and study to create a graphic analysis technique that incorporates the Spencer 2005). As its power output begins to plateau at a fre- electrofishing principles asserted by power-transfer theory (Kolz quency setting of 60 Hz, a duty cycle of about 18% is achieved 1989; Miranda and Dolan 2003; Reynolds and Kolz 2013). This (Martinez and Kolz 2009). This potential limitation in promoting inclusive visual approach simultaneously establishes the rela- electrotaxis with the GPP 5.0 may be overcome by electrofish- tionship between the variation of resistance for any electrode ing at 120 Hz, whereby the voltage control within the first half array and the required electrofishing power, as well as between of its adjustment includes duty cycles exceeding 20% (Martinez the voltage and current at any measured value of ambient wa- and Kolz 2009). While duty cycle can be adjusted on the VVP ter conductivity. These analyses can be performed graphically 15B, its power limitation restricted our tests to 10% duty cycle, without reference tables or manual calculations. which would not be expected to elicit electrotaxis. However, Our study contributes to the recent emphasis for standardiz- field observations suggest that at lower ambient water conduc- ing fish sampling using electrofishing (Martinez and Kolz 2009; tivities around 100–200 µS/cm, a duty cycle of about 20% may Miranda 2009; Reynolds and Kolz 2013). It is suggested that be sustained with VVP-15 units (Hall 1986), facilitating elec- the equipment measurements performed in this study be seri- trotaxis and fish capture (C. Walford, Larval Fish Laboratory, ously considered as part of the development of protocols for Colorado State University, personal communication). standardized electrofishing. This information should improve the understanding of the role of electrode resistance in dictating Instrumentation power demand from electrofishers. It should also aid selection The availability of accurate electrical meters on electrofishers of a boat electrofisher model based on its capacity to sustain is an important feature for monitoring power output and elec- standardized power levels in accordance with the system resis- trofishing performance (Dean and Temple 2011). It was not the tance of the electrofishing craft in which they are used and over intent of this study to evaluate the meters supplied with the boat the range of water conductivities encountered. Downloaded by [Department Of Fisheries] at 19:26 20 January 2013 electrofishers; however, boat electrofishers deserving recogni- tion for being designed with accurate peak voltage and peak current meters include the ETS MBS 1D (Miranda 2009) and ACKNOWLEDGMENTS the MLES Infinity. Peak reading meters provide the advantage This work would not have been possible without the equip- of applying the precepts of constant power transfer by using ei- ment, assistance, and cooperation provided by Anita Martinez, ther peak voltage or current measurements as indices of power Lori Martin, and Estevan Vigil of the Colorado Parks and output or by using these values to calculate peak power. The Wildlife and Bob Burdick and Michael Gross at the U. S. Fish VVP 15B is instrumented with a peak voltmeter that provided and Wildlife Service in Grand Junction, Colorado. We thank only marginal scale resolution (Coffelt VVP 15 meters are not Angela Kantola of the Upper Colorado River Endangered Fish peak-reading; Van Zee et al. 1996), and the GPP current me- Recovery Program and the National Fish and Wildlife Founda- ter is considered too inaccurate for standardized measurements tion for their financial support. Burke O’Neal (ETS Electrofish- (Pope et al. 2001, Miranda and Spencer 2005; Martinez and ing, LLC) and Tom Lehman (Midwest Lake Management, Inc.) Kolz 2009). graciously provided loans for the boat electrofishers used in our Recommendations to overcome metering deficiencies of evaluation. Reference to trade names does not imply endorse- some boat electrofisher models have included portable oscil- ment of commercial products by the federal government. EVALUATION OF FOUR BOAT ELECTROFISHERS 43

REFERENCES Kolz, A. L., and J. B. Reynolds. 1989. Determination of power threshold re- Bayley, P. B., and D. J. Austen. 2002. Capture efficiency of a boat electrofisher. sponse curves. U.S. Fish and Wildlife Service Technical Report 22:15–24. Transactions of the American Fisheries Society 131:435–451. Maret, T. R., D. S. Ott, and A. T. Herlihy. 2007. Electrofishing effort required Beaumont, W. R. C., M. J. Lee, and G. Peirson. 2005. The equivalent resistance to estimate biotic condition in southern Idaho rivers. North American Journal and power requirements of electric fishing electrodes. Fisheries Management of Fisheries Management 27:1041–1052. and Ecology 12:37–43. Martinez, P. J., and A. L. Kolz. 2009. Evaluating the power output of the Smith- Beaumont, W. R. C., G. Peirson, and M. J. Lee. 2006. Factors affecting the Root GPP 5.0 electrofisher to promote electrofishing fleet standardization. characteristics and propagation of voltage gradient fields from electric fishing North American Journal of Fisheries Management 29:570–575. anodes. Fisheries Management and Ecology 13:47–52. Miranda, L. E. 2005. Refining boat electrofishing equipment to improve con- Bohl, R. J., T. B. Henry, R. J. Strange, and P. L. Rakes. 2009. Effects sistency and reduce harm to fish. North American Journal of Fisheries Man- of electroshock on cyprinid embryos: implications for threatened and en- agement 25:609–618. dangered fishes. Transactions of the American Fisheries Society 138:768– Miranda, L. E. 2009. Standardizing electrofishing power for boat electrofishing. 776. Pages 223–230 in S. A. Bonar, W. A. Hubert, and D. W. Willis, editors. Burkhardt, R. W., and S. Gutreuter. 1995. Improving electrofishing catch con- Standard methods for sampling North American freshwater fishes. American sistency by standardizing power. North American Journal of Fisheries Man- Fisheries Society, Bethesda, Maryland. agement 15:375–381. Miranda, L. E., and J. Boxrucker. 2009. Warmwater fish in large standing waters. Burnet, A. M. R. 1959. Electric fishing with pulsatory direct current. New Pages 29–42 in S. A. Bonar, W. A. Hubert, and D. W. Willis, editors. Standard Zealand Journal of Science 2:45–56. methods for sampling North American freshwater fishes. American Fisheries Dean, J. C., and A. J. Temple. 2011. Maximum output of peak power for Society, Bethesda, Maryland. two backpack electrofishers operated at various pulsed direct current duty Miranda, L. E., and C. R. Dolan. 2003. Test of a power transfer model for cycles and water conductivity levels. North American Journal of Fisheries standardized electrofishing. Transactions of the American Fisheries Society Management 31:520–529. 132:1179–1185. Dolan, C. R., and L. E. Miranda. 2003. Immobilization thresholds of elec- Miranda, L. E., and C. R. Dolan. 2004. Electrofishing power requirements in trofishing relative to fish size. Transactions of the American Fisheries Society relation to duty cycle. North American Journal of Fisheries Management 132:969–976. 24:55–62. Dolan, C. R., and L. E. Miranda. 2004. Injury and mortality of warmwater Miranda, L. E., and R. H. Kidwell. 2010. Unintended effects of electrofishing on fishes immobilized by electrofishing. North American Journal of Fisheries nongame fishes. Transactions of the American Fisheries Society 139:1315– Management 24:118–127. 1321. Dumont, S. C., and J. A. Dennis. 1997. Comparison of day and night electrofish- Miranda, L. E., and M. Kratochv´ıl. 2008. Boat electrofishing relative to anode ar- ing in Texas reservoirs. North American Journal of Fisheries Management rangement. Transactions of the American Fisheries Society 137:1358–1362. 17:939–946. Miranda, L. E., and A. B. Spencer. 2005. Understanding the output of a Smith- Habera, J. W., M. A. Kulp, S. E. Moore, and T. B. Henry. 2010. Three-pass Root GPP electrofisher. North American Journal of Fisheries Management depletion sampling accuracy of two electric fields for estimating trout abun- 25:848–852. dance in a low-conductivity stream with limited habitat complexity. North Novotny, D. W., and G. R. Priegel. 1974. Electrofishing boats: improved designs American Journal of Fisheries Management 30:757–766. and operational guidelines to increase the effectiveness of boom shockers. Hall, T. J. 1986. Electrofishing catch per hour as an indicator of Largemouth Wisconsin Department of Natural Resources Technical Bulletin 73. Bass density in Ohio impoundments. North American Journal of Fisheries Pope, K. L., B. E. Van Zee, M. C. Mayo, and M. Rahman. 2001. Assessment of Management 6:397–400. outputs from Smith-Root model-5.0 GPP and model-7.5 GPP electrofishers. Hill, T. D., and D. W. Willis. 1994. Influence of water conductivity on pulsed North American Journal of Fisheries Management 21:353–357. AC and pulsed DC electrofishing catch rates for Largemouth Bass. North Reynolds, J. B. 1996. Electrofishing. Pages 221–253 in B. R. Murphy and D. W. American Journal of Fisheries Management 14:202–207. Willis, editors. Fisheries techniques, 2nd edition. American Fisheries Society, Holliman, F. M., J. B. Reynolds, and T. J. Kwak. 2003. A predictive risk model Bethesda, Maryland. for electroshock-induced mortality of the endangered Cape Fear Shiner. North Reynolds, J. B., and A. L. Kolz. 2013. Electrofishing. Pages 305–361 in A. V. American Journal of Fisheries Management 23:905–912. Zale, D. L. Parrish, and T. M. Sutton, editors. Fisheries techniques, 3rd edition. Kolz, A. L. 1989. A power transfer theory for electrofishing. U.S. Fish and American Fisheries Society, Bethesda, Maryland. Wildlife Service Fish and Wildlife Technical Report 22:1–11. Speas, D. W., C. J. Walters, D. L. Ward, and R. S. Rogers. 2004. Effects of in-

Downloaded by [Department Of Fisheries] at 19:26 20 January 2013 Kolz, A. L. 1993. In-water electrical measurements for evaluating electrofishing traspecific density and environmental variables on electrofishing catchability systems. U.S. Fish and Wildlife Service Biological Report 11. of Brown and Rainbow Trout in the Colorado River. North American Journal Kolz, A. L. 2006. Electrical conductivity as applied to electrofishing. Transac- of Fisheries Management 24:586–596. tions of the American Fisheries Society 135:509–518. Van Zee, B. E., T. D. Hill, D. W. Willis, L. F. Brown, J. B. Reynolds, Kolz, A. L. 2008. Understanding Smith-Root LR-24 equipment speci- N. G. Sharber, and J. Sharber. 1996. Comment: clarification of the outputs fications. North American Journal of Fisheries Management 28:1563– from a Coffelt VVP-15 electrofisher. North American Journal of Fisheries 1567. Management 16:477–478. This article was downloaded by: [Department Of Fisheries] On: 20 January 2013, At: 19:26 Publisher: Taylor & Francis Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

North American Journal of Fisheries Management Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/ujfm20 Economic Values and Regional Economic Impacts of Recreational Fisheries in Mississippi Reservoirs Clifford P. Hutt a , Kevin M. Hunt a , Susan F. Steffen a d , Stephen C. Grado b & L. E. Miranda c

a Law Enforcement Laboratory, Human Dimensions and Conservation, Department of Wildlife and Fisheries, Mississippi State University, Box 9690, Mississippi State, Mississippi, 39762, USA b Law Enforcement Laboratory, Human Dimensions and Conservation, Department of Forestry, Mississippi State University, Mississippi State, Mississippi, 39762, USA c U.S. Geological Survey, Mississippi Cooperative Fish and Wildlife Research Unit, Box 9691, Mississippi State, Mississippi, 39762, USA d Kansas Department of Wildlife, Parks, and Tourism, Emporia Research and Survey Office, Post Office Box 1525, Emporia, Kansas, 66801, USA Version of record first published: 07 Jan 2013.

To cite this article: Clifford P. Hutt , Kevin M. Hunt , Susan F. Steffen , Stephen C. Grado & L. E. Miranda (2013): Economic Values and Regional Economic Impacts of Recreational Fisheries in Mississippi Reservoirs, North American Journal of Fisheries Management, 33:1, 44-55 To link to this article: http://dx.doi.org/10.1080/02755947.2012.739986

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material. North American Journal of Fisheries Management 33:44–55, 2013 C American Fisheries Society 2013 ISSN: 0275-5947 print / 1548-8675 online DOI: 10.1080/02755947.2012.739986

ARTICLE

Economic Values and Regional Economic Impacts of Recreational Fisheries in Mississippi Reservoirs

Clifford P. Hutt,* Kevin M. Hunt, and Susan F. Steffen1 Law Enforcement Laboratory, Human Dimensions and Conservation, Department of Wildlife and Fisheries, Mississippi State University, Box 9690, Mississippi State, Mississippi 39762, USA Stephen C. Grado Law Enforcement Laboratory, Human Dimensions and Conservation, Department of Forestry, Mississippi State University, Mississippi State, Mississippi 39762, USA L. E. Miranda U.S. Geological Survey, Mississippi Cooperative Fish and Wildlife Research Unit, Box 9691, Mississippi State, Mississippi 39762, USA

Abstract This study estimated the economic benefit of recreational fisheries on two trophy crappie Pomoxis spp. fisheries in Mississippi. We accomplished this by estimating economic impacts of angler expenditures and angler willingness-to- pay for fishing trips above current expenditures. Anglers spent 91,811 activity days on Sardis Reservoir in 2006 and 46,036 on Grenada Reservoir in 2007. The total economic impacts to the state were estimated at US$5.83 million on Sardis Reservoir and $2.15 million on Grenada Reservoir, supporting 75 and 51 full- and part-time jobs, respectively. Total consumer surplus was $5.57 million and $4.46 million on Sardis and Grenada reservoirs, respectively. Last, we used a model that predicted the effect of average spring water levels on total angling effort to estimate resulting changes in economic benefits. Spring angling effort and associated economic benefits peaked at intermediate water levels and declined at both low and high water levels. Future changes in precipitation patterns in the southeastern United States may require altering reservoir guide curves to maximize economic benefits associated with reservoir fisheries. Downloaded by [Department Of Fisheries] at 19:26 20 January 2013

Recreational activities such as fishing have the potential to anglers receive from the resource (i.e., the value anglers place make important contributions to local and state economies in the on fishing above their actual angling-related expenditures, their USA; however, these economic contributions are often over- net “willingness-to-pay” (WTP) for the resource) (Connelly and looked or underestimated because of their nonmarket nature. Brown 1991; Edwards 1991; Criddle et al. 2003; Loomis 2006). Researchers looking to determine the economic benefits of fish- The first, economic impact, is a measure of the monetary ben- eries resources are faced with two options: (1) determination of efits of angling to the regional economy and is generated by the economic impacts of angling-related expenditures within a angling expenditures, which are the costs of fishing to the an- region (i.e., their effects on state income and employment), or (2) gler. As such, economic impact does not represent benefits to the determination of the net economic value, or consumer surplus, angler, only the regional economy (Connelly and Brown 1991;

*Corresponding author: [email protected] 1Present address: Kansas Department of Wildlife, Parks, and Tourism, Emporia Research and Survey Ofﬁce, Post Ofﬁce Box 1525, Emporia, Kansas 66801, USA. Received June 25, 2012; accepted October 3, 2012

44 RECREATIONAL FISHERIES IN MISSISSIPPI RESERVOIRS 45

Edwards 1991). Net economic value, or consumer surplus, on that keeping water levels at full pool for an additional month the other hand represents the economic benefit anglers receive within the summer would result in an additional 87,000 angler from fishing because it is a measure of the value they place trips taken within the study region. Hamel et al. (2002) devel- on the activity over and above their personal costs (Connelly oped a model demonstrating the effect of catch expectations and Brown 1991; Edwards 1991). Respectively, these estimates on angler participation levels and the resulting effect on the represent two distinct types of economic benefits of the fishery economic impact of the fishery. Criddle et al. (2003) expanded that can be compared with the economic benefits of competing on this analysis to include the effects of catch expectations on market goods and services. angler consumer surplus benefits, or welfare. In this study, we Despite the nonmarket nature of fisheries resources, they gen- sought to establish a relationship between reservoir water lev- erate great economic impact and value as millions of Americans els and angler participation to demonstrate the effect fluctuating pursue them every year, spend billions of dollars in the process, water levels can have on regional economic impacts and benefits and, in most cases, would be willing to spend even more money associated with recreational fishing. than they already do to pursue them (USFWS and USCB 2008). In this study, estimates were made of both the economic im- In 2005, there were 626,000 anglers in the state of Mississippi pact and value of recreational fisheries derived from the Sardis alone that spent an estimated US$264 million on fishing trips and and Grenada reservoirs. Located in northwestern Mississippi, fishing-related equipment (USFWS and USCB 2008), generat- and managed by USACE for flood control, these reservoirs pro- ing a total economic impact of approximately $490 million and vide two of the best crappie Pomoxis spp. fisheries in the state of supporting 3,731 jobs (Posadas 2009). During the same time Mississippi, and they draw considerable attention from nonres- period it was estimated that anglers in the southeastern USA ident anglers and fishing tournament organizations due to their spent $16 billion on fishing trips and fishing-related equipment reputation for producing trophy-sized crappies. Like all flood generating nearly $21.4 billion in total economic impact and control reservoirs operated by USACE, operation schedules for supporting 163,541 jobs (Munn et al. 2010). While these num- Sardis and Grenada reservoirs are dictated by federally man- bers are impressive, they are not a full measure of the total dated guide curves that provide seasonal targets for reservoir economic benefit of fishing to the region because they do not water levels (Miranda et al. 2012). These guide curves call for include a measure of the totality of benefits anglers derive from fall water releases to lower reservoir water levels to a conser- fishing. vation pool to provide storage capacity for winter and spring Knowledge about the economic benefits derived from runoff. Water levels are then raised to normal pool in March– fisheries resources is of vital importance to fisheries agencies May and held constant through summer to facilitate greater seeking to manage these resources in the face of competing reservoir use for fishing and boating activities. Temporal vari- demands and expected changes in hydrological regimes due ability in water level deviations from the mandated guide curves to anthropocentric climate change (Mulholland et al. 1997; result from annual variability in precipitation within a reservoir’s Wood et al. 1997; Daugherty et al. 2011). For example, many catchment. popular freshwater fisheries in the southern USA are on In recent years, fisheries in Sardis and Grenada reservoirs flood control reservoirs operated by the U.S. Army Corps of have been faced with the potential impacts of increased angling Engineers (USACE) or Tennessee Valley Authority. Several pressure from fishing tournaments and nonresident anglers, as studies conducted on reservoirs have demonstrated the effect well as extended periods of drought that reduced reservoir water of water level fluctuations on angler visitation with higher levels, angler access, and available habitat (Hunt et al. 2008). water levels generally resulting in increased angler visitation These forces have motivated the Mississippi Department of so long as they remain below flood stage (Ward et al. 1996; Wildlife, Fisheries, and Parks (MDWFP) to adopt more strin- Downloaded by [Department Of Fisheries] at 19:26 20 January 2013 Platt and Munger 1999). Reductions in angling effort due to gent regulations on these fisheries and increased MDWFP inter- low reservoir water levels have been attributed to a variety of est in the economic impact and value of the fisheries and factors causes ranging from declines in reservoir aesthetics and access that might influence angler visitation and thus economic impact to increased crowding of remaining reservoir users (Ward et al. and value. This paper focused primarily on the effect of spring 1996; Daugherty et al. 2011; Miranda et al. 2012). (March–May) water levels on recreational fishing effort and Any reductions in angling activity caused by fluctuating wa- resultant economic benefits, as most (56–78%) fishing effort on ter levels would in turn result in reductions in economic impact these reservoirs occurs during the spring season when crappie and recreational value due to trip reductions and related ex- are spawning (Hunt et al. 2008). penditures and benefits. Thus, models estimating the effect of The goal of this study was to estimate the economic benefits exogenous variables (i.e., reservoir water level) on angler de- of recreational fisheries derived from two Mississippi reser- mand and participation could be paired with economic impact voirs. Four objectives were identified to achieve this goal. Ob- and valuation models to assess the costs of different manage- jective 1 was to determine angling expenditures and activity ment actions (Jakus et al. 2000; Hamel et al. 2002; Criddle et al. day totals on Sardis and Grenada reservoirs by resident and 2003). Jakus et al. (2000) demonstrated the effect of water level nonresident anglers. Objective 2 was to quantify economic im- drawdowns in Tennessee flood control reservoirs and estimated pacts and multipliers of angling on Mississippi’s economy from 46 HUTTETAL.

angling expenditures on Sardis and Grenada reservoirs by res- the sole purpose of recruiting participants for the mail survey. In ident and nonresident anglers. Objective 3 was to determine all, 436 anglers (260 residents, 176 nonresidents) were recruited the WTP above expenditures, or consumer surplus, of resident for the mail survey at Sardis Reservoir, and 481 anglers (395 and nonresident anglers utilizing Sardis and Grenada reservoirs. residents, 86 nonresidents) were recruited for the mail survey at Last, objective 4 was to use an angler demand model combined Grenada Reservoir. with estimates developed with objectives 1–3 to assess the ef- Mail survey methods.—Data for the economic assessment of fects of spring reservoir water level fluctuations on the regional recreational fishing on Sardis and Grenada reservoirs were col- economic effects of recreational fisheries on Sardis and Grenada lected via a self-administered mail survey. Mail surveys were reservoirs. preferred over on-site surveys for collecting expenditure data because they allow respondents to report the full expenditures of their trips and do not compromise the integrity of the creel sur- METHODS vey (Riechers et al. 1991; Dillman 2007). The self-administered Access creel survey.—Access-point creel surveys were con- questionnaires collected data on angler characteristics, trip char- ducted on each reservoir to estimate angler activity days and acteristics, activity days, trip expenditures, WTP, and trip satis- collect angler contact information for a follow-up mail survey faction. Anglers that were recruited to participate in the mail sur- designed to collect data on trip expenditures, angler WTP, and vey during access site creel surveys were first sent a cover letter, angler trip characteristics. Surveys were conducted for 1 year on questionnaire, and business reply envelope (complete packet) each reservoir (March 2006 to February 2007 on Sardis Reser- within 3 months of their original on-site contact (day 1), fol- voir, and March 2007 to February 2008 on Grenada Reservoir). lowed by a postcard reminder or thank you (day 10). Three ad- Sampling was stratified according to four quarters (December– ditional complete packets were sent as necessary to nonrespon- February, March–May, June–August, September–November). dents (dependent on returns) on days 21, 35, and 49 (Dillman Interviews were conducted during 6-h days over 24 sampling 2007). All questions and research procedures were approved days selected at random within each quarter (96 total per year), by the MSU Institutional Review Board’s Committee for the 12 on weekdays and 12 on weekend days, and approximately Protection of Human Subjects (Docket No. 06-061). evenly over access points distributed throughout each reservoir. Nonresponse bias analysis.—The potential for nonresponse Interviews recorded party size, hours fished, and selected char- bias was examined using the method developed by Fisher acteristics of the catch. Effort was estimated as the product of (1996). Fisher’s method uses logistic regression to determine if mean trip length, mean party size, and number of trips. Trip respondents and nonrespondents differ significantly on known length and party size were recorded during the access creel sur- variables and, if so, calculates response probabilities to be used vey. The number of trips and daily effort were estimated through for developing nonresponse adjustment weights. A separate instantaneous counts of boat trailers at access sites surrounding analysis was conducted for Sardis and Grenada reservoirs to the reservoirs using methods described in Hunt et al. 2008. The determine if residency status or gender had a significant influ- number of activity days (i.e., total number of days anglers spent ence on response rate. In both cases, no significant differences fishing on each reservoir over the course of the year sampled) were found between respondents and nonrespondents, and it was estimated by dividing total effort (h) by the average daily was determined that no adjustment for nonresponse was needed effort per angler (Sardis = 5.0 h/d; Grenada = 5.1 h/d) (Hunt for this study. et al. 2008). Economic impact analysis.—Average daily expenditures re- In addition, creel clerks asked one participant from each lated to fishing were calculated for use in an input-output (I-O) fishing party to be a part of a follow-up angler survey follow- model of the state economy to determine the economic impact Downloaded by [Department Of Fisheries] at 19:26 20 January 2013 ing methods described by Ditton and Hunt (2001). Specifically, of fishing on Sardis and Grenada reservoirs in Mississippi. The creel clerks told members of the fishing party that Mississippi Impact Analysis for Planning (IMPLAN) model software was State University (MSU) was conducting an angler survey and used to perform these analyses. This software, originally de- needed responses from one randomly selected adult member veloped by the U.S. Department of Agriculture Forest Service from each group. Creel clerks then asked for the adult party to evaluate forest management plan impacts, consists of both member with the most recent birthday and presented that angler national and county level data for industrial and commercial with an information flier with the study specifics. If the angler sectors (currently 440 sectors as described by the U.S. Depart- agreed, the creel clerk obtained name and address information as ment of Labor). The IMPLAN software and the most current well as supplemental characteristics of their current trip. In the data on the Mississippi economy at the time of the study (i.e., case of guided trips (asked at start of interview), only customers 2007) was used to build a model of the Mississippi economy and were selected. Supplemental information included trip origin, generate direct and secondary (i.e., indirect, induced) impacts whether it was a guided trip, telephone number, and gender. resulting from in-state participant expenditures. Direct impacts To achieve a sufficient number of anglers to participate in the are those dollars spent directly with businesses and kept in the mail survey, 8–12 supplemental sampling days (6-h days at one economy, and secondary impacts include supporting businesses randomly selected boat ramp) were conducted each quarter for that resupply direct businesses. Induced impacts are the ensuing RECREATIONAL FISHERIES IN MISSISSIPPI RESERVOIRS 47

employee purchasing tied to both direct and indirect business and activity day estimates. In all, the I-O model was used to wages. Regional purchase coefficients generated during model calculate the influence of the fishery on the output of goods and development determine the economy’s ability to absorb trip ex- services and the generation of value-added, income, tax, and penses. As a result, in-state expenditures made on behalf of employment impacts on the Mississippi economy (Bohnsack angler trip activities were viewed as final demands on current et al. 2002). state industries and businesses. In addition to providing expenditure data, resident anglers Respondents were asked to itemize their expenditures for were also asked to address a hypothetical question asking them the angling trip during which they were initially recruited for to estimate what percentage of their annual expenditures for the study (e.g., gas, lodging, food, bait). Respondents were in- fishing on Sardis or Grenada reservoirs would have been spent structed to only report expenditures made in the state of Missis- out of state if they could no longer fish that reservoir. The specific sippi. Respondents were asked to provide itemized expenditures wording of the question was as opposed to total expenditures so expenditures could be inserted into the appropriate economic sectors within the IMPLAN For residents of Mississippi only, given the hypothetical situation model. Trip expenditures were divided by trip length (days), per- that you would not be able to fish at Sardis [Grenada] Lake, what percent of the money you currently spend per year in Mississippi for cent of days spent fishing, and by number of anglers in the fishing fishing Sardis [Grenada] Lake would you spend out-of-state to fish party the respondent paid for to calculate daily trip expenditures or participate in any other activity (fishing or nonfishing related)? per angler. Daily expenditures per angler (US$/angler/activity day) were then averaged for each sector in the model for resi- The average percentage provided by the above question was dent and nonresident anglers, respectively, regardless of whether used to calculate adjusted economic impacts for resident an- they purchased the items or not, to create expenditure profiles glers to provide a more conservative estimate of resident angler (i.e., US$/angler/activity day for residents and nonresidents) for economic impacts. Many economists insist that resident expen- economic impact analysis. As anglers were initially contacted ditures should be excluded from economic impact analysis as using an on-site creel survey, all expenditure estimates were their money would still circulate in the economy in question corrected for avidity bias using the method outlined by Loomis and, therefore, would not provide an additional impact to the (2007). This involved weighting angler expenditure data by the economy (Crompton et al. 2001; Chen et al. 2003; Stoll and inverse of the number of days they had fished in the previous 12 Ditton 2006). However, if resident anglers were to spend their months on the lake on which they were contacted. money out of state in the absence of the fisheries on Sardis and Economic impact was determined using the 2007 IMPLAN Grenada reservoirs, then these expenditures could be viewed as I-O model for the state of Mississippi (Olson and Lindall 2000). a leakage from the Mississippi economy and would thus legiti- Previous studies have utilized the IMPLAN model to estimate mately count as a source of economic impact (Steinback 1999; the economic impact of marine charter boat fisheries (Steinback Loomis 2006; Grado et al. 2011). Therefore, a more accurate 1999; Grado et al. 2005), Bluefin Tuna Thunnus thynnus fish- picture of the economic impact of resident angler expenditures eries (Bohnsack et al. 2002), and individual freshwater fisheries can be illustrated by adjusting these economic impacts by the (Schorr et al. 1995; Chen et al. 2003). In this study, economic percentage of expenditures they would spend out of state, on ei- impacts of resident and nonresident anglers were determined for ther fishing or other activities, if fishing on Sardis and Grenada each reservoir in the IMPLAN model by inputting average daily reservoirs were no longer available (Grado et al. 2011). expenditures (US$/angler/activity day) by economic sector, and Estimation of consumer surplus.—To estimate net eco- number of activity days within the given year. From this infor- nomic value, or consumer surplus, for the fisheries on Sardis mation, the IMPLAN model was used to calculate the direct, and Grenada reservoirs, respondents were presented with a Downloaded by [Department Of Fisheries] at 19:26 20 January 2013 indirect, and induced economic impacts of the recreational fish- dichotomous-choice contingent valuation (CV) question to de- eries associated with the Sardis and Grenada reservoirs. Direct termine how much money anglers would be willing to pay above impacts constituted local income and employment generated by their current trip costs before they would have no longer been the immediate expenditures of anglers paid to local businesses, willing to make the fishing trip. Each respondent was presented industries, and services within the region of interest (i.e., the with a single, noniterative CV question with a randomly se- state). Indirect impacts were the result of expenditures paid lected bid value from a preselected range of values from $3to to local businesses, industries, and services resupplying these $400 for Sardis Reservoir, and $3to$1,200 for Grenada Reser- economic enterprises. Finally, induced impacts were the result voir. Respondents were asked if they would have still taken the of household expenditures made by the individuals that were fishing trip in question if their trip costs had increased by the employed in businesses, industries, and services resulting from amount of the bid due to increases in the costs of goods and direct and indirect impacts. The sum of direct, indirect, and services. Additional bid amounts above $400 were added in induced impacts constituted the total economic impact of the the second year of the study because of the high percentage of fishery. Finally, upper and lower bound estimates of economic Sardis Reservoir anglers that responded in the affirmative to the impact were estimated by running the IMPLAN model with higher bid amounts, which could lead to an underestimation of both the upper and lower bound estimates of trip expenditures angler WTP. 48 HUTTETAL.

Additionally, respondents were asked several questions parameters listed by Miranda et al. (2012). Estimates of total related to trip satisfaction and their angling behavior that was effort were modeled for a broad range of water levels for hypothesized would influence their level of WTP. Respondents Y = 2006 in Sardis and Y = 2007 in Grenada and converted were asked to indicate whether fishing ranked first, second, to activity days by dividing E by average daily effort per angler third, or lower among their most important outdoor activities, as for each reservoir (Sardis = 5.0 h/d, Grenada = 5.1 h/d; Hunt it was hypothesized that WTP would be higher for individuals et al. 2008). The resulting estimates of March–May angler that placed greater importance on fishing (IMPFISH; scale: activity days per water level were then used to estimate percent 1 = most important outdoor activity, 2 = second most important, change in activity days and resulting changes in estimated 3 = third most important, 4 = none of above). Respondents expenditures, economic impact, employment, and consumer were also asked if they were a member of any fishing clubs surplus as compared with those estimated for each reservoir (CLUB: 1 = yes, 0 = no) as this is considered an indicator of during their respective study year. dedication and increased involvement to the sport (Ditton et al. 1992). Respondents were also asked if they felt the trip had been RESULTS well worth the money they spent on their trip (WORTH) and if they had considered any of the fish they had caught on their trip Response Rates and Angler Demographics to be trophy fish (TROPHY), with both questions being asked Creel clerks contacted 512 fishing parties on Sardis Reser- on a 5-point Likert type scale (1 = strongly disagree, 2 = agree, voir and succeeded in recruiting 436 anglers to participate 3 = neutral, 4 = agree, 5 = strongly agree). in the mail survey as 70 parties were repeat contacts and 6 Separate estimates of resident and nonresident angler con- refused to participate. Creel clerks contacted 614 parties on sumer surplus were estimated for each reservoir with a logistic Grenada Reservoir and recruited 481 to the mail survey, with regression model, as the dependent variable was a binary vari- 80 repeat anglers and 53 refusals. Nonresidents comprised able (yes = 1, no = 0) (Loomis 2006). Reservoir-specific values 40% (n = 176) and 18% (n = 86) of the recruited anglers on were estimated using a probit model in SAS with dollar amount Sardis and Grenada reservoirs, respectively. Seventy-four per- (BID), CLUB, IMPFISH, WORTH, TROPHY, INCOME (mea- cent of nonresident anglers contacted on Sardis Reservoir were sured in $10,000 intervals ranging from $5,000 to $105,000), from Tennessee, while the largest portion (31%) of nonresident and SQINC (i.e., income squared, to test the assumption of anglers on Grenada Reservoir was from Missouri, with another marginal utility of income) serving as the independent vari- 18% from Tennessee and Illinois. Questionnaires were returned ables. The probit model was estimated twice, first with all the by 331 anglers from Sardis Reservoir (9 undeliverable) and 345 independent variables included, and again with only the statis- anglers from Grenada Reservoir (15 undeliverable) for adjusted tically significant (P = 0.05) independent variables from the response rates of 78% and 74%, respectively. These response first model. The second model was used to estimate mean WTP rates were high by mail survey standards and were comparable and the marginal effects of the significant independent variables to other mail surveys of anglers that recruited anglers on site for on WTP. Total consumer surplus was estimated by multiplying economic assessments (Chen et al. 2003; Loomis 2006). mean WTP by the number of fishing trips taken on each reser- Resident and nonresident anglers on both reservoirs were al- voir during the surveyed year. Finally, confidence intervals for most exclusively white males with an average age ranging from mean WTP were estimated by using a bootstrapping procedure 52 to 54 years old (Table 1). Median household income was (Bateman et al. 2002). Bootstrapping was conducted in SAS to $50,000–59,999 for residents and $70,000–79,999 for nonres- generate 1,000 resamples of the WTP data. Probit models were idents on both reservoirs, respectively. The percentage of resi- then estimated for each resample and were used to estimate dent and nonresident anglers on Sardis Reservoir with at least Downloaded by [Department Of Fisheries] at 19:26 20 January 2013 mean WTP for each resample. The resample estimates of mean some college education was 43.7% and 50.7%, respectively. WTP were then sorted and the 2.5 and 97.5 percentile estimates The percentage of resident and nonresident anglers on Grenada were used to define the 95% confidence interval of the original Reservoir with at least some college education was 43.8% and mean WTP estimates. 54.8%, respectively. Water level effects on angling activity, impact, and value.— The effect of water level on economic impacts and valuations Activity Days, Angler Expenditure, and Input-Output was estimated for Sardis and Grenada reservoirs with models Model Results linking fishing effort to March—May water levels (Miranda Average trip length for resident anglers was 1.2 d on Sardis et al. 2012). Those models related angling effort (h) to water Reservoir with 45,140 trips taken by resident anglers for a total levels as estimate of 54,168 activity days (ad) (Table 2). Average trip length for nonresident anglers was 3.7 ad on Sardis Reservoir 2 E = β0 + β1(WL) − β2(WL ) + β3(Y), (1) with 10,174 trips for a total estimate of 37,643 ad. Average daily expenditures for trip-related goods and services by resident an- where E is total March–May angling effort (h), WL is average glers on Sardis Reservoir were $43.63/angler/ad and average March–May water level (m), Y is year, and β0-3 are regression total trip expenditures (average daily expenditures multiplied RECREATIONAL FISHERIES IN MISSISSIPPI RESERVOIRS 49

TABLE 1. Socioeconomic characteristics of residents and nonresidents of Mississippi responding to a mail survey of anglers ﬁshing Sardis and Grenada reservoirs in Mississippi during March 1, 2006, to February 28, 2007, and March 1, 2007, to February 29, 2008, respectively.

Median Median Gender Race household education Reservoir Resident status Average age (% male) (% White) income ($) level Sardis Resident 53 97 94 50,000 – 59,999 High school Nonresident 54 95 92 70,000 – 79,999 Some college Grenada Resident 52 99 97 50,000 – 59,999 High school Nonresident 54 98 100 70,000 – 79,999 Some college

by average trip length) were $52.36/angler/trip (Table 3). Av- each dollar spent. Unadjusted total economic impacts of resident erage daily expenditures for trip-related goods and services by anglers on Sardis Reservoir were $4.16 million and supported 37 nonresident anglers on Sardis Reservoir were $55.02/angler/ad full- and part-time jobs in Mississippi. However, this estimate of and average total trip expenditures were $203.57/angler/trip. total economic impact was adjusted to $1.98 million as resident Average trip length for resident anglers was 1.2 ad on anglers indicated that they would spend an average of 47.5% Grenada Reservoir with 31,074 trips taken by resident anglers (SE = 3.8%) of their Sardis Reservoir–related angling expendi- for a total estimate of 37,289 ad by resident anglers (Table 2). Av- tures outside of Mississippi if they could no longer fish on the erage trip length for nonresident anglers was 4.1 ad on Grenada reservoir. Therefore, 52.5% of resident expenditures were not Reservoir with 2,133 trips by nonresident anglers for a total es- considered economic impacts. The SAM multiplier for resident timate of 8,747 ad. Average daily expenditures for trip-related impacts on Sardis Reservoir was 1.76. Altogether, we estimated goods and services by resident anglers on Grenada Reservoir total economic impact of recreational fishing on Sardis Reser- were $58.77/angler/ad with average total trip expenditures of voir to be $5.83 million supporting 75 full- and part-time jobs $70.52/angler/trip (Table 3). Average daily expenditures for trip- with an overall SAM multiplier of 1.83. related goods and services by nonresident anglers on Grenada The total economic impacts of nonresident anglers on Reservoir were $66.10/angler/ad with average total trip expen- Grenada Reservoir were $1.09 million and supported 12 full- ditures of $271.01/angler/trip. and part-time jobs in the state of Mississippi. The SAM multi- The IMPLAN model estimated overall economic impacts in plier was 1.88. Unadjusted total economic impacts of resident 2007 dollars from fishing expenditures by resident and nonresi- anglers on Grenada Reservoir were $3.99 million and supported dent anglers fishing on Sardis and Grenada reservoirs (Table 4). 39 full- and part-time jobs in the state. The adjusted impact of Total economic impacts of nonresident anglers on Sardis Reser- resident anglers was $1.06 million as Grenada anglers indicated voir were $3.85 million and were estimated to support 38 full- they would spend an average of 26.5% (SE = 2.8%) of their and part-time jobs in the state of Mississippi. The social ac- fishing-related expenditures on Grenada Reservoir out of state counting matrix (SAM) multiplier, the total economic impacts if they could no longer fish there. Therefore, 73.5% of resident divided by the direct economic impacts, was 1.86, meaning that expenditures were not considered economic impacts. The SAM for every dollar spent in Mississippi by nonresident anglers fish- multiplier was 1.81. Altogether, we estimated the total economic ing Sardis Reservoir there was an economic return of $1.86 for impact of recreational fishing on Grenada Reservoir to be $2.15

Downloaded by [Department Of Fisheries] at 19:26 20 January 2013 million, supporting 51 full- and part-time jobs with an overall TABLE 2. Total number (n) of fishing trips, average trip length, and total SAM multiplier of 1.86. activity days (ad) of participation at Sardis (March 1, 2006, to February 28, 2007) and Grenada (March 1, 2007, to February 29, 2008) reservoirs in Mississippi. Contingent Valuation Estimates of WTP by Reservoir Bid amount and the angler’s evaluation of whether the trip Average was worth the money spent (WORTH) were significant predic- Residence Trips trip length Activity tors of WTP for both resident and nonresident anglers on Sardis Reservoir status (n) (ad) days (n) Reservoir, while only the bid amount was a significant predictor Sardis Resident 45,140 1.2 54,168 of WTP for resident and nonresident anglers on Grenada Reser- Nonresident 10,174 3.7 37,643 voir (Table 5). The coefficient for the bid amount was negative Combined 55,314 1.7 91,811 in all models indicating that anglers were less likely to pay Grenada Resident 31,074 1.2 37,289 for increasing trip costs, while the coefficient for WORTH was Nonresident 2,133 4.1 8,747 positive indicating that anglers were willing to pay more if they Combined 33,207 1.4 46,036 felt the trip had been worth the money. These results were consistent with economic theory indicating that the model is valid 50 HUTTETAL.

TABLE 3. Average expenditures incurred for goods and services by residents and nonresidents per activity day (ad) by anglers ﬁshing Sardis and Grenada reservoirs in Mississippi during March 1, 2006, to February 28, 2007, and March 1, 2007, to February 29, 2008, respectively, and adjusted for avidity bias (in 2007 USD).

Sardis Grenada Residents Nonresidents Residents Nonresidents Expenditure items ($/ad) ($/ad) ($/ad) ($/ad) Transportation Airfare 0.00 1.49 0.00 0.00 Automobile, boat gas and oil 21.50 13.44 21.91 14.72 Rental vehicle 0.00 0.07 0.00 0.00 Lodging Lodging at state park hotels or cabins 0.48 1.20 1.32 7.01 Lodging at hotel or motel 0.24 3.13 0.80 6.71 Public or private campground fees 0.10 0.49 1.95 0.50 Vacation home rental 0.09 0.04 0.02 0.15 Food and beverages Adult beverages 0.49 0.50 0.79 1.08 Groceries, ice, and nonalcoholic beverages 8.60 4.49 8.49 5.88 Restaurant or take-out meals 3.46 9.27 2.06 15.82 Other shopping, services, and entertainment Bait and tackle 2.93 7.52 8.42 2.86 Boat and equipment rental 0.04 0.00 0.01 0.00 Boat launch and daily use fees 2.16 2.02 3.77 0.68 Casinos, movies, other entertainment 0.21 0.00 0.08 0.16 Fishing guide fees 0.00 0.00 0.00 2.67 Fishing license 3.34 11.35 8.80 7.77 Propane and cooking fuel 0.01 0.01 0.02 0.10 Anything else in Mississippi 0.00 0.02 0.31 0.00 Total expenditures/angler-day 43.63 55.02 58.77 66.10 Total expenditures/angler trip 53.36 203.57 70.52 271.01

TABLE 4. Economic impacts (direct, indirect, induced, and total) generated from resident and nonresident trip-related fishing expenditures on Sardis and Grenada reservoirs in 2006 and 2007, respectively. All economic impacts are estimated in 2009 US$. Resident economic impacts were adjusted for the average percentage of expenditures anglers anticipated spending out of state in the event they could no longer fish Sardis or Grenada. Lower and upper bound estimates were generated by rerunning the economic impact model with the lower and upper confidence intervals of the expenditure and activity day estimates.

Impacts

Downloaded by [Department Of Fisheries] at 19:26 20 January 2013 Residence Expenditure Reservoir status level Direct Indirect Induced Total Grenada Nonresident Mean 578,264 175,453 334,294 1,088,011 (n = 52) Lower 237,932 75,166 120,454 433,552 Upper 1,005,644 300,428 605,606 1,911,678 Resident Mean 580,543 161,634 314,827 1,057,003 (n = 240) Lower 322,215 91,204 144,763 558,181 Upper 931,897 256,349 552,076 1,740,322 Sardis Nonresident Mean 2,071,871 638,408 1,140,454 3,850,732 (n = 110) Lower 1,107,953 347,629 585,477 2,041,059 Upper 3,335,444 1,017,691 1,865,187 6,218,322 Resident Mean 1,123,106 314,005 539,566 1,976,676 (n = 163) Lower 594,616 168,854 253,157 1,016,626 Upper 1,856,534 514,610 954,039 3,325,182 RECREATIONAL FISHERIES IN MISSISSIPPI RESERVOIRS 51

TABLE 5. Final probit model summaries for willingness-to-pay (WTP) estimation of anglers utilizing Sardis and Grenada reservoirs in 2006 and 2007, respectively. Mean WTP represents angler WTP for a fishing trip on Sardis or Grenada reservoirs above their travel expenditures. Mean trip WTP was calculated by dividing the product of each variables mean and model coefficient by the bid coefficient and summing across variables. Mean daily WTP was calculated by dividing mean trip WTP by average trip length. Lower and upper bounds on all WTP estimates were calculated using a bootstrapping procedure with 1,000 iterations. All models were weighted to adjust for avidity bias as described by Loomis (2007).

Model Bounds Variable Mean β p β(X) WTP Lower Upper Sardis Resident (n = 163) CONSTANT 1.000 −0.1957 0.354 −0.1957 −38.75 BID −0.0051 < 0.001 WORTH 0.662 0.9449 < 0.001 0.6255 123.86 Mean Trip WTP = 85.11 12.36 215.76 Mean Daily WTP = 70.93 10.30 179.80 Nonresident (n = 110) CONSTANT 1.000 0.2744 0.213 0.2744 45.89 BID −0.0060 < 0.001 WORTH 0.657 1.1309 < 0.001 0.7427 124.19 Mean Trip WTP = 170.08 104.05 301.12 Mean Daily WTP = 45.97 28.12 81.38 Grenada Resident (n = 240) CONSTANT 1.000 0.7718 < 0.001 0.7718 127.99 BID −0.0060 < 0.001 Mean Trip WTP = 127.99 71.37 177.65 Mean Daily WTP = 106.66 59.48 148.04 Nonresident (n = 52) CONSTANT 1.000 0.877 < 0.001 0.877 226.06 BID −0.004 0.006 Mean Trip WTP = 226.06 78.10 532.82 Mean Daily WTP = 55.14 19.05 129.97

(Swanson and McCollum 1991). Whether the angler caught surplus of $1.73 million. Mean WTP of Grenada Reservoir resi- what they considered a trophy ﬁsh, whether they were a mem- dent anglers was estimated to be $127.99 per trip (LB = $71.37, ber of a ﬁshing club, the level of importance they placed on UB = $177.65; Table 5). Multiplied by the estimated number of

Downloaded by [Department Of Fisheries] at 19:26 20 January 2013 fishing as an outdoor activity, their income, and their income resident fishing trips taken to Grenada Reservoir in 2007, total squared (i.e., to test the assumption of marginal utility of in- consumer surplus was $3.98 million. Mean WTP of Grenada come) proved to be insignificant in all models and were thus Reservoir nonresident anglers was estimated to be $226.06 per dropped from the final probit models. trip (LB = $78.10, UB = $532.82; Table 5). Multiplied by the Based on the probit model results, mean WTP of Sardis estimated number of nonresident fishing trips taken to Grenada Reservoir resident anglers was estimated to be $85.11 per trip Reservoir in 2007, total consumer surplus was $0.48 million. with bootstrapping providing lower (LB) and upper bounds (UB) of $12.36 and $215.76 per trip, respectively (Table 5). Water Level Effects on Angling Activity, Impact, and Value Multiplying the average WTP by the estimated number of res- Water levels at Sardis and Grenada reservoirs during March– ident fishing trips taken to Sardis Reservoir in 2006 produced May of the study years averaged 74.9 and 61.2 m, respectively. an estimated total consumer surplus of $3.84 million. Mean Water levels during the 1986–2007 period modeled by Miranda WTP of Sardis Reservoir nonresident anglers was estimated to et al. (2012) ranged from 73.9 to 82.4 m at Sardis Reservoir be $170.08 per trip (LB = $104.05, UB = $301.12; Table 5). and from 60.7 to 65.5 m at Grenada Reservoir. The models gen- Multiplied by the estimated number of nonresident fishing trips erated by Miranda et al. (2012) demonstrated that water level taken to Sardis Reservoir in 2006, this produced a total consumer and year were strong predictors of fishing effort on both Sardis 52 HUTTETAL.

TABLE 6. Estimated changes in March–May ﬁshing effort, regional economic impact, and consumer surplus relative to reservoir water levels during March–May on Sardis and Grenada reservoirs in 2006 and 2007, respectively. Change was deﬁned relative to estimates made in the study years.

Change in angler Change in Change in % Change in expenditures Change in total employment consumer surplus Water Level (m) activity days (US$) impact (US$) (jobs) (US$) Sardis 73 −32.60 −499,884 −911,752 −24 −871,478 74 −5.31 −81,478 −148,610 −4 −142,045 75 15.37 235,761 430,010 12 411,016 76 29.46 451,832 824,109 22 787,706 77 36.95 566,736 1,033,686 28 988,026 78 37.85 580,473 1,058,741 28 1,011,974 79 32.15 493,042 899,274 24 859,551 80 19.85 304,444 555,285 15 530,757 81 0.96 14,679 26,774 1 25,592 82 −24.53 −376,253 −686,258 −18 −655,945 83 −56.62 −868,352 −1,583,812 −42 −1,513,852 Grenada 60 −42.27 −377,126 −698,090 −22 −1,451,476 61 8.02 71,549 132,443 4 275,378 62 35.77 319,197 590,859 18 1,228,521 63 41.00 365,817 677,157 21 1,407,953 64 23.69 211,410 391,338 12 813,674 65 −16.14 −144,023 −266,599 −8 −554,316 66 −78.50 −700,485 −1,296,654 −40 −2,696,017

(R2 = 0.84; P < 0.01) and Grenada reservoirs (R2 = 0.80; age spring water levels on each reservoir were used to illustrate P = 0.06). Angling effort and associated economic benefits were how variability in spring precipitation and alterations to reser- found to peak at intermediate water levels and decline at both voir guide curves could affect the economic benefits provided low and high water levels in a parabolic fashion (Table 6). Con- by these fisheries. sidering the relationship between water level and fishing effort Minimal differences in daily expenditures of resident and in equation (1), we estimated percent changes in March–May nonresident anglers were found on Sardis and Grenada reser- angler activity days ranging from –56.6% to 37.9% (Table 6). voirs. While average trip expenditures were much higher for These estimated changes in activity days could potentially pro- nonresidents than resident anglers, average daily expenditures duce changes in total economic impact ranging from –$1.58 to for goods and services were only slightly higher for nonresident $1.06 million at Sardis Reservoir and from –$1.30 to $0.68 mil- anglers than resident anglers due to nonresidents taking signif-

Downloaded by [Department Of Fisheries] at 19:26 20 January 2013 lion at Grenada Reservoir. Similarly, consumer surplus was esti- icantly longer trips. Given these similarities in average daily mated to range from –$1.51 to $1.01 million at Sardis Reservoir expenditures, differences in economic impacts between resi- and from –$2.70 to $1.41 million at Grenada Reservoir (Table 6). dent and nonresident anglers on the two reservoirs were mainly a function of the difference in number of activity days between the two groups. As such, unadjusted economic impacts of res- DISCUSSION ident anglers were far greater than the impacts of nonresident The economic benefits of the recreational fisheries were anglers. However, many researchers have excluded expenditures quantified for the Sardis and Grenada reservoirs in Mississippi of resident anglers from their economic impact analyses on the by estimating the economic impacts of trip-related angler ex- grounds that resident expenditures represent money that would penditures and angler consumer surplus, or net economic value, most likely still circulate within the economy in question, and over and above their current trip costs. Respectively, these esti- would just be spent on other substitute products or services mates provided information on the effects of these fisheries on within the economy if the fishery in question was not available the Mississippi economy and angler welfare, and they provided for use (Crompton et al. 2001; Chen et al. 2003; Stoll and Ditton a picture of the economic benefits of each fishery. Furthermore, 2006). Conversely, Steinback (1999), Loomis (2006), and Grado models that predicted spring fishing effort as a function of aver- et al. (2011) argued that it is valid to consider the economic RECREATIONAL FISHERIES IN MISSISSIPPI RESERVOIRS 53

impact of resident expenditures if those expenditures would be resulted in fewer angler contacts than we would have normally spent out of state in the event that the fishery under examination expected. Indeed, of the four groups for which consumer surplus was no longer available. estimates were generated, the estimated bounds were widest for To address this issue, resident anglers were asked what per- the group with the smallest sample size and narrowest for the centage of their trip-related expenditures would have been spent group with the largest sample size. out of state if they were no longer able to fish on Sardis or While most researchers argue that nonresident economic im- Grenada reservoirs (Grado et al. 2011). We used the average per- pacts are more important than resident impacts because they centage estimated from this question to adjust resident economic represent new money within the economy (Chen et al. 2003), impacts for each fishery. It was recognized that this method of the exact opposite emphasis is placed on nonresident consumer adjustment, while purely hypothetical in nature, still offers a surplus. Because consumer surplus represents benefits anglers reasonable starting point for estimating the potential leakage to receive from the resource, consumer surplus of resident anglers the economy that would result from the loss of a fishery, which is of greater importance to regional fisheries managers than the may not be as rare an event as previously thought given recent consumer surplus of nonresident anglers. Specifically, if nonres- events triggered by industrial accidents, regulations, and natural ident angler exploitation of the resource is such that it negatively disasters. Furthermore, a much lower expenditure leakage was impacts the benefits resident anglers receive from the fishery estimated for Grenada Reservoir than Sardis Reservoir. This then it may negate the benefit of nonresident expenditures on was reasonable considering Sardis Reservoir is much closer to the local economy (Swanson and McCollum 1991). the state borders and potential substitute sites such as Lake Lastly, models that predicted the effect of reservoir water lev- Chicot in Arkansas and Kentucky Lake in Tennessee. Alterna- els on angler effort during March–May were used to estimate tively, Grenada Reservoir is located closer to central Missis- how economic impact and consumer surplus could be affected sippi and is close to several substitute sites located within the by factors such as variation in winter and spring precipitation state. While not as conservative as studies that exclude resident and reservoir operation decisions (i.e., guide curve changes). impacts altogether, it was felt this study provided a moderate The analyses presented in this paper could be used to help guide estimate of resident economic impact that serves as a good decisions pertaining to water level targets on flood-control reser- compromise between those factions that would exclude resi- voirs like those examined in this study. Predicted gains and dent impacts altogether and those that would argue for their losses in fishery economic benefits as a result of changes in inclusion. reservoir water levels could be weighed against other costs and In addition to representing the benefit to a state economy, benefits associated with water storage to identify target water expenditures also represent the cost of fishing to the anglers levels that will optimize overall regional economic benefits. in question (Connelly and Brown 1991; Edwards 1991; Stoll Specifically, it was found that fishing effort, and resulting eco- and Ditton 2006). These costs to the angler do represent a form nomic benefits, were maximized at intermediate water levels of value in the sense that they represent a minimum amount in both reservoirs. However, the USACE’s ability to maintain of money the angler is willing to forego to facilitate their par- the water levels assigned by the guide curve was dependent on ticipation in the activity, but they do not represent the benefit highly variable and unpredictable precipitation patterns, which the resource provides to the angler. Consumer surplus repre- according to current climate models may change considerably sents the benefit, or additional utility, anglers receive from the in the coming decades (Mulholland et al. 1997; Sun et al. 2005). resource that is not forfeited via expenditures (Connelly and Mulholland et al. (1997) estimated the effects of a doubling of Brown 1991; Loomis 2006). We estimated per-trip consumer atmospheric carbon dioxide from preindustry levels on seasonal surplus separately for resident and nonresident anglers on Sardis precipitation in the southeastern United States. They estimated Downloaded by [Department Of Fisheries] at 19:26 20 January 2013 and Grenada reservoirs. Our estimates of average daily WTP that precipitation in northern Mississippi would decline by 10% were much higher than those generated by Dorr et al. (2002) over the winter while increasing by the same percentage in sum- for Sardis Reservoir using the travel cost method; however, mer. While this would simply represent a redistribution of annual their study was conducted before the two reservoirs developed precipitation, such a scenario would likely delay the attainment national reputations as trophy crappie fisheries. Furthermore, of the summer conservation pool on Sardis and Grenada reser- our estimates of WTP were comparable to those generated by voirs and lead to a decrease in average March–May water lev- other studies of reservoir and other quality freshwater fisheries els, which would negatively affect fishing effort and associated (Chizinski et al. 2005; Loomis 2006). A bootstrapping method economic benefits. Lower March–May water levels could be was utilized to generate lower and upper bounds for our WTP avoided on each reservoir by raising the water level target for estimates with a 95% confidence interval (Bateman et al. 2002). each reservoir’s winter conservation pool. Additional research Ranges of these bounds indicated that our estimates were highly will be needed to determine if alterations to existing reservoir variable. It was believed this was largely the result of small sam- guide curves are needed to account for changes in precipitation ple sizes resulting from splitting the sample into resident and due to anthropogenic climate change, associated costs due to nonresident groups and the fact that our study was conducted temporal shifts in flooding threats, and the economic benefits of on two low-water years that suffered from low fishing effort and recreational angling and other reservoir uses. 54 HUTTETAL.

While many studies choose to estimate only economic im- Ditton, R. B., and K. M. Hunt. 2001. Combining creel intercept and mail survey pacts or consumer surplus, this study estimated both to provide methods to understand the human dimensions of local freshwater fisheries. a more complete picture of the economic effects of these fish- Fisheries Management and Ecology 8:295–301. Ditton, R. B., D. K. Loomis, and S. Choi. 1992. Recreation specialization: eries on the Mississippi economy and the benefits they provide re-conceptualization from a social worlds perspective. Journal of Leisure to anglers that utilize them. Furthermore, it was demonstrated Research 24:33–51. how economic impacts and valuation data can be used to assess Dorr, B., I. A. Munn, and K. O. Meals. 2002. A socioeconomic and biological reservoir management decisions by linking economic values to evaluation of current and hypothetical crappie regulations in Sardis Lake, reservoir water levels. Future research can improve upon this Mississippi: an integrated approach. North American Journal of Fisheries Management 22:1376–1384. analysis by considering other factors that can influence angler Edwards, S. F. 1991. A critique of three “economics” arguments commonly participation levels including angler catch expectations, travel used to influence fishery allocations. North American Journal of Fisheries costs, and provision of additional on-site amenities. However, Management 11:121–130. our application provides an example of how economic impacts Fisher, M. R. 1996. Estimating the effect of nonresponse bias on angler surveys. and valuation data can be used to inform reservoir management Transactions of the American Fisheries Society 125:118–126. Grado, S. C., K. M. Hunt, C. P. Hutt, X. T. Santos, and R. M. Kaminski. decisions. 2011. Economic impacts of waterfowl hunting in Mississippi derived from a state-based mail survey. Human Dimensions of Wildlife 16:100–113. Grado, S. C., J. C. Jones, S. T. Earles, and W. D. Jones. 2005. Fishing activities ACKNOWLEDGMENTS and economic impacts of charter boat businesses on the Mississippi Gulf Coast. Proceedings of the Annual Conference Southeastern Association of We appreciate the funding support provided by U.S. Fish and Fish and Wildlife Agencies 57(2003):112–123. Wildlife Service Federal Assistance in Sport Fish Restoration Hamel, C., M. Herrmann, S. T. Lee, K. R. Criddle, and H. T. Geier. 2002. Linking Mississippi Grant Number F-138 through the MDWFP. We ap- sportfishing trip attributes, participation decisions, and regional economic preciate the contributions of MDWFP biologists K. Meals and impacts in lower and central Cook Inlet, Alaska. Annals of Regional Science A. Dunn. Also, we thank the students and technicians in the 36:247–264. Hunt, K., S. Grado, L. E. Miranda, and S. F. Baker. 2008. A social and economic Human Dimensions and Conservation Law Enforcement Labo- analysis of the recreational fisheries in Mississippi flood control reservoirs. ratory at MSU who worked on this research. Finally, we thank Mississippi Department of Wildlife, Fisheries and Parks, Freshwater Fisheries the anonymous reviewers for their comments and suggestions Report 250, Jackson. that helped to improve to this manuscript. Jakus, P. M., P. Dowell, and M. N. Murray. 2000. The effect of fluctuating water levels on reservoir fishing. Journal of Agricultural and Resource Economics 25:520–532. Loomis, J. 2006. Use of survey data to estimate economic value and regional REFERENCES economic effects of fishery improvements. North American Journal of Fish- Bateman, I. J., R. T. Carson, B. Day, W. M. Hanemann, N. Hanley, T. Hett, eries Management 26:301–307. M. Jones Lee, G. Loomes, S. Mourato, E. Ozdemiroglu,¨ D. W. Pearce, R. Loomis, J. 2007. Correcting for on-site visitor sampling bias when estimating Sugden, and J. Swanson. 2002. Economic valuation with stated preference the regional economic effects of tourism. Tourism Economics 13:41–47. techniques: a manual. Edward Elgar, Northampton, Massachusetts. Miranda, L. E., J. D. Dagel, L. Kaczka, E. B. Mower, S. L. Wigen, and K. Bohnsack, B. L., R. B. Ditton, J. R. Stoll, R. J. Chen, R. Novak, and L. S. Smutko. O. Meals. 2012. Promoting crappie recruitment in northwest Mississippi 2002. The economic impacts of the recreational Bluefin Tuna fishery in reservoirs. Mississippi Department of Wildlife, Fisheries and Parks, Jackson. Hatteras, North Carolina. North American Journal of Fisheries Management Mulholland, P. J., G. R. Best, C. C. Coutant, G. M. Hornberger, J. L. Meyer, P. 22:165–176. J. Robinson, J. R. Stenberg, R. E. Turner, F. Vera-Herrera, and R. G. Wetzel. Chen, R. J., K. M. Hunt, and R. B. Ditton. 2003. Estimating the economic 1997. Effects of climate change on freshwater ecosystems of the south-eastern impacts of a trophy Largemouth Bass fishery: issues and applications. North United States and the Gulf Coast of Mexico. Hydrological Processes 11:949– American Journal of Fisheries Management 23:835–844. 970.

Downloaded by [Department Of Fisheries] at 19:26 20 January 2013 Chizinski, C. J., K. L. Pope, D. B. Willis, G. R. Wilde, and E. J. Rossman. 2005. Munn, I. A., A. Hussain, S. Spurlock, and J. E. Henderson. 2010. Economic Economic value of angling at a reservoir with low visitation. North American impact of fishing, hunting, and wildlife-associated recreation expenditures Journal of Fisheries Management 25:98–104. on the southeast U.S. regional economy: an input–output analysis. Human Connelly, N. A., and T. L. Brown. 1991. Net economic value of the freshwater Dimensions of Wildlife 15:433–449. recreational fisheries of New York. Transactions of the American Fisheries Olson, D., and S. Lindall. 2000. IMPLAN professional R , 2nd edition. Minnesota Society 120:770–775. IMPLAN Group, Stillwater. Criddle, K. R., M. Herrmann, S. T. Lee, and C. Hamel. 2003. Participation Platt, J., and D. Munger. 1999. Impact of fluctuating reservoir elevation on recre- decisions, angler welfare, and the regional economic impact of sportfishing. ational use and value. U.S. Department of Interior, Bureau of Reclamation, Marine Resource Economics 18:291–312. Technical Memorandum EC-1999-1, Denver. Crompton, J. L., S. Lee, and T. J. Shuster. 2001. A guide for undertaking Posadas, B. C. 2009. Economic impact of the Mississippi recreational fishing economic impact studies: the Springfest example. Journal of Travel Research industry at the year 2006. Coastal Research and Extension Center, Mississippi 40:79–87. State University, Biloxi. Daugherty, D. J., D. L. Buckmeier, and P. K. Kokkanti. 2011. Sensitivity of Riechers, R. K., G. C. Matlock, and R. B. Ditton. 1991. A dual-survey ap- recreational access to reservoir water level variation: an approach to iden- proach for estimating the economic aspects of fishing. Pages 344–355 in D. tify future access needs in reservoirs. North American Journal of Fisheries Guthrie, J. M. Hoenig, M. Holliday, C. M. Jones, M. J. Mills, S. A. Moberly, Management 31:63–69. K. H. Pollock, and D. R. Talhelm, editors. Creel and angler surveys in fish- Dillman, D. A. 2007. Mail and internet surveys: the tailored design method, 2nd eries management. American Fisheries Society, Symposium 12, Bethesda, edition. Wiley, Hoboken, New Jersey. Maryland. RECREATIONAL FISHERIES IN MISSISSIPPI RESERVOIRS 55

Schorr, M. S., J. Sah, D. F. Schreiner, M. R. Meador, and L. G. Hill. 1995. M. Hoenig, M. Holliday, C. M. Jones, M. J. Mills, S. A. Moberly, K. H. Regional economic impact of the Lake Texoma (Oklahoma-Texas) Striped Pollock, and D. R. Talhelm, editors. Creel and angler surveys in fisheries Bass fishery. Fisheries 20(5):14–18. management. American Fisheries Society, Symposium 12, Bethesda, Mary- Steinback, S. R. 1999. Regional economic impact assessments of recreational land. fisheries: an application of the IMPLAN modeling system to marine party USFWS (U.S. Fish and Wildlife Service) and USCB (U.S. Census Bureau). and charter boat fishing in Maine. North American Journal of Fisheries Man- 2008. 2006 national survey of fishing, hunting, and wildlife-associated agement 19:724–736. recreation: Mississippi. USFWS and USCB, FHW/06-MS, Washington, Stoll, J. R., and R. B. Ditton. 2006. Understanding anglers’ willingness to D.C. pay under alternative management regimes. Human Dimensions of Wildlife Ward, F. A., B. A. Roach, J. B. Loomis, R. C. Ready, and J. E. Hen- 11:27–42. derson. 1996. Regional recreation demand models for large reservoirs: Sun, G., S. G. McNulty, J. Lu, D. M. Amatya, Y. Liang, and R. K. Kolka. 2005. database development, model estimation, and management applications. Regional annual water yield from forest lands and its response to potential U.S. Army Corps of Engineers, Technical Report R-96-2, Washington, deforestation across the southeastern United States. Journal of Hydrology D.C. 308:258–268. Wood, A. W., D. P. Lettenmaier, and R. N. Palmer. 1997. Assessing cli- Swanson, C. S., and D. W. McCollum. 1991. Application of economics to recre- mate change implications for water resources planning. Climatic Change 37: ational fisheries management: an overview. Pages 299–315 in D. Guthrie, J. 203–228. Downloaded by [Department Of Fisheries] at 19:26 20 January 2013 This article was downloaded by: [Department Of Fisheries] On: 20 January 2013, At: 19:29 Publisher: Taylor & Francis Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

North American Journal of Fisheries Management Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/ujfm20 Movement and Habitat Differentiation among Adult Shoal Bass, Largemouth Bass, and Spotted Bass in the Upper Flint River, Georgia Matthew R. Goclowski a , Adam J. Kaeser b & Steven M. Sammons a a Department of Fisheries, Auburn University, 203 Swingle Hall, Auburn, Alabama, 36849, USA b U.S. Fish and Wildlife Service, Panama City Fish and Wildlife Conservation Office, 1601 Balboa Avenue, Panama City, Florida, 32405, USA Version of record first published: 09 Jan 2013.

To cite this article: Matthew R. Goclowski , Adam J. Kaeser & Steven M. Sammons (2013): Movement and Habitat Differentiation among Adult Shoal Bass, Largemouth Bass, and Spotted Bass in the Upper Flint River, Georgia, North American Journal of Fisheries Management, 33:1, 56-70 To link to this article: http://dx.doi.org/10.1080/02755947.2012.741555

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material. North American Journal of Fisheries Management 33:56–70, 2013 C American Fisheries Society 2013 ISSN: 0275-5947 print / 1548-8675 online DOI: 10.1080/02755947.2012.741555

Movement and Habitat Differentiation among Adult Shoal Bass, Largemouth Bass, and Spotted Bass in the Upper Flint River, Georgia

Matthew R. Goclowski* Department of Fisheries, Auburn University, 203 Swingle Hall, Auburn, Alabama 36849, USA Adam J. Kaeser U.S. Fish and Wildlife Service, Panama City Fish and Wildlife Conservation Ofﬁce, 1601 Balboa Avenue, Panama City, Florida 32405, USA Steven M. Sammons Department of Fisheries, Auburn University, 203 Swingle Hall, Auburn, Alabama 36849, USA

Abstract The Shoal Bass Micropterus cataractae is a ﬂuvial specialist endemic to the Apalachicola River drainage in Alabama, Florida, and Georgia that has experienced declines throughout much of its range. The Flint River, Georgia, represents the largest remaining intact ecosystem for Shoal Bass in their native range. Spotted Bass M. punctulatus have recently been introduced into this system, causing concern about the potential negative impacts the species may have on the native populations of Shoal Bass and Largemouth Bass M. salmoides. To assess the symmetry and strength of competition and gain the greatest perspective on the interrelationships among these sympatric, congeneric species, we compared the movement patterns and habitat use of all three species of black bass present in this system. Fifteen Shoal Bass, 10 Largemouth Bass, and 6 Spotted Bass were implanted with radio transmitters in the Flint River and tracked for a period of 1 year (2008). Daily and hourly movements did not vary among species or season, though individuals of each species were observed moving >5 km to shoal complexes during spring. Habitat overlap varied between species during the study; overlap was highest between Spotted Bass and Largemouth Bass, intermediate between Spotted Bass and Shoal Bass, and lowest between Shoal Bass and Largemouth Bass. Shoal Bass tended to select coarse rocky habitat, while Largemouth Bass tended to select depositional habitat. Spotted Bass exhibited the widest niche breadth and generally used habitat in proportion to its availability. Use of similar habitats by these three species during the spring spawning period highlights the potential risk of genetic introgression of the two native species by introduced Spotted Bass. Physical barriers that restrict access to habitat during long-distance seasonal Downloaded by [Department Of Fisheries] at 19:29 20 January 2013 movements, as observed for several Shoal Bass in this study, may negatively impact populations of this species.

The introduction of nonnative fishes is recognized as a prin- 2011). The Shoal Bass Micropterus cataractae is endemic to cipal cause of the imperilment of native fishes throughout North the Apalachicola–Chattahoochee–Flint (ACF) River basin in America (Jelks et al. 2008). Interactions between nonnative and Alabama, Florida, and Georgia (Williams and Burgess 1999). resident fishes are varied and include direct or indirect com- Populations of Shoal Bass are declining throughout much of petition for resources such as shelter, food, or spawning ar- their native range; as a result, the Shoal Bass is a species of eas, in addition to hybridization among closely related species high conservation concern in Alabama (Mirarchi et al. 2004), (Whitmore 1983; Scoppettone 1993; Cucherousset and Olden considered threatened in Florida (Gilbert 1992), and listed as

*Corresponding author: [email protected] Received March 31, 2012; accepted October 12, 2012

56 BASS MOVEMENT AND HABITAT DIFFERENTIATION 57

vulnerable over its entire range by the American Fisheries So- highest relative abundances in shoal and riffle habitat in Al- ciety (Williams et al. 1989; Jelks et al. 2008). Cited threats to abama streams, where Spotted Bass have replaced Shoal Bass Shoal Bass include habitat loss or degradation and the potential as the dominant bass species (Sammons and Maceina 2009). for competition and hybridization with introduced bass species Although Spotted Bass in their native range have been shown (Williams and Burgess 1999; Wheeler and Allen 2003; Stormer to exist sympatrically with other black bass species through re- and Maceina 2008). Significant declines in the distribution of source partitioning (Vogele 1975; Scott and Angermeier 1998; Shoal Bass in Chattahoochee River tributaries (Stormer and Ma- Sammons and Bettoli 1999), introductions of Spotted Bass into ceina 2008) prompted the Alabama Department of Conservation streams and reservoirs have been implicated in the decline of and Natural Resources to close the harvest of Shoal Bass in Al- several native bass species, including Shoal Bass (Koppelman abama waters, the first such closure of a Micropterus fishery in 1994; Pierce and Van Den Avyle 1997; Barwick et al. 2006; the state (Maceina et al. 2007). Stormer and Maceina 2008). Shoal Bass are described as habitat specialists that associate In recent years, populations of introduced Spotted Bass have with high-velocity, riffle-and-run habitats containing boulder become established in the upper Flint and Ocmulgee rivers in and bedrock substrates in lotic systems (i.e., shoals; Wheeler Georgia. Spotted Bass were first collected in the upper Flint and Allen 2003; Stormer and Maceina 2009). Although Shoal River in 2005, and their numbers have increased rapidly through- Bass are capable of surviving and even reproducing in ponds out the river (J. Evans, Georgia Department of Natural Re- (Smitherman and Ramsey 1972), they do not persist in natu- sources, personal communication). The recent introduction of ral systems that have been impounded (Williams and Burgess Spotted Bass in the upper Flint River has caused great concern 1999). Johnston and Kennon (2007) suggested that Shoal Bass among anglers and fisheries managers about the potential neg- require complex shoal habitats with a variety of microhabitats at ative impacts attributable to competition and hybridization of various life stages in order to persist. Shoal Bass in unrestricted Spotted Bass with native black bass species. Thus, the objective rivers appear to congregate in large shoal complexes to spawn of this study was to assess the potential for resource compe- (Sammons and Goclowski 2012), and high densities of age-0 tition among the three bass species by examining the spatial Shoal Bass are typically found in these areas during the sum- and temporal patterns of movement and habitat use of adult fish mer (Goclowski 2010), indicating that these habitats are also co-occurring in a reach of the upper Flint River. At the time of important nursery areas. the study, the relative abundances of all three bass species were Spotted Bass M. punctulatus were first found in the ACF similar, making it an ideal location at which to examine their basin around 1941; these fish were found below the fall line resource use. defining the boundary between the Piedmont and Coastal Plain physiographic provinces (Williams and Burgess 1999). How- ever, the species was slow to spread within the system, likely METHODS owing to the presence of numerous upstream dams, and spec- Study site.—The Flint River flows 565 km from its head- imens were not recorded above the fall line until 1968, possi- waters near Atlanta, Georgia, to its confluence with the Chat- bly from a separate stocking event. As of 1999, Spotted Bass tahoochee River at Lake Seminole. This study was conducted were distributed throughout the Chattahoochee River system but along a 33.8-km stretch of the upper third of the Flint River, had not been found above the Albany Dam on the Flint River Georgia, from Flat Shoals (rkm 457.8) downstream to Sprewell (Williams and Burgess 1999). Bluff State Park (rkm 424.0; Figure 1). This section of the river Although Shoal Bass did not occur naturally with Spotted flows through the Piedmont physiographic province of Georgia Bass, they do occur sympatrically with the Largemouth Bass M. and is characterized by a series of wide (up to 250 m), granite Downloaded by [Department Of Fisheries] at 19:29 20 January 2013 salmoides throughout their entire range (Williams and Burgess shoal areas with shallow water and higher current velocity inter- 1999). Wheeler and Allen (2003) found that adult Shoal Bass spersed with runs and pools exhibiting deeper water and lower and Largemouth Bass in the Chipola River, Florida, had very velocity. The Flint River ranges from approximately 50–250 m similar diets; however, they used different habitats and appeared in width throughout this section, with a mean annual discharge to coexist through habitat partitioning. The diets of Shoal Bass of 63.3 m3/s at the U.S. Geological Survey (USGS) gauge lo- and Largemouth Bass in the Flint River, Georgia, showed little cated at Carsonville, Georgia (USGS site 02347500), approxi- overlap, whereas the diets of Shoal Bass and Spotted Bass were mately 40 rkm downstream of the study area. Major sport fishes more similar (Sammons 2012). Shoal Bass were more abundant occurring in the upper Flint River include Largemouth Bass, in the shallower, faster-moving shoal areas in the Chipola River, Shoal Bass, Spotted Bass, Channel Catfish Ictalurus puncta- Florida, while Largemouth Bass were more abundant in the tus, Flathead Catfish Pylodictus olivarus, and Redbreast Sunfish deeper, slower-moving pools and backwater areas. The Spot- Lepomis auritus. ted Bass has been described as a habitat generalist that often Telemetry and movement.—In December 2007, 11 Shoal inhabits a variety of areas, including shallow rocky riffles and Bass, 6 Spotted Bass, and 10 Largemouth Bass were collected shoals (Hurst 1969; Vogele 1975; Tillma et al. 1998; Horton using a boat-mounted DC electrofishing unit from the Flint River and Guy 2002). However, Shoal Bass and Spotted Bass had the in the vicinity of the Georgia Highway 18 Bridge (Figure 1) 58 GOCLOWSKI ET AL. Downloaded by [Department Of Fisheries] at 19:29 20 January 2013

FIGURE 1. Map of the study site on the upper Flint River showing the area in which ﬁsh were tagged, along with the locations of major shoals, Sprewell Bluff State Park, and the USGS gauging station at the Georgia Highway 18 Bridge. [Figure available in color online.] BASS MOVEMENT AND HABITAT DIFFERENTIATION 59

and surgically implanted with an Advanced Telemetry Systems TABLE 1. Classification scheme developed for the upper Flint River substrate (ATS) radio transmitter as described by Maceina et al. (1999). In map. June 2008, four additional Shoal Bass were captured in the same Substrate class Definition river section and implanted with transmitters recovered from fish that had died or shed their transmitters. Two size-groups of Sand ≥75% of area composed of particles <2mm fish were used in this study. Fish weighing 180–700 g received in diameter (sand, silt, clay, or fine organic 3.6-g radio transmitters (ATSModel F1580) with a 258-d battery detritus) life expectancy, whereas fish >700 g received 14-g transmitters Rocky fine >25% of area composed of rocks >2mm (ATS Model F1835) with a 502-d battery life expectancy. Im- but <500 mm in diameter across the planted transmitters were <2% of fish body weight in order to longest axis ensure that movement and behavior would not be affected (Win- Rocky boulder An area that includes >3 boulders, each ter 1996). The large transmitters were equipped with mortality >500 mm in diameter across the longest sensors. The small transmitters did not have a mortality sensor, axis, each boulder within 1.5 m of the so mortality was assumed for any fish that did not move during adjacent boulder. Any area meeting these three consecutive tracking events. Because Spotted Bass had criteria, regardless of underlying substrate, become established in the Flint River only a few years prior to is classified as rocky boulder. this study, no fish were collected that could be implanted with Bedrock ≥75% of area composed of bedrock or an the larger-sized tags. Thus, only small Spotted Bass were used outcropping with relatively smooth texture in this study. (not fractured into blocks >500 mm in Tracking activities commenced 2 weeks after surgery in Jan- diameter) uary 2008 and were conducted approximately every 14 d until Mixed rocky An area comprising two or more substrates December 2008. To assess diel movements, fish were located ev- classes (at least one being rocky) arranged ery 6 h over a 24-h period during every other site visit (Sammons such that no homogeneous portion is et al. 2003). Tagged fish were located by manually tracking the >10 m2 study reach by boat with an ATS Model R2000 receiver and di- No data An area beyond the sonar range but within rectional yagi antenna. Fish locations were defined as the points the boundaries of the river channel at which the signal was strongest when the antenna was pointed directly at the water. The location of each fish was recorded using a Lowrance iFinder H20 Global Positioning System (GPS) (species, seasons, diel periods) and random effects (individual unit, and time of day, water depth, velocity, and temperature fish) (Maceina et al. 1994; Wilkerson and Fisher 1997; Bunnell were recorded at each location. During May 2008, an aerial re- et al. 1998; SAS 2002; Sammons and Maceina 2005). A Bonfer- connaissance of approximately 120 rkm was conducted using roni correction (P < 0.05/n) was used for multiple comparisons a fixed-wing aircraft fitted with two wing-mounted directional among species, seasons, and diel periods. yagi antennae to locate fish that had left the study area. River Habitat mapping and analysis.—During March 2009, a stage and discharge data during tracking activities were obtained single-pass sonar survey of the study area was conducted us- from the USGS gauge located at the Georgia Highway 18 Bridge ing a Humminbird 981c Side Imaging system at a frequency near Molena (USGS site 02344872; Figure 1). of 455 kHz and a range of 26 m (85 ft) per side, as described The daily movement rate of each fish was calculated in in Kaeser and Litts (2010). The survey began at the base of terms of minimum displacement per day by dividing the dis- Flat Shoals (Figure 1) and continued 22 km downstream, en- Downloaded by [Department Of Fisheries] at 19:29 20 January 2013 tance moved in meters between locations by the time elapsed compassing all of the known locations of radio-tagged bass. (Colle et al. 1989; Wilkerson and Fisher 1997; Sammons et al. These data were geoprocessed following the methods described 2003). The diel movement rate of each fish located during the in Kaeser and Litts (2010) to create a map of habitat features 24-h tracking sessions was similarly calculated in terms of min- that included the predominant substrate types, river banks, and imum displacement per hour. Fish movement was compared large woody debris. The substrates present in the surveyed reach among four seasons as defined by water temperature: winter, were classified according to a scheme used during mapping of <12◦C; spring, increasing from 12◦Cto22◦C; summer, greater the lower 124 km of the Flint River (Kaeser et al., in press), with than >22◦C; and fall, decreasing from 22◦Cto12◦C (Todd one modification; a general bedrock class replaced the two lime- and Rabeni 1989; Wilkerson and Fisher 1997; Sammons et al. stone bedrock classes present in the lower Flint River scheme 2003). Diel movement observations were pooled into two pe- (Table 1). riods: day (the hours between sunrise and sunset) and night Habitat associations among the three species were investi- (the hours between sunset and sunrise). Mean daily and diel gated using a reduced data set that included only a single day- movement rates were log10-transformed to normalize the data time observation (midday) per tagged individual per relocation and compared among seasons, species, and diel periods using survey in order to keep the time intervals consistent between a repeated-measures analysis of variance (ANOVA) with fixed observations. Observed fish locations were overlaid with the 60 GOCLOWSKI ET AL.

substrate, riverbank, and large woody debris layers in a GIS to BLOSSOM (Cade and Richards 2005) was used to execute the extract habitat data from the map relevant to fish locations. The MRPP; test statistics were based on Euclidean distances to re- Euclidean distance from each fish location to the edge of the duce the influence of outlying observations (Mielke and Berry nearest polygon in each substrate class was calculated using the 2001), and variables were commensurated using the mean Eu- NEAR tool in ArcGIS 10 (ESRI 2007). A distance value of 0 was clidean distance for each variable. Following a significant MRPP entered for the substrate class that contained the fish location. A test, differences in multivariate dispersions were tested using a similar approach was taken to calculate the distance of each fish permutation version of a modified Van Valen’s test in BLOS- location to the riverbank and the nearest piece of large woody SOM (Van Valen 1978; Atkinson et al. 2010). Using MRPP, debris. A distance-based approach was chosen for the habitat the global differences among all groups were tested first, fol- use analysis instead of a classification-based approach because lowed by pairwise comparisons between species groups; tests some positional error (average, <10 m), attributable to the GPS were considered significant at a Bonferroni-corrected α level of equipment used during the study, was inherent in both the fish 0.0125. The chance-corrected within-group (A) was calculated locations and map data (Kaeser and Litts 2010). A distance- as described by McCune and Grace (2002), and reported as a based approach is not only robust to positional error (Conner measure of effect size. The Euclidean distance between comet al. 2003) but also preserves the complexity of the information mensurated, multivariate medians was calculated and used as provided by a spatially complete (i.e., full-census) map of the a measure of the difference in central tendency between two habitat features throughout the study area. An index of substrate species habitat use distributions; the average within-group dis- complexity within the vicinity of a fish location (hereafter re- tance to the multivariate median was used as a measure of dis- ferred to as the “edge”) was generated by creating a 15-m buffer persion (i.e., the breadth of habitat use). around each location and calculating the length of all substrate Discriminant analysis was conducted to describe the pat- boundaries (i.e., lines) captured by each buffer. A count of the tern and degree of habitat partitioning along gradients defined number of pieces of large woody debris present within each by the habitat variables examined in this study and to com- 15-m buffer was calculated in a similar manner. To conduct an pare ecological niches and niche overlap among species (SAS inventory of the available habitat, a regular grid of points spaced 2002). Given the descriptive purposes of this analysis, the ob- 3 m apart was generated and habitat data were extracted from served locations of fish in the reduced data set were simply the map for each point as described above for all fish locations. pooled by species. Discriminant analysis generates linear com- Depth and flow data were not incorporated into the mul- binations of the environmental variables (i.e., discriminant fac- tivariate habitat data matrix because several gaps in the data tors) that maximize the variance among species while mini- record occurred due to gear malfunction. Instead, differences mizing within-species variance, thereby identifying variables between flow and depth were tested among species, seasons, that best separate or uniquely define the niche of each species and diel periods (fixed effects; the random effects consisted of (McGarigal et al. 2000). As a single variable that represents individual fish) using mixed-model ANOVAs (SAS 2002). A a composite of multiple habitat variables, the first discrimi- Bonferroni correction was applied during all multiple compar- nant factor was used to calculate niche indices based on habitat isons of movement and habitat use. association and availability. Hurlbert’s B, an index that con- Multiple-response permutation procedures (MRPP) were siders the variation in habitat availability, was used to describe used to test for overall differences in multivariate habitat use niche breadth, and Hurlbert’s niche overlap (L), a metric that de- among the three species. This procedure is a nonparametric ap- scribes the probability of interspecific overlap in habitat use rel- proach that does not require the distributional assumptions of ative to the frequency distribution of available habitat was used multivariate normality and homogeneity of variances (Mielke to describe potential niche overlap among the species (Hurl- Downloaded by [Department Of Fisheries] at 19:29 20 January 2013 1984; Zimmerman et al. 1985; Mielke and Berry 2001). Us- bert 1978). Chi-square analysis was used to determine whether ing MRPP, the null hypothesis of no difference between two or species used habitats in proportion to their relative abundance by more groups is tested while simultaneously testing for differ- comparing the observed habitat use distribution of each species ences in central tendency and dispersion. To avoid the issue of with the distribution of total available habitat as defined by the pseudoreplication inherent in treating telemetry locations as the first discriminant factor. sampling unit (Rogers and White 2007), median values were calculated for each habitat variable per individual, and these RESULTS individual-based, median habitat vectors were examined during MRPP tests. Medians rather than means were selected for Telemetry and Movement this analysis to reduce the influence of outlying observations of A total of 27 tracking events were conducted over the duration habitat use. To limit the analysis to individuals observed over of the study, resulting in a total of 677 fish locations. The reduced the duration of the study, we included only those fish that had data set used to analyze multivariate fish habitat use included been located on >10 occasions in the MRPP data set. Sample 120 Largemouth Bass locations, 67 Shoal Bass locations, and sizes by species in the MRPP analysis were as follows: 8 Large- 83 Spotted Bass locations between January 4 and September mouth Bass, 6 Shoal Bass, and 6 Spotted Bass. The program 28, 2008. At the end of the study, 3 Largemouth Bass and 1 BASS MOVEMENT AND HABITAT DIFFERENTIATION 61

TABLE 2. Species (Largemouth Bass [LMB], Shoal Bass [SHB], and Spotted Bass [SPB]), total length, weight, number of days at large, number of locations, and fate of ﬁsh tracked in the upper Flint River during 2008.

Species TL (mm) Weight (g) Days at large No. locations Fate LMB 380 660 273 36 Tag expired 337 461 146 13 Harvested by angler 275 222 168 13 Consumed by bird 324 367 253 35 Died 247 180 168 22 Died 394 850 370 46 Study ended 478 1,389 370 45 Study ended 426 937 369 45 Study ended 392 770 167 16 Missing 462 1,594 225 27 Died SHB 258 212 273 17 Tag expired 294 300 273 17 Harvested by angler 345 471 168 18 Died 279 252 25 1 Died 342 520 273 30 Tag expired 300 300 226 15 Harvested by angler 508 2,040 156 4 Missing 467 1,483 155 15 Died 493 1,562 25 1 Died 492 1,409 145 10 Died 409 863 301 18 Missing 392 875 126 13 Harvested by angler 499 1,872 224 18 Died 354 470 102 15 Tag expired 415 851 169 19 Study ended SPB 323 447 273 31 Tag expired 261 210 224 26 Died 281 202 302 41 Tag expired 286 240 272 17 Tag expired 260 207 271 36 Tag expired 290 254 272 17 Tag expired

Shoal Bass with active transmitters remained in the study area. ping Rock Shoals, a large shoal complex that began about 5 km Of the remaining fish, 10 died or expelled transmitters, 9 were downstream from the Georgia Highway 18 Bridge (Figure 1). Downloaded by [Department Of Fisheries] at 19:29 20 January 2013 tracked until their transmitter battery failed, 4 were known to Two additional Shoal Bass tags with mortality sensors engaged be harvested by anglers, 1 was consumed by a bird, and 3 were were found in close proximity to this shoal. During the aerial missing and never relocated during the study (Table 2). survey, one Shoal Bass was located in a shoal near Sprewell All telemetered fish remained in close proximity to their Bluff State Park, approximately 20 km downstream of its tag- original capture locations for most of the winter season. In late ging location (Figure 1); this individual returned to the tagging February and early March, individuals of all three species (3 of area within 10 d of location by the aerial survey. Two emigrating 10 Largemouth Bass, 10 of 11 Shoal Bass, and 3 of 6 Spotted Shoal Bass were unaccounted for until they returned to the study Bass) emigrated from this region of the study area. Subsequent area in early May and June. All of the nonemigrating fish and telemetry surveys and the aerial survey revealed that most of all returning fish generally remained in close proximity to their these fish had migrated 5–8 km toward large shoal complexes. tagging areas throughout the study. Three Largemouth Bass moved upstream to the base of Flat The mean annual daily movement of tagged fish ranged from Shoals, a large shoal complex located 9 km upstream from the 178 to 430 m/d (Table 3) and was similar among species (F = Georgia Highway 18 Bridge (Figure 1); one Spotted Bass was 2.77; df = 2, 25; P = 0.082) or among seasons for each species located approximately 1 km below this shoal complex. Five (F ≤ 1.36; df ≥ 3, 12; P ≥ 0.21; Figure 2). The mean annual diel Shoal Bass and two Spotted Bass moved downstream into Drip- movement of all species ranged from 76 to 119 m/h (Table 3) 62 GOCLOWSKI ET AL.

TABLE 3. Annual mean movement rates, depth use, and ﬂow use of radio-tagged black bass in the Flint River, Georgia during 2008.

Largemouth Bass Shoal Bass Spotted Bass Variable Mean (n) SE Mean (n) SE Mean (n)SE Diel movement (m/h) 119 (10) 43.6 97 (11) 30.8 76 (6) 17.6 Daily movement (m/d) 288 (10) 74 430 (11) 210 178 (6) 83.7 Depth (m) 1.69 (10) 0.06 1.64 (11) 0.10 1.86 (6) 0.01 Flow (m/s) 0.05 (10) 0.01 0.15 (11) 0.02 0.08 (6) 0.01 Downloaded by [Department Of Fisheries] at 19:29 20 January 2013

FIGURE 2. Seasonal mean (A) daily and (B) diel (hourly) movement rates of Largemouth Bass, Shoal Bass, and Spotted Bass in the upper Flint River in 2008. No Shoal Bass were observed in 24-h tracking events during the spring because they had all migrated out of the diel tracking area. The error bars represent SEs. BASS MOVEMENT AND HABITAT DIFFERENTIATION 63

central tendency between the distributions for Largemouth Bass and Spotted Bass was 1.54, that between the distributions for Shoal Bass and Spotted Bass was 1.81. The first discriminant factor had a relatively low eigenvalue 2 (0.335) and low squared canonical correlation value (Rc = 0.25), indicating the relatively low power of this function to discriminate among species based on the habitat variables examined. Likewise, the plot of species frequency distributions along the first discriminant axis illustrated that habitat overlap occurred among the fish observed in this study (Figure 4). Group means along the first discriminant axis were closely spaced (Largemouth Bass, –0.56; Spotted Bass, 0.11; Shoal Bass, 0.87), with the greatest separation occurring between the distributions of Largemouth Bass and Shoal Bass. The observed frequency of habitat use by Largemouth Bass and Shoal Bass differed from the frequency distribution of total available habitat in the study χ2 = = < FIGURE 3. Movement rates of black bass in diel periods in the Flint River area (Largemouth Bass: 130; df 28; P 0.0001; Shoal 2 during 2008. Different lowercase letters denote significant differences between Bass: χ = 64; df = 28; P = 0.0001). The observed frequency diel periods within species (P ≤ 0.05); the error bars represent SEs. of habitat use by Spotted Bass did not differ statistically from the distribution of total available habitat (χ2 = 34; df = 28; and did not differ among species (F = 0.31; df = 2, 24; P = P = 0.2200). 0.7329) or among seasons for any species (F ≤ 4.58; df ≥ 2, 3; The first discriminant axis represented an ecological gradient P ≥ 0.06; Figure 2). Diel mean movement was greater during generally defined as depositional and woody on one end, and daylight hours than at night for Largemouth Bass (F = 6.44; coarse rocky with less wood on the other (i.e., shoals with limited df = 1, 9; P = 0.03; Figure 3) but was similar between periods LWD; Table 4). A set of Largemouth Bass (40%) and Spotted for Shoal Bass and Spotted Bass (F ≤ 3.20; df ≥ 1, 6; P ≥ 0.10; Bass (20%) locations were associated with depositional areas Figure 3). comprised of sandy substrate and LWD at greater distances from rocky shoals (i.e., low scores on the discriminant axis); this multivariate habitat type was not used by Shoal Bass. Half Habitat Differentiation among Bass Species of all observed locations of Largemouth Bass were within 3 m of The results of sonar habitat mapping indicated that the 99.7- sandy substrate (Table 4). Likewise, a set of Shoal Bass (22%) ha study area had an overall substrate composition of 49% sand, and Spotted Bass (7%) locations were associated with coarse 19% rocky boulder, 8% rocky fines, 14% bedrock, and 3% mixed rocky habitat, low abundance of LWD, and greater distance rocky substrate. Missing data accounted for the remaining 7% from depositional habitats (i.e., high scores on the discriminant of the area. A total of 3,117 pieces of large woody debris (LWD) axis); these locations occurred in the large downstream shoal were mapped throughout the study area. Most LWD was located used by Shoal and Spotted Bass during the spring spawning near the stream margins, and LWD density was noticeably lower period (Figure 4; Table 4). Although the variable “distance to in coarse rocky areas than in runs and deeper reaches of the study bedrock” did not contribute to the discrimination of species area. along the first axis, Shoal Bass clearly exhibited the strongest Downloaded by [Department Of Fisheries] at 19:29 20 January 2013 Habitat use differed among the three species (MRPP: A = affinity for bedrock substrate (Table 4). Half of all observed 0.079; P = 0.0065; Van Valen’s test: P = 0.028). The within- Shoal Bass locations were within 15.6 m of bedrock substrate, group dispersion for each species (Largemouth Bass = 1.52; compared with median distances of 56.8 m for Spotted Bass and Shoal Bass = 3.22; Spotted Bass = 2.08) showed that Shoal 102.9 m for Largemouth Bass. Bass had greater variation in habitat use among individuals Hurlbert’s niche breadth (B) was highest for Spotted Bass than Spotted Bass and Largemouth Bass. Pairwise comparisons (0.749), next highest for Shoal Bass (0.611), and lowest for between species further suggested that the significant global Largemouth Bass (0.435). Niche overlap between Largemouth MRPP test result was primarily attributable to differences in the and Shoal Bass was less than would be expected (L = 0.785) if central tendency of habitat use between Largemouth Bass and both species used habitat in proportion to its availability; overlap Shoal Bass (MRPP: A = 0.131; P = 0.0011; Van Valen’s test: P was greater than expected (L = 1.372) between Largemouth = 0.023); the Euclidean difference in central tendency between and Spotted Bass. When Shoal Bass and Spotted Bass used the two species was 3.02. Pairwise comparisons between Large- the same habitats, their use was directly proportional to habitat mouth Bass and Spotted Bass (MRPP: A = 0.036; P = 0.116), availability (L = 1.007). and Shoal Bass and Spotted Bass habitat use (MRPP: A = 0.018; The distributions of the depths used were similar among P = 0.236) were not statistically significant. The difference in species throughout the study (Kolmogorov–Smirnov test: KSa 64 GOCLOWSKI ET AL.

TABLE 4. Within-species and total available habitat, multivariate median coordinates for habitat variables with respect to the reduced data set of 270 fish locations, and total canonical structure coefficients for habitat variables used to define the first discriminant factor (CAN1). All distance metrics, including the variable edge, are in meters. See Table 2 for species abbreviations.

Habitat variable LMB SHB SPB Total habitat CAN1 Distance to sand 2.9 24.4 10.1 58.1 0.665 Distance to rocky ﬁnes 41.3 34.5 105.6 89.6 0.508 Distance to rocky boulders 189.2 149.4 140 192.8 –0.451 Distance to mixed rocky substrate 276.9 238.3 223.8 308.9 0.615 Distance to bedrock 102.9 15.6 56.8 92.2 –0.133 Distance to bank 9.8 16.4 11.9 14.6 0.131 Distance to LWD 10.1 18.2 10.5 23.2 0.473 LWD within 15-m buffera 3.11 2.0 2.2 1.73 –0.562 Edge 36.6 43.3 37.1 34.6 –0.103

aNumber of pieces.

< 1.31; P ≥ 0.0646; Figure 5), and mean depth did not vary range of flow values than Largemouth Bass and Spotted Bass among Shoal Bass (1.64 m), Largemouth Bass (1.69 m), and during this study (KSa > 1.62; P < 0.0106; Figure 5). Similarly, Spotted Bass (1.86 m) (F = 0.45; df = 2, 26; P = 0.6454). the mean flow used by Shoal Bass (0.15 m/s) was higher than The mean depths used did not vary among diel periods for any that of Spotted Bass (0.08 m/s) and Largemouth Bass (0.05 m/s) species (F ≤ 1.68; df ≥ 3, 15; P ≥ 0.17). Shoal Bass were (F = 10.99; df = 2, 26; Bonferroni-adjusted P < 0.0413), which more commonly found in flows >0.2 m/s and used a wider were similar (Bonferroni-adjusted P = 0.6067). However, the Downloaded by [Department Of Fisheries] at 19:29 20 January 2013

FIGURE 4. Frequency distributions of observed fish locations and total available habitat in the study area as defined by the first discriminant factor (canonical axis 1). BASS MOVEMENT AND HABITAT DIFFERENTIATION 65 Downloaded by [Department Of Fisheries] at 19:29 20 January 2013

FIGURE 5. Distributions of depth and ﬂow values used by radio-tracked Largemouth Bass (LMB), Shoal Bass (SHB), and Spotted Bass (SPB) in the upper Flint River during 2008. Distributions with the same letter were similar among species (Komogorov–Smirnov test; P ≤ 0.05). 66 GOCLOWSKI ET AL.

mean flow at fish locations did not vary among diel periods for summer. Each of these fish moved back to approximately the any species (F ≤ 3.29; df ≥ 3, 9; P ≥ 0.0520). area where it had been located before emigrating; one of them actually moved back to the exact same rocky outcrop that it DISCUSSION had inhabited before leaving and remained there for the duration of the study. Similar homing behavior has been observed Movement for Smallmouth Bass in other studies (Todd and Rabeni 1989; Mean daily movement rates did not vary among species or Langhurst and Schoenike 1990; VanArnum et al. 2004). seasons during this study; however, the highest observed dis- As the Flint River experienced atypical low-flow conditions placements occurred during the spring, when individuals of during 2008, the observed fish movements may have been less all three species moved distances in excess of 5 km, presum- than normal. Low-water conditions during droughts can reduce ably to spawning areas. Seasonal differences in movement rates habitat connectivity and restrict fish movements (Lake 2003; have been observed for several species of stream-dwelling black Schrank and Rahel 2006). Tagged Shoal Bass in the Flint River bass, including Shoal Bass (Stormer and Maceina 2009), Spot- were commonly recaptured by anglers 60–100 km away from ted Bass (Horton and Guy 2002), and Smallmouth Bass Mi- where they were tagged, and many of these large migrations cropterus dolomieu (Montgomery et al. 1980; Todd and Rabeni were associated with high-discharge events (Sammons and Go- 1989; Langhurst and Schoenike 1990). Spawning migrations clowski 2012). The return rate of these fish to their former home have been reported in some stream-dwelling basses, notably areas is unknown, but the low flows observed in summer 2008 Smallmouth Bass (Montgomery et al. 1980; Todd and Rabeni may have restricted Shoal Bass movements and reduced the 1989). In this study, all eight Shoal Bass and two of the Spotted likelihood of additional fish exhibiting homing behavior during Bass that were found outside of the study area during spring this study. A long-term telemetry study of a greater number of were located within, or in very close proximity to, major shoal Shoal Bass may provide more insight into the seasonal move- complexes. Our observations suggest that large shoal complexes ment patterns and homing behavior of Shoal Bass in the Flint serve as important spawning and nursery areas for Shoal Bass River. and Spotted Bass in the upper Flint River. Similar spawning aggregations of Shoal Bass have been observed in shoal habi- Habitat Use tats during spring elsewhere in the Flint River (Sammons and Within our Flint River study area, distinct habitat differen- Goclowski 2012; T. Ingram, Georgia Department of Natural tiation was only evident between Largemouth and Shoal Bass. Resources [GDNR], unpublished data) as well as in several The habitat use of Largemouth and Shoal Bass differed from the tributary streams of the Chattahoochee River, Georgia (Sam- distribution of available habitat, and niche overlap was lowest mons 2011; J. Slaughter, Georgia Power Company, unpublished between the two species. Both Largemouth and Shoal Bass are data). The co-occurrence of Shoal and Spotted Bass in this habi- native to the Flint River, so that partitioning may have evolved tat during the spawning season in the Flint River suggests that to support their coexistence (Wheeler and Allen 2003). Miller competition for nesting areas, genetic introgression, and inter- (1975) documented a similar occurrence of habitat partitioning actions between these species at early life stages are potential among three sympatric black bass species in which Spotted Bass conservation concerns. showed habitat preferences between those of Largemouth Bass The largest observed displacement in this study involved a and Smallmouth Bass: Largemouth Bass inhabited deep pools Shoal Bass that was located approximately 20 km downstream and quiet backwaters, Smallmouth Bass inhabited fast-moving from the study area during an aerial tracking survey. Within waters, and Spotted Bass were found in intermediate areas in 10 d, this individual had returned to the study area, very close to shallow pools near fast-moving water. Downloaded by [Department Of Fisheries] at 19:29 20 January 2013 its point of departure. This fish had not been located on tracking Spotted Bass are often described as habitat generalists (Hurst trips conducted through the same area as recently as 2 weeks 1969; Vogele 1975; Tillma et al. 1998), and our results broadly earlier, indicating that it was further downstream. A related support this characterization. Although Horton and Guy (2002) study of Shoal Bass on the Flint River documented spring up- reported that Spotted Bass used pools more often than runs and stream movements as great as 200 km (Sammons and Goclowski riffles in Kansas streams and suggested that Spotted Bass prefer 2012). Similar upstream movements of 70–120 km have been low-velocity environments, a very different pattern of habitat use found for Shoal Bass in the lower Flint River downstream of by Spotted Bass was observed in the upper Flint River. Spotted Albany, Georgia (T. Ingram, GDNR, unpublished data). Long- Bass were found both in pools and in shoals, over depositional range movements (i.e., >50 km) have been documented for areas and within shallow rocky reaches of the upper Flint River. Smallmouth Bass in other river systems (Montgomery et al. Spotted Bass (and Shoal Bass) exhibited high variation in habitat 1980; Langhurst and Schoenike 1990), but spring spawning mi- use, but only Spotted Bass used habitats in proportion to their grations of Shoal Bass may be unusually large for Micropterus availability. Spotted Bass habitat use broadly overlapped that spp. Although most of the Shoal Bass that emigrated from the of both Largemouth Bass and Shoal Bass, indicating that direct study site remained near the shoals to which they moved, four or indirect competition for resources may be imposed on both of them returned to the study area during late spring or early native species by the introduced Spotted Bass. Hurlbert’s niche BASS MOVEMENT AND HABITAT DIFFERENTIATION 67

overlap was highest between Spotted Bass and Largemouth Bass the use of flows by Largemouth Bass and Spotted Bass were because the adults of both species sometimes used a type of surprisingly similar throughout the study. Prevailing drought habitat that was not particularly abundant in the study area. conditions resulted in abnormally low discharge levels (up to This habitat type was represented by low scores on the first 87% below mean monthly discharge at USGS gauge 02347500) discriminant axis and could be described as sandy or rocky– in the Flint River during the summer and fall of 2008. Low dis- fine reaches with high LWD abundance, far removed from shoal charge levels may have reduced the heterogeneity of velocity areas. Although Spotted Bass were associated with habitats used levels throughout our study area, making it more difficult to de- by both Largemouth Bass and Shoal Bass, these results suggest tect differences in flow use among species during these seasons. that the strength of competition is greater between Spotted and Johnston and Kennon (2007) found that Shoal Bass in Little Largemouth Bass in the upper Flint River. Uchee Creek, Alabama, used lower water velocities in summer Large woody debris is an important component of many during a dry year than they did during a wet year. If the Flint stream ecosystems. Instream wood provides stable substrate for River study had been conducted during a wet year, more pro- aquatic invertebrate production (Angermeier and Karr 1984; nounced differences in flow use among species may have been Benke et al. 1985), offers fish refuge from strong current veloc- observed, particularly during summer and fall. ities (Crook and Robertson 1999), provides fish with cover in which to hide from predators and ambush prey (Angermeier and Management Implications Karr 1984; Crook and Robertson 1999), and can play a large Few Spotted Bass >350 mm were found in the Flint River role in stream channel formation (Abbe and Montgomery 1996). during this study, and collecting fish >700 g for the larger tag Many studies have documented the importance of large woody sizes was not possible; thus, a low number of Spotted Bass in debris to black bass in lotic environments (Todd and Rabeni a small size range were used in this study. A different study 1989; Tillma et al. 1998; Horton and Guy 2002), although the design, one that either restricted all species to the available sizes relative importance of woody debris habitat does not appear to of Spotted Bass or tagged larger Spotted Bass, may have resulted be consistent across species. Wheeler and Allen (2003) found in different findings. However, we feel that this study provides that Largemouth Bass in the Chipola River, Florida, were as- managers with useful preliminary data with which to evaluate sociated with areas of higher than average woody debris index the effects of introduced Spotted Bass on congeneric species in scores, whereas Shoal Bass presence was not related to woody a lotic system. debris index scores. Scott and Angermeier (1998) found that Drought conditions, as experienced in this study, physically Spotted Bass in the New River, Virginia, occupied areas that fea- reduce the habitat available for partitioning and intensify in- tured woody cover, overhanging bank vegetation, and undercut teractions between and within species (Matthews and Marsh- banks, while Smallmouth Bass occupied rocky areas that lacked Matthews 2003). We observed that the habitat use of Spotted woody cover. We observed that all three species of black bass Bass overlapped with that of both Shoal Bass and Largemouth were located closer to, and near greater quantities of, LWD than Bass and suspect that the intensity of overlap and concomitant would be expected from the total available habitat. Although competition for resources may have been elevated through an the association between Shoal Bass and LWD was weaker than ecological crunch as described by Wiens (1977). Thus, under that for Largemouth and Spotted Bass, the association appeared normal flow conditions, habitat overlap among these species, to be context dependent. For example, when Shoal Bass were especially in terms of water velocity, may be much lower than located within 5 m of rocky boulder habitat (a class of habitat was found during our study. However, agricultural, industrial, that provides multidimensional cover), the median distance of and residential uses of water in the Flint River watershed are fish to large woody debris was 62 m; when they were located high and increasing (Richter et al. 2003; Opsahl et al. 2007). The Downloaded by [Department Of Fisheries] at 19:29 20 January 2013 near bedrock and all other habitats, the median distance of Shoal cumulative effects of water use can alter low-flow periodicity Bass to woody debris was 13 m, similar to that of Largemouth and longevity and either mimic or exacerbate natural drought Bass and Spotted Bass. We infer from these results that Shoal conditions, leading to increased competition among aquatic or- Bass associate with LWD as cover when not closely associ- ganisms for reduced habitat and food and ultimately resulting in ated with rocky boulder habitat. Shoal Bass farther downstream fish assemblage shifts and species declines (Richter et al. 2003; in the coastal plain of the Flint River are commonly found in Freeman and Marcinek 2006; Johnston and Maceina 2009). The close association with woody debris (J. Evans, GDNR, personal results from this study may indicate that the competitive effects communication); thus, large woody debris can be important for among native and introduced black bass species are intensified the maintenance of Shoal Bass populations in these areas. The during droughts. Although the long-term consequences of ele- association of black bass with large woody debris emphasizes vated competition could not be examined in this study, managers the importance of conserving this habitat feature and maintain- should be aware that continued increases in water use throughout ing the processes responsible for recruiting woody debris to the this basin may favor Spotted Bass over Shoal Bass, particularly stream system. in smaller tributary streams (Stormer and Maceina 2008). Thus, As expected, Shoal Bass were typically found using higher Shoal Bass and Spotted Bass abundances throughout the Flint velocities than the other two species in the Flint River. However, River basin should be monitored closely in the future. 68 GOCLOWSKI ET AL.

Shoals are a critical habitat type for Shoal Bass in the Flint ACKNOWLEDGMENTS River, and they may serve as important spawning or nursery Funding was provided by the Georgia Department of Nat- areas. Management efforts to protect the endemic Shoal Bass ural Resources. Ryan Hunter, Jonathan Brown, Tyler Thomas, should focus on conserving shoal habitat and preserving con- and Benjamin Hutto provided field assistance. Laurie Earley nectivity throughout this unimpounded river reach. Shoal Bass provided GIS expertise to create Figure 1. form spawning aggregations in shoals during the spring, further emphasizing the need to protect these habitats and connectivity. Because both Shoal Bass and Spotted Bass were ob- REFERENCES served moving into shoal areas during the spawning season, Abbe, T. B., and D. R. Montgomery. 1996. Large woody debris jams, channel there is the potential for introgressive hybridization between hydraulics and habitat formation in large rivers. Regulated Rivers: Research and Management 12:201–221. these species, which may alter the Shoal Bass gene pool and Angermeier, P. L., and J. R. Karr. 1984. Relationships between woody debris threaten the persistence of native stocks of the species. Native and fish habitat in a small warmwater stream. Transactions of the American stocks of Guadalupe Bass M. treculii in Texas were threatened Fisheries Society 113:716–726. after the introduction of nonnative Smallmouth Bass into sev- Atkinson, C. L., S. P. Opsahl, A. P. Covich, S. W. Golladay, and L. M. Conner. eral streams. Researchers identified introgressive hybridization 2010. Stable isotopic signatures, tissue stoichiometry, and nutrient cycling (C and N) of native and invasive freshwater bivalves. Journal of the North between the two species as the primary threat to the contin- American Benthological Society 29:496–505. ued persistence of Guadalupe Bass (Whitmore 1983; Littrell Barwick, D. H., K. J. Oswald, J. M. Quattro, and R. D. Barwick. 2006. Red- et al. 2007). Similarly, native stocks of Redeye Bass M. coosae eye Bass (Micropterus coosae) and Alabama Spotted Bass (M. punctulatus in South Carolina are threatened by introgressive hybridiza- henshalli) hybridization in Keowee Reservoir. Southeastern Naturalist 5:661– tion with nonnative Alabama Bass M. henshalli (Barwick et al. 668. Benke, A. C., R. L. Henry III, D. M. Gillespie, and R. J. Hunter. 1985. Impor- 2006). Introduced Spotted Bass may pose a similar threat to tance of snag habitat for animal production in southeastern streams. Fisheries Shoal Bass, as they have been found to hybridize with several 10(5):8–13. black bass species in tributaries of the Chattahoochee River, Bunnell, D. B., Jr., J. J. Isely, K. H. Burrell, and D. H. Van Lear. 1998. Diel including Shoal Bass and Largemouth Bass (Maceina et al. movement of Brown Trout in a southern Appalachian river. Transactions of 2007). Further research should be conducted in the upper Flint the American Fisheries Society 127:630–636. Cade, B. S., and J. D. Richards. 2005. User manual for BLOSSOM statistical River to assess the potential for long-term impacts related to software. U.S. Geological Survey, Midcontinent Ecological Science Center, hybridization. Open File Report 2005-1353, Fort Collins, Colorado. The large-scale movements documented in this study, cou- Colle, D. E., R. L. Cailteux, and J. V. Shireman. 1989. Distribution of Florida pled with observations from an exploitation study and additional Largemouth Bass in a lake after elimination of all submersed aquatic vegeta- telemetry data on Shoal Bass from the coastal plain section of tion. North American Journal of Fisheries Management 9:213–218. Conner, L. M., M. D. Smith, and L. W. Burger. 2003. A comparison of distance- the Flint River (Sammons and Goclowski 2012), highlight the based and classification-based analyses of habitat use. Ecology 84:526– importance of connectivity throughout the Flint River for Shoal 531. Bass. Shoal Bass in the river below Albany, Georgia, have been Crook, D. A., and A. I. Robertson. 1999. Relationships between riverine fish observed to move as much as 100 km to reach a spawning and woody debris: implications for lowland rivers. Marine and Freshwater shoal complex located a few kilometers below a dam (T. In- Research 50:941–953. Cucherousset, J., and J. D. Olden. 2011. Ecological impacts of non-native fresh- gram, GDNR, personal communication). Above this dam is an water fishes. Fisheries 36:215–230. extensive set of shoals that was likely the main spawning area ESRI (Environmental Systems Research Institute). 2007. ArcGIS, release 9.2. for these fish before the dam was constructed. The GDNR has ESRI, Redlands, California. stocked Shoal Bass in this section of the river for the past two Freeman, M. C., and P. A. Marcinek. 2006. Fish assemblage responses to water Downloaded by [Department Of Fisheries] at 19:29 20 January 2013 decades to bolster the low natural recruitment likely stemming withdrawals and water supply reservoirs in Piedmont streams. Environmental Management 38:435–450. from loss of connectivity and altered flow regimes. Currently, Gilbert, C. R., editor. 1992. Rare and endangered biota of Florida, volume II: authorization of new dams on the upper Flint River is being fishes. University Press of Florida, Gainesville. discussed as a possible solution to the ongoing water allocation Goclowski, M. R. 2010. Relations between Shoal Bass and sympatric congeneric conflicts in the ACF River basin among Georgia, Alabama, and basses in the Flint River, Georgia. Master’s thesis. Auburn University, Auburn, Florida (Jones 2008). Measures should be taken to prevent any Alabama. Horton, T. B., and C. S. Guy. 2002. Habitat use and movement of Spotted Bass new impoundments from being constructed on the Flint River, in Otter Creek, Kansas. Pages 161–171 in D. P. Philipp and M. S. Ridg- as the likely consequences would be less availability of prime way, editors. Black bass: ecology, conservation, and management. American spawning and nursery habitat, along with a concomitant reduc- Fisheries Society, Symposium 31, Bethesda, Maryland. tion in gene flow. Evidence of the effects of impoundments on Hurlbert, S. H. 1978. The measurement of niche overlap and some relatives. Ecology 59:67–77. Shoal Bass has been seen in the Chattahoochee River system, Hurst, H. N. 1969. Comparative life history of the Redeye Bass, Micropterus where Spotted Bass thrive in the reservoir habitat and Shoal coosae (Hubbs and Bailey), and the Spotted Bass, Micropterus p. punctu- Bass persist in fragmented, relict populations below the dams latus (Rafinesque), in Halawakee Creek, Alabama. Master’s thesis. Auburn (Sammons and Maceina 2009). University, Auburn, Alabama. BASS MOVEMENT AND HABITAT DIFFERENTIATION 69

Jelks, H. L., S. J. Walsh, N. M. Burkhead, S. Contreras-Balderas, E. D´ıaz-Pardo, Opsahl, S. P., S. E. Chapal, D. W. Hicks, and C. K. Wheeler. 2007. Evaluation D. A. Hendrickson, J. Lyons, N. E. Mandrak, F. McCormick, J. S. Nelson, of ground-water and surface-water exchanges using streamflow difference S. P. Platania, B. A. Porter, C. B. Renaud, J. J. Schmitter-Soto, E. B. Taylor, analyses. Journal of the American Water Resources Association 43:1132– and M. L. Warren Jr. 2008. Conservation status of imperiled North American 1141. freshwater and diadromous fishes. Fisheries 33:372–407. Pierce, P. C., and M. J. Van Den Avyle. 1997. Hybridization between intro- Johnston, C. E., and R. A. Kennon. 2007. Habitat use of the Shoal Bass Mi- duced Spotted Bass and Smallmouth Bass in reservoirs. Transactions of the cropterus cataractae, in an Alabama stream. Journal of Freshwater Ecology American Fisheries Society 126:939–947. 22:493–498. Richter, B. D., R. Mathews, D. L. Harrison, and R. Wigington. 2003. Ecolog- Johnston, C. E., and M. J. Maceina. 2009. Fish assemblage shifts and species ically sustainable water management: managing river flows for ecological declines in Alabama, USA streams. Ecology of Freshwater Fish 18:33–40. integrity. Ecological Applications 13:206–224. Jones, L. B. 2008. Resolving disputes over the management of federal reservoirs: Rogers, K. B., and G. C. White. 2007. Analysis of movement and habitat use evaluating the litigation alternative. Pages 63–71 in M. S. Allen, S. Sammons, from telemetry data. Pages 625–676 in C. S. Guy and M. L. Brown, editors. and M. J. Maceina, editors. Balancing fisheries management and water uses Analysis and interpretation of freshwater fisheries data. American Fisheries for impounded river systems. American Fisheries Society, Symposium 62, Society, Bethesda, Maryland. Bethesda, Maryland. Sammons, S. M. 2011. Habitat use, movement, and behavior of Shoal Bass, Mi- Kaeser, A. J., and T. L. Litts. 2010. A novel technique for mapping habi- cropterus cataractae, in the Chattahoochee River near Bartletts Ferry Reser- tat in navigable streams using low-cost side scan sonar. Fisheries 35:163– voir. Final Report to Georgia Power Company, Atlanta. 174. Sammons, S. M. 2012. Diets of juvenile and sub-adult size classes of three Kaeser, A. J., T. L. Litts, and T. W. Tracy. In press. Using low-cost side-scan Micropterus spp. in the Flint River, Georgia: potential for trophic competition. sonar for benthic mapping throughout the lower Flint River, Georgia, USA. Southeastern Naturalist 11:387–404. River Research and Applications. DOI: 10.1002/rra.2556. Sammons, S. M., and P. W. Bettoli. 1999. Spatial and temporal variation in Koppelman, J. B. 1994. Hybridization between Smallmouth Bass, Micropterus electrofishing catch rates of three species of black bass (Micropterus spp.) dolomieu, and Spotted Bass, M. punctulatus, in the Missouri River system, from Normandy Reservoir, Tennessee. North American Journal of Fisheries Missouri. Copeia 1994:204–210. Management 19:454–461. Lake, P.S. 2003. Ecological effects of perturbation by drought in flowing waters. Sammons, S. M., and M. R. Goclowski. 2012. Relations between Shoal Bass Freshwater Biology 48:1161–1172. and sympatric congeneric black bass species in Georgia rivers with emphasis Langhurst, R. W., and D. L. Schoenike. 1990. Seasonal migration of Smallmouth on movement patterns, habitat use, and recruitment. Final Report to Georgia Bass in the Embarrass and Wolf rivers, Wisconsin. North American Journal Department of Natural Resources, Social Circle, Montgomery. of Fisheries Management 10:224–227. Sammons, S. M., and M. J. Maceina. 2005. Activity patterns of Largemouth Littrell, B. M., D. J. Lutz-Carrillo, T. H. Bonner, and L. T. Fries. 2007. Status of Bass in a subtropical U.S. reservoir. Fisheries Management and Ecology an introgressed Guadalupe Bass population in a central Texas stream. North 12:331–339. American Journal of Fisheries Management 27:785–791. Sammons, S. M., and M. J. Maceina. 2009. Conservation status of Shoal Bass in Maceina, M. J., P. W. Bettoli, and D. R. DeVries. 1994. Use of a split-plot analy- Alabama: distribution, abundance, stocking efficacy, and possible effects of sis of variance design for repeated-measures fishery data. Fisheries 19(3):14– sympatric congeneric black bass in selected tributaries of the Chattahoochee 20. River, Alabama. Final Report to the Alabama Division of Wildlife and Fresh- Maceina, M. J., C. E. Johnston, and D. G. Stormer. 2007. Distribution, abun- water Fisheries, Montgomery. dance, biology, and habitat use of Shoal Bass and species associates in selected Sammons, S. M., M. J. Maceina, and D. G. Partridge. 2003. Changes in behav- tributaries of the Chattahoochee River, Alabama. Final Report to the Alabama ior, movement, and home ranges of Largemouth Bass following large-scale Department of Conservation and Natural Resources, Montgomery. hydrilla removal in Lake Seminole, Georgia. Journal of Aquatic Plant Man- Maceina, M. J., J. W. Slipke, and J. M. Grizzle. 1999. Effectiveness of agement 41:31–38. three barrier types for confining Grass Carp in embayments of Lake Semi- SAS (Statistical Analysis Systems). 2002. SAS system for linear models, release nole, Georgia. North American Journal of Fisheries Management 19:968– 9.1. SAS Institute, Cary, North Carolina. 976. Schrank, A. J., and F. J. Rahel. 2006. Factors influencing summer movement Matthews, W. J., and E. Marsh-Matthews. 2003. Effects of drought on fish patterns of Bonneville Cutthroat Trout (Oncorhynchus clarkii utah). Canadian across axes of space, time and ecological complexity. Freshwater Biology Journal of Fisheries and Aquatic Sciences 63:660–669. 48:1232–1253. Scoppettone, G. G. 1993. Interactions between native and nonnative fishes of the

Downloaded by [Department Of Fisheries] at 19:29 20 January 2013 McCune, B., and J. B. Grace. 2002. Analysis of ecological communities. MjM upper Muddy River, Nevada. Transactions of the American Fisheries Society Software Design, Gleneden Beach, Oregon. 122:599–608. McGarigal, K., S. Cushman, and S. Stafford. 2000. Multivariate statistics for Scott, M. C., and P. L. Angermeier. 1998. Resource use by two sympatric black wildlife and ecology research. Springer-Verlag, New York. basses in impounded and riverine sections of the New River, Virginia. North Mielke, P. W., Jr. 1984. Meteorological applications of permutation techniques American Journal of Fisheries Management 18:221–235. based on distance functions. Pages 813–830 in P. R. Krishnaiah and P. K. Smitherman, R. O., and J. S. Ramsey. 1972. Observations on spawning and Sen, editors. Handbook of statistics, volume 4. Elsevier, New York. growth of four species of basses (Micropterus) in ponds. Proceedings of the Mielke, P. W., Jr., and K. J. Berry. 2001. Permutation methods: a distance Annual Conference Southeastern Association of Game and Fish Commis- function approach. Springer-Verlag, New York. sioners 25(1971):357–365. Miller, R. J. 1975. Comparative behavior of centrarchid basses. Pages 85–94 in Stormer, D. G., and M. J. Maceina. 2008. Relative abundance, distribution, and R. H. Stroud and H. E. Clepper, editors. Black bass biology and management. population metrics of Shoal Bass in Alabama. Journal of Freshwater Ecology Sport Fishing Institute, Washington, D.C. 23:651–661. Mirarchi, R. E., J. T. Garner, M. F. Mettee, and P. E. O’Neil, editors. 2004. Stormer, D. G., and M. J. Maceina. 2009. Habitat use, home range, and move- Alabama wildlife, volume two: imperiled aquatic mollusks and ﬁshes. Uni- ment of Shoal Bass in Alabama. North American Journal of Fisheries Man- versity of Alabama Press, Montgomery. agement 29:604–613. Montgomery, J. C., D. H. Fickeisen, and C. D. Becker. 1980. Factors inﬂuencing Tillma, J. S., C. S. Guy, and C. S. Mammoliti. 1998. Relations among habitat and Smallmouth Bass production in the Hanford area, Columbia River. Northwest population characteristics of Spotted Bass in Kansas streams. North American Science 54:296–305. Journal of Fisheries Management 18:886–893. 70 GOCLOWSKI ET AL.

Todd, B. L., and C. F. Rabeni. 1989. Movement and habitat use by stream- Wilkerson, M. L., and W. L. Fisher. 1997. Striped Bass distribution, movements, dwelling Smallmouth Bass. Transactions of the American Fisheries Society and site ﬁdelity in Robert S. Kerr Reservoir, Oklahoma. North American 118:229–242. Journal of Fisheries Management 17:677–686. VanArnum, C. J. G., G. L. Buynak, and J. R. Ross. 2004. Movement of Small- Williams, J. D., and G. H. Burgess. 1999. A new species of bass, Micropterus mouth Bass in Elkhorn Creek, Kentucky. North American Journal of Fisheries cataractae (Teleostei: Centrarchidae), from the Apalachicola River basin in Management 24:311–315. Alabama, Florida, and Georgia. Bulletin of the Florida Museum of Natural Van Valen, L. 1978. The statistics of variation. Evolutionary Theory 4:33–43. History 42:80–114. Vogele, L. E. 1975. The Spotted Bass. Pages 34–45 in R. H. Stroud and H. E. Williams, J. E., J. E. Johnson, D. A. Hendrickson, S. Contreras-Balderas, J. D. Clepper, editors. Black bass biology and management. Sport Fishing Institute, Williams, M. Navarro-Mendoza, D. E. McAllister, and J. E. Deacon. 1989. Washington, D.C. Fishes of North America endangered, threatened, or of special concern: 1989. Wheeler, A. P., and M. S. Allen. 2003. Habitat and diet partitioning between Fisheries 14(6):2–20. Shoal Bass and Largemouth Bass in the Chipola River, Florida. Transactions Winter, J. 1996. Advances in underwater biotelemetry. Pages 555–590 in B. R. of the American Fisheries Society 132:438–449. Murphy and D. W.Willis, editors. Fisheries techniques, 2nd edition. American Whitmore, D. H. 1983. Introgressive hybridization of Smallmouth Bass (Mi- Fisheries Society, Bethesda, Maryland. cropterus dolomieui) and Guadalupe Bass (M. treculi). Copeia 1983:672–679. Zimmerman, G. M., H. Goetz, and P.W. Mielke Jr. 1985. Use of an improved sta- Wiens, J. A. 1977. On competition and variable environments. American Sci- tistical method for group comparisons to study effects of prairie ﬁre. Ecology entist 65:590–597. 66:606–611. Downloaded by [Department Of Fisheries] at 19:29 20 January 2013 This article was downloaded by: [Department Of Fisheries] On: 20 January 2013, At: 19:29 Publisher: Taylor & Francis Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

North American Journal of Fisheries Management Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/ujfm20 Multicompartment Gravel Bed Flume to Evaluate Swim- Up Success of Salmonid Larvae N. L. Pilgrim a , R. Royer a , L. Ripley b , J. Underwood b , J. B. Rasmussen a & A. Hontela a a Water Institute for Sustainable Environments (WISE), University of Lethbridge, 4401 University Drive West, Lethbridge, Alberta, T1K 3M4, Canada b Allison Creek Brood Trout Station, 6076 Allison Creek Road, Coleman, Alberta, T0K 0M0, Canada Version of record first published: 09 Jan 2013.

To cite this article: N. L. Pilgrim , R. Royer , L. Ripley , J. Underwood , J. B. Rasmussen & A. Hontela (2013): Multicompartment Gravel Bed Flume to Evaluate Swim-Up Success of Salmonid Larvae, North American Journal of Fisheries Management, 33:1, 71-78 To link to this article: http://dx.doi.org/10.1080/02755947.2012.741556

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material. North American Journal of Fisheries Management 33:71–78, 2013 C American Fisheries Society 2013 ISSN: 0275-5947 print / 1548-8675 online DOI: 10.1080/02755947.2012.741556

MANAGEMENT BRIEF

Multicompartment Gravel Bed Flume to Evaluate Swim-Up Success of Salmonid Larvae

N. L. Pilgrim and R. Royer Water Institute for Sustainable Environments (WISE), University of Lethbridge, 4401 University Drive West, Lethbridge, Alberta T1K 3M4, Canada L. Ripley and J. Underwood Allison Creek Brood Trout Station, 6076 Allison Creek Road, Coleman, Alberta T0K 0M0, Canada J. B. Rasmussen and A. Hontela* Water Institute for Sustainable Environments (WISE), University of Lethbridge, 4401 University Drive West, Lethbridge, Alberta T1K 3M4, Canada

(Barlaup et al. 1994; Brann¨ as¨ 1995), swim-up success is a key Abstract reproductive endpoint and a potential bottleneck for recruitment Swim-up success, the proportion of fry emerging from a of populations (Coleman and Fausch 2007). It is difficult to gravel redd, is difficult to quantify in the field, and currently capture newly emerged fry in the field and relate these to the available laboratory systems are limited. We used custom-built gravel bed flumes to assess the swim-up success of Rainbow number of eggs laid, and systems used in the laboratory to Trout Oncorhynchus mykiss, Brook Trout Salvelinus fontinalis, estimate swim-up success are limited by the small number of and Cutthroat Trout Oncorhynchus clarkii. Flumes were built with clutches they can accommodate and the flow characteristics. compartments to separate individual egg batches, eggs were buried We describe here a custom-built, multicompartment gravel bed under gravel, and oxygenated water was supplied to simulate up- flume that was used to estimate swim-up success of up to 18 welling through the gravel. Temperature, dissolved oxygen, pH, and flow were constant between compartments and flumes oper- egg clutches simultaneously and simulate conditions optimal ating in parallel. Swim-up success was scored both by the emer- for developing salmonid larvae. gence of fry relative to the number of eggs placed in the flume In the wild, salmonids locate suitable spawning sites based (59–85% depending on species) and the hatching success (propor- on species-specific habitat characteristics, including adequate tion of eggs that hatch and resorb their yolk sac) of a sample from stream flow, groundwater upwelling, and substrate (Crisp and the same clutch reared in a vertical incubator (62–84% depending on species). The gravel bed flumes could be used to estimate the Carling 1989; Snucins et al. 1992; Curry et al. 1995; Malcolm Downloaded by [Department Of Fisheries] at 19:29 20 January 2013 swim-up success of salmonid egg clutches from hatchery stocks et al. 2003). Fertilized eggs are deposited into a gravel depres- or fish from the wild or under experimental regimes of relevance sion created by the spawning pair and covered by flushed up to fishery or environmental assessment, including changes in pH, gravel (Barlaup et al. 1994; Blanchfield et al. 2003). Eggs hatch siltation, temperature, toxicants, or events simulating floods or into larvae (alevins) which remain under the gravel until most droughts. or all of their yolk has been absorbed, and the “buttoned-up” fry swim up to the surface through the interstitial spaces of the Reproductive endpoints are used to compare the performance gravel (Godin 1980; Brann¨ as¨ 1989; Fraser et al. 1994). This phe- of individuals, genetic strains, or populations and to evaluate nomenon, referred to as “swim-up”, is an important test of larval the impacts of environmental stressors (Barlaup et al. 1994; fitness (Brann¨ as¨ 1989). Many factors can impede swim-up of Curry et al. 1995; Rudolph et al. 2008; Granier et al. 2011). For fry, including fine sediments clogging interstitial spaces, larval fish species that bury their eggs under gravel, leaving them to deformities, and edema (Kondolf 2000; Sternecker and Geist develop into larvae that swim to the surface of the gravel bed 2010). Moreover, prior to emergence many eggs and alevins

*Corresponding author: [email protected] Received July 12, 2012; accepted October 11, 2012 71 72 PILGRIM ET AL.

die from lack of oxygen, fungus, predation, or disease (Snucins vertical incubator tray to estimate hatching success (proportion et al. 1992; Rubin 1995; Dumas and Marty 2006). of eggs that hatch and resorb their yolk sac, i.e. “button-up”). Swim-up is challenging to study in the field because fry are Swim-up success then represents the proportion of the eggs that difficult to capture, swim-up can occur over an extended period, hatched and emerged above gravel, divided by the proportion and the number of eggs buried by spawning fish is not known of fry that hatched and resorbed their yolk sac in the vertical (Rubin 1995; Groves et al. 2008). Researchers have simulated incubator. redds in the laboratory, where a known number of eggs is placed Flume design.—Two replicate flumes were made from sheets in an artificial redd and all fry that emerge are captured. One of of plexiglass bonded together with methylene chloride. The di- the first designs was a gravel-filled box, with water entering from mensions of the flumes were 200 cm (length) × 40 cm (width) × the top and trickling down through the gravel (Dill 1970). Eggs 30 cm (height), with 8-mm notches (4-mm deep) cut in the side or alevins were inserted under the gravel using a pipe. Godin panel every 10 cm along the length of the flumes, for insertion (1980) modified this design by switching the inflow and outflow of removable dividers (Figure 1) to create 20 compartments. pipes to create upwelling, and Fraser et al. (1994) placed this Dividers were made by bonding aluminum window screen be- “redd box” in a flume with an outflow at the back of the flume and tween two plexiglass frames, 40 cm (length) × 25 cm (height) space behind the box to retain swim-up. Other designs covered × 4.76 mm (width) and 3.175 mm (width), with a 5-cm border. the gravel with plastic, leaving larvae to swim up through the The outflow hole (diameter = 1.9 cm) was drilled 44 mm from pipes (Godin 1980; Brann¨ as¨ 1989; Mirza et al. 2001; Sundstrom¨ the top center of the back panel and a bulkhead was placed et al. 2005). Palm et al. (2009) buried eggs under gravel at the in the hole to streamline water flow out of the downstream end bottom of an aquarium and used a propeller at the front of the of the flume (Figure 1). tank to increase water velocity. Treated dechlorinated (City of Lethbridge) water was deliv- To allow separation of individual clutches to compare swim- ered at 32 pounds per square inch gauge through a 2.54-cm- up success of egg batches from different females under the same diameter hose connected to a PVC pipe (diameter = 6.35 cm) conditions, Fudge et al. (2008) divided the box into compart- that ran the length of each flume. From this pipe, five PVC pipes ments separated by mesh screens and created upwelling through (diameter = 1.27 cm) entered the flume at compartment 2, 6, 10, pipes in the bottom. A pump was installed at one side of the box 14, and 18 and connected to forks with four capped prongs that to increase water flow and an intake at the other end allowed ran along the bottom of the flume to supply water to sections recirculation of water. This design separated egg batches and A–E of the flume (Figure 1). Five evenly spaced perforations provided upwelling, but it did not simulate natural stream flow, (diameter = 0.16 cm) were drilled into each prong running along as the water was funnelled towards the outflows located at the the bottom of each compartment, ensuring each compartment back of the central compartment. Sternecker and Geist (2010) had its own water flow upwelling through the gravel (see Gravel used four-compartment boxes arranged in series in a flume, with bed simulation, below), as shown in Figure 1. Ball valves were water delivered through the front of the flume, entering the boxes placed in between the pipe connected to the hose and the main through mesh screen at the bottom and exiting through a screen pipe running the length of the flume and also on each pipe that at the back of the box. ran into the flume from the main pipe. Ball valves were par- The objective of the present study was to create artificial tially closed at the top of the flume, closest to the water source, redds in a gravel bed flume that had separate compartments for and left open at the bottom of the flume to create uniform wa- multiple (n = 18) egg batches to be tested simultaneously, with ter pressure and flow throughout the entire flume, as pressure constant upwelling and flow through each compartment. The decreased along the main pipe due to distance from the water flume was designed to make the collection of swim-up larvae source. Flumes were flushed for several days with treated water Downloaded by [Department Of Fisheries] at 19:29 20 January 2013 easy and accurate, to allow for flexibility in the size of the to ensure any residue from the bonding agents as well as air compartments, to increase ease of dismantling and storage, to pockets, which could cause interference in flow, were removed provide potential for use in the field, and to be built with low cost from the pipes. Each compartment was then filled with gravel, materials. To test the performance of our artificial redd flume, and flushed with processed water for a minimum of 24 h, be- we compared swim-up success in three species, Rainbow Trout fore fertilized eggs were deposited at a specific depth under the Oncorhynchus mykiss, Brook Trout Salvelinus fontinalis, and gravel (see Gravel bed simulation). Cutthroat Trout Oncorhynchus clarkii. Vertical incubator.—A 16 tray MariSource vertical incubator was used in this study to compare hatching success ([number of buttoned-up fry/number of eggs deposited in incubator trays] × METHODS 100) in the incubator and emergence of swim-up fry in the flumes Experimental approach.—A known number of eggs from ([number of fry emerged from gravel/number of eggs deposited each clutch was deposited into compartments of two custom- in flume redd compartment] × 100). The incubator was also built gravel flumes to estimate the proportion of larvae that used to compare lag time between development of larvae in the emerged from the gravel (swim-up), and simultaneously a incubator and appearance of fish larvae above the gravel in the subsample of the eggs from each clutch was placed into a flumes. MANAGEMENT BRIEF 73

FIGURE 1. Flume schematic presented in sections (A–E) to demonstrate chronological time in the experiment: A–empty flume with PVC pipes, B–compartments separated by screens with 5 cm of gravel (for Rainbow Trout) added to the compartments, C–eyed eggs added in oval-shaped “redd”, D–compartments have gravel topped up to 20 cm, and E–emerged fry appear above the gravel in each compartment. (Two replicates flumes were used in each experiment; only one flume is shown in the schematic).

Downloaded by [Department Of Fisheries] at 19:29 20 January 2013 Water parameters.—Water quality parameters were recorded of washed round rock (28 mm and 14 mm) purchased from a in three areas of each flume; the first section (top), midway local landscaping center were mixed 1:1. These sizes were cho- (middle), and just before outflow (bottom). Subtle adjustments sen since similar sizes have been reported for salmonid redds to flow rate (3.5–4.0 L/min) or ball valve positions were executed in the literature (Crisp and Carling 1989; Kondolf 2000). The in order to maintain equal flow conditions in all compartments of gravel was rinsed with water to remove any fines, disinfected the flumes. Dissolved oxygen and temperature were measured with a 1% Ovadine solution (Syndel Laboratories), and thor- using a YSI-85 probe, and pH was measured using a CEL850 oughly rinsed with processed water. Gravel was added to the Sension1 portable pH meter with Platinum Series electrode. The flumes to a depth of 5 cm for Rainbow Trout (see Figure 1) and optimal conditions required for swim-up were 8–10◦C, water 10 cm for Brook Trout and Cutthroat Trout. Eggs were gently flow of 3.5–4 L/min in the flume and 18–20 L/min in the vertical deposited (see Placement of eggs and collection of emerged fry, incubator, pH of 8, and 8–10 mg O2/L. See Table 1 for water below) onto the gravel, followed by adding 15 cm of gravel for quality data. Rainbow Trout, and 10 cm of gravel for Brook and Cutthroat Gravel bed simulation.—Trout spawning redds were simu- Trout for a total gravel depth of 20 cm (Figure 1). Thus Rainbow lated in the experimental flumes to create standardized, optimal Trout eggs were placed deeper in the gravel than the Brook and conditions for the fry to swim up through the gravel. Two sizes Cutthroat Trout eggs. The mature Rainbow Trout were much 74 PILGRIM ET AL.

TABLE 1. Water quality in the gravel bed flume and in the vertical incubator. Temperature, O2, and pH are given as mean ± SE. No significant differences were detected between different sections in the flumes tested at the same time (Student’s t-test, P < 0.05). Flow rate Area Temperature ◦ Species System (L/min) measured ( C) O2 (mg/L) pH Rainbow Flume 1 3.7Top8.51 ± 0.34 9.53 ± 0.36 8.02 ± 0.09 Middle 7.93 ± 0.21 9.72 ± 0.20 8.06 ± 0.08 Bottom 7.68 ± 0.13 10.00 ± 0.27 8.11 ± 0.06 Flume 2 3.7Top8.06 ± 0.32 9.19 ± 0.30 8.19 ± 0.09 Middle 7.66 ± 0.17 9.76 ± 0.17 8.24 ± 0.08 Bottom 7.47 ± 0.15 9.81 ± 0.11 8.26 ± 0.07 Incubator 18.1 Bottom 7.32 ± 0.10 11.06 ± 0.08 8.24 ± 0.04 Brook Flume 1 4 Top 9.43 ± 0.02 10.82 ± 0.15 8.34 ± 0.05 Middle 9.31 ± 0.01 10.79 ± 0.15 8.33 ± 0.05 Bottom 9.22 ± 0.03 10.75 ± 0.15 8.32 ± 0.06 Flume 2 4 Top 9.59 ± 0.04 10.76 ± 0.18 8.02 ± 0.11 Middle 9.38 ± 0.09 10.55 ± 0.18 8.18 ± 0.07 Bottom 9.22 ± 0.02 10.58 ± 0.18 8.25 ± 0.06 Incubator 20 Bottom 10.10 ± 0.01 8.53 ± 0.14 8.35 ± 0.02 Cutthroat Flume 1 3.5Top9.75 ± 0.09 7.85 ± 0.50 8.21 ± 0.05 Middle 9.39 ± 0.09 8.09 ± 0.40 8.19 ± 0.08 Bottom 9.33 ± 0.08 8.33 ± 0.38 8.20 ± 0.03 Flume 2 3.5Top9.93 ± 0.08 8.66 ± 0.35 8.21 ± 0.09 Middle 9.56 ± 0.06 8.72 ± 0.35 8.17 ± 0.08 Bottom 9.14 ± 0.06 8.71 ± 0.34 8.22 ± 0.06 Incubator 20 Bottom 10.53 ± 0.32 8.96 ± 0.10 8.30 ± 0.10

larger than the other two species and had larger eggs, and the identified as ready for spawning were injected with Ovaprim depth to which we decided to bury the eggs was based on the (Syndel Laboratories) at 0.5 mL/kg body weight to induce ovu- work of Crisp and Carling (1989) and Steen and Quinn (1999), lation. A week later fish were euthanized with tricaine methane- who showed that larger-sized females dug deeper redds. sulfonate (MS-222; 0.1 g/L) and gently squeezed to extrude Adult female fish.—Adult females were selected from the eggs into labeled, individual spawning pails. Milt pooled from existing broodstock at Allison Creek Brood Trout Station multiple males was used to fertilize the eggs; saline solution (Coleman, Alberta) and held at the hatchery for the duration of (0.6% NaCl) was added to improve sperm motility and aid mix- the experiment. Five-year-old Rainbow Trout (n = 9, mean ± ing. Eggs were rinsed after 2 min and placed into a pail of water SE fork length 65.17 ± 1.25 cm, mean ± SE body weight for 2 h to water harden. When the eggs were firm to the touch, a 3890 ± 33 g), 3-year-old Brook Trout (n = 9, mean ± SE fork subsample from each clutch was placed single file on a V-shaped

Downloaded by [Department Of Fisheries] at 19:29 20 January 2013 length 39.33 ± 0.76 cm, mean ± SE body weight 937 ± 83 g), trough that was 30 cm × 2 cm and counted. The number of eggs and 3-year-old Cutthroat Trout (n = 9, mean ± SE fork length per liter and average size of eggs was determined for each adult 35.23 ± 0.87 cm, mean ± SE body weight 497 ± 33 g) were female using the von Bayer egg chart (von Bayer 1950). Total used in the experiments. Adult Rainbow Trout were held from volume of eggs produced per female was determined using glass August 2009 to December 2009 until they were spawned and beakers before eggs were placed in an incubator tray. Trays with eggs were collected. Brook Trout were held from July 2010 the eggs were placed into a tub with Ovadine (1%) for 5 min to November 2010 until spawned, while Cutthroat Trout were and then set in the vertical incubator. Iodophor treatment is a held from December 2010 to April 2011. All adult females standard fish hatchery practice to reduce the bacterial load on were fed EWOS Salmonid Brood Feed trout chow throughout the eggs. Eggs were incubated at the Allison Creek Brood Trout the entire period. Experimental protocols were approved by Station until the eyed stage (eye pigment spots were visible), the Animal Welfare Committee (University of Lethbridge) in and then a subsample from each female was transported to the accordance with Canadian Council on Animal Care guidelines. Aquatic Research Facility at the University of Lethbridge for the Placement of eggs and collection of emerged fry.—As the flume swim-up experiment. Eggs were transported by ground predicted spawning season neared, fish were anaesthetized and (150 km) in plastic mesh tubes, wrapped in wet foam, in a cooler examined for signs of maturation (a soft underside). Females with ice. MANAGEMENT BRIEF 75

FIGURE 2. Emergence success ([number of emerged fry /number of eggs placed in gravel redd compartment] × 100; mean ± SE) and hatching success ([number of buttoned-up fry /number of eggs placed in vertical incubator] × 100; mean ± SE) scored for individual females from each species (Rainbow Trout n = 7, Brook Trout n = 9, Cutthroat Trout n = 9). Different letters indicate a signiﬁcant difference; small letters for hatching success, capital letters for emergence (Student’s t-test, P < 0.05).

Upon arrival, the tubes were soaked in a 1% Ovadine so- RESULTS lution for 5 min then rinsed with processed water. Eggs from There were no significant differences in dissolved oxygen, individual females were placed in a beaker to measure the to- temperature, or pH between different sections (top, middle, tal volume and ensure that a known number of eyed eggs from bottom) of individual flumes (Table 1) or between flumes individual females were deposited into labeled compartments tested at the same time within the incubation period (Decem- of the two experimental gravel bed flumes. Approximately ber 2009–March 2010 for Rainbow Trout eggs, November 1000 eggs/compartment redd were placed for Rainbow Trout 2010–March 2011 for Brook Trout eggs, and April 2011–June (December 2009) and Brook Trout (November 2010), and 450 2011 for Cutthroat Trout eggs). Water quality and flow remained eggs/compartment redd were placed for Cutthroat Trout (April constant even when all flume compartments were filled with 2011) due to lower numbers of eggs spawned per female. Eggs eggs. were gently poured in an oval-shape formation onto the gravel A significant difference in emergence success ([number of bed in each compartment (Figure 1) to simulate a redd. Gravel swim-up fry collected/number of eggs buried] × 100) was de- Downloaded by [Department Of Fisheries] at 19:29 20 January 2013 was then gently added by hand until the eggs were fully covered, tected between species (Figure 2; Student’s t-test: t = 2.07, df = and the remaining gravel was slowly poured on to reach a total 24, P < 0.05). While Rainbow Trout and Cutthroat Trout had gravel depth of 20 cm. Eggs were not added to compartment 1 mean ± SE emergence success of 58.1 ± 9.7% and 52.9 ± or compartment 20, where gravel lies between a flow-through 5.6%, respectively, the emergence success of Brook Trout was divider and a solid pane (Figure 1). The remaining eggs at the significantly higher at 85.2 ± 6.4%. The success of hatching Allison Creek Trout Brood station were kept in the vertical in- and survival to the fry stage (buttoned-up or yolkless fry) in cubator until alevin had absorbed their yolk sac. The fish were the vertical incubator was similar; the mean ± SE was 61.5 ± then euthanized with MS-222 and stored in Davidson’s solution 9.5% for Rainbow Trout, 61.9 ± 4.8% for Cutthroat Trout, and (50% formalin, 10% acetic acid, 10% glycerol, 30% ethanol). 82.7 ± 6.9% for Brook Trout and not significantly different Fish that appeared above the gravel were collected every day from emergence success for any of the species. The swim-up using a small net and a large pipette. Fish were euthanized with success, calculated as emergence success/hatching success, was MS-222, counted, and stored in jars of Davidson’s solution. Fry ∼100% (85.2/82.7) for Brook Trout, 85% (52.9/61.9) for Cut- collected from each compartment were stored in one jar. Data throat Trout, and 94% (58.1/61.5) for Rainbow Trout. These were analyzed using JMP software from SAS, and ANOVAs estimates were not significantly different from each other or with a Student’s t-test done post hoc were used on the data. from 100%. 76 PILGRIM ET AL.

140

120

100

80 Rainbow 60 Brook

Percent Emerged 40 Cuhroat

0 33.544.555.56 Diameter of Egg (mm)

FIGURE 3. Emerged fry collected in relation to average egg size of clutches from individual females for each species. Solid shape indicates species average and whiskers indicate SE.

Mean egg diameter was 5.31 mm, 4.45 mm, and 4.32 mm for The timing of emergence, which closely corresponded to Rainbow Trout, Brook Trout, and Cutthroat Trout, respectively; the appearance of buttoned-up fry in the vertical incubator, is however, no correlation was observed for emergence success illustrated by S-shaped curves (Figure 4) with different slopes based on egg size (Figure 3). and ranges and appears species-speciﬁc, with differences in the Downloaded by [Department Of Fisheries] at 19:29 20 January 2013

FIGURE 4. Number of emerged fry collected (cumulative % of total) for each egg clutch per week after egg placement. MANAGEMENT BRIEF 77

onset of emergence, the numbers of fish collected each week, females dig deeper redds (Crisp and Carling 1989), and since and the duration of the emergence period. Cutthroat Trout fry Rainbow Trout females in this study were larger than Brook appeared in large numbers above gravel only 3 weeks after Trout and Cutthroat Trout females, eggs from Rainbow Trout placement at the eyed stage. Rainbow Trout started to appear at were placed under 15 cm of gravel while eggs from the other 3 weeks post burial but in low numbers, until 7–8 weeks post two species were placed under 10 cm of gravel. Since for all burial when the emergence noticeably increased. The emergence three species the hatching success and the emergence success period lasted the longest for Rainbow Trout compared with the were similar, the depth of egg burial of 10–15 cm appeared to other two species. Brook Trout fry appeared at 7 weeks after have had little or no influence on either emergence or swim-up placement, but numbers increased rapidly over the next 4 weeks success. The effect of gravel depth on emergence success can with emergence continuing until week 11 (Figure 4). be further investigated using the experimental system described here. To accommodate species that bury their eggs deeper than 15 cm, a deeper flume would be required to provide sufficient DISCUSSION space for the gravel and for water flowing over the gravel. We custom built and tested a multicompartment gravel bed Emergence of all three species from the gravel followed an redd system designed to estimate swim-up success of salmonid S-shaped curve, with species-specific patterns of timing of ap- larvae. Swim-up success of Rainbow Trout, Brook Trout, and pearance of the first larvae, numbers collected, and duration Cutthroat Trout larvae was assessed under controlled pH, tem- of the emergence period. These patterns closely mimicked the perature, dissolved oxygen, and flow rate of water upwelling differences in developmental time between the three species through the gravel within different sections of individual flumes observed in the vertical incubator, including the earlier emer- and two different flumes used in parallel. gence of Cutthroat Trout larvae (Sadler et al. 1986; Wagner The emergence success ([number of swim-up fry collected/ et al. 2006), providing additional validation for the use of the number of eggs buried] × 100) was compared in the three gravel bed flume. It was noted that some of the early emerging salmonid species. Brook Trout had significantly higher emer- Rainbow Trout and Brook Trout fry collected above the gravel gence success (85%) than Rainbow Trout (58%) and Cut- still had partial yolk sacs. In our experiment there was no food throat Trout (53%). Other swim-up studies with salmonids have given to entice the fry to the surface, so it is not clear why some shown variable rates of success. In laboratory experiments us- fry appeared above the gravel at such an early stage. The environ- ing various experimental systems, 87.5% emergence has been mental cues used by fry in the timing of the emergence could be reported for Rainbow Trout (Fudge et al. 2008), 85% for Brown investigated in the experimental gravel bed flume described here. Trout Salmo trutta, 93.5% for Danube Salmon Hucho hucho Our results indicate that the experimental gravel bed flumes (Sternecker and Geist 2010), and 60% for Coho Salmon On- can be used in the laboratory for a quantitative assessment of corhynchus kisutch (Sundstrom¨ et al. 2005). Field studies have the swim-up success in salmonids. Uniform water quality (pH, reported 63.6% swim-up success for Brown Trout (Dumas et al. temperature, dissolved oxygen) was achieved throughout each 2007) and 66–90% for Atlantic salmon Salmo salar (Malcolm flume and it was possible to replicate these conditions in two dif- et al. 2003; Dumas and Marty 2006; Heywood and Walling ferent flumes running at the same time. A major improvement of 2007). The emergence success rates recorded in the experimen- our design, compared with the most recently developed systems tal gravel bed flume in our study were within the range reported (Fudge et al. 2008; Palm et al. 2009) is the delivery of water in the literature and indicate that species differences exist. using five supply pipes branching into a total of 20 perforated The simultaneous use of the vertical incubator to evaluate prongs, lying below the gravel and supplying upwelling water hatching success with a sample of eggs from each female pro- to individual compartments. The upwelling system delivered Downloaded by [Department Of Fisheries] at 19:29 20 January 2013 vided a unique approach to estimate the true swim-up success, water at a fast enough rate (3.5–4.0 L/min) to maintain constant by dividing emergence success in the flume by hatching success oxygen concentrations and temperature in different sections of in the vertical incubator. Since the estimates for emergence the flume, despite an overall flow from the upstream (top section success and hatching success were not significantly different for E) to downstream section (bottom section A). The flumes have any of the species (Figure 2), the estimates of swim-up success not been tested for consistency of delivery of toxicants or other were not significantly different from 100%. In order to achieve stressors; additional validation studies would be required. The more precise estimates, more clutches would be required. The flumes were affordable (∼$1000/flume) and easy to build. At average SD among clutches for both emergence and hatching the completion of the experiment, the flumes, pipes, and screens success for the three species was 30%. Thus if two flumes were rinsed with 1% Ovadine solution and dechlorinated water, were used√ at full capacity (n = 36), then SE estimates of 5% all the PVC pipes were disconnected, the screens were removed, (= 30/ 36) could be expected for both measures and SE of and the flumes were stacked for storage. The flumes could be ∼7% would be expected for swim-up success. If greater preci- used in the field with field-collected eggs; however, minor struc- sion were required, more flumes could be operated in parallel. tural modifications would be required. An important aspect of the flume was the depth of gravel in The design can be improved in future experiments by includ- the artificial redds. To simulate natural conditions where larger ing a sturdy supporting base to hold the flumes at a waist height, 78 PILGRIM ET AL.

rather than having them on the floor as in the current study, to Dumas,J.,J.G.Bassenave,M.Jarry,L.Barriere,` and S. Glise. 2007. Effects of improve ergonomics while collecting swim-up. Stainless steel fish farm effluents on egg-to-fry development and survival of Brown Trout in screens, rather than aluminum mesh, would also improve the artificial redds. Journal of Fish Biology 70:1734–1758. Dumas, J., and S. Marty. 2006. A new method to evaluate egg-to-fry survival in durability of the divider mesh screens. Use of flexible plastic salmonids, trials with Atlantic Salmon. Journal of Fish Biology 68:284–304. tubing (i.e., Tygon), instead of PVC, for the forks leading into Fraser, N. H. C., F. A. Huntingford, and J. E. Thorpe. 1994. The effect of light ◦ the flumes would eliminate the need for a large number of 90 intensity on the nightly movements of juvenile Atlantic Salmon alevins away elbow connections. Positioning of the perforated bottom pipes from the redd. Journal of Fish Biology 45(Supplement A):143–150. could be also modified to avoid having to cut a corner out of Fudge, T. S., K. G. Wautier, R. E. Evans, and V. P. Palace. 2008. Effect of different levels of fine-sediment loading on the escapement success of some of the dividers (Figure 1) where foam had to be inserted to Rainbow TroutTrout fry from artificial redds. North American Journal of block movement of the fry between compartments. The experi- Fisheries Management 28:758–765. mental gravel bed flume described in this study offers exciting Godin, J. G. J. 1980. Temporal aspects of juvenile Pink Salmon (Oncorhynchus possibilities for future experiments to investigate the effects gorbuscha Walbaum) emergence from a simulated gravel redd. Canadian of temperature, oxygen saturation, gravel depth, siltation, tox- Journal of Zoology 58:735–744. Granier, S., C. Audet, and L. Bernatchez. 2011. Heterosis and outbreeding de- icants, and flow rates on the swim-up success of larvae and to pression between strains of young-of-the-year Brook TroutTrout (Salvelinus determine species-specific requirements. The experimental sys- fontinalis). Canadian Journal of Zoology 89:190–198. tem described here would be a valuable tool for assessment of Groves, P.A., J. A. Chandler, and T. J. Richter. 2008. Comparison of temperature swim-up success of hatchery brood stocks or wild stocks and data collected from artificial Chinook Salmon redds and surface water in the for predicting the impact of environmental stressors on the re- Snake River. North American Journal of Fisheries Management 28:766–780. Heywood, M. J. T., and D. E. Walling. 2007. The sedimentation of salmonid productive success of salmonids and other fish species that bury spawning gravels in the Hampshire Avon catchment, UK: implications for their eggs in gravel. the dissolved oxygen content of intragravel water and embryo survival. Hy- drological Processes 21:770–788. Kondolf, G. M. 2000. Assessing salmonid spawning gravel quality. Transactions of the American Fisheries Society 129:262–281. ACKNOWLEDGMENTS Malcolm, I. A., A. F. Youngson, and C. Soulsby. 2003. Survival of salmonid We thank H. Bird, J. Fearns, and K. Neilson for their tireless eggs in a degraded gravel-bed stream: effects of groundwater–surface water collection of swim-up, the Allison Creek Brood Trout Station interactions. River Research and Applications 19:303–316. staff for their expertise on spawning fish and egg incubation, Mirza, R. S., D. P. Chivers, and J. G. J. Godin. 2001. Brook Charr alevins alter and Sarah Bogart for the technical drawing of the flume. This timing of nest emergence in response to chemical cues from fish predators. Journal of Chemical Ecology 27:1775–1785. project was funded by the Alberta Conservation Association and Palm, D., M. Lindberg, E. Brann¨ as,¨ H. Lundqvist, J. Ostergren,¨ and U. Metals in the Human Environment (MITHE)-NSERC Strategic Carlsson. 2009. Influence of European sculpin, Cottus gobio, on Atlantic Network. Salmon Salmo salar, recruitment and the effect of gravel size on egg predation: implications for spawning habitat restoration. Fisheries Management and Ecology 16:501–507. Rubin, J. F. 1995. Estimating the success of natural spawning of salmonids in REFERENCES streams. Journal of Fish Biology 46:603–622. Barlaup, B. T., H. Lura, H. Sægrov, and R. C. Sundt. 1994. Inter- and intra- Rudolph, B. L., I. Andreller, and C. J. Kennedy. 2008. Reproductive success, specific variability in female salmonid spawning behaviour. Canadian Journal early life stage development, and survival of Westslope Cutthroat TroutTrout of Zoology 72:636–642. (Oncorhynchus clarki lewisi) exposed to elevated selenium in an area of active Blanchfield, P. J., M. S. Ridgway, and C. C. Wilson. 2003. Breeding success of coal mining. Environmental Science and Technology 42:3109–3114. male Brook TroutTrout (Salvelinus fontinalis) in the wild. Molecular Ecology Sadler, S. E., G. W. Friars, and P. E. Ihssen. 1986. The influence of temperature 12:2417–2428. and genotype on the growth rate of hatchery-reared salmonids. Canadian

Downloaded by [Department Of Fisheries] at 19:29 20 January 2013 Brann¨ as,¨ E. 1989. The use of a simulated redd for incubating Baltic salmon Journal of Animal Science 66:599–606. (Salmo salar) and trout (Salmo trutta) alevins. Aquaculture 83:261–267. Snucins, E. J., R. A. Curry, and J. M. Gunn. 1992. Brook TroutTrout (Salvelinus Brann¨ as,¨ E. 1995. First access to territorial space and exposure to strong preda- fontinalis) embryo habitat and timing of alevin emergence in a lake and a tion pressure: a conﬂict in early emerging Atlantic Salmon (Salmo salar L.) stream. Canadian Journal of Zoology 70:423–427. fry. Evolutionary Ecology 9:411–420. Steen, R. P., and T. P. Quinn. 1999. Egg burial depth by Sockeye Salmon (On- Coleman, M. A., and K. D. Fausch. 2007. Cold summer temperature regimes corhynchus nerka): implications for survival of embryos and natural selection cause a recruitment bottleneck in age-0 Colorado River Cutthroat TroutTrout on female body size. Canadian Journal of Zoology 77:836–841. reared in laboratory streams. Transactions of the American Fisheries Society Sternecker, K., and J. Geist. 2010. The effects of stream substratum composition 136:639–654. on the emergence of salmonid fry. Ecology of Freshwater Fish 19:537–544. Crisp, D. T., and P. A. Carling. 1989. Observations on siting, dimensions and Sundstrom,L.F.,M.L¨ ohmus,¨ and R. H. Devlin. 2005. Selection on increased structure of salmonid redds. Journal of Fish Biology 34:119–134. intrinsic growth rates in Coho Salmon, Oncorhynchus kisutch. Evolution Curry, R. A., D. L. G. Noakes, and G. E. Morgan. 1995. Groundwater and 59:1560–1569. the incubation and emergence of Brook TroutTrout (Salvelinus fontinalis). von Bayer, H. 1950. A method of measuring ﬁsh eggs. Progressive Fish-Culturist Canadian Journal of Fisheries and Aquatic Sciences 52:1741–1749. 12:105–107. Dill, P. A. 1970. On the development of diel activity rhythms in Atlantic Salmon Wagner, E. J., R. E. Arndt, and R. Roubidoux. 2006. The effect of temperature and Rainbow TroutTrout. Doctoral dissertation. University of Western changes and transport on Cutthroat TroutTrout eggs soon after fertilization. Ontario, London. North American Journal of Aquaculture 68:235–239. This article was downloaded by: [Department Of Fisheries] On: 20 January 2013, At: 19:30 Publisher: Taylor & Francis Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

North American Journal of Fisheries Management Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/ujfm20 Sampling Efficiency of the Moore Egg Collector Thomas A. Worthington a , Shannon K. Brewer b , Timothy B. Grabowski c & Julia Mueller d a Oklahoma Cooperative Fish and Wildlife Research Unit, Oklahoma State University, 007 Agriculture Hall, Stillwater, Oklahoma, 74078, USA b U.S. Geological Survey, Oklahoma Cooperative Fish and Wildlife Research Unit, Oklahoma State University, 007 Agriculture Hall, Stillwater, Oklahoma, 74078, USA c U.S. Geological Survey, Texas Cooperative Fish and Wildlife Research Unit, Texas Tech University, 218 Agricultural Sciences, Lubbock, Texas, 79409-2120, USA d Texas Cooperative Fish and Wildlife Research Unit, Texas Tech University, 218 Agricultural Sciences, Lubbock, Texas, 79409, USA Version of record first published: 09 Jan 2013.

To cite this article: Thomas A. Worthington , Shannon K. Brewer , Timothy B. Grabowski & Julia Mueller (2013): Sampling Efficiency of the Moore Egg Collector, North American Journal of Fisheries Management, 33:1, 79-88 To link to this article: http://dx.doi.org/10.1080/02755947.2012.741557

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material. North American Journal of Fisheries Management 33:79–88, 2013 American Fisheries Society 2013 ISSN: 0275-5947 print / 1548-8675 online DOI: 10.1080/02755947.2012.741557

ARTICLE

Sampling Efﬁciency of the Moore Egg Collector

Thomas A. Worthington Oklahoma Cooperative Fish and Wildlife Research Unit, Oklahoma State University, 007 Agriculture Hall, Stillwater, Oklahoma 74078, USA Shannon K. Brewer* U.S. Geological Survey, Oklahoma Cooperative Fish and Wildlife Research Unit, Oklahoma State University, 007 Agriculture Hall, Stillwater, Oklahoma 74078, USA Timothy B. Grabowski U.S. Geological Survey, Texas Cooperative Fish and Wildlife Research Unit, Texas Tech University, 218 Agricultural Sciences, Lubbock, Texas 79409-2120, USA Julia Mueller Texas Cooperative Fish and Wildlife Research Unit, Texas Tech University, 218 Agricultural Sciences, Lubbock, Texas 79409, USA

Abstract Quantitative studies focusing on the collection of semibuoyant fish eggs, which are associated with a pelagic broadcast-spawning reproductive strategy, are often conducted to evaluate reproductive success. Many of the fishes in this reproductive guild have suffered significant reductions in range and abundance. However, the efficiency of the sampling gear used to evaluate reproduction is often unknown and renders interpretation of the data from these studies difficult. Our objective was to assess the efficiency of a modified Moore egg collector (MEC) using field and laboratory trials. Gear efficiency was assessed by releasing a known quantity of gellan beads with a specific gravity similar to that of eggs from representatives of this reproductive guild (e.g., the Arkansas River Shiner Notropis girardi) into an outdoor flume and recording recaptures. We also used field trials to determine how discharge and release location influenced gear efficiency given current methodological approaches. The flume trials indicated that gear efficiency ranged between 0.0% and 9.5% (n = 57) in a simple 1.83-m-wide channel and was positively related to discharge. Efficiency in the field trials was lower, ranging between 0.0% and 3.6%, and was negatively related to bead release distance from the MEC and discharge. The flume trials indicated that the gellan beads were not distributed uniformly across the channel, although aggregation was reduced at higher discharges. This clustering of passively Downloaded by [Department Of Fisheries] at 19:30 20 January 2013 drifting particles should be considered when selecting placement sites for an MEC; further, the use of multiple devices may be warranted in channels with multiple areas of concentrated flow.

Sampling eggs and early life history stages is an important timing of reproduction and environmental conditions related component of the development of conservation and management to spawning (Munk et al. 2009; Wakeﬁeld 2010); identify the plans for marine and estuarine ﬁshes (see Beck et al. 2001). The location and relative quality or importance of spawning sites results from ichthyoplankton surveys have the potential to in- (Conway et al. 1997; MacKenzie et al. 2000; Lorenzo et al. dicate the presence or absence of spawning populations of a 2007; Nicholson et al. 2008; Munk et al. 2009); serve as a given species (Bilkovic et al. 2002; Connell 2010); reveal the proxy for spawning stock biomass (see Bacheler et al. 2009;

*Corresponding author: [email protected] Received August 8, 2012; accepted October 12, 2012 79 80 WORTHINGTON ET AL.

Ralston and MacFarlane 2010); and provide data on the biol- cluding the Arkansas River Shiner Notropis girardi (1.00589 ± ogy, ecology, and transport of early life history stages (Scott 0.00011), and tested artificial products that could mimic the and Nielsen 1989). Although ichthyoplankton sampling is an semibuoyant characteristic of the fish eggs. Gellan beads, tradi- important tool for assessing estuarine and marine stocks, it is tionally used in food and drink products, were found to have a used less frequently for freshwater fishes. Lack of planktonic shape and specific gravity similar to those of fish eggs and have eggs and the short duration of the planktonic larval phases of been used in egg transport studies of Striped Bass Morone sax- most species make surveys in freshwater systems less common. atilis and Pecos Bluntnose Shiner N. simus pecosensis (Reinert However, ichthyoplankton sampling is important in the man- et al. 2004; Medley et al. 2007). The underlying assumption in agement and conservation of the freshwater species that release the use of gellan beads as egg surrogates is that they mimic the buoyant and semibuoyant eggs, such as Striped Bass Morone spatial distribution of fish eggs in river systems and all beads saxatilis (Bilkovic et al. 2002) and pelagic broadcast-spawning (and eggs) respond similarly to environmental conditions. cyprinids (Moore 1944; Bottrell et al. 1964; Durham and Wilde We hypothesized that the uncertainty as to the length of free- 2008; Wilde and Durham 2008). flowing river required for the development of pelagic broadcast- Pelagic broadcast-spawning cyprinids represent a reproduc- spawning cyprinids is, in part, due to the lack of a thorough tive guild of at least 12 species found in the rivers of the U.S. assessment of the efficiency of the sampling gears. Specifically, Great Plains (Platania and Altenbach 1998; Dudley and Platania there are two largely untested assumptions inherent in the use of 1999; Perkin and Gido 2011). These species produce eggs that a passive gear type in the collection of drifting particles. The first are negatively buoyant upon release but experience a reduction assumption is that eggs have a uniform vertical and horizontal in specific gravity as they take on water after fertilization (Plata- distribution within the river channel. The second assumption nia and Altenbach 1998; Dudley and Platania 1999). The eggs do is that geomorphological features within the stream channel not achieve neutral buoyancy in freshwater (Dudley and Platania and the turbulence they create do not significantly affect this 1999) but instead require water movement to remain in suspen- uniform distribution of eggs. Both of these assumptions have sion during development (Platania and Altenbach 1998; Dudley been questioned based on anecdotal evidence from field studies and Platania 1999). Thus, discharge, current velocity, and the (Reinert et al. 2004; Widmer et al. 2007), but they have not been distance available to drift are important factors in determin- evaluated systematically under controlled conditions. In relation ing reproductive success (Platania and Altenbach 1998; Dudley to conservation efforts for the federally listed Arkansas River and Platania 2007; Perkin and Gido 2011). One current debate in Shiner, our objectives were to assess the capture efficiency of the conservation and management of threatened and endangered the Moore egg collector (MEC) and to examine the effects of pelagic broadcast-spawning cyprinids is over the degree of flow turbulence and discharge on the spatial distribution of fish eggs and the length of uninterrupted river reach required to maintain under controlled conditions. The MEC has been used to study populations of these fishes. For example, surveys to estimate the the ichthyoplankton of pelagic broadcast-spawning cyprinids. It habitat requirements of pelagic broadcast-spawning cyprinids has been used to collect the eggs of Rio Grande Silvery Minnow in the Rio Grande and Pecos rivers suggest that these species Hybognathus amarus and Sturgeon Chub Macrhybopsis gelida require, on average, a distance of approximately 140 km to suc- (Hoagstrom et al. 2006; Osborne et al. 2006); assess the effect cessfully complete development (Dudley and Platania 2007). of river regulation on particle transport velocity and distance Perkin and Gido (2011) used presence/absence data to estimate (Dudley and Platania 2007); and examine egg drift and retention that a minimum of 100–300 km of unfragmented habitat (de- in relation to flow (Widmer et al. 2012). pending on the species) was required to maintain stable, viable populations. However, these data and their interpretation have Downloaded by [Department Of Fisheries] at 19:30 20 January 2013 not been without controversy. Other studies have proposed that METHODS eddies and other circulation features might retain eggs within Our experimental studies were conducted in an outdoor con- a stream reach and thereby dramatically shorten the length of crete flume (29.26 m long, 1.83 m wide, and 2.44 m deep) at river and flows needed for successful reproduction (Medley et al. the U.S. Department of Agriculture–Agricultural Research Ser- 2007; Kinzli and Myrick 2010; Widmer et al. 2012). The results vice’s National Hydraulic Engineering Research Unit in Still- of these studies have become contentious (see Medley et al. water, Oklahoma. Water entered the channel by gravity via a 2009; Zymonas and Propst 2009), with major implications for series of channels from Lake Carl Blackwell, with the water conservation and management planning in watersheds through- depth in the flume regulated by a downstream gate of variable out the Great Plains. height. Because the Arkansas River Shiner is federally listed, The gear efficiency of the devices or nets used to assess the re- large numbers of eggs from wild fish could not be used and it productive success of pelagic broadcast-spawning cyprinids has was not possible to produce a sufficient quantity from captive largely been untested because of difficulties in collecting suf- populations. Therefore, this study used gellan beads, which have ficient numbers of larvae or eggs (Franzin and Harbicht 1992; a shape and specific gravity similar to those of Arkansas River Reinert et al. 2004). However, Dudley and Platania (1999) calcu- Shiner eggs (Dudley and Platania 1999). The gellan beads have lated the specific gravity of the eggs of six cyprinid species, in- a specific gravity of 1.0015–1.0246 when soaked in tap water MOORE EGG COLLECTOR 81

FIGURE 1. Schematic of the Moore egg collector. Water passes through the opening at the front of the device and eggs are washed up the mesh screen to be collected at the back of the device.

for 0–48 h, and salinity does not change the physical condi- allowed for the 0.14 and 0.28 m3/s treatments and 3 min for the tion of the beads for up to3d(Reinertetal.2004). The beads 2.12 m3/s treatment. Trial length was determined from prelim- Downloaded by [Department Of Fisheries] at 19:30 20 January 2013 are manufactured by Technology Flavors and Fragrances, Inc. inary releases and based on the time taken to observe all the (Amityville, New York). gellan beads encountered and included a minimum of 1 min Gear efficiency: flume trials.—We tested the efficiency of the with no captures. At the end of each trial period, all gellan MEC (Figure 1), a sluice box–like gear designed for the non- beads captured by the MEC were counted and the screen was destructive capture of semibuoyant fish eggs (Altenbach et al. removed and cleaned of debris and any remaining beads. A 2000), under five discharge scenarios (0.14, 0.28, 0.71, 1.42, and total of 57 trials was undertaken, 6 each at the two lowest dis- 2.12 m3/s). The MEC was mounted in the center of the flume charges and 15 each at the three greatest discharges. Equal num- 28 m from the upstream end of the channel. The entire opening bers of trials were conducted per discharge at each of the three was submerged, and the upper crossbar was approximately at horizontal-release points. Water velocity was measured 300 mm surface water level. The mean water depth in the channel was upstream at approximately the center of the MEC opening 102.6 cm (SE, 0.51) across the five different treatments. using an electromagnetic flowmeter (Marsh-McBirney Model Each trial consisted of 1,000 gellan beads being released at 2000 Portable Flowmeter, ACG Technology Ltd., Woodbridge, one of three equally spaced points across the channel (center Ontario). channel, left, or right) 4 m downstream from the start of the Differences in capture efficiency under the different dis- channel. Trial length was dependent on discharge, 5 min being charge scenarios and between the release positions were 82 WORTHINGTON ET AL.

FIGURE 2. Schematic of the setup for the spatial distribution experiments showing the locations (from right to left) of the honeycomb ﬂow straightener (downstream position [DS] = 4 m), PVC cylinders (DS = 8 m), release point (DS = 9 m), and net grid (DS = 28 m). The schematic is not to scale.

examined using a full-factorial analysis of variance (ANOVA). ing between three equally-spaced release points (center channel, Tukey’s honestly significant difference (HSD) post hoc test was left, or right). Gellan beads were captured using 12 nets (mesh, used to identify differences between groups when the overall 0.79 mm) held in place by a wire grid that split the water column ANOVA was significant. Bootstrapping was used (1,000 itera- into 0.139 m3 sections. A total of 60 trials was conducted, 15 per tions; SPSS, SPSS version 20.0.0, Chicago) when data trans- treatment at each discharge. Trials lasted 5 min at 0.28 m3/s and formation (e.g., Box–Cox; Box and Cox 1964) still resulted in 4 min at 0.71 m3/s. The net grid was removed from the channel the violation of test assumptions (normality and homogeneity of following each trial to allow residual beads to be washed out variance). Bootstrapping was used to provide robust confidence and all captured beads to be counted. intervals of the mean differences between groups. To assess the complete spatial randomness (CSR) of the gel- Downloaded by [Department Of Fisheries] at 19:30 20 January 2013 Spatial distribution.—The two-dimensional spatial distribu- lan beads, a chi-square statistic was calculated for each trial (n = tion of fish eggs within the channel was tested under two dis- 60) based on the number of beads observed in each grid net and charges (0.28 and 0.71 m3/s), with each discharge consisting of the number expected (total number captured per trial divided by two experimental treatments (Figure 2). The first treatment rep- the number of grid nets). Complete spatial randomness implies resented laminar-flow conditions, created using a honeycomb that the frequency distribution of bead captures will follow a flow straightener (depth, 100 mm; cell size, 7 mm; Tubus Bauer Poisson distribution. Uniformly distributed captures would have GmbH, Bad Sackingen,¨ Germany). The flow straightener was a χ2 value of zero, whereas randomly distributed captures would placed across the width of the channel 4 m from the upstream have a χ2 value close to one; as captures become more clustered, end. The second treatment represented turbulent conditions, the size of the χ2 statistic increases (Thomas 1977). The indi- with four equally-spaced 114-mm polyvinyl chloride (PVC) vidual trial chi-square statistics were then analyzed using a mul- cylinders placed vertically in the water column. The PVC cylin- tifactorial ANOVA to determine whether differences occurred ders were located 8 m from the upstream end of the channel among treatments (i.e., discharge, turbulent versus nonturbulent but downstream of the flow straightener. For each trial, 100 flow, and release position). Again, where transformation (e.g., gellan beads were released into the experimental channel 1 m Box–Cox; Box and Cox 1964) of the data resulted in violation of downstream of the cylinder location, with trials again alternat- the test assumptions, bootstrapping was used (1,000 iterations; MOORE EGG COLLECTOR 83

SPSS 20.0.0; SPSS 2011) to provide robust confidence intervals May, 21 May, 31 May, and 11 July 2012). The sampling dates of the mean differences between groups. Tukey’s HSD post hoc were chosen to assess the MEC efficiency under a range of river test was used to identify differences between groups when the discharges (4.90, 8.38, 10.22, 16.28, and 17.61 m3/s). Appro- overall ANOVA was significant. priate discharges were selected using a gauge on the Cimarron Gear efficiency: field trials.—The efficiency of the MEC was River near Ripley, Oklahoma (U.S. Geological Survey gauge also assessed using field trials. The MEC was deployed at a site 07161450). on the Cimarron River upstream of State Highway 108 near Rip- We constructed a generalized linear model with a negative bi- ley, Oklahoma (36◦0128.69N, 96◦5453.83W). Great Plains nomial distribution (log-link function) using SPSS. The general rivers, including the Cimarron, are generally characterized by linear model accounted for overdispersion in the gellan bead wide, shallow, braided channels dominated by coarse sand and counts. Standardized deviance residuals were plotted against silt substrates (Friedman et al. 1998). The MEC was placed in predicted values of the mean to examine the fit of the negative an area of relatively high velocity within the channel, although binomial model (Hilbe 2007). We examined the relationship the exact location of the device varied between sampling occa- between the mean velocity of the water passing through the de- sions due to the dynamic nature of the river’s sand substrate. vice and the discharge recorded by the gauge in order to allow The device was placed so that it faced the direction of flow comparisons with published studies (some studies have reported and the entire opening was submerged just below the water sur- discharge, while others have reported velocity). face. One thousand gellan beads were released upstream of the MEC, and we recorded the number of beads collected. Dur- RESULTS ing each sampling trip, five releases of 1,000 beads were made 45.72 and 91.44 m directly upstream from the MEC (n = 10). Gear Efficiency: Flume Trials The mean velocity of the water passing through the device was The capture efficiency of the MEC in the flume trials ranged calculated from two readings using a digital, mechanical, low- from 0.0% to 9.5% (Figure 3). Capture efficiency was related velocity flowmeter (General Oceanics, Inc., Miami, Florida) to discharge (F4, 42 = 70.74, P < 0.001) but not release point attached to a frame seated at the mouth of the MEC. The sam- (F2, 42 = 0.12, P = 0.89) or the interaction between discharge pling procedure was completed on five occasions (27 April, 11 and release point (F8, 42 = 0.49, P = 0.86). Tukey’s HSD post Downloaded by [Department Of Fisheries] at 19:30 20 January 2013

FIGURE 3. Mean number of gellan beads recaptured at different discharges under experimental ﬂume conditions. The error bars represent SEs. 84 WORTHINGTON ET AL.

FIGURE 4. Percentage of beads in each of the grid nets at two different discharges.

hoc test indicated that there were differences between two main Spatial Distribution groups, one group encompassing the three lowest discharges and We captured 75.6% (lower discharge [0.28 m3/s]) and 96.2% the other encompassing the two highest discharges. No gellan (higher discharge [0.71 m3/s]) of the beads released during our beads were captured at the two lowest discharges; however, spatial distribution trials. The locations of the captures also var- at 0.71 m3/s efﬁciency ranged between 0.0% and 3.3%. At the ied with discharge, with 95.7% of the beads being caught in the two highest discharges, captures were considerably higher (2.1– bottom row of the grid at the lower discharge and 51.4% being 9.5% at 1.42 m3/s and 4.5–6.6% at 2.12 m3/s). As expected, the caught in the middle and upper rows at the higher discharge (Fig- velocity of water passing through the MEC was almost perfectly ure 4). At the lower discharge, captures were concentrated in the correlated with the discharge setting for the ﬂume (r = 0.9995, center of the channel (Figure 4) but also highly related to the ini- n = 5, P < 0.001). tial release position (Figure 5). However, at the higher discharge Downloaded by [Department Of Fisheries] at 19:30 20 January 2013

FIGURE 5. Percentage of beads caught in each vertical set of grid nets (x-axis) in relation to initial release location (shading) at two different discharges. MOORE EGG COLLECTOR 85

FIGURE 6. Number of beads captured by the MEC relative to discharge at different release locations (circles represent the 45.72-m release location, squares the 91.44-m release location). Predictions from the negative binomial generalized linear model are also shown for the two release locations (the solid line represents the 45.72-m release location and the dashed line the 91.44-m release location).

the horizontal capture position was more evenly distributed and χ2 = 21.564, df = 2, P < 0.001). Bead captures were approxi- less inﬂuenced by release location (Figure 5). The magnitude of mately 270% higher at the 45.72-m release point, and captures the CSR statistic was signiﬁcantly related to discharge (F1, 48 = decreased by 13.6% for every one-unit increase in discharge 140.2, P < 0.001) but not to release position, turbulence treat- (Figure 6). ment, or the interaction between turbulence treatment and release location. The Tukey’s HSD post hoc test result is not re- DISCUSSION

Downloaded by [Department Of Fisheries] at 19:30 20 January 2013 ported, as the significant predictor variable had only two groups. The calculated χ2 value was lower at the higher discharge than Sampling efficiency for fishes in early life stages is most at the lower discharge, though it was greater than 9.0 across all often evaluated by comparing the relative efficiency between trials, indicating high levels of aggregation among beads. gear types (Jessop 1985; Somerton and Kobayashi 1989; Pepin and Shears 1997). Studies evaluating gear-specific sampling efficiencies are rarely attempted due to the difficulty of obtaining Gear Efficiency—Field Trials.—The velocity of the water pass- and marking large quantities of eggs or larvae (Reinert et al. ing through the MEC ranged from 0.50 m/s to 0.85 m/s, which 2004; but see Franzin and Harbicht 1992). However, our study falls within the range of velocities of the three highest dis- demonstrated the importance in carrying out assessments of gear charges tested in the flume experiments. Water velocity was not efficiency when an understanding of population dynamics such correlated with the gauge reading for the site; however, this was as spawning stock biomass is required. We found that factors driven by a suspect reading at the highest discharge (r = 0.865, such as discharge, velocity, and release distance resulted in dra- n = 5, P = 0.058). The efficiency of the MEC during the field matically different capture rates in both laboratory and field tri- trials varied from 0.0% to 3.6%. The number of gellan beads als. These results should be considered when using the MEC to caught by the MEC was negatively related to discharge (gauge monitor the reproductive success of pelagic broadcast-spawning reading) and release distance from the MEC (likelihood ratio cyprinids as part of conservation management strategies, 86 WORTHINGTON ET AL.

as variability in gear efficiency will have direct effects on pop- semipelagic fish eggs. From an ecological standpoint, discharge ulation estimates and observed trends. and velocity appear to be important in the vertical positioning of The capture efficiency of the MEC was highly variable when gellan beads or fish eggs. Impoundment of Great Plains rivers tested under experimental conditions. Efficiency was zero at the has led to dramatic changes in the timing and magnitude of two lowest discharges, suggesting that the gellan beads were not high-flow events (Graf 1999; Costigan and Daniels 2012). This subject to gear capture (i.e., not traveling high enough in the wa- reduction in discharge has been linked to the decline of Arkansas ter column); however, as discharge increased, so did efficiency. River Shiner and other pelagic broadcast spawners (Bonner and At the higher discharges, the mean capture rates (0.8–6.5%) Wilde 2000). Platania and Altenbach (1998) suggest that veloc- in our experiments were much greater than those in a study ities as low as 0.01 m/s are sufficient to keep Arkansas River using gellan beads to assess the efficiency of a surface push Shiner eggs in suspension. However, our results indicated that net. Reinert et al. (2004) recorded recaptures between 0.0003% at velocities >0.01 m/s the gellan beads still traveled near the and 0.0088%, although the distance between sampling sites and channel floor, which in the sand bed rivers of the species’ dis- release location was far greater in their study. This link be- tribution may make them vulnerable to abrasion or smothering tween release distance, discharge, and capture efficiency was (Moore 1944; Bestgen et al. 1989; Osborne et al. 2006). also apparent in trials estimating the efficiency of drift nets to The MEC provides a more effective method for the nonde- recapture a known number of hatchery-reared Walleyes Sander structive capture of semibuoyant fish eggs than seines or drift vitreus (Franzin and Harbicht 1992). nets (see Altenbach et al. 2000); however, an understanding of As in our flume experiments, the recapture rates of gellan efficiency and the relationship between captures, discharge, and beads in our field trials were highly variable between sampling release point is important to an accurate assessment of the re- dates. Overall, however, the capture rates in the field trials were productive success of endangered pelagic broadcast-spawning lower than those in the flume experiments. This reduction in cyprinids. Experiments in both a controlled setting (flume trials) capture efficiency between the simplified cross-sectional flume and field tests were affected by discharge and velocity, although channel and the complex braided river site could be explained by the direction of this relationship was different between the ex- greater retention of the gellan beads in the natural environment. periments. We believe that this disparity is related to the natural Studies in the Pecos and Rio Grande rivers, New Mexico, sug- variability in flow patterns caused by channel geomorphology. gest that complex geomorphic reaches (e.g., sand bars and veg- To our knowledge, no other studies have directly quantified the etated islands) entrain the eggs of pelagic broadcast-spawning efficiency of the MEC. However, in this study the highest re- cyprinids (Medley et al. 2007; Widmer et al. 2012). Unlike in the capture percentage for the MEC was <10%, and this should be controlled flume experiments, the capture rate was negatively considered when using the device to sample semipelagic fish related to discharge and was also highly influenced by release eggs. Further, the spatial aggregation of the gellan beads sug- distance from the MEC. At lower discharges, the water within gests that careful consideration be given to the placement of the river was confined to a limited number of channels, making the MEC in the river. Additionally, the use of multiple MECs areas of higher velocity and the predicted downstream route of may be necessary in streams with braided channels or multiple the gellan beads easier to identify. Conversely, at higher dis- areas of concentrated flow if the observed patterns are to be charges, the irregular and complex nature of the channel profile ecologically meaningful. provided multiple routes for the gellan beads to travel. These results are important from a management standpoint because they clearly indicate that channel complexity has a strong influence ACKNOWLEDGMENTS on the efficiency of the MEC. This research is a contribution of the Oklahoma Coopera- Downloaded by [Department Of Fisheries] at 19:30 20 January 2013 The link between discharge, the vertical position of the gellan tive Fish and Wildlife Research Unit (U.S. Geological Survey, beads in the water column, and the sampling efficiency of the Oklahoma Department of Wildlife Conservation, Oklahoma MEC was reinforced by assessment of the gellan beads’ spatial State University, and Wildlife Management Institute cooper- distribution. The gellan beads were highly aggregated and con- ating) with collaboration by the Texas Cooperative Fish and centrated at the bottom of the channel at the lower discharge. Wildlife Research Unit (U.S. Geological Survey, Texas Depart- However, despite the less aggregated captures at the higher dis- ment of Parks and Wildlife, Texas Tech University, and Wildlife charge, the gellan beads were still highly clumped and therefore Management Institute cooperating). Funding was provided by failed a significant assumption for the use of the MEC. Under the U.S. Fish and Wildlife Service, Great Plains Landscape Con- a uniform or random spatial distribution, the section of water servation Cooperative (U.S. Fish and Wildlife Service agree- sampled by the MEC would be irrelevant; however, our results ment F11AP00112). The use of trade, firm, or product names is indicate that eggs are likely clumped during transport. These for descriptive purposes only and does not imply endorsement results suggest that appropriate placement of the MEC in the by the U.S. Government. We thank B. Brewer, S. Maichak, channel is critical to obtaining an accurate and precise sample. and J. Powers for field assistance; G. Fox and S. Lovern for From a practical perspective, our results should be consid- technical support and equipment; G. Hanson, S. Hunt, K. Ka- ered when using the MEC to monitor the downstream drift of davy, T. Selvey, and B. Sappington of the U.S. Department of MOORE EGG COLLECTOR 87

Agriculture–Agricultural Research Service National Hydraulic Graf, W. L. 1999. Dam nation: a geographic census of American dams and their Engineering Research Unit and K. Graves and the staff of the large-scale hydrologic impacts. Water Resources Research 35:1305–1311. U.S. Fish and Wildlife Service for logistical support; C. Jen- Hilbe, J. M. 2007. Negative binomial regression. Cambridge University Press, Cambridge, UK. nings for providing gellan beads; and S. Platania and associates Hoagstrom, C. W., C. A. Hayer, J. G. Kral, S. S. Wall, and C. R. Berry Jr. 2006. for technical and logistical assistance. This manuscript was Rare and declining fishes of South Dakota: a river drainage scale perspective. improved by comments from P. Bettoli and three anonymous Proceedings of the South Dakota Academy of Science 85:171–211. reviewers. Jessop, B. M. 1985. Influence of mesh composition, velocity, and run time on the catch and length composition of juvenile Alewives (Alosa pseudoharengus) and Blueback Herring (A. aestivalis) collected by pushnet. Canadian Journal of Fisheries and Aquatic Sciences 42:1928–1939. REFERENCES Kinzli, K. D., and C. A. Myrick. 2010. Bendway weirs: could they create Altenbach, C. S., R. K. Dudley, and S. P. Platania. 2000. A new device for col- habitat for the endangered Rio Grande Silvery Minnow. River Research and lecting drifting semibuoyant fish eggs. Transactions of the American Fisheries Applications 26:806–822. Society 129:296–300. MacKenzie, B. R., H. H. Hinrichsen, M. Plikshs, K. Wieland, and A. S. Zezera. Bacheler, N. M., K. M. Bailey, L. Ciannelli, V. Bartolino, and K. S. Chan. 2009. 2000. Quantifying environmental heterogeneity: habitat size necessary for Density-dependent, landscape, and climate effects on spawning distribution successful development of cod Gadus morhua eggs in the Baltic Sea. Marine of Walleye Pollock Theragra chalcogramma. Marine Ecology Progress Series Ecology Progress Series 193:143–156. 391:1–12. Medley, C. N., J. W. Kehmeier, O. B. Myers, and R. A. Valdez. 2007. Simulated Beck, M. W., K. L. Heck Jr., K. W. Able, D. L. Childers, D. B. Eggleston, transport and retention of pelagic fish eggs during an irrigation release in the B. M. Gillanders, B. Halpern, C. G. Hays, K. Hoshino, T. J. Minello, R. J. Orth, Pecos River, New Mexico. Journal of Freshwater Ecology 22:499–513. P. F. Sheridan, and M. P. Weinstein. 2001. The identification, conservation, Medley, C. N., J. W. Kehmeier, O. B. Myers, and R. A. Valdez. 2009. A re- and management of estuarine and marine nurseries for fish and invertebrates. analysis of data and critique of Medley et al.—“Simulated transport and BioScience 51:633–641. retention of pelagic fish eggs during an irrigation release in the Pecos River, Bestgen, K. R., S. P. Platania, J. E. Brooks, and D. L. Propst. 1989. Dispersal and New Mexico”—authors’ response. Journal of Freshwater Ecology 24:680– life history traits of Notropis girardi (Cypriniformes: Cyprinidae), introduced 683. into the Pecos River, New Mexico. American Midland Naturalist 122:228– Moore, G. A. 1944. Notes on the early life history of Notropis girardi. Copeia 235. 1944:209–214. Bilkovic, D. M., J. E. Olney, and C. H. Hershner. 2002. Spawning of Amer- Munk, P., C. J. Fox, L. J. Bolle, C. J. G. Van Damme, P. Fossum, and G. Kraus. ican Shad (Alosa sapidissima) and Striped Bass (Morone saxatilis)inthe 2009. Spawning of North Sea fishes linked to hydrographic features. Fisheries Mattaponi and Pamunkey rivers, Virginia. U.S. National Marine Fisheries Oceanography 18:458–469. Service Fishery Bulletin 100:632–640. Nicholson, G., G. P. Jenkins, J. Sherwood, and A. Longmore. 2008. Physical Bonner, T. H., and G. R. Wilde. 2000. Changes in the Canadian River fish environmental conditions, spawning and early-life stages of an estuarine fish: assemblage associated with reservoir construction. Journal of Freshwater climate change implications for recruitment in intermittently open estuaries. Ecology 15:189–198. Marine and Freshwater Research 59:735–749. Bottrell, C. E., R. H. Ingersol, and R. W. Jones. 1964. Notes on the embryology, Osborne, M. J., M. A. Benavides, D. Alo,` and T. F. Turner. 2006. Genetic effects early development, and behavior of Hybopsis aestivalis tetranemus (Gilbert). of hatchery propagation and rearing in the endangered Rio Grande Silvery Transactions of the American Microscopical Society 83:391–399. Minnow, Hybognathus amarus. Reviews in Fisheries Science 14:127–138. Box, G. E. P., and D. R. Cox. 1964. An analysis of transformations. Journal of Pepin, P., and T. H. Shears. 1997. Variability and capture efficiency of bongo and the Royal Statistical Society B 26:211–252. Tucker trawl samplers in the collection of ichthyoplankton and other macro- Connell, A. D. 2010. A 21-year ichthyoplankton collection confirms sardine zooplankton. Canadian Journal of Fisheries and Aquatic Sciences 54:765– spawning in KwaZulu-Natal waters. African Journal of Marine Science 773. 32:331–336. Perkin, J. S., and K. B. Gido. 2011. Stream fragmentation thresholds for a Conway, D. V. P., S. H. Coombs, and C. Smith. 1997. Vertical distribution of reproductive guild of Great Plains fishes. Fisheries 36:371–383. fish eggs and larvae in the Irish Sea and southern North Sea. ICES Journal of Platania, S. P., and C. S. Altenbach. 1998. Reproductive strategies and egg types Marine Science 54:136–147. of seven Rio Grande basin cyprinids. Copeia 1998:559–569.

Downloaded by [Department Of Fisheries] at 19:30 20 January 2013 Costigan, K. H., and M. D. Daniels. 2012. Damming the prairie: human al- Ralston, S., and B. R. MacFarlane. 2010. Population estimation of Bocaccio teration of Great Plains river regimes. Journal of Hydrology 444–445:90– (Sebastes paucispinis) based on larval production. Canadian Journal of 99. Fisheries and Aquatic Sciences 67:1005–1020. Dudley, R. K., and S. P. Platania. 1999. Imitating the physical properties of Reinert, T. R., T. A. Will, C. A. Jennings, and W. T. Davin. 2004. Use of egg drifting semibuoyant fish (Cyprinidae) eggs with artificial eggs. Journal of surrogates to estimate sampling efficiency of Striped Bass eggs in the Savan- Freshwater Ecology 14:423–430. nah River. North American Journal of Fisheries Management 24:704–710. Dudley, R. K., and S. P. Platania. 2007. Flow regulation and fragmentation Scott, M. T., and L. A. Nielsen. 1989. Young fish distribution in backwaters and imperil pelagic-spawning riverine fishes. Ecological Applications 17:2074– main-channel borders of the Kanawha River, West Virginia. Journal of Fish 2086. Biology 35(Supplement A):21–27. Durham, B. W., and G. R. Wilde. 2008. Composition and abundance of drifting Somerton, D. A., and D. R. Kobayashi. 1989. A method for correcting catches fish larvae in the Canadian River, Texas. Journal of Freshwater Ecology of fish larvae for the size selection of plankton nets. U.S. National Marine 23:273–280. Fisheries Service Fishery Bulletin 87:447–455. Franzin, W. G., and S. M. Harbicht. 1992. Tests of drift samplers for estimating Thomas, R. W. 1977. An introduction to quadrat analysis. Geo Abstracts, abundance of recently hatched Walleye larvae in small rivers. North American Norwich, UK. Journal of Fisheries Management 12:396–405. Wakefield, C. B. 2010. Annual, lunar and diel reproductive periodicity of a Friedman, J. M., W. R. Osterkamp, M. L. Scott, and G. T. Auble. 1998. Down- spawning aggregation of snapper Pagrus auratus (Sparidae) in a marine stream effects of dams on channel geometry and bottomland vegetation: embayment on the lower west coast of Australia. Journal of Fish Biology regional patterns in the Great Plains. Wetlands 18:619–633. 77:1359–1378. 88 WORTHINGTON ET AL.

Widmer, A. M., J. J. Fluder III, J. W. Kehmeier, C. N. Medley, and R. A. Valdez. Wilde, G. R., and B. W. Durham. 2008. Daily survival rates for juveniles of six 2012. Drift and retention of pelagic spawning minnow eggs in a regulated Great Plains cyprinid species. Transactions of the American Fisheries Society river. River Research and Applications 28:192–203. 137:830–833. Widmer, A. M., J. W. Kehmeier, and J. J. Fluder III. 2007. Characterization of Zymonas, N. D., and D. L. Propst. 2009. A re-analysis of data and critique of Rio Grande Silvery Minnow egg and larval drift and retention in the middle Medley et al.—“Simulated transport and retention of pelagic ﬁsh eggs during Rio Grande. SWCA Environmental Consultants, Report to New Mexico In- an irrigation release in the Pecos River, New Mexico.” Journal of Freshwater terstate Stream Commission, Albuquerque. Ecology 24:671–679. Downloaded by [Department Of Fisheries] at 19:30 20 January 2013 This article was downloaded by: [Department Of Fisheries] On: 20 January 2013, At: 19:30 Publisher: Taylor & Francis Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

North American Journal of Fisheries Management Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/ujfm20 Comparing Habitat Suitability Indices (HSIs) Based on Abundance and Occurrence Data Pey-Yi Lee a & Jian-Ping Suen a a Department of Hydraulic and Ocean Engineering, National Cheng Kung University, Number 1 University Road, Tainan, 701, Taiwan Version of record first published: 09 Jan 2013.

To cite this article: Pey-Yi Lee & Jian-Ping Suen (2013): Comparing Habitat Suitability Indices (HSIs) Based on Abundance and Occurrence Data, North American Journal of Fisheries Management, 33:1, 89-96 To link to this article: http://dx.doi.org/10.1080/02755947.2012.743933

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material. North American Journal of Fisheries Management 33:89–96, 2013 C American Fisheries Society 2013 ISSN: 0275-5947 print / 1548-8675 online DOI: 10.1080/02755947.2012.743933

ARTICLE

Comparing Habitat Suitability Indices (HSIs) Based on Abundance and Occurrence Data

Pey-Yi Lee and Jian-Ping Suen* Department of Hydraulic and Ocean Engineering, National Cheng Kung University, Number 1 University Road, Tainan 701, Taiwan

Abstract Suitability-based physical habitat modeling is sensitive to the choice of habitat suitability index. We compared the effects of using abundance versus occurrence data on suitability indices along environmental gradients for four fish species collected by means of prepositioned electrofishing in a mountain stream in southern Taiwan. The results indicate that abundance data provide more sensitive outcomes than occurrence data in terms of both mean values and habitat suitability. The mean values of flow velocity and water depth from the abundance data separated the four fish species into four distinct flow groups and two depth groups; the occurrence data did not produce similar groupings. Habitat suitability curves from the abundance data provided clearer outcomes than those from the occurrence data. The differences were greater for species with differential densities along environmental gradients that were related to species characteristics. Advancements in point abundance sampling methodology, such as the development of prepositioned electrofishing, provide novel opportunities to collect abundance data that can enhance suitability results.

Habitat suitability modeling can be seen as an operational ap- (Parasiewicz and Walker 2007; Conallin et al. 2010). Both di- plication of the niche theory (Grinnell 1917; Hutchinson 1959). rect studies (i.e., those expressing the inﬂuences of HSI values The concept of the ecological niche relates a set of environmen- on instream model analysis; Moir et al. 2005; Conallin et al. tal variables to the ﬁtness of a species, while habitat suitability 2010) and indirect studies (i.e., those exploring the transferabil- modeling relates environmental variables to the likelihood of ity of HSIs between streams; Thomas and Bovee 1993; Garner occurrence or the abundance of the species (Hirzel and Le Lay 1995; Freeman et al. 1997; Glozier et al. 1997; Moir et al. 2005; 2008). The concept of the habitat suitability index (HSI) was Conallin et al. 2010) suggest that the particular HSI values used

Downloaded by [Department Of Fisheries] at 19:30 20 January 2013 developed by the U.S. Fish and Wildlife Service for habitat eval- can greatly affect the outcomes of instream flow models and uation in the 1970s. Since then, it has been incorporated into the influence management policy. instream flow incremental methodology (Bovee and Milhous Bovee (1986) formulated the basic approach and describes 1978) to simulate the influences of human impacts on aquatic three types of HSI. Among them is the preference his—the fre- ecosystems (Bovee 1986; Chou and Chuan 2011). The theoret- quency of habitat utilization divided by the frequency of habitat ical foundations and applicability of HSI have been studied and availability—which has been the most accepted and widely ap- discussed; the topics considered include the transferability of plied index in suitability modeling (Vilizzi et al. 2004; Moir et al. the HSI (Thomas and Bovee 1993; Garner 1995; Freeman et al. 2005; Conallin et al. 2010). However, the frequency of habi- 1997; Glozier et al. 1997; Moir et al. 2005; Conallin et al. 2010), tat utilization has been calculated using either occurrence data the effects of interactions between the physical variables (Vadas (such as number of occupied sites; e.g., Garner 1995; Glozier and Orth 2001; Vismara et al. 2001; Jowett and Davey 2007), et al. 1997; Chou and Chuang 2011) or abundance data (such ecological interactions (Orth 1987), and modeling techniques as the number of individuals; e.g., Thomas and Bovee 1993;

*Corresponding author: [email protected] Received April 15, 2012; accepted October 17, 2012 89 90 LEE AND SUEN

Freeman et al. 1997; Vadas and Orth 2001; Vismara et al. 2001; Parasiewicz and Walker 2007), and the different data sets have produced different HSI values. Suitability-based physical habitat modeling in streams is sensitive to the choice of HSI used in the modeling process. Different HSI values will change the instream flow increment modeling output results in weighted usable areas (Moir et al. 2005; Conallin et al. 2010). With the continuing development of prepositioned electrofishing (Bain et al. 1985; Schwartz and Herricks 2004) and other point abundance sampling methods (Copp 2010) to collect abundance data, understanding the effects of the type of data on HSIs is critical for accurate results in suitability modeling. Our objective was to compare the effect of using occurrence versus abundance data on HSIs based on flow velocity and water depth. Comparisons were made for four fish species—the Formosan River Loach Hemimyzon formosanum, Nantai Goby Rhinogobius nantaiensis, Deep-Body Shovelnose Minnow Ony- chostoma alticorpus, and Taiwan Striped Barb Acrossocheilus paradoxus—to understand the variation among species. First, we compared the mean HSI values for flow velocity and water depth for given species obtained from occurrence and abundance data using Student’s t-test. Then we used an F-test and the Bonferroni post hoc test to compare the differences in flow velocity and water depth HSIs among species to examine their groupings on microhabitat usage. Last, we compared the HSIs from occurrence and abundance data using Spearman’s rank correlation and illustrations of HSI curves. FIGURE 1. Locations of the study sites, the Min-Chuan and Twelfth Bridge areas, on Cishan Stream. The stream flows from north to south.

METHODS Fieldwork was conducted in the upper part of the Cishan measured for weight and total length (to the nearest g and mm). Stream (Figure 1), a major tributary of the Kaoping River, the Fish were not return to the water until the end of the daily sur- largest river in southern Taiwan. This tributary maintains a natu- vey to avoid recapture on the same day. Five fish species were ral flow regime of 30 m3/s (annual mean flow; Nan-Feng Bridge recorded during the surveys; however, only the four most abun- Hydrologic Station; WRA 2009). The stream meanders through dant species provided a large enough sample size for analysis the valley floor, and riparian vegetation seldom shades the water. (Lee and Suen 2012). The sampling plots were located inside a wildlife sanctuary with Microhabitat characteristics such as flow velocity and water excellent water quality. Fishing regulations in the area protect depth were measured at nine evenly distributed points (3 m × populations of endemic fish and limit human disturbance. 3 m) within the electrode frames after fish sampling. The aver- Downloaded by [Department Of Fisheries] at 19:30 20 January 2013 Field surveys of fish abundance and microhabitat were con- ages of the nine measurements of water depth and flow velocity ducted at five riffles in two stream segments, the Min-Chuan represent the microhabitat characteristics of the sampling plot. Bridge (58 plots) and the Twelfth Bridge areas (38 plots), to The 96 randomly sampled plots reflect the availability of phys- include more variation in instream habitat. All the fish surveys ical microhabitats (i.e., flow velocity and water depth) in the were in the daytime during the dry season between November study area (Figure 2). 2008 and March 2009. Bain et al.’s (1985) prepositioned elec- We followed Bovee (1986) and Vismara et al. (2001) to define trofishing sampling methodology was adopted to sample rect- suitability curves. First, each variable was divided into classes. angular electrode frames of 1-m × 1-m (n = 88) and 1-m × Flow velocity was divided into increments of 10 cm/s, and water 2-m (n = 8) areas for 1 min. This method collects multiple depth was divided into increments of 10 cm. After calculating individuals per sample plot and is suitable for comparing occur- the frequencies of utilization and availability of every class, we rence and abundance data. To minimize the effects of sampling used those values to calculate the preferences for each interval disturbance on fish behavior, the electrodes were not charged of the measured variable. Preferences were then normalized to for at least 11 min after the placement of the electrofishing de- a maximum value of 1.0 by dividing by the maximum preferred vices. Once the electrodes were charged, immobilized fish were values. Frequencies of utilization were obtained from the num- collected immediately with dip nets, identified to species, and ber of individuals and the number of occupied plots of each fish HSIs BASED ON ABUNDANCE AND OCCURRENCE 91

FIGURE 2. Frequency distributions of available microhabitat in terms of (a) ﬂow velocity and (b) water depth during this study.

species to compare the suitability results between abundance and calculated from occupied plots (t = 5.5, df = 264, P = 0.02) occurrence data. Statistical analyses, such as Student’s t-test, and the mean water depth calculated from individual fish counts F-test, Bonferroni post hoc test, and Spearman’s rank correla- was significantly greater than that calculated from occupied tion, were performed with SPSS Statistics 17.0. plots (t = 6.6, df = 264, P = 0.01). For both the Nantai Goby and Taiwan Striped Barb, there were no significant differences in either flow velocity or water depth between the occurrence = = = = RESULTS and abundance data (t 0.01, df 185, P 0.97 and t 1.8, df = 70, P = 0.19, respectively). Mean Value Comparison The Student’s t-test results indicated that the influence of data type on the mean values of microhabitat utilization was species Species Grouped by Velocity specific (Table 1). For the Formosan River Loach, the mean The four fish species grouped differently depending on velocity calculated from individual fish counts was significantly the mean velocity at individual locations and occupied plots higher than that calculated from occupied plots (t = 14.7, df = (Table 1). Based on the locations of individual fish, the F-test and 856, P = 0.001) and the mean water depth calculated from Bonferroni post hoc test indicated four groups distinguished by individual fish counts was significantly less than that calculated flow velocity. Formosan River Loaches preferred microhabitats from occupied plots (t = 5.8, df = 856, P = 0.001). For the with the highest flow velocity, followed by Nantai Goby, Deep- Deep-Body Shovelnose Minnow, the mean velocity calculated Body Shovelnose Minnows, and Taiwan Striped Barbs. Based from individual fish counts was significantly lower than that on site occupation, on the other hand, the F- and Bonferroni

TABLE 1. Mean velocity and water depth for four species as calculated from fish and plot data. Different lowercase letters indicate significant differences (P < 0.05) among species according to the Bonferroni post hoc test and F-test. Asterisks indicate significant differences in mean values between abundance and

Downloaded by [Department Of Fisheries] at 19:30 20 January 2013 occurrence data; n = sample size.

Fish Plot Species n Mean SE n Mean SE Velocity Formosan River Loach* 770 93.4 0.9 z 88 82.5 3.0 z Nantai Goby 128 82.5 2.0 y 59 82.7 3.2 z Deep-Body Shovelnose Minnow* 203 63.5 2.0 x 63 73.3 3.6 zy Taiwan Striped Barb 45 52.4 4.1 w 27 60.9 4.7 y Depth Nantai Goby 128 31.4 1.1 y 59 32.4 1.7 y Formosan River Loach* 770 31.4 0.5 y 88 35.1 1.8 zy Deep-Body Shovelnose Minnow* 203 47.4 1.4 z 63 40.1 2.3 zy Taiwan Striped Barb 45 48.1 3.1 z 27 44.3 4.4 z 92 LEE AND SUEN

FIGURE 3. Habitat suitability indices based on abundance and occurrence and density as functions of ﬂow velocity for four ﬁsh species.

post hoc tests indicated two velocity groups. Nantai Goby and with different members: Nantai Goby preferred microhabitats Downloaded by [Department Of Fisheries] at 19:30 20 January 2013 Formosan River Loaches preferred microhabitats with higher in shallower water and Taiwan Striped Barbs preferred micro- flow velocities, while Taiwan Striped Barbs preferred lower habitat with deeper water. However, the mean water depths for flow velocities. However, the mean flow velocity of Deep-Body Formosan River Loaches and Deep-Body Shovelnose Minnows Shovelnose Minnows was not significantly different from the were not significantly different from the depths of the shallower mean velocities of the other three species. and deeper groups.

Species Grouped by Water Depth Suitability of Velocity Water depth usage grouped fish differently depending on Abundance, occurrence, and density provided different HSIs whether we used individual fish locations or occupied plots. with respect to flow velocity for the four fish species (Figure 3). Based on location, the F- and Bonferroni post hoc tests indi- Formosan River Loaches strongly preferred higher-velocity ar- cated two distinct groups: Formosan River Loaches and Nantai eas based on abundance and density but not occurrence. Deep- Goby preferred microhabitats with shallower water, while Deep- Body Shovelnose Minnows preferred lower-velocity areas based Body Shovelnose Minnows and Taiwan Striped Barbs preferred on all three indicators. Nantai Goby preferred moderate-velocity microhabitats with deeper water. Based on occupied sites, the areas according to all three indicators. Taiwan Striped Barbs F- and Bonferroni post hoc tests also indicated two groups, but preferred lower-velocity areas according to all three indicators. HSIs BASED ON ABUNDANCE AND OCCURRENCE 93

TABLE 2. Coefﬁcients and probabilities (P) of Spearman’s rank correlation analysis between HSIs from abundance and occurrence data; *P < 0.05, **P < 0.01.

Velocity Depth Species Coefﬁcient nPCoefﬁcient nP Formosan River Loach 0.425 16 0.101 −0.312 9 0.414 Deep-Body Shovelnose Minnow 0.905 15 0.001** 0.529 9 0.143 Nantai Goby 0.673 11 0.004** 0.948 6 0.001** Taiwan Striped Barb 0.989 9 0.001** 0.714 8 0.031*

Spearman’s rank correlation analysis showed significant corre- those based on abundance data being clearer than those based on lations between abundance HSI and occurrence HSI for Deep- occurrence data. Formosan River Loaches could not be clearly Body Shovelnose Minnows, Nantai Goby, and Taiwan Striped grouped according to water depth by occurrence data, and Deep- Barbs but not for Formosan River Loaches (Table 2), suggesting Body Shovelnose Minnows could not be clearly grouped accord- that the slopes of the trend lines are different. The differences ing to either flow velocity or water depth by occurrence data. were caused by the different density gradients along flow ve- The correlation analysis indicated the different shapes or slopes locity (Figure 2). For example, the density of Formosan River between HSI curves. The occurrence HSI was different from Loaches was higher at relatively high velocity sites. the abundance HSI for Formosan River Loaches for flow velocity and water depth, while the occurrence HSI was different from the abundance HSI for Deep-Body Shovelnose Minnows Suitability of Depth for water depth only. The two less abundant fish species showed The four fish species had different HSIs with respect to water no influence on mean values and HSI curves of microhabitat depth based on the different types of data: abundance, occur- usage according to data type. Furthermore, the inconsist use rence, or density (Figure 4). Formosan River Loaches strongly of abundance and occurrence data on microhabitat utilization preferred areas with shallow water based on abundance and may increase the uncertainty of instream model applications in density but only weakly preferred those areas based on occur- stream management and restoration (Moir et al. 2005; Conallin rence. Deep-Body Shovelnose Minnows preferred areas with et al. 2010). Those using just occurrence data in model predic- moderate water depth (60 cm) based on abundance and den- tions should be cautious of this discrepancy. sity but deeper-water areas based on occurrence. Nantai Goby When all three analyses are integrated, it becomes apparent strongly preferred shallower water areas based on abundance that two factors may influence the inconsistency between results and occurrence but only weakly preferred those areas based on based on abundance data and those based on occurrence data: density. Taiwan Striped Barbs preferred areas with moderate distribution and density. Abundance is basically the product of water depth (60 cm) based on abundance and density but pre- density and occurrence. Density can provide additional differ- ferred deeper-water areas based on occurrence. Spearman’s rank ence in abundance HSIs, especially when occurrence HSIs fail correlation analysis indicated a significant correlation between to provide a meaningful pattern (e.g., the occurrence HSIs for abundance HSIs and occurrence HSIs for Nantai Goby and Tai- flow velocity and water depth for Formosan River Loaches), and wan Striped Barbs but not for Formosan River Loaches and therefore the abundance HSI is a better indicator than density Downloaded by [Department Of Fisheries] at 19:30 20 January 2013 Deep-Body Shovelnose Minnows (Table 2). These differences or electivity alone (Lee and Suen 2012). For species that were were caused by the different density gradient along water depth. evenly distributed in occurrence but produced skewed density For example, the density of Formosan River Loaches was higher curves along environmental gradients (such as the Formosan at relatively shallow water sites, and the densities of Deep-Body River Loach with respect to flow velocity and water depth Shovelnose Minnows and Taiwan Striped Barbs were higher at and the Deep-Body Shovelnose Minnow with respect to water areas of moderate water depth. depth), the influence of density is maximal. For species whose density is relatively evenly distributed along environmental gradients (such as the Nantai Goby and Taiwan Striped Barb with DISCUSSION respect to water depth), the influence of density is minimal. With Our results show the species-specific influences on the mean the additional information provided by density, the abundance values and HSIs of microhabitat utilization between data sets of HSI is a better indicator than the occurrence HSI. different types. Student’s t-test indicated the different mean val- Our results also support the view that the mean values of ues resulting from abundance and occurrence for the two most environmental gradients such as flow velocity and water depth abundant species, Formosan River Loach and Deep-Body Shov- are capable of providing reliable indicators of niche separation elnose Minnow. The F-test indicated different groupings, with (Chuang et al. 2004), even for relatively nonnormally distributed 94 LEE AND SUEN

FIGURE 4. Habitat suitability indices based on abundance and occurrence and density as functions of water depth for four ﬁsh species.

data. By comparison, the mean values from abundance data pro- to be herbivorous and nonterritorial and therefore exhibit ideal Downloaded by [Department Of Fisheries] at 19:30 20 January 2013 vided clearer niche separation than the mean values from occur- free distribution in aggregations. On the contrary, Nantai Goby rence data for separating the fishes into distinct groups (Table 1). and Taiwan Striped Barb tend to be insectivorous and territorial Furthermore, suitability curves provided detailed descriptions and therefore display ideal despotic distribution. This trait of of preferences along environmental gradients and were better natural history explains the noncorrelation between abundance indicators than mean values, especially for relatively nonnor- and occurrence HSIs in both Formosan River Loach and Deep- mally distributed data. Altogether, abundance data should be Body Shovelnose Minnow (Table 2, Figures 3 and 4), whose used instead of occurrence data whenever possible because of densities were higher than those of Nantai Goby and Taiwan the enhanced results for mean values and suitability curves. Striped Barb. Natural history of species may affect the density distribu- Social interaction, such as conspecific attraction, also in- tion along environmental gradients (Fretwell and Lucas 1969). creases density and may influence habitat selection (Stamps Lee and Suen (2012) studied the microhabitat preference of the 1988; Scott and Lee, in press). This phenomenon is especially same four fish species and illustrated differential density along significant in young fish, and the aggregation of such fish may environmental gradients that related to ideal free distribution affect the results of microhabitat preferences when those pref- and ideal despotic distribution (Fretwell and Lucas 1969). For- erences differ between old and young fish (Rosenfeld 2003). If mosan River Loach and Deep-Body Shovelnose Minnow tend this is the case, two HSIs should be developed, one for adults HSIs BASED ON ABUNDANCE AND OCCURRENCE 95

and one for younger fish. However, the occurrence HSI, which Nien Lai, Chi-Hsien Chiu, and Chin-Yuan Lin for field assis- does not count individuals, avoids this kind of bias. Younger life tance and Ruey Lin and Cathy Marcinkevage for proofreading stages usually display temporal fluctuations, with high densities and language editing. after breeding season and low densities before breeding season. This should be considered in fish and microhabitat surveys. Our REFERENCES study did not find large numbers of young in the surveyed lo- Bain, M. B., J. T. Finn, and H. E. Booke. 1985. A quantitative method for cations (Lee and Suen 2012), so the microhabitat utilizations in sampling riverine microhabitats by electrofishing. North American Journal this application were not biased toward young fish. of Fisheries Management 5:489–493. Grain size (the size of the sampling plots) and individual Bovee, K. D. 1986. Development and evaluation of habitat suitability criteria distances between organisms may affect the number of indi- for use in the instream flow incremental methodology. U.S. Fish and Wildlife viduals per sample (Wiens 1989). There will be no difference Service Biological Report 86(7). Bovee, K. D., and R. T. Milhous. 1978. Hydraulic simulation in instream flow between abundance and occurrence when the density equals studies: theory and techniques. U.S. Fish and Wildlife Service Biological one individual per sampling plot. Therefore, the methods of fish Services Program FWS/OBS-78/33. sampling may affect the data used in HSI modeling and other Chou, W. C., and M. D. Chuang. 2011. Habitat evaluation using suitability in- analyses (Copp 1991, 2010). The focal fish methods, such as dex and habitat type diversity: a case study involving a shallow forest stream bank and diving observations, retrieve a single individual each in central Taiwan. Environmental Monitoring and Assessment 172:689– 704. time; the frequency of habitat utilization is calculated from ob- Chuang, L. C., S. H. Liang, and Y.S. Lin. 2004. Habitat use of two benthic fishes, servations of individual fish. However, the frequency of habitat Crossostoma lacustre and Rhinogobius candidianus, in the Hapen Creek of availability has to be sampled separately. Point abundance sam- northern Taiwan. Taiwania 49:166–174. pling (PAS) methods can get multiple fish at sampling plots, so Conallin, J., M. Olsen, E. Boegh, J. K. Jensen, and S. Pedersen. 2010. Habitat that the frequency of habitat utilization can be calculated from suitability indices development in Denmark: are international indices applicable under small lowland stream conditions? International Journal of River individual fish and sampling plots. In addition, the random sites Basin Management 8:151–160. of PAS provide information for calculating habitat availabil- Copp, G. H. 1991. Sampling strategies and capture methods in the fisheries ity. Therefore, abundance data from PAS methods can provide management of large river systems. Pages 261–270 in K. T. O’Grady, density-enhanced results that are more accurate than those of A. J. B. Butterworth, P. B. Spillett, and J. C. J. Domaniewski, editors. Fish- nonquantitative methods. However, the size of the PAS frame eries in the year 2000: proceedings of the 21st anniversary conference of the Institute of Fisheries Management. Institute of Fisheries Management, is critical. In this study, the individual lengths of the four target Nottingham, UK. species were shorter than 1 m. Therefore, a 1-m × 1-m PAS Copp, G. H. 2010. Patterns of diel activity and species richness in young and frame is large enough to catch multiple individuals and small small fishes of European streams: a review of 20 years of point abundance enough to include little variation in environmental conditions. sampling by electrofishing. Fish and Fisheries 11:439–460. We demonstrated the advantage of using abundance data in Freeman, M. C., Z. H. Bowen, and J. H. Crance. 1997. Transferability of habitat suitability criteria for fishes in warmwater streams. North American Journal determining the mean values and HSIs for microhabitat utiliza- of Fisheries Management 17:20–31. tion. Habitat suitability indices were first developed by using oc- Fretwell, S. D., and H. L. Lucas. 1969. On territorial behavior and other factors currence data (i.e., presence/absence) for fast assessment of fish influencing habitat distribution in birds, I. Theoretical development. Acta habitat in streams (Bovee 1986). However, with recent devel- Biotheoretica 19:16–36. + Rutilus rutilus opments and the promotion of advanced PAS techniques (Bain Garner, P. 1995. Suitability indices for juvenile 0 Roach ( L.) using point abundance sampling data. Regulated Rivers: Research and et al. 1985; Copp 1991, 2010; Schwartz and Herricks 2004), Management 10:99–104. abundance data can be acquired more easily and quickly than Glozier, N. E., J. M. Culp, and G. J. Scrimgeour. 1997. Transferability of habitat before. Furthermore, the concept of an ecological niche is or- suitability curves for a benthic minnow, Rhinichthys cataractae. Journal of Downloaded by [Department Of Fisheries] at 19:30 20 January 2013 ganism oriented (Grinnell 1917; Hutchinson 1959); therefore, Freshwater Ecology 12:379–393. habitat suitability modeling, which is an operational applica- Grinnell, J. 1917. The niche-relationships of the California Thrasher. Auk 34:427–433. tion of the niche concept, should also be organism oriented. Hirzel, A. H., and G. Le Lay. 2008. Habitat suitability modelling and niche The quantitative measurement (i.e., abundance data) should be theory. Journal of Applied Ecology 45:1372–1381. adopted over the conventional semiquantitatively measurement Hutchinson, G. E. 1959. Homage to Santa Rosalia, or why are there so many (i.e., occurrence data) whenever possible. kinds of animals? American Naturalist 93:145–159. Jowett, I. G., and A. J. H. Davey. 2007. A comparison of composite habitat suitability indices and generalized additive models of invertebrate abundance and fish presence–habitat availability. Transactions of the American Fisheries Society 136:428–444. ACKNOWLEDGMENTS Lee, P. Y., and J. P. Suen. 2012. Niche partitioning of fish assemblages in The authors gratefully acknowledge the research support a mountain stream with frequent natural disturbances: an examination of of the National Science Council, Taiwan, under grant num- microhabitat in riffle areas. Ecology of Freshwater Fish 21:255–265. Moir, H. J., C. N. Gibbins, C. Soulsby, and A. F. Youngson. 2005. PHABSIM bers NSC 99-2628-E-006-159 and NSC 99-2811-E-006-054. modelling of Atlantic Salmon spawning habitat in an upland stream: testing We also thank Wei-Che Su, Shih-Kai Chou, Tsung-Chi Chang, the influence of habitat suitability indices on model output. River Research Ying-Sheng Lyu, Ruei-Hong Wang, Ting-Hsiang Huang, Hung- and Applications 21:1021–1034. 96 LEE AND SUEN

Orth, D. J. 1987. Ecological considerations in the development and application of Thomas, J. A., and K. D. Bovee. 1993. Application and testing of a procedure instream flow-habitat models. Regulated Rivers: Research and Management to evaluate transferability of habitat suitability criteria. Regulated Rivers: 1:171–181. Research and Management 8:285–294. Parasiewicz, P., and J. D. Walker. 2007. Comparison of mesoHABSIM with Vadas, R. L., Jr., and D. J. Orth. 2001. Formulation of habitat suitability models two microhabitat models (PHABSIM and HARPHA). River Research and for stream fish guilds: do the standard methods work? Transactions of the Applications 23:904–923. American Fisheries Society 130:217–235. Rosenfeld, J. 2003. Assessing the habitat requirements of stream fishes: an Vilizzi, L., G. H. Copp, and J. M. Roussel. 2004. Assessing variation in suit- overview and evaluation of different approaches. Transactions of the Ameri- ability curves and electivity profiles in temporal studies of fish habitat use. can Fisheries Society 132:953–968. River Research and Applications 20:605–618. Schwartz, J. S., and E. E. Herricks. 2004. Use of prepositioned areal electrofish- Vismara, R., A. Azzellino, R. Bosi, G. Crosa, and G. Gentili. 2001. Habitat ing devices with rod electrodes in small streams. North American Journal of suitability curves for Brown Trout (Salmo trutta fario L.) in the River Adda, Fisheries Management 24:1330–1340. northern Italy: comparing univariate and multivariate approaches. Regulated Scott, T. A., and P. Y. Lee. In press. Reconciling social interaction with habitat Rivers: Research and Management 17:37–50. selection in territorial species. Theory in Biosciences. DOI: 10.1007/s12064- Wiens, J. A. 1989. Spatial scaling in ecology. Functional Ecology 3:385–397. 012-0166-z. WRA (Water Resources Agency). 2009. Hydrological yearbook of Taiwan, Stamps, J. A. 1988. Conspecific attraction and aggregation in territorial species. Republic of China: part II—river stage and discharge. WRA, Ministry of American Naturalist 131:329–347. Economic Affairs, Taipei, Taiwan. Downloaded by [Department Of Fisheries] at 19:30 20 January 2013 This article was downloaded by: [Department Of Fisheries] On: 20 January 2013, At: 19:31 Publisher: Taylor & Francis Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

North American Journal of Fisheries Management Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/ujfm20 Reproductive Behavior and Success of Captive-Reared Chinook Salmon Spawning under Natural Conditions David A. Venditti a , Chris A. James a c & Paul Kline b d a Idaho Department of Fish and Game, Nampa Fish Research Office, 1414 East Locust Lane, Nampa, Idaho, 83686, USA b Idaho Department of Fish and Game, Eagle Fish Hatchery, 1800 Trout Road, Eagle, Idaho, 83616, USA c Oregon Department of Fish and Wildlife, Post Office Box 9, John Day, Oregon, 97845, USA d Idaho Department of Fish and Game, 600 South Walnut, Boise, Idaho, 83712, USA Version of record first published: 18 Jan 2013.

To cite this article: David A. Venditti , Chris A. James & Paul Kline (2013): Reproductive Behavior and Success of Captive- Reared Chinook Salmon Spawning under Natural Conditions, North American Journal of Fisheries Management, 33:1, 97-107 To link to this article: http://dx.doi.org/10.1080/02755947.2012.746244

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material. North American Journal of Fisheries Management 33:97–107, 2013 C American Fisheries Society 2013 ISSN: 0275-5947 print / 1548-8675 online DOI: 10.1080/02755947.2012.746244

ARTICLE

Reproductive Behavior and Success of Captive-Reared Chinook Salmon Spawning under Natural Conditions

David A. Venditti* and Chris A. James1 Idaho Department of Fish and Game, Nampa Fish Research Ofﬁce, 1414 East Locust Lane, Nampa, Idaho 83686, USA Paul Kline2 Idaho Department of Fish and Game, Eagle Fish Hatchery, 1800 Trout Road, Eagle, Idaho 83616, USA

Abstract In response to declining returns of Pacific salmon Oncorhynchus spp., captive-rearing programs have emerged as one approach to maintain natural-origin stocks while mechanisms responsible for their decline are addressed. However, it remains unclear if observed differences in reproductive behavior between captive-origin (captive) and natural-origin (natural) Chinook Salmon O. tshawytscha in laboratory experiments accurately predict reproductive success of captive salmon in the natural environment. Over a 4-year period, 859 natural Chinook Salmon juveniles were collected for full-term rearing in captivity. From these, we returned 247 maturing adults to their natal stream over 2 years, assessed trends in habitat use, monitored their courtship and spawning behaviors, and sampled their redds to determine if fertilized eggs were present. Habitat use followed a logical trend toward spawning areas as the season progressed. Captive fish were smaller and spawned later than did natural fish, but their size did not prevent females from constructing redds or prevent males from courting and fertilizing eggs. Captive fish were observed participating in 16 spawning events over two seasons. Captive and natural males displayed the same courtship behaviors, but natural males were more aggressive. Captive females selected redd sites similar to those of natural females and displayed digging behavior consistent with published observations. Eggs were collected from 22 of the 26 redds sampled, and survival to organogenesis was 68.3% in 2001 and 34.6% in 2002. The decline in 2002 was at least partially due to an apparent failure of females from brood year 1997 to produce viable eggs. We estimated captive fish contributed 19,000 eyed eggs to the population. If the 859 juveniles reared in captivity had been left in the stream, fewer than one adult would be expected to have returned. Growth, timing of spawning, and egg survival issues remain in captive fish, which illustrates a need to focus future research on improved culture techniques. Downloaded by [Department Of Fisheries] at 19:31 20 January 2013

In response to declining returns of Paciﬁc salmon collecting naturally produced adults or juveniles from their na- Oncorhynchus spp. over much of their range during the past tal environment and rearing them to sexual maturity in artiﬁcial several decades, hatchery conservation programs have emerged culture conditions. At this point, the two techniques diverge. as one approach to maintain local populations while the under- In a broodstock program, the adults are spawned in captivity lying mechanisms responsible for their decline are addressed and most of the resulting progeny are released at one or more (Flagg et al. 1995; Sayre 1995; Berejikian et al. 2004; Hebdon life history stages (e.g., eyed egg, parr, or smolt) into their na- et al. 2004). These programs commonly take one of two general tal waters. A small number of progeny are typically retained forms, and are generally referred to as captive broodstocking in culture to continue the broodstock (see Flagg et al. 2004). and captive rearing. These programs are typically initiated by Conversely, in a captive-rearing program, adults are returned to

*Corresponding author: [email protected] 1Present address: Oregon Department of Fish and Wildlife, Post Ofﬁce Box 9, John Day, Oregon 97845, USA. 2Present address: Idaho Department of Fish and Game, 600 South Walnut, Boise, Idaho 83712, USA. Received July 21, 2011; accepted October 26, 2012 97 98 VENDITTI ET AL.

their natal waters and allowed to spawn with other captive-origin reproduce in the natural environment, (2) estimate embryo (captive) or natural-origin (natural) individuals (see Kuligowski survival to organogenesis (developmental stage characterized et al. 2005). The subsequent brood group is again collected from by the formation of internal organs and circulatory system) the natural population, resulting in a hatchery program that is in eggs from captive females under natural conditions, and fully integrated with the natural population. (3) describe habitat use by captive Chinook Salmon prior to The captive-rearing concept has several theoretical and prac- spawning in a natural stream environment. tical advantages over broodstocking. By including naturally produced juveniles in the program each year, natural selection is allowed to function as the parents compete for spawning access METHODS and mates (Berejikian et al. 2010, 2011) as well as during the This study took place in the West Fork Yankee Fork Salmon period eggs or juveniles are in the stream environment. Hatch- River (WFYF), located in central Idaho (Figure 1). Habitat con- ery selection is also reduced since artificial spawning is not ditions in the WFYF are nearly pristine, with little or no human needed (Hankin et al. 2009), although some selection undoubt- development. Hatchery influence has also been minimal. Prior edly occurs during rearing. Facility demands are also reduced to this study, the stream had received three relatively small rein a captive-rearing program because it is not necessary to rear leases of Chinook Salmon parr in 1977 (N = 57,000), 1978 (N = large numbers of juveniles to a specific life stage before release. 51,000), and 1994 (N = 25,000; Idaho Department of Fish and For example, in this program we only needed to rear approxi- Game, unpublished data). mately 300 individuals per brood year to adulthood, when fish Fish contributing to this study were collected as juveniles left the facilities between the ages 2 and 5. from brood years 1996–1999. Our goal was to collect 300 juve- Before captive-rearing programs can be considered more niles from each brood year, but this goal was not met in 3 of the than experimental in nature, fundamental questions about the 4 years. Brood year 1996 juveniles (N = 120) were collected reproductive potential of these fish in the natural environment as parr (87%) and smolts (13%) using a rotary screw trap (E.G. must be answered (Fleming and Gross 1992, 1993; Flagg and Solutions, Corvallis, Oregon) located in the WFYF near the con- Mahnken 1995; Berejikian et al. 2001a, 2001b). Previous work fluence with the Yankee Fork Salmon River. Those from brood comparing the spawning behavior of natural- and hatchery- years 1997 (N = 210) and 1998 (N = 229) were collected as origin Chinook Salmon O. tshawytscha and Coho Salmon O. parr by beach seining in the lower WFYF, and those from brood kisutch in artificial channels (Berejikian et al. 1997, 2001a, year 1999 were collected as eyed eggs by hydraulic extraction 2001b) and in the natural environment (Chebanov and Riddell (McNeil 1964; Collins et al. 2000; Berejikian et al. 2011) from 1998) reported measurable differences between the two groups. redds within the lower 4 km of the WFYF. However, due to the In these studies, natural males usually dominated hatchery males 3-year maturation window for Chinook Salmon, not all of the when competing for mates, and when dominant, hatchery males maturing individuals from any single brood year were available generally displayed lower courtship frequencies prior to spawn- for this 2-year study. ing than did natural males. Similar results have been reported After collection, the juveniles were reared at two hatch- for Atlantic Salmon Salmo salar (Fleming et al. 1996, 2000) of ery facilities until approximately 1 month before spawning. natural and production-hatchery-origin fish spawning together Juveniles were reared in freshwater from the time of collec- in artificial channels. However, it remains unclear how these tion through smoltification at the Idaho Department of Fish differences may affect the ability of captive fish to reproduce in and Game (IDFG) Eagle Fish Hatchery (Eagle), Eagle, Idaho. the natural environment. Venditti et al. (2003a) provide additional detail on fish rearing at Because captive Chinook Salmon have not previously been Eagle. During the time natural-origin smolts were migrating to Downloaded by [Department Of Fisheries] at 19:31 20 January 2013 evaluated in the natural environment, it was also not known if the Pacific Ocean, we transferred the majority of fish to the Na- these fish would make appropriate habitat selections prior to tional Oceanic and Atmospheric Administration (NOAA) Fish- spawning. In addition, habitat use by maturing Chinook Salmon eries, Manchester Marine Experimental Station (Manchester), between the time of entry into spawning streams and the onset Manchester, Washington, for rearing in saltwater to a point ap- of spawning has not been described well in the literature. Pools, proximately 1 to 2 months prior to sexual maturity. Maynard cut banks, and large woody debris (LWD) can provide protec- et al. (2003) provide additional detail on fish rearing at Manch- tion from predators and probably help conserve depleted energy ester. Approximately 20% of each cohort remained at Eagle for reserves for spawning, and we have observed natural-origin Chi- their entire life cycle to provide a measure of protection against nook Salmon staging in these habitat types before spawning in possible catastrophic loss at Manchester. Near the beginning of other systems. the sexual maturation process, fish were transported to Eagle for In this study, we released sexually maturing, captive a period of freshwater maturation before being returned to the Chinook Salmon into their natal stream and monitored their WFYF to reproduce naturally. reproductive behavior and habitat use. Our objectives were to Maturing individuals were identified into two maturation (1) determine how behavioral differences observed in captive sorts approximately 6 weeks apart using one of two techniques. Chinook Salmon in previous studies affected their ability to In 2001, maturing fish were identified at both facilities by visual CAPTIVE-REARED CHINOOK SALMON REPRODUCTIVE SUCCESS 99 Downloaded by [Department Of Fisheries] at 19:31 20 January 2013

FIGURE 1. Location of the West Fork Yankee Fork Salmon River (WFYF) and major tributaries. Captive-reared Chinook Salmon were released at two locations in the WFYF in August 2001 and 2002, and allowed to distribute volitionally. The study section was divided into six reaches of approximately equal length (1.6 km) to facilitate systematic sampling upstream from a ﬁsh weir. The boundary between reaches is represented by Bound. 100 VENDITTI ET AL.

inspection and tactile manipulation of the gonads through the each day, which allowed the entire study section to be monitored body wall. Maturing fish from Manchester were transferred to every 1–2 d. Observers walked slowly upstream watching for Eagle immediately after both maturation sorts. In 2002, matur- Chinook Salmon and, when observed, recorded habitat associ- ing fish at Manchester were identified using portable ultrasound ation as well as fish behaviors. We did not quantify the relative equipment (Aloka SSD-500V with an Aloka Electronic Linear availability of the different habitat types, but assumed this re- Probe UST-556L-7.5) and transferred to Eagle. A second matu- mained constant with the stable flows during the time period ration sort was conducted at Manchester using visual inspection from late August through October each year, so trends in use and tactile manipulation, and maturing fish were transported to reflect real shifts throughout the spawning season. Eagle in a second shipment. The degree of maturation of fish When an individual Chinook Salmon was located, the fish held at Eagle was determined by visual and tactile inspection. was observed for 5 min and we recorded its general behavior Fish not determined to be maturing remained in culture at both and associated habitat type (Table 1). This technique provided a locations until their fifth year. Those not maturing at age 5 were standardized measurement of trends in fish behavior and habitat culled. use over time. If multiple fish were observed simultaneously, We fitted all maturing adults with color-coded and numbered behavior and habitat information were recorded separately for disk tags (2.54 cm diameter) before release for volitional spawn- each individual. ing. Color codes identified fish to brood year and the color code We determined from previous work that the postrelease be- and number combination allowed observers to identify individ- havior of captive Chinook Salmon could be divided into three ual fish. Prior to tagging, fish were anesthetized in a solution of general time periods of roughly equal duration centered on MS-222 (tricaine methanesulfonate, buffered to neutrality with peak spawning activity. We classified these as early, peak, and sodium bicarbonate), weighed to the nearest 1 g, measured to late spawning periods. The timing of peak spawning varied be- the nearest 1 mm FL, and scanned for a PIT tag, which identi- tween years, so general behaviors exhibited by Chinook Salmon fied its brood year. We attached disk tags to the fish by passing during the standardized 5-min observation periods were exam- a stainless steel pin through a hole in the center of the tag and ined to determine when spawning-related activities represented then through the dorsal musculature of the fish just ventral to the the highest proportion of observations. This became the peak midline of the dorsal fin. A corresponding tag (with the same spawning period, and the early and late periods were then as- number and color code) was then slipped onto the pin on the op- signed. We then plotted observed behaviors and habitat associ- posite side of the fish. The pin was trimmed to length, and the ations during these three periods to examine how behaviors and second tag was secured in place by forming a loop at the end of habitat associations changed throughout the study. the pin with needle-nose pliers. When we observed courtship behaviors during the standard- The study section on the WFYF was approximately 9.7 km ized observation period, additional observation protocols were in length and contained high quality Chinook Salmon spawning initiated. If a new redd was present, we flagged it and recorded habitat. To prevent emigration of captive fish we constructed a the disk tag code of the female. During this second set of ob- blocking weir at the downstream end of the section. The compo- servations, we made detailed records of fish behavior (Table 1) nents of the weir were flown to the site via helicopter and assem- in 10-min intervals to predict the time until spawning and to bled on site. Trap boxes built into the weir allowed technicians determine how the frequency of courtship behaviors changed to pass natural fish that entered the traps in either direction or during the time leading up to and immediately after spawning. return captive fish that attempted to move downstream back into If, based on these frequencies, the observer judged spawning the study section. Shortly after the weir was assembled, captive would occur within 1–2 h they remained with that spawning fish were flown to one of two release areas (in two 60-L coolers), group (typically one female along with one or more competing Downloaded by [Department Of Fisheries] at 19:31 20 January 2013 which contained abundant holding habitat and consisted of three males) and continued to record all courtship behaviors in 10-min to five closely spaced pools. The first release site was near the blocks until 30 min after spawning. downstream end of the study section and the second was near From these records, we compared the frequency of courtship the middle (Figure 1). We transported no more than six captive behaviors (behavior events/10-min interval) by natural and cap- adults at a time, and total flight time (from loading to release) tive males. Male courtship behaviors of interest included quiv- was approximately 10–15 min depending on release site. We ers, crossovers, and aggression. In a quiver, the male darts toward did not construct a migration barrier at the upstream limit of the the female at a 30–45◦ angle and just before contact performs a study reach because habitat conditions (abundant bedrock, large high frequency, low amplitude undulation of the body (Tautz and cobble, and stream size) above the confluence of the WFYF and Groot 1975). A crossover involves the movement of the male Cabin Creek made spawning above this point unlikely. from one side of the female to the other with the male’s head Data collection began approximately 24 h after fish were passing over the female’s caudal peduncle (Hartman 1969; Tautz released. The study section was divided into six reaches, each and Groot 1975). Aggression was an attempt by one male to approximately 1.6 km in length, to permit systematic obser- chase another male away from the vicinity of the redd. Courtship vations of Chinook Salmon activities upstream from the weir. and aggression frequencies were computed for both male types Observers were assigned one to four stream reaches to survey during each 10-min observation period. Average values for each CAPTIVE-REARED CHINOOK SALMON REPRODUCTIVE SUCCESS 101

TABLE 1. Habitat and behavior variables recorded during observations of on the errors were computed for each observation period (pre- captive-reared Chinook Salmon released into the West Fork Yankee Fork Salmon and postspawning) to construct approximate 90% confidence in- River for volitional spawning, August–October, 2001–2002. tervals (CIs)√ around the point estimates using the formula CI = Variable Definition X ± 1.6(SD/ n), where X is the mean number of behaviors observed during that 10-min interval pre- or postspawning, SD is Habitat types the standard deviation for the behavior in that interval, and n is Vegetation Aquatic macrophytes or overhanging the number of spawning events with males of that origin (Neter terrestrial vegetation et al. 1988). Bounds for all observation periods overlapped for Cut bank Stream extends under the terrestrial surface both years, so behavioral data from both years were pooled. Pool Area of low gradient, increased water depth, We also performed a similar analysis to assess the digging and low velocity behavior (see Tautz and Groot 1975) of captive females spawn- > > Large woody Woody material 1 m in length and ing with natural and captive males. We assumed all digs before debris 10 cm in diameter egg deposition were nest digs and those after deposition were (LWD) cover digs. Dig frequencies (number of digs/10 min) and CIs Riffle–run Shallow, swiftly flowing, stream section were computed for each 10-min observation period as described Pool tail-out Gradient break at the downstream extent of a above. Average values and approximate 90% CIs for each obser- pool vation period relative to spawning were computed and plotted General behaviors (both sexes) for captive females being courted by natural and captive males. Staging Remaining in one position, not associated The mean estimates of dig frequency were also compared with with a redd values from the published literature. Milling Movement not resulting in longitudinal After spawning, we hydraulically sampled eggs from a por- displacement tion of redds produced by captive females to verify egg depo- Moving (A) Movement in an upstream direction sition, estimate embryo survival to organogenesis, and estimate Moving (B) Movement in a downstream direction potential contribution of eyed eggs to the population. We used Aggression Aggression between Chinook Salmon of thermographs located within the study section to estimate the undetermined sex number of Celsius temperature units (CTU; sum of mean daily Holding Maintaining position on or near a redd temperatures in ◦C) developing embryos received. We collected Courting Active male and receptive female eggs after they had received a minimum of 200 CTU to en- Spawn Observed release of eggs and milt sure the embryo had reached organogenesis (Velsen 1980; Piper Male courtship et al. 1989) and generally before the embryos reached the eyed Quiver Dart toward female ending with body stage. Opaque eggs or those having fungal growth were con- vibrations sidered dead. Clear eggs were classified as viable and fixed in Crossover Movement to opposite side, head passing Stockard’s solution (Velsen 1980), which causes embryos in over peduncle early organogenesis to become visible. Eggs in this category Aggression (A) Male on male aggression were further determined to be fertilized or unfertilized depend- Aggression (B) Male on female aggression ing on the presence or absence of an embryo in the fixed egg. Aggression (C) Male on other species aggression The number of eggs in each category was enumerated and the Following Female present, no redd present percentage in each computed. Finally, we estimated the potential number of eyed eggs (EN) produced by captive females as:

Downloaded by [Department Of Fisheries] at 19:31 20 January 2013 Satellite Holding away or downstream from a courting pair E = N × V × 0.5F Female courtship N R Aggression (A) Female on female aggression where N is the number of redds spawned by captive females, Aggression (B) Female on male aggression R V is the mean proportion of fertilized eggs in the redds, and F Aggression (C) Female on other species aggression is the mean fecundity of captive females (Venditti et al. 2003a, Test dig 2–6 body flexures, not concentrated 2003b). Fecundity estimates were adjusted for observed egg Nest dig 5–8 body flexures in a concentrated area retention by captive Chinook Salmon in a spawning channel Cover dig 8–12 body flexures along redd perimeter (Berejikian et al. 2001b). Nesting Redd grooming, trenching, or defense In 2002, we recorded the number of live and dead eggs in redds spawned by natural females, while collecting the observation period, relative to the time of spawning (designated next program cohort. We provide these data as a baseline to as time zero), were computed for natural and captive males and compare egg survival from captive females. Redds selected plotted to determine how similar (or dissimilar) overall behavior for this purpose were located in similar habitat in the WFYF patterns and frequencies were between the two groups. Bounds below the study section. We timed our sampling with the 102 VENDITTI ET AL.

TABLE 2. Number and average FL (mm) of captive-reared Chinook Salmon by brood year (BY) released into the West Fork Yankee Fork Salmon River in August 2001–2002.

FL (mm) Release year BY Average (SD) Range N 2001 1996 422.3 (76.3) 350–502 4 1997 542.0 (66.2) 328–630 42 1998 538.4 (79.0) 335–655 43 All 536.5 (76.1) 328–655 89 2002 1997 532.9 (48.6) 453–620 25 1998 557.8 (54.2) 443–665 54 1999 361.7 (20.1) 297–400 77 All 462.3 (102.9) 297–665 158a FIGURE 2. Behaviors of captive-reared Chinook Salmon released to spawn aTwo fish released in 2002 had lost their PIT tags, making it impossible to determine volitionally in the West Fork Yankee Fork Salmon River in the summers of 2001 the brood year to which they belonged. and 2002. In 2001 the early period extended from August 19 to August 30, the peak period was August 31–September 11, and the late period extended from September 12 to September 23. In 2002, the early period extended from August temperature data described above, except we sampled these 9 to August 23, the peak period was August 24–September 8, and the late period redds at approximately 300 CTU to ensure the embryos had extended from September 9 to September 23. Values above the bars represent reached the eyed stage of development. the number of observations. Habitat use observed in captive Chinook Salmon also re- RESULTS flected the increasing trend toward spawning-related activities. Maturing, captive Chinook Salmon released into the WFYF We made a total of 1,478, 1,760, and 665 observations of habitat to spawn volitionally were substantially smaller than natural associations during the early, middle, and late spawning periods, adults returning to the WFYF and other nearby locations. Fish respectively. During the early spawning period, pools, LWD, and released on August 17, 2001 (N = 89) were from brood years cut banks were all heavily used habitat types (Figure 3). Riffles 1996–1998 and averaged 536.5 mm FL (Table 2). Captive Chi- and runs were also frequently used at this time, but this may nook Salmon released on August 8, 2002 (N = 158) averaged be because fish are more visible in these habitats than they are 462.3 mm FL and consisted of fish from brood years 1997–1999 in pools or when they are associated with LWD. Pool tail-outs (Table 2). Natural-origin carcasses collected from throughout and riffles and runs became the most-used habitat types during the WFYF during 2001–2002 (N = 16) averaged 784 mm FL the peak spawning period (Figure 3). These two habitat types (SD = 111; range, 680–1,040 mm; IDFG, unpublished data), (tail-outs and riffle–run) became even more heavily used during and natural-origin adults returning to the nearby Sawtooth Fish the late spawning period. Captive Chinook Salmon were rarely Hatchery (N = 1,539) during the same years averaged 808 mm FL (SD = 115; range, 430–1,090 mm; see Snider et al. 2003, 2004). General behaviors observed in captive Chinook Salmon dur- Downloaded by [Department Of Fisheries] at 19:31 20 January 2013 ing standardized observation periods were similar during both years of the study and followed a consistent pattern as sexual maturation progressed. Captive fish distributed themselves throughout the study section with relatively few fish (0 in 2001 and approximately 10% in 2002) entering the downstream trap box. (One natural male entered the upstream trap box in 2002 and was passed into the study section). We made a total of 1,467, 2,050, and 665 observations of behavior in the early, middle, and late spawning periods, respectively. The dominant behavior during the early spawning periods was staging (Figure 2). During the peak spawning period, staging and courting became the dominant behaviors as fish competed for mates FIGURE 3. Habitat use of captive-reared Chinook Salmon released to spawn volitionally in the West Fork Yankee Fork Salmon River in the summers of 2001 and prepared and defended redds. Courting and holding on or be- and 2002. Early, peak, and late periods are the same as described in Figure 2. low redds became the dominant activities during the late spawn- LWD = large woody debris. Values above the bars represent the number of ing period (Figure 2). observations. CAPTIVE-REARED CHINOOK SALMON REPRODUCTIVE SUCCESS 103

observed associated with aquatic vegetation, although this habitat type was rare in the WFYF. We documented 16 separate spawning events during the 2 years of study (eight each in 2001 and 2002) in which captive Chinook Salmon participated. In 2001, three of the observed spawning events involved captive females and natural males, and both fish were captive in the remaining five. In 2002, one mating involved a captive female and a natural male, and both sexes were of captive origin in the remaining seven matings. Subdominant or precocial males, or both, of natural origin were typically present, but their participation was not assessed. We observed no natural females spawning in the study reach in either year, although a small number of natural-origin redds (N ≤ 3 each year) were present prior to releasing captive fish. Captive males displayed the same courtship behaviors as natural males, and their courtship frequencies did not differ from natural males during the time leading up to spawning. The frequency of quivers (N = 473 observed) and crossovers (N = 568 observed) by natural males generally increased as spawning approached with a pronounced spike in quivers immediately before spawning (Figure 4A, B). Courtship frequencies by captive males remained constant or declined slightly during the period leading up to spawning (N = 1,465 quivers and N = 987 crossovers observed), but these fish exhibited the spike in quivers immediately before spawning (Figure 4A, B). Despite the different trajectories of the point estimates, all 90% CIs overlapped except at 60 min before spawning when captive males quivered at a higher rate than natural males (Figure 4A). The largest difference between the two types of males was that captive males were significantly less aggressive than natural males before spawning. Natural males typically displayed between 10 and 15 aggressive acts within each 10-min observa- = FIGURE 4. Frequency (behavior events/10-min interval) of courtship and tion period (N 520 observed), while captive males only aver- aggression (mean ± approximate 90%CI) observed in male captive-reared and aged around five aggressive acts (N = 462 observed). The 90% natural-origin Chinook Salmon during observed spawning events (N = 16) in the CIs did not overlap during any time period prior to spawning West Fork Yankee Fork Salmon River in 2001 and 2002. Time zero is spawning (Figure 4C). and negative and positive numbers are minutes before and after spawning, Captive females dug and covered redds at similar rates be- respectively. Values in the legend represent the number of observations for that male type. tween years and their behavior was not affected by male origin. We observed captive females making a total of 962 combined Downloaded by [Department Of Fisheries] at 19:31 20 January 2013 nest and cover digs during the 2 years of study. Captive females One redd contained only dead eggs but appeared to have been made approximately two to four nest digs during each 10-min constructed in high quality habitat and was well developed. observation period up until time of egg deposition (Figure 5). Sampling revealed that this redd was constructed on a thin (ap- After spawning, females proceeded to cover dig almost con- proximately 7 cm) layer of gravel–cobble armoring over a large, tinuously for about 10 min and maintained elevated digging decayed log. Omitting this redd from the analysis increased the frequencies for at least 30 min (Figure 5). average egg viability to 68.3%. Based on these data (68.3% via- We collected eggs from a portion of the redds in which captive bility) and an estimated fecundity of 1,221 eggs/female (Venditti females spawned each year to estimate egg survival to organo- et al. 2003a, 2003b), potential production by captive females in genesis. No redd superimposition occurred in either year, so fe- 2001 was estimated at approximately 7,500 eyed eggs. On Oc- males spawning on top of previously constructed redds was not tober 8–9, 2002, we sampled 18 of the 33 redds constructed a factor in egg survival. On October 15–16, 2001, we attempted by captive females. We found eggs in 17 of the 18 redds sam- to hydraulically collect eggs from 8 of 18 redds constructed by pled, and nine contained viable eggs. None of the redds sampled captive females and found eggs in five of the eight redds sam- (N = 5) that were known to have been used for spawning by pled. The percentage of viable eggs in these redds ranged from brood year 1997 females contained live eggs. The percentage of 0 to 89% and averaged 54.7%. All clear eggs were fertilized. live eggs in redds sampled in 2002 ranged from 0 to 100%, and 104 VENDITTI ET AL.

et al. 2003; Swanson et al. 2008). These remain areas where further research is needed to improve culture practices for Chinook Salmon. The behavior and habitat use of captive Chinook Salmon after release in this study changed over time in a manner that reflected their changing requirements as the spawning season progressed. Initially, study fish were generally observed holding position or moving, often in association with pools and LWD, which is consistent with behavior documented in prespawning Atlantic Salmon (Bardonnet and Bagliniere` 2000). Selecting habitats with generally low water velocity and complex structures may be of benefit by helping conserve energy reserves for spawning activities (Torgersen et al. 1999) or by providing FIGURE 5. Frequency (digs/10-min interval) of digging (mean ± approxi- refuge from predators. As the season progressed, spawning- mate 90%CI) by captive-reared female (F) Chinook Salmon during observed related behaviors became dominant and fish moved onto pool spawning events (N = 16) with captive-reared (N = 12) and natural-origin (N = tail-outs, although pools and LWD remained important as rest- 4) males (M) in the West Fork Yankee Fork Salmon River in 2001 and 2002. ing and staging areas. This is consistent with the observations Time zero is spawning and negative and positive numbers are minutes before documented in Atlantic Salmon by Bardonnet and Bagliniere` and after spawning, respectively. Values in the legend represent the number of digs observed. (2000) and natural Chinook Salmon (D. A. Venditti, personal observation). averaged 34.6% overall and 65.3% for those redds that contained Several lessons can be learned from our detailed spawning live eggs. Based on these data (34.6% viable) and an estimated observations and from the behavior patterns that emerged. In fecundity of 2,011 eggs/female, production by captive females males, the pattern of courtship and frequency of aggression dif- in 2002 was estimated to be approximately 11,500 eyed eggs. fered between captive fish and those of natural origin, but captive males displayed the full suite of expected behaviors. Given the competitive advantage larger size provides male salmon DISCUSSION (Fleming and Gross 1993; Berejikian et al. 1997), our results Captive fish were substantially smaller and spawned several are neither surprising nor unexpected. In an evaluation of the weeks later than natural Chinook Salmon returning to the WFYF ability of captive Chinook Salmon to spawn and fertilize eggs and other nearby populations. For example, natural adults pass- in a natural stream environment, it is not necessary (or perhaps ing the Sawtooth Hatchery weir (approximately 40 km upstream even desirable if the technique is implemented for augmen- from the mouth of the WFYF on the Salmon River) ranged from tation or recovery) for captive males to outcompete males of 430 to 1,090 mm FL in 1991 (Snider et al. 2003) and from 450 natural origin. Captive males courted appropriately and suffi- to 1,090 mm in 1992 (Snider et al. 2004). However, small size ciently to induce oviposition in females and they fertilized eggs did not prevent captive females from constructing redds in what under natural conditions. The frequency of prespawn digging appeared to be appropriate locations (e.g., pool tail-outs with by captive females was similar to the average of one dig every clean gravel) or prevent captive males from courting, spawning, 2.5 min (4 digs/10 min) reported for natural Chinook Salmon and successfully fertilizing eggs. Also, captive females spawned (Schroder et al. 2008), who also reported elevated cover dig- later than their natural counterparts, such that natural males were ging for approximately 30 min postspawning. This indicates Downloaded by [Department Of Fisheries] at 19:31 20 January 2013 only available to spawn with the earliest-spawning captive fe- that captive females retained the innate drive to pair, construct males in both years, which may result in lower survival in their redds, spawn, and provide final redd grooming. It is unlikely our progeny if later spawn timing results in later emergence and tagging methods affected the performance of captive Chinook smaller size going into their first winter. These two conditions Salmon, as other studies have used disk tags to evaluate adult appear to be common in Chinook Salmon captive rearing pro- salmonids and have not noted any effects of the tags on either grams (Joyce et al. 1993; Schiewe et al. 1997; Berejikian et al. survival or behavior (Quinn and Foote 1994; Foote et al. 1997; 2003). Cowen et al. 2007; Doctor et al. 2010). We obtained brood stocks annually from the naturally spawn- Our egg collections revealed challenges that program man- ing population in the WFYF, so changes in size and spawn agers will need to address to maximize the utility of the captive- timing in these fish were probably due to the culture envi- rearing technique for Chinook Salmon. A number of redds sam- ronment. Steelhead O. mykiss and Sockeye Salmon O. nerka pled contained low numbers of eggs or eggs that were not viable. reared in captivity frequently attain body sizes that equal or Based on our observations of progressing redd development exceed that of their natural cohorts (Pollard and Flagg 2004), (e.g., upstream expansion as new egg pockets were added), it and gonadotropin-releasing hormone implants have been used is likely that some of the redds where no eggs were collected to advance spawning in captive Chinook Salmon (Berejikian actually contained eggs that we simply did not locate with our CAPTIVE-REARED CHINOOK SALMON REPRODUCTIVE SUCCESS 105

hydraulic equipment (Berejikian et al. 2011). The fact that a may be more desirable than a high risk of local extirpation number of redds contained a large number of dead eggs, com- (Fraser 2008; Neff et al. 2011). A key unanswered question re- pared with their natural counterparts, suggests some breakdown garding hatchery conservation programs thus remains untested: occurred in the physical or physiological processes required for How quickly can fish from a conservation hatchery program successful spawning in the wild. Identifying and correcting the move back toward their natural fitness optimum? causative factor responsible for the low egg survival should re- We have demonstrated that captive Chinook Salmon reared main a priority for future research. However, despite the low to near maturity in a hatchery are capable of successfully spawn- survival observed, captive Chinook Salmon were responsible ing in the natural environment and can contribute to natural pro- for constructing successful redds and producing approximately duction in depressed populations. Females constructed redds 19,000 eyed eggs in the WFYF over the 2 years of the study. and deposited eggs. Natural males actively competed for and We cannot be certain how many of the eggs collected were spawned with captive females, and in their absence captive fertilized by captive males, but evidence suggests at least a por- males demonstrated the full suite of reproductive behaviors and tion were. Female Chinook Salmon deposit eggs in several nest appeared to successfully fertilize eggs. Growth and spawn tim- pockets within a redd and can have multiple partners, so the ori- ing differ between captive and natural fish, and egg survival to gin of the male was not known for the eggs we collected, even in organogenesis is lower in captive than in natural fish. However, redds where captive males were observed spawning. However, captive rearing has the potential to ensure a continuum of year- captive males were observed participating in three-quarters of to-year reproductive success, in the habitat, for populations of the spawning events we observed (12 of 16), and we believe this spring–summer run Chinook salmon at risk. For example, the accurately reflects their overall contribution to spawning during 859 juveniles brought into the program (brood year 1996–1999 this study and should be reflected in our egg collections. Addi- collections above) produced 247 adults that contributed about tionally, experimental activities at Eagle using WFYF captive 19,000 eyed eggs to the population during the 2 years of study. males from the brood years released demonstrated their milt was Actual eyed-egg production from captive fish was probably even of high quality and viable. Average milt motility was found to higher, since a portion of fish from each brood year matured be about 98.4% for these fish (Venditti et al. 2002), and in fac- outside of the study period. If left in the stream, based on es- torial crosses (where individual males were artificially spawned timated life-stage-specific survival estimates for Salmon River with sublots of eggs from multiple females) egg survival var- tributaries (25.2% egg–parr and 19.4% parr–smolt, Kiefer and ied widely and was dependent primarily on maternal influence Lockhart 1997), average survival of smolts from the WFYF to (Venditti et al. 2003b). Therefore, it seems reasonable to assume Lower Granite Dam (43.8%, Venditti et al. 2007), and the aver- that male gamete quality was not a limiting factor. age Sawtooth Hatchery smolt-to-adult survival for brood years To collect additional juveniles for this program, we removed 1996–1999 (0.8%, IDFG, unpublished data), fewer than one a total of 272 eggs from six redds from natural females in 2001, adult would be expected to have returned from those 859 indi- and 321 eggs from five redds used for spawning by natural-origin viduals. Even if one assumes an optimistic return of one adult females in 2002. Dead eggs were not enumerated in 2001, but (and that adult is female), only about 4,500 eggs could have been their frequency was similar to results obtained the following produced. Future research should focus on developing culture year (D. A. Venditti, personal observation). In 2002, 308 of the methods that enable managers to develop full-term, hatchery- 321 eggs were live indicating natural egg survival to the eyed reared Chinook Salmon adults that more closely approximate stage was approximately 96% in the WFYF (IDFG, unpublished the natural template with respect to size at release and timing of data). The high survival of eggs in redds of natural females in spawning. the WFYF suggests the lower survival in redds spawned in by Downloaded by [Department Of Fisheries] at 19:31 20 January 2013 captive females, in the same years, was an effect of origin and not due to habitat conditions. ACKNOWLEDGMENTS Numerous studies have shown that salmonids in a hatch- We thank the staff at Eagle Hatchery who assisted with fish ery program can quickly respond to artificial selection, thereby collection and rearing. We also thank the staff at Manchester reducing their fitness when released back into the natural envi- Hatchery for rearing most of the fish used in this study. Along ronment (see reviews in Berejikian and Ford 2004; Araki et al. with the hatchery personnel, our field crews made this study pos- 2008). The general interpretation of these results has been that sible. We also thank the biologists with the Shoshone–Bannock fitness lost is fitness that can never be regained. However, natu- Tribes’ fisheries program who operated the screw trap where ral selection can also be strong, and captive-reared individuals a portion of the juveniles were collected, and Paul Bunn for sourced from the local population should be capable of respond- making the study site map. Tim Copeland, Bill Schrader, Dan ing. In a reanalysis of Hood River steelhead data from Araki et al. Schill, and six anonymous reviewers made many suggestions (2007a, 2007b, 2009), Kitada et al. (2011) suggests the effect of that greatly improved earlier versions of the manuscript. Fund- captive rearing could be eliminated in the first generation after ing for this program was provided by the Bonneville Power Ad- reintroducing hatchery fish. For highly imperiled populations, ministration (Project Number 1997-00-100) through the North- paying a fitness penalty from short-term hatchery intervention west Power and Conservation Council’s Fish and Wildlife 106 VENDITTI ET AL.

Program. The use of trade names does not imply endorsement River system, Bristol Bay, Alaska. Transactions of the American Fisheries by the Idaho Department of Fish and Game. Society 139:80–91. Flagg, T. A., and C. V.W. Mahnken. 1995. An assessment of the status of captive broodstock technology for Pacific salmon. Final Report to the Bonneville Power Administration, Project 93-56, Portland, Oregon. REFERENCES Flagg, T. A., C. V. W. Mahnken, and K. A. Johnson. 1995. Captive brood- Araki, H., W. R. Ardren, E. Olsen, B. Cooper, and M. S. Blouin. 2007a. Re- stocks for recovery of Snake River Sockeye Salmon. Pages 81–90 in H. L. productive success of captive-bred Steelhead Trout in the wild: evaluation Schramm Jr. and R. G. Piper, editors. Uses and effects of cultured fishes in of three hatchery programs in the Hood River. Conservation Biology 21: aquatic ecosystems. American Fisheries Society, Symposium 15, Bethesda, 181–190. Maryland. Araki, H., B. A. Berejikian, M. J. Ford, and M. S. Blouin. 2008. Fitness of Flagg, T. A., W. C. McAuley, P. A. Kline, M. S. Powell, D. Taki, and J. C. hatchery-reared salmonids in the wild. Evolutionary Applications 1:342–355. Gislason. 2004. Application of captive broodstocks to preservation of ESA- Araki, H., B. Cooper, and M. S. Blouin. 2007b. Genetic effects of captive breed- listed stocks of Pacific salmon: Redfish Lake Sockeye Salmon case exam- ing cause a rapid, cumulative fitness decline in the wild. Science 318:100–103. ple. Pages 387–400 in M. J. Nickum, P. M. Mazik, J. G. Nickum, and D. Araki, H., B. Cooper, and M. S. Blouin. 2009. Carry-over effect of captive D. MacKinlay, editors. Propagated fish in resource management. American breeding reduces reproductive fitness of wild-born descendants in the wild. Fisheries Society, Symposium 44, Bethesda, Maryland. Biology Letters 5:621–624. Fleming, I. A., and M. R. Gross. 1992. Reproductive behavior of hatchery Bardonnet, A., and J. L. Bagliniere.` 2000. Freshwater habitat of Atlantic Salmon and wild Coho Salmon (Oncorhynchus kisutch): does it differ? Aquaculture (Salmo salar). Canadian Journal of Fisheries and Aquatic Sciences 57:497– 103:101–121. 506. Fleming, I. A., and M. R. Gross. 1993. Breeding success of hatchery and wild Berejikian, B. A., W. T. Fairgrieve, P. Swanson, and E. P. Tezak. 2003. Current Coho Salmon (Oncorhynchus kisutch) in competition. Ecological Applica- velocity and injection of GnRHa affect reproductive behavior and body com- tions 3:230–245. position of captively reared offspring of wild Chinook Salmon (Oncorhynchus Fleming, I. A., K. Hindar, I. B. Mjølnerød, B. Jonsson, T. Balstad, and A. tshawytscha). Canadian Journal of Fisheries and Aquatic Sciences 60: Lamberg. 2000. Lifetime success and interactions of farm salmon invading a 690–699. native population. Proceedings of the Royal Society of London B 267:1517– Berejikian, B. A., T. Flagg, and P. Kline. 2004. Release of captively reared adult 1523. anadromous salmonids for population maintenance and recovery: biological Fleming, I. A., B. Jonsson, M. R. Gross, and A. Lamberg. 1996. An experimental trade-offs and management considerations. Pages 233–245 in M. J. Nickum, study of the reproductive behaviour and success of farmed and wild Atlantic P. M. Mazik, J. G. Nickum, and D. D. MacKinlay, editors. Propagated fish in Salmon (Salmo salar). Journal of Applied Ecology 33:893–905. resource management. American Fisheries Society, Symposium 44, Bethesda, Foote, C. J., G. S. Brown, and C. C. Wood. 1997. Spawning success of males Maryland. using alternative mating tactics in Sockeye Salmon, Oncorhynchus nerka. Berejikian, B. A., and M. J. Ford. 2004. Review of relative fitness of hatchery Canadian Journal of Fisheries and Aquatic Sciences 54:1785–1795. and natural salmon. NOAA Technical Memorandum NMFS-NWFSC-61. Fraser, D. J. 2008. How well can captive breeding programs conserve biodiver- Berejikian, B. A., J. T. Gable, and D. T. Vidergar. 2011. Effectiveness and sity? a review of salmonids. Evolutionary Applications 1:535–586. trade-offs associated with hydraulic egg collections from natural salmon and Hankin, D. G., J. Fitzgibbons, and Y. Chen. 2009. Unnatural random mating Steelhead redds for conservation hatchery programs. Transactions of the policies select for younger age at maturity in hatchery Chinook Salmon American Fisheries Society 140:549–556. (Oncorhynchus tshawytscha) populations. Canadian Journal of Fisheries and Berejikian, B. A., E. P. Tezak, L. Park, E. LaHood, S. L. Schroder, and E. Beall. Aquatic Sciences 66:1505–1521. 2001a. Male competition and breeding success in captively reared and wild Hartman, G. F. 1969. Reproductive biology of the Gerrard stock Rainbow Trout. Coho Salmon (Oncorhynchus kisutch). Canadian Journal of Fisheries and Pages 53–67 in T. G. Northcote, editor. Symposium on salmon and trout in Aquatic Sciences 58:804–810. streams. Institute of Fisheries, University of British Columbia, Vancouver. Berejikian, B. A., E. P. Tezak, and S. L. Schroder. 2001b. Reproductive behavior Hebdon, J. L., P. Kline, D. Taki, and T. A. Flagg. 2004. Evaluating reintroduc- and breeding success of captively reared Chinook Salmon. North American tion strategies for Redfish Lake Sockeye Salmon captive broodstock progeny. Journal of Fisheries Management 21:255–260. Pages 401–413 in M. J. Nickum, P. M. Mazik, J. G. Nickum, and D. D. Berejikian, B. A., E. P. Tezak, S. L. Schroder, C. M. Knudsen, and J. J. MacKinlay, editors. Propagated fish in resource management. American Fish- Hard. 1997. Reproductive behavioral interactions between wild and cap- eries Society, Symposium 44, Bethesda, Maryland.

Downloaded by [Department Of Fisheries] at 19:31 20 January 2013 tively reared Coho Salmon (Oncorhynchus kisutch). ICES Journal of Marine Joyce, J. E., R. M. Martin, and F. P. Thrower. 1993. Successful maturation and Science 54:1040–1050. spawning of captive Chinook Salmon brood stock. Progressive Fish-Culturist Berejikian, B. A., D. M. Van Doornik, R. C. Endicott, T. L. Hoffnagle, E. P. 55:191–194. Tezak, M. E. Moore, and J. Atkins. 2010. Mating success of alternative Kiefer, R. B., and J. N. Lockhart. 1997. Intensive evaluation and monitoring of male phenotypes and evidence for frequency-dependent selection in Chi- Chinook Salmon and Steelhead Trout production, Crooked River and upper nook Salmon, Oncorhynchus tshawytscha. Canadian Journal of Fisheries and Salmon River sites: annual progress report January 1, 1994 to December Aquatic Sciences 67:1933–1941. 31, 1994. Report to the Bonneville Power Administration, Project 91-073, Chebanov, N. A., and B. E. Riddell. 1998. The spawning behavior, selec- Portland, Oregon. tion of mates, and reproductive success of Chinook Salmon (Oncorhynchus Kitada, S., H. Kishino, and K. Hamasaki. 2011. Bias and signiﬁcance of rela- tshawytscha) spawners of natural and hatchery origins under conditions of tive reproductive success estimates based on Steelhead Trout (Oncorhynchus joint spawning. Journal of Ichthyology 38:517–526. mykiss) data: a Bayesian meta-analysis. Canadian Journal of Fisheries and Collins, K. M., E. L. Brannon, L. L. Moulton, M. A. Cronin, and K. R. Parker. Aquatic Sciences 68:1827–1835. 2000. Hydraulic sampling protocol to estimate natural embryo mortality of Kuligowski, D. R., M. J. Ford, and B. A. Berejikian. 2005. Breeding struc- Pink Salmon. Transactions of the American Fisheries Society 129:827–834. ture of Steelhead inferred from patterns of genetic relatedness among nests. Cowen, L., N. Trouton, and R. E. Bailey. 2007. Effects of angling on Chinook Transactions of the American Fisheries Society 134:1202–1212. Salmon for the Nicola River, British Columbia, 1996–2002. North American Maynard, D. J., W. C. McAuley, and T. A. Flagg. 2003. Snake River Journal of Fisheries Management 27:256–267. spring/summer Chinook captive broodstock rearing and research, 2002: an- Doctor, K. K., R. Hilborn, M. Rowse, and T. Quinn. 2010. Spatial and temporal nual report. Report to the Bonneville Power Administration, Project 96-067- patterns of upriver migration by Sockeye Salmon populations in the Wood 00, Portland, Oregon. CAPTIVE-REARED CHINOOK SALMON REPRODUCTIVE SUCCESS 107

McNeil, W. J. 1964. A method of measuring mortality of Pink Salmon eggs and Snider, B. R., J. Heindel, M. Hughes, J. Seggerman, and D. Munson. 2003. larvae. U.S. Fish and Wildlife Service Fishery Bulletin 63:575–588. Sawtooth fish hatchery and East Fork satellite 2001 spring Chinook brood Neff, B. D., S. R. Garner, and T. E. Pitcher. 2011. Conservation and enhancement year report, 2002 Steelhead brood year report. Idaho Department of Fish and of wild fish populations: preserving genetic quality versus genetic diversity. Game, Report 03-45, Boise. Canadian Journal of Fisheries and Aquatic Sciences 68:1139–1154. Swanson, P., B. Campbell, K. Shearer, J. Dickey, B. Beckman, D. Larsen, L. Neter, J., W. Wasserman, and G. A. Whitmore. 1988. Applied statistics, 3rd Park, and B. Berejikian. 2008. Application of reproductive technologies to edition. Allyn and Bacon, Newton, Massachusetts. captive breeding programs for conservation of imperiled stocks of Pacific Piper, R. G., I. B. McElwain, L. E. Orme, J. P. McCraren, L. G. Fowler, and salmon. Cybium 32:279–282. J. R. Leonard. 1989. Fish hatchery management, 4th edition. U.S. Fish and Tautz, A. F., and C. Groot. 1975. Spawning behavior of Chum Salmon (On- Wildlife Service, Washington, D.C. corhynchus keta) and Rainbow Trout (Salmo gairdneri). Journal of the Fish- Pollard, H. A., II, and T. A. Flagg. 2004. Guidelines for use of captive brood- eries Research Board of Canada 32:633–642. stocks in recovery efforts for Pacific salmon. Pages 333–345 in M. J. Nickum, Torgersen, C. E., D. M. Price, H. W. Li, and B. A. McIntosh. P. M. Mazik, J. G. Nickum, and D. D. MacKinlay, editors. Propagated fish in 1999. Multiscale thermal refugia and stream habitat associations of resource management. American Fisheries Society, Symposium 44, Bethesda, Chinook Salmon in northeastern Oregon. Ecological Applications 9: Maryland. 301–319. Quinn, T. P., and C. J. Foote. 1994. The effects of body size and sexual di- Velsen, F. P. J. 1980. Embryonic development in eggs of Sockeye Salmon, morphism on the reproductive behaviour of Sockeye Salmon, Oncorhynchus Oncorhynchus nerka. Canadian Special Publication of Fisheries and Aquatic nerka. Animal Behaviour 48:751–761. Sciences 49. Sayre, J. A. 1995. Terminal fisheries and captive broodstock enhancement by Venditti, D. A., J. Lockhart, A. Kohler, A. Brimmer, K. A. Apperson, B. nonprofit groups. Pages 303–306 in H. L. Schramm Jr. and R. G. Piper, Bowersox, and C. Bretz. 2007. Idaho supplementation studies: brood year editors. Uses and effects of cultured fishes in aquatic ecosystems. American 2004 cooperative report, August 1, 2004 to July 31, 2006. Report to the Fisheries Society, Symposium 15, Bethesda, Maryland. Bonneville Power Administration, Report 07-24, Portland, Oregon. Schiewe, M. H., T. A. Flagg, and B. A. Berejikian. 1997. The use of captive Venditti, D. A., C. Willard, T. Giambra, D. Baker, and P. Kline. 2003a. Captive broodstocks for gene conservation of salmon in the western United States. rearing program for Salmon River Chinook Salmon: annual progress report, Bulletin of the National Research Institute of Aquaculture Supplement 3: January 1, 2001 to December 31, 2001. Report to the Bonneville Power 29–34. Administration, Report 03-33, Portland, Oregon. Schroder, S. L., C. M. Knudsen, T. N. Pearsons, T. W. Kassler, S. F. Young, C. A. Venditti, D. A., C. Willard, C. James, P. Kline, and D. Baker. 2003b. Captive Busack, and D. E. Fast. 2008. Breeding success of wild and first-generation rearing program for Salmon River Chinook Salmon: annual progress report, hatchery female spring Chinook Salmon spawning in an artificial stream. January 1, 2002 to December 31, 2002. Report to the Bonneville Power Transactions of the American Fisheries Society 137:1475–1489. Administration, Report 03-57, Portland, Oregon. Snider, B. R., R. Elmore, M. Hughes, H. Lehman, and D. Munson. 2004. Venditti, D. A., C. Willard, C. Looney, P. Kline, and P. Hassemer. 2002. Captive Sawtooth fish hatchery and East Fork satellite 2002 spring Chinook brood rearing program for Salmon River Chinook Salmon: project progress report, year report, 2003 Steelhead brood year report. Idaho Department of Fish and January 1, 2000 to December 31, 2000. Report to the Bonneville Power Game, Report 04-34, Boise. Administration, Report 02-22, Portland, Oregon. Downloaded by [Department Of Fisheries] at 19:31 20 January 2013 This article was downloaded by: [Department Of Fisheries] On: 28 February 2013, At: 22:28 Publisher: Taylor & Francis Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

North American Journal of Fisheries Management Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/ujfm20 Otolith Microstructure during the Early Life-History Stages of Brown Trout: Validation and Interpretation Julian J. Dodson a , Pascal Sirois b , Gaétan Daigle c , Philippe Gaudin d & Agnès Bardonnet e a Département de Biologie, Université Laval, 1045 Avenue de la Médecine, Quebec, Quebec, G1V 0A6, Canada b Laboratoire des Sciences Aquatiques, Département des Sciences Fondamentales, Université du Québec à Chicoutimi, 555 Boulevard de l’Université, Chicoutimi, Quebec, G7H 2B1, Canada c Département de Mathématique et de Statistique, Université Laval, 1045 Avenue de la Médecine, Quebec, Quebec, G1V 0A6, Canada d Unité Mixte de Recherche Ecobiop, Université de Pau et des Pays de l'Adour, Unité de Formation et de Recherche des Sciences de la Côte Basque, Anglet, F-64600, France e Institut National de Recherche Agronomique, Unité Mixte de Recherche Ecobiop, Pôle Hydrobiologie Institut National de Recherche Agronomique, St Pée-sur-Nivelle, F-64310, France Version of record first published: 26 Jan 2013.

To cite this article: Julian J. Dodson , Pascal Sirois , Gaétan Daigle , Philippe Gaudin & Agnès Bardonnet (2013): Otolith Microstructure during the Early Life-History Stages of Brown Trout: Validation and Interpretation, North American Journal of Fisheries Management, 33:1, 108-116 To link to this article: http://dx.doi.org/10.1080/02755947.2012.746245

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material. North American Journal of Fisheries Management 33:108–116, 2013 C American Fisheries Society 2013 ISSN: 0275-5947 print / 1548-8675 online DOI: 10.1080/02755947.2012.746245

MANAGEMENT BRIEF

Otolith Microstructure during the Early Life-History Stages of Brown Trout: Validation and Interpretation

Julian J. Dodson Departement´ de Biologie, Universite´ Laval, 1045 Avenue de la Medecine,´ Quebec, Quebec G1V 0A6, Canada Pascal Sirois* Laboratoire des Sciences Aquatiques, Departement´ des Sciences Fondamentales, UniversiteduQu´ ebec´ a` Chicoutimi, 555 Boulevard de l’Universite,´ Chicoutimi, Quebec G7H 2B1, Canada Gaetan´ Daigle Departement´ de Mathematique´ et de Statistique, Universite´ Laval, 1045 Avenue de la Medecine,´ Quebec, Quebec G1V 0A6, Canada Philippe Gaudin Unite´ Mixte de Recherche Ecobiop, Universite´ de Pau et des Pays de l’Adour, Unite´ de Formation et de Recherche des Sciences de la Coteˆ Basque, Anglet F-64600, France Agnes` Bardonnet Institut National de Recherche Agronomique, Unite´ Mixte de Recherche Ecobiop, Poleˆ Hydrobiologie Institut National de Recherche Agronomique, St Pee-sur-Nivelle´ F-64310, France

tions and to document the hatching, dispersal, growth, and survival Abstract of the early life history stages of Brown Trout in nature. We examined the extent to which otolith microstructure provides an accurate estimate of age, growth, and early life history The analysis of age and growth information contained in the transitions during the period between hatching and 1 week after otoliths of ﬁshes has become a highly useful tool for ﬁsheries emergence in Brown Trout Salmo trutta exposed to natural variations in ambient water temperature. All fry analyzed possessed a scientists and managers that provides information about age,

Downloaded by [Department Of Fisheries] at 22:28 28 February 2013 prominent check on the observed date of hatching. After hatching, growth rate, life history, and recruitment in exploited fish stocks daily growth increments were visible on sagittal otoliths. There (Stevenson and Campana 1992; Panfili et al. 2002). When ap- was no evidence for the formation of an emergence check mark plied to the otoliths of the early life history stages of fishes, one and no statistically significant evidence that emergence and daily can estimate the age of individuals as growth increments are temperature fluctuations interacted to form check marks. How- ever, daily temperature fluctuations may influence the formation of typically deposited on a daily basis (Pannella 1971). In many check marks, largely based on an observed increase in the propor- species, the timing of major life history transitions such as hatch- tion of fish possessing checks on the days following the two largest ing, first feeding, or metamorphosis can be determined, because temperature fluctuations observed during the experiment. There such transitions are often accompanied by the deposition of was no evidence that feeding or stressing emergent fish contributed distinctive increments, or check marks, within otoliths. Fur- to the formation of an emergence check mark. The observed proportionality of somatic and otolith growth in conjunction with daily thermore, individual growth trajectories may be reconstructed growth increments and the formation of a prominent hatch mark from the relative spacing between growth increments as there provides the opportunity to back-calculate somatic length distribu- is generally a proportional relationship between the growth

*Corresponding author: pascal [email protected] Received May 3, 2012; accepted October 29, 2012

108 MANAGEMENT BRIEF 109

of the otolith and the somatic growth of the fish (Campana somatic length distributions at otolith check marks associated 1990). with known life history transitions or environmental effects. The interpretation of otolith microstructure relies on two Although the Brown Trout is chiefly a European species, it assumptions. Firstly, deposition rates of increments must be has been widely spread by humans. Self-sustaining populations of regular and known periodicity, and secondly, the back- have been established in 26 countries beyond the species’ na- calculation of growth rates assumes that there is a proportional tive range. In North America, introductions began in the late relationship between growth of the otolith and somatic growth 19th century, and naturalized populations now occur in 34 of of the fish (Campana and Jones 1992). Previous studies on the 50 states of the USA and in 9 of the 10 provinces of Canada salmonids have demonstrated that otolith microstructure pro- (Jonsson and Jonsson 2011). Because of its role in reducing vides a reliable record of age between hatching and shortly after native fish populations (particularly those of other salmonids), emergence in Arctic Char Salvelinus alpinus and Brown Trout the Brown Trout is considered as one of the “100 worst invasive Salmo trutta (Mosegaard and Titus 1987), Atlantic Salmon S. alien species” by the Invasive Species Specialist Group (Lowe salar (Meekan et al. 1998), and Chinook Salmon Oncorhynchus et al. 2000). In North America, the ecology of the early life his- tshawytscha (Neilson and Geen 1982). On the other hand, sev- tory stages of Brown Trout is poorly understood, and such basic eral authors have attempted to demonstrate that the relationship information is necessary for the management of this important between otolith and fish size at the time of emergence is weak or alien species. More specifically, the ability to back-calculate absent (Arctic Char, Mosegaard et al. 1988; Brown Trout, Titus size at age, based on validated otolith microstructure, during and Mosegaard 1991; Atlantic Salmon, Wright et al. 1990; Met- the weeks and months after emergence provides estimates of calfe et al. 1992). The lack of correlation between otolith and growth rate and a measure of competitive success in the pres- fish size at emergence is quite different from that observed only ence of sympatric native salmonids. Back-calculated growth tra- a few weeks later, when there is a significant concordance be- jectories also provide estimates of size-selective mortality and tween otolith and fish size (Titus and Mosegaard 1991; Wright the development of alternative life history tactics (Aubin-Horth et al. 1991; Metcalfe et al. 1992). et al. 2005). Size at age early in life is an important determinant To account for this apparent contradiction, some workers of whether salmonids remain resident in freshwater streams or have hypothesized that variation in otolith size reflects differ- adopt the migratory life style (Aubin-Horth et al. 2005), a life ences in metabolic rate among individuals, so that fish with history decision of critical importance in predicting the long- high metabolic rates tend to have larger otoliths, but are not distance dispersal capabilities of invasive salmonids (Thibault necessarily larger in somatic size (Mosegaard 1990; Titus and et al. 2010). Mosegaard 1991; Metcalfe et al. 1992). As individuals with To promote the utility of exploiting otolith microstructure higher metabolic rates grow faster, otolith size and fish size be- in understanding the ecology of the early life history stages of come correlated once relatively large size differences develop Brown Trout, we examined the extent to which otoliths pro- after a few weeks of life in the stream habitat. The hypothe- vide an accurate estimate of age, growth, and early life history sized uncoupling of otolith and fish growth early in life pro- transitions during the period between hatching and 1 week af- posed by these studies has done little to encourage the use of ter emergence in fish exposed to natural variations in ambient otolith microstructure in reconstructing early life history events temperature. A previous attempt at validating the microstruc- in salmonids. ture of the otoliths of Brown Trout alevins under controlled Alternatively, the weak relationship between otolith and fish temperature conditions reported daily growth increments, but size at emergence recorded in previous studies may simply be an provided no supporting data, and concluded that otolith growth artifact of measurement error and the truncation of size ranges rate was not coupled to somatic growth rate, when expressed Downloaded by [Department Of Fisheries] at 22:28 28 February 2013 in regression analyses associated with taking brief “snapshots” as wet mass (Mosegaard and Titus 1987). In the present exper- at some specific point in time along the developmental trajec- iment, we aimed to validate (1) the formation of a hatch mark tory (Meekan et al. 1998). These latter authors examined the by associating known dates of hatching with the formation of development of the relationship between otolith size and body distinctive increments (check marks), (2) daily deposition by size in Atlantic Salmon between hatching and emergence by re- comparing counts of increments within otoliths with the known peatedly measuring individual fish. Weak but significant linear age of individuals, (3) the formation of an emergence mark regressions were found between the back-calculated radius of by associating known dates of emergence with the formation the sagittal otoliths and standard length (SL) of alevins measured of check marks, (4) that exogenous feeding and experimen- on five separate occasions. However, when pooled among sam- tal manipulation contribute to the formation of the emergence pling times, these variables were highly correlated, with varia- check mark, (5) that fluctuations in ambient water temperature tion in otolith size explaining almost all of the variation in SL of influence the formation of check marks, because strong daily alevins (Meekan et al. 1998). Thus, the examination of otolith temperature fluctuations are known to generate check marks on microstructure provides the means to reconstruct the growth otoliths (Volk et al. 1999), and (6) the proportionality of somatic history of individual fish since hatching and to back-calculate and otolith growth in Brown Trout. 110 DODSON ET AL.

METHODS sacrificed on March 11. These fish thus provided a temporal sequence of emergence over a period of 8 d. Laboratory Protocol To validate that exogenous feeding and experimental manip- Experimental group 1.—To identify the hatch mark (objec- ulation contribute to the formation of the emergence check mark tive 1), validate daily increment deposition between hatching (objective 4), two experiments were conducted. On February 29, and emergence (objective 2), and validate the proportionality 60 emergent fry from experimental group 2 were divided into of somatic and otolith growth between hatching and emergence two groups of 30 and held in basins located within the stream (objective 6), eyed embryos were obtained from the controlled channels. One group was fed ad libitum for a period of 7 d crosses of two female and five male Brown Trout obtained from with live Daphnia sp. The second group was not fed over the Bertiz River, a tributary of the Bidassoa River in northwestern ◦ ◦ 7-d period. We predicted that the stress of the transition from Spain (1 37 W, 43 10 N); date of fertilization was December 13, endogenous to exogenous feeding would accentuate the emer- 2007. Several hours after fertilization, eggs were transported to gence check mark. On March 1, 100 emergent fry were divided the Lapitxuri field station (Station d’hydrobiologie INRA, Saint- into two groups of 50 and held in separate basins for 7 d. Group Pee-sur-Nivelle)´ located on a tributary of the Nivelle River in ◦ ◦ 1 was subjected to stress by subjecting them to a 15-min period southwestern France (1 29 W, 43 17 N). The eggs were incu- of anesthetization (with phenoxyethanol), individual manipu- bated together in a vertical flow incubator until the eyed stage lation with forceps, and transfer to a petri dish for recovery and then in stream channels. Both structures were supplied with before being moved to the holding basin. The second group water diverted from the Lapitxuri Brook. Temperature thus mir- was transferred directly to the holding basin with no additional rored the natural temperature regime. Photoperiod was adjusted manipulation. We predicted that the stress of handling would to follow the natural photoperiod (10 h light :14 h dark until accentuate the emergence check mark. mid-February, and then 11 h light : 13 h dark). On January 18, Before being sacrificed, all fish were anesthetized with phe- 2008, we placed 25 eyed embryos in each of four small incuba- noxyethanol in a small petri dish, photographed with a Canon tors (10 cm height, 6.3 cm in diameter; filled with gravel) that S70 camera connected to a Leica M76 binocular microscope were buried in the gravel of two parallel stream channels (50 cm along with a millimeter scale, and then placed in 95% ethanol wide, 3 m long). Fish could not emerge from these incubators. in microcentrifuge tubes for subsequent otolith extraction. Fish One incubator was removed every 7 d posthatch (February 5, were measured to the nearest 0.01 mm using the Image J (ver- 12, 19, and 26, 2008), and alevins were sacrificed to identify sion 1.42) software prior to fixation. Preserved alevins were the hatch mark and to validate daily growth increments and the measured once again before otolith extraction. proportionality of somatic and otolith growth between hatching and emergence. A sample of 25 eyed embryos was held in Otolith Microstructure Analysis a control incubator located in a third channel, free of gravel, Sagittal otoliths were removed from the pre- and poste- and monitored daily to record the date of hatching, and hence, mergent fry under a dissecting microscope using fine forceps. the beginning of the sampling schedule of the four additional Otoliths were mounted on a microscope slide with thermoplastic incubators. These embryos all hatched on January 30, 2008, glue and those of larger individuals (postemergent) were pol- following 12 d of incubation in the stream tank. ished with a 3- or 5-µm lapping film. Otoliths were measured Experimental group 2.—To validate daily increment deposi- using an image-analysis system (SigmaScanPro 5.0) connected tion after emergence (objective 2), validate the formation of an to a light microscope at 400 × to 1,000 × magnification. Be- emergence mark (objective 3), validate that fluctuations in am- cause of the presence of several accessory primordia, measure- bient water temperature influence the formation of check marks ments were taken along a postrostral axis and included the hatch

Downloaded by [Department Of Fisheries] at 22:28 28 February 2013 (objective 5), and extend the relationship between otolith and mark radius (µm), otolith postrostral radius (µm), and increment fish size to include postemergent fry (objective 6), a second counts (Figure 1). All otoliths were analyzed twice by the same experimental protocol was established. On January 18, 2008, reader at an interval of >1 month, and each count estimate was 345 Brown Trout eyed embryos were placed in two incubators ranked according to the confidence of the reading. If the CV (20 cm in height, 8 cm in diameter) buried in gravel in two (SD/mean × 100) in the number of increments counted be- tanks supplied continuously with the same water as the stream tween the first and the second reading did not exceed 10%, the channels. Incubators were equipped with emergence traps such single best increment estimate was used as the age estimation that newly emerged fry could leave the gravel but were retained (Campana and Jones 1992). for daily sampling. The emergence traps were monitored daily until no emergence was observed for seven consecutive days. Statistical Analysis Emergence of fry started on February 28, 2008. The last fish The formation of daily increments was validated by regress- emerged on March 5. All fry sampled at emergence were held ing the number of days after the date of hatching (calendar age) for a period of 7 d in incubators without food before being with the number of increments distal to the hatch mark (esti- sacrificed to ensure that any mark laid down at emergence was mated age) and then comparing the intercept and the slope of succeeded by seven daily growth increments. The last fry was the relationship to a value of 0 and 1, respectively, using the MANAGEMENT BRIEF 111 Downloaded by [Department Of Fisheries] at 22:28 28 February 2013

FIGURE 1. The sagittal otoliths of a Brown Trout fry aged (A) 7 d posthatch (400 × magniﬁcation) and (B) 37 d posthatch (200 × magniﬁcation). The hatch mark (January 30, 2008) is indicated on both otoliths. The horizontal line on both otoliths represents the postrostral radius. [Figure available in color online.] 112 DODSON ET AL.

usual t-test statistics on regression parameters. Each of these RESULTS two statistical tests followed a Student distribution with n − 2 Brown trout eyed embryos were transferred to the stream degrees of freedom, where n is equal to the sample size. channels on January 18, 2008, at a temperature of 10.5◦C From the same regression model, confidence intervals (CIs) (Figure 2). Temperatures subsequently fluctuated around a mean for the mean value and prediction intervals for a new observation of 9.4◦C(SD= 1.2), with several daily fluctuations surpassing were calculated. This was done first to estimate the calendar age 2◦C. from a specific number of growth increments on the otolith. In this case, the confidence and prediction intervals are related to Validation of Hatch Marks and Check Formation variation on the y-axis. This was also done to estimate the age All embryos held in the control incubator hatched on January read on otoliths from a specific calendar age, and in that case, 30, 2008. A total of 249 otoliths were examined but 57 (23%) the confidence and prediction intervals are related to variation were rejected due to their unreadability or to the 10% crite- on the x-axis. The latter estimates were calculated using the rion. The remaining 192 fry all possessed a prominent check inverse regression technique (Neter et al. 1985). on January 30 that we assumed to correspond to the hatch mark Embryos in the control incubator hatched on January 30 and (Figures 1, 3). The hatch mark occurred at a postrostral radius of all analyzed fry had a prominent otolith check mark on this between 60 and 75 µm in 80% of the fish examined (Figure 4). date. Thus, we assumed that the hatch mark of all analyzed fish A variable proportion of fry possessed check marks throughout corresponded to January 30 ± the calculated margin of error the duration of the experiment, and a large proportion possessed (see below). As numerous other check marks were observed, we check marks shortly after the transfer of eyed embryos to the assessed the contribution of emergence and daily temperature stream tanks 12 d before hatching (Figure 3). fluctuations to the formation of these check marks by calculating the proportion of fish possessing a check mark between Validation of Daily Increment Deposition the day after hatching and the end of the experiment 42 d later After hatching, daily growth increments were detected on (January 31 to March 12). To do so, we employed a logistic sagittal otoliths (Figure 5). The relationship between the number regression model. The response was a binary variable corre- of growth increments and known age was significant (r2 = 0.99). sponding to the presence (1) or absence (0) of a check mark on The intercept was not significantly different from 0 (t =−0.47, each fish and for each of the 42 d of the experiment. However, df = 190, P = 0.638). The slope, however, was significantly because check formation on any given day was not independent different from a slope of 1 (t = 2.57, df = 190, P = 0.011). Age of check formation on previous days, the generalized estimating based on the otolith growth increments slightly underestimated equation (GEE) estimation approach was used with an autore- the true age after fry were 6 d old. However, the 95% CI for the gressive correlation structure. Two independent variables were estimation of the mean of calendar age varied from ± 0.164 to considered in this model: the date of emergence was entered as ± 0.357 d and in all cases was less than 1 d for the duration of a binary variable (date of emergence scored as 1, all other dates the experiment. scored as 0) and daily temperature fluctuations were entered as a continuous variable (maximum daily temperature minus min- Emergence, Temperature Fluctuations, and Check imum daily temperature). As the estimated age was not without Formation error (see Results), we used another GEE logistic regression Emergence occurred between February 28 and March 5, model with a new binary response variable corresponding to the but there was no evidence for the formation of an emergence presence (1) or absence (0) of at least one check mark on each check mark (Table 1). Furthermore, there was little evidence fish for each day of the experiment ± 2 d. The first and the last that daily temperature fluctuations influenced the formation of Downloaded by [Department Of Fisheries] at 22:28 28 February 2013 2 d of the experiment were removed for this analysis to take into check marks, or that emergence and daily temperature fluctua- account the 2-d margin of error. Thus, the time series for this tions interacted to form check marks (Table 2). Considering the model was 38 d long, beginning on February 2 and ending on margin of error of ± 2 d in the estimation of a new observation March 10. for calendar age (Table 1), our analyses revealed that 37.6% We also used χ2 tests to test the hypothesis that a greater (SE = 3.2%) of fry possessed check marks during emergence proportion of fed (or stressed) fish possessed a check mark on compared with 42.9% (SE = 1.2%) of fish with check marks the date of emergence (or the date of emergence ± 2 d) than formed on all other days between hatching and the end of the the proportion of unfed (or unstressed) fish possessing a check experiment, but excluding the dates of emergence ( ± 2 d). If we mark on the date of emergence (or the date of emergence ± exclude the margin of error, 7.6% (SE = 2.8%) of fry possessed 2 d). To ensure that fish provided with food in fact did feed, we check marks formed during emergence compared with 12.8% compared the wet mass of unfed fish and fed fish after 7 d of (SE = 0.4%) of fish with check marks formed on days excluding exposure to food using a t-test with unequal variances. Finally, the dates of emergence (Table 1). we tested the proportionality of somatic and otolith growth by When including the margin of error, daily temperature fluc- regressing log SL on log postrostral radius of the otolith. All tuations appeared to have influenced the formation of check of the foregoing statistical analyses were conducted with SAS marks, although the P-value was marginally nonsignificant (P = (SAS version 9.2; SAS Institute, Cary, North Carolina). 0.087; Table 2). This relationship appears to be largely related MANAGEMENT BRIEF 113

FIGURE 2. Mean (middle solid line), maximum (upper dotted line), and minimum (lower dotted line) daily water temperatures measured during the experiment. The date of hatching (January 30) and the dates of emergence (February 28–March 5) are indicated. 1 = the largest daily temperature ﬂuctuation observed ◦ ◦ (maximum − minimum = 2.8 C) recorded on March 4, 2008. 2 = the second largest daily temperature ﬂuctuation observed (maximum − minimum = 2.2 C)

Downloaded by [Department Of Fisheries] at 22:28 28 February 2013 recorded on February 6, 2008.

to an increase in the proportion of fish possessing checks on the difference was observed in the proportion of fed and unfed fry days following the two largest temperature fluctuations observed possessing check marks on the date of emergence (excluding 2 during the experiment (1 and 2, Figure 2; Figure 3). How- the margin of error: χ = 1.00, df = 1, P = 0.316; including the ever, the paucity of large daily temperature fluctuations during margin of error: χ2 = 0.048, df = 1, P = 0.826). Similarly, no the course of the experiment weakens the statistical power of significant difference was observed in the proportion of stressed the regression analysis to adequately estimate the significance and unstressed fry possessing check marks on the date of emer- of this relationship. gence (excluding the margin of error: χ2 = 0.0, df = 1, P = 1.0; There was no evidence that feeding or stressing emergent including the margin of error: χ2 = 0.186, df = 1, P = 0.666). fish contributed to the formation of an emergence check mark. Although a significant increase in wet mass occurred among fry Proportionality of Somatic and Otolith Growth fed for 7 d (unfed fry [n = 30] mean wet mass = 0.119 g [SE = There was a significant relationship between postrostral 0.001 g]; fed fry [n = 30] mean wet mass = 0.139 g [SE = length and SL of the fish during the course of the experiment 0.003 g]; t =−6.19, df = 41.45, P < 0.0001), no significant (Figure 6). Variation in natural log otolith size explained 89% 114 DODSON ET AL.

and to back-calculate somatic length distributions. Hatching was marked by a strong check mark on the otoliths of all fish and growth increments were formed on a daily basis thereafter, permitting the age determination of Brown Trout fry and the determination of hatching date of fry captured in nature. There was also evidence that daily growth increments were formed prior to hatching, with prominent check marks forming in the days after the transfer of eyed embryos to the stream tanks (Figure 3). The proportionality of somatic and otolith growth provides the opportunity to back-calculate somatic length distributions and reconstruct individual growth trajectories early in the life of Brown Trout using such back-calculation models as the biological intercept method (Campana 1990) or the time-varying growth method (Sirois et al. 1998). Surprisingly, we failed to validate the formation of an emergence check mark. The proportion of fish with check marks formed on the day of emergence was no different from the proportion of fish exhibiting check marks outside the period of emergence. The manipulations designed to accentuate the emergence mark (feeding and stressing fish at emergence) failed to FIGURE 3. The proportion of Brown Trout fry possessing daily otolith check generate consistent check marks on the otolith. However, a pre- marks throughout the duration of the study. Hatching occurred on January 30, cipitous drop in temperature that occurred at the end of the 2008. Open bars identify the prehatching period (embryos were transferred to period of emergence (starting on March 4, Figure 2) coincided the laboratory on January 18, 2008). Filled bars identify the posthatching period. with an increase in the percentage of fish with emergent check marks to between 50% and 60% (Table 1). Therefore, a strong of the variation in the natural log SL of Brown Trout between 1 temperature fluctuation may have coincided with emergence on and 42 d of age. these dates to create the appearance of an emergence check mark. A similar increase in the proportion of fish with a check DISCUSSION The otoliths of pre- and postemergent Brown Trout provide the means to reconstruct the growth history of individual fish Downloaded by [Department Of Fisheries] at 22:28 28 February 2013

FIGURE 5. Relationship between the number of days since hatching (calendar age and date) and the number of otolith growth increments (age read on otolith) for Brown Trout fry (n = 192). Solid lines define the 95% confidence interval for the estimation of the mean of calendar age and the dashed lines define the FIGURE 4. The distribution of hatch marks along the postrostral radius of 95% prediction interval. Diameter of data points is proportional to the number Brown Trout otoliths. Numbers above bars represent actual percentage values. of observations. MANAGEMENT BRIEF 115

TABLE 1. The mean proportion of Brown Trout fry possessing check marks TABLE 2. Results of the GEE logistic regression model to relate check for- on the date of their emergence and on the date of their emergence ± 2d. mation with emergence date of Brown Trout and daily temperature fluctuations. The mean proportion for all emergence dates combined is a weighted mean, Two regression models are presented: one with the margin of error of the pre- accounting for variation in sample sizes among dates of emergence. diction of a new observation of calendar age included (check marks on day i ± 2 d), and one without (check marks on day i). Z = results of Z-test, max– Emergence date min = difference between the maximum and minimum recorded temperature, (day/month/year) N Mean proportion SD SE emerge = date of emergence, max–min × emerge = interaction between daily temperature fluctuation and emergence. On date of emergence 28/02/08 6 0.167 0.373 0.152 Parameter Estimate SE ZP a 29/02/08 39 0.103 0.303 0.049 Margin of error not included b 01/03/08 30 0.067 0.249 0.046 Intercept −1.887 0.178 −10.63 <0.0001 02/03/08 12 0.000 0 0 Max–min −0.022 0.115 −0.19 0.846 03/03/08 16 0.063 0.242 0.061 Emerge 0.975 0.858 1.14 0.256 04/03/08 6 0.000 0 0 Max–min × emerge −1.052 0.707 −1.49 0.137 05/03/08 5 0.600 0.490 0.219 All emergence dates 114 0.096 0.295 0.028 Margin of error included − − < On date of emergence ± 2d Intercept 0.384 0.071 5.41 0.0001 28/02/08 6 0.333 0.471 0.152 Max–min 0.065 0.038 1.71 0.087 Emerge 0.282 0.342 0.83 0.409 29/02/08a 39 0.359 0.480 0.049 Max–min × emerge −0.340 0.263 −1.29 0.197 01/03/08b 30 0.233 0.423 0.046 02/03/08 12 0.333 0.471 0 03/03/08 16 0.500 0.5 0.061 04/03/08 6 0.500 0.5 0 studies using tetracycline antibiotics (e.g., Meerbeek and Bet- 05/03/08 5 0.600 0.490 0.219 toli 2005) and Alizarin fluorochrome compounds (e.g., Baer and All emergence dates 114 0.360 0.480 0.045 Rosch¨ 2008) for marking large numbers of young fish, but the aIncludes fed and unfed fry. use of such compounds are either restricted or banned in many bIncludes stressed and unstressed fry. European and North American jurisdictions. Thermal marking is free of such concerns and, in combination with microstructure mark occurred a day after the second most important daily tem- analysis, has the potential to be a useful tool for studying the perature fluctuation, leading to the marginally nonsignificant hatching, dispersal, growth, and survival of the early life history relationship revealed by the regression analysis (Table 2). stages of Brown Trout. The coincidence of emergence and strong daily temperature fluctuations may explain the apparent contradiction between the present results and those reported by Titus and Mosegaard (1991). Those authors reported the formation of an emergence check mark during an experiment with Brown Trout fry. During their experiment, however, fish were held at 4◦Cas preemergence-stage fry and then acclimated to a temperature of 14◦C at the time of emergence. A temperature increase of Downloaded by [Department Of Fisheries] at 22:28 28 February 2013 10◦C at the time of emergence most probably contributed to the formation of the check mark. Such high temperature changes induce more distinguishable check marks than the smaller temperature changes (on the order of 2◦C) recorded here for Brown Trout and for other salmonids (Volk et al. 1999). The susceptibility of Brown Trout otoliths to form reliable thermal otolith check marks needs validation, but this is a promising avenue of pursuit as the use of thermal marking is a simple and relatively unobtrusive way of marking large numbers of fish. Patterns of light and dark bands akin to bar codes can be created on the otolith by exposing fish to relatively short periods of cold or warm water relative to ambient temperature (Volk et al. 1999). The technique has been used successfully in Atlantic Salmon (e.g., Letcher and Terrick 1998; Aubin-Horth FIGURE 6. The proportionality of somatic and otolith growth in Brown Trout and Dodson 2002), but not in Brown Trout to the best of our fry. A simple regression relates the natural log of postrostral otolith radius to knowledge. Brown Trout have been the subject of several recent the natural log of standard length: y = 0.316x + 1.495, n = 192, r2 = 0.89. 116 DODSON ET AL.

ACKNOWLEDGMENTS Mosegaard, H. 1990. What is reflected by otolith size at emergence: a reeval- This experiment was conducted during the senior author’s uation of the results. Canadian Journal of Fisheries and Aquatic Sciences sabbatical leave at the Station d’hydrobiologie, Pole Hydro- 47:225–228. Mosegaard, H., H. Svedang,¨ and K. Taberman. 1988. Uncoupling of somatic biologie INRA, UMR Ecobiop, St Pee-sur-Nivelle,´ France. and otolith growth rates in Arctic Char (Salvelinus alpinus) as an effect We thank Anne-Lise Fortin and Isabelle Poirier, Universitedu´ of differences in temperature response. Canadian Journal of Fisheries and Quebec´ a` Chicoutimi, for otolith microstructure analysis and Aquatic Sciences 45:1514–1524. Stephane´ Glise and François Gueraud,´ Station d’hydrobiologie, Mosegaard, H., and R. Titus. 1987. Daily growth rates of otoliths in yolk sac INRA, for the experimental design and egg production. fry of two salmonid species at five different temperatures. Pages 221–227 in S. O. Kullander and B. Farnholm, editors. Proceedings of the V congress of European ichthyologists. Swedish Museum of Natural History, Stockholm. REFERENCES Neilson, J. D., and G. H. Geen. 1982. Otoliths of Chinook Salmon (On- Aubin-Horth, N., and J. J. Dodson. 2002. Impact of differential energy allocation corhynchus tshawytscha): daily growth increments and factors influenc- in Atlantic Salmon (Salmo salar) precocious males on otolith-somatic size ing their production. Canadian Journal of Fisheries and Aquatic Sciences proportionality: a longitudinal approach. Canadian Journal of Fisheries and 39:1340–1347. Aquatic Sciences 59:1575–1583. Neter, J., W. Wasserman, and M. H. Kutner. 1985. Applied linear statistical Aubin-Horth, N., D. A. J. Ryan, S. P. Good, and J. J. Dodson. 2005. Balancing models, 2nd edition. Irwin, Homewood, Illinois. selection on size: effects on the incidence of an alternative reproductive tactic. Panfili, J., H. de Pontual, H. Troadec, and P. J. Wright, editors. 2002. Manuel Evolutionary Ecology Research 7:1171–1182. de sclerochronologie´ des poissons. Ifremer-IRD, Brest, France. Baer, J., and R. Rosch.¨ 2008. Mass-marking of Brown Trout (Salmo trutta L.) Pannella, G. 1971. Fish otoliths: daily growth layers and periodical patterns. larvae by alizarin: method and evaluation of stocking. Journal of Applied Science 173:1124–1127. Ichthyology 24:44–49. Sirois, P., F. Lecomte, and J. J. Dodson. 1998. An otolith-based back-calculation Campana, S. E. 1990. How reliable are growth back-calculations based on method to account for time-varying growth rate in Rainbow Smelt (Osmerus otoliths? Canadian Journal of Fisheries and Aquatic Sciences 47:2219–2227. mordax) larvae. Canadian Journal of Fisheries and Aquatic Sciences 55:2662– Campana, S. E., and C. M. Jones. 1992. Analysis of otolith microstructure 2671. data. Pages 73–100 in D. K. Stevenson and S. E. Campana, editors. Otolith Stevenson, D. K., and S. E. Campana, editors. 1992. Otolith microstructure microstructure examination and analysis. Canadian Special Publication of examination and analysis. Canadian Special Publication of Fisheries and Fisheries and Aquatic Sciences 117. Aquatic Sciences 117. Jonsson, B., and N. Jonsson. 2011. Ecology of Atlantic Salmon and Brown Thibault, I., R. D. Hedger, J. J. Dodson, J. C. Shiao, Y. Iizuka, and W. N. Tzeng. Trout: habitat as a template for life histories. Springer, New York. 2010. Anadromy and the dispersal of an invasive fish species (Oncorhynchus Letcher, B. H., and T. D. Terrick. 1998. Thermal marking of Atlantic Salmon mykiss) in eastern Quebec, as revealed by otolith microchemistry. Ecology otoliths. North American Journal of Fisheries Management 18:406–410. of Freshwater Fish 19:348–360. Lowe, S., M. Browne, S. Boudjelas, and M. De Poorter. 2000. 100 of the world’s Titus, R. G., and H. Mosegaard. 1991. Selection for growth potential among worst invasive alien species: a selection from the global invasive species migratory Brown Trout (Salmo trutta) fry competing for territories: evidence database. Invasive Species Specialist Group, Auckland, New Zealand. from otoliths. Canadian Journal of Fisheries and Aquatic Sciences 48:19– Meekan, M. G., J. J. Dodson, S. P. Good, and D. A. J. Ryan. 1998. Otolith and 27. fish size relationships, measurement error, and size-selective mortality during Volk, E. C., S. L. Schroder, and J. J. Grimm. 1999. Otolith thermal marking. the early life of Atlantic Salmon (Salmo salar). Canadian Journal of Fisheries Fisheries Research 43:205–219. and Aquatic Sciences 55:1663–1673. Wright, P. J., N. B. Metcalfe, and J. E. Thorpe. 1990. Otolith and somatic growth Meerbeek, J. R., and P. W. Bettoli. 2005. Mass-marking Rainbow Trout and rates in Atlantic Salmon parr, Salmo salar L.: evidence against coupling. Brown Trout fingerlings with oxytetracycline. Journal of Freshwater Ecology Journal of Fish Biology 36:241–249. 20:555–559. Wright, P. J., D. Rowe, and J. E. Thorpe. 1991. Daily growth increments Metcalfe, N. B., P. J. Wright, and J. E. Thorpe. 1992. Relationships between in the otoliths of Atlantic Salmon parr, Salmo salar L., and the influence social status, otolith size at first feeding and subsequent growth in Atlantic of environmental factors on their periodicity. Journal of Fish Biology 39: Salmon (Salmo salar). Journal of Animal Ecology 61:585–589. 103–113. Downloaded by [Department Of Fisheries] at 22:28 28 February 2013 This article was downloaded by: [Department Of Fisheries] On: 28 February 2013, At: 22:29 Publisher: Taylor & Francis Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

North American Journal of Fisheries Management Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/ujfm20 Eradication of Nonnative Brook Trout with Electrofishing and Antimycin-A and the Response of a Remnant Bull Trout Population Mark W. Buktenica a , David K. Hering a , Scott F. Girdner a , Brian D. Mahoney b & Bruce D. Rosenlund c d a U.S. National Park Service, Crater Lake National Park, Post Office Box 7, Crater Lake, Oregon, 97604, USA b Confederated Tribes of the Umatilla Indian Reservation, Water and Environmental Center, Walla Walla Community College, 500 Tausick Way, Walla Walla, Washington, 99362, USA c U.S. Fish and Wildlife Service, Colorado Fish and Wildlife Management Assistance Office, Post Office Box 25486, Denver Federal Center, Denver, Colorado, 80225, USA d Retired Version of record first published: 25 Jan 2013.

To cite this article: Mark W. Buktenica , David K. Hering , Scott F. Girdner , Brian D. Mahoney & Bruce D. Rosenlund (2013): Eradication of Nonnative Brook Trout with Electrofishing and Antimycin-A and the Response of a Remnant Bull Trout Population, North American Journal of Fisheries Management, 33:1, 117-129 To link to this article: http://dx.doi.org/10.1080/02755947.2012.747452

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material. North American Journal of Fisheries Management 33:117–129, 2013 C American Fisheries Society 2013 ISSN: 0275-5947 print / 1548-8675 online DOI: 10.1080/02755947.2012.747452

ARTICLE

Eradication of Nonnative Brook Trout with Electroﬁshing and Antimycin-A and the Response of a Remnant Bull Trout Population

Mark W. Buktenica,* David K. Hering, and Scott F. Girdner U.S. National Park Service, Crater Lake National Park, Post Office Box 7, Crater Lake, Oregon 97604, USA Brian D. Mahoney Confederated Tribes of the Umatilla Indian Reservation, Water and Environmental Center, Walla Walla Community College, 500 Tausick Way, Walla Walla, Washington 99362, USA Bruce D. Rosenlund1 U.S. Fish and Wildlife Service, Colorado Fish and Wildlife Management Assistance Office, Post Office Box 25486, Denver Federal Center, Denver, Colorado 80225, USA

Abstract A remnant population of native Bull Trout Salvelinus confluentus was threatened with extirpation by competition and hybridization with introduced Brook Trout S. fontinalis in Sun Creek, a second-order headwater stream in Crater Lake National Park, Oregon. Between 1992 and 2005, artificial barriers were installed to exclude nonnative fish, and multiple applications of electrofishing and the piscicide antimycin-A were used to remove Brook Trout from 14.6 km of stream. Several novel methods were employed, including diver-directed and trap-net electrofishing and the use of a portable raceway to hold Bull Trout during piscicide treatments. Electrofishing likely eradicated Brook Trout from a small headwater section of the stream but required more effort (54 person-days/km) than antimycin treatments (17 person-days/km) in the same reach to ensure eradication. For eradication of Brook Trout from larger downstream reaches, antimycin treatments applied in consecutive years were more successful than multiple treatments applied within a single year. Brook Trout have not been detected by annual surveys in the project area since 2005. The total effort expended to eradicate Brook Trout from 14.6 km of stream was approximately 138 person-days/km. Between 1989 and 2010, Bull Trout abundance increased approximately tenfold, and distribution increased from approximately 1.9 km to 11.2 km. These results exemplify the response of an imperiled Bull Trout population after removal of Brook Trout. The large investment of time and resources required to restore small populations like this Downloaded by [Department Of Fisheries] at 22:29 28 February 2013 one may be warranted only for critical population segments with special status or local management importance.

Introduced species pose threats to native stream ﬁshes world- are listed as threatened under the U.S. Endangered Species Act wide (Allan and Flecker 1993; Cucherousset and Olden 2011), (USFWS 1998a, 1999), have declined in abundance and distri- and introduction of nonnative salmonids has caused the de- bution throughout most of the species’ native range. Competi- cline of native trout and char in western North America through tion and hybridization with introduced Brook Trout Salvelinus competition, predation, and hybridization (Krueger and May fontinalis have been implicated in the decline or extirpation 1991; Rahel 2000). Bull Trout Salvelinus conﬂuentus, which of many Bull Trout populations (e.g., Nelson 1965; Markle

*Corresponding author: mark [email protected] 1Retired. Received November 10, 2011; accepted October 31, 2012 117 118 BUKTENICA ET AL.

1992; Ratliff and Howell 1992; Leary et al. 1993). Introduced were eradicated with a piscicide; however, we were concerned Brook Trout are widespread in the Upper Klamath River basin of that the small extant Bull Trout population could not with- southern Oregon, and 40% of historical Bull Trout populations stand incidental mortality from salvage actions. Accordingly, in the basin have experienced recent extirpation (Buchanan et al. we adopted a strategy with the following steps: (1) suppression 1997). of Brook Trout abundance with electrofishing and limited, tar- Several mechanisms are responsible for the displacement of geted applications of antimycin so that Bull Trout abundance Bull Trout by Brook Trout, including competition, hybridiza- might increase; (2) creation of Bull Trout refuge populations tion, and life history differences between the species. Bull Trout in nearby stream reaches; and (3) temporary removal of Bull and Brook Trout use similar habitat and exhibit similar feeding Trout from Sun Creek, followed by eradication of Brook Trout behavior (Dambacher et al. 1992; Gunckel et al. 2002), causing (Buktenica 1997). This strategy took many years to implement, direct competition for resources. Hybridization between the two and techniques evolved during the course of the project to min- species is also common (Leary et al. 1983). Although reduced imize harm to the remaining Bull Trout population while maxi- fertility of hybrids appears to prevent widespread introgression mizing Brook Trout removal efficiency. The long-term, adaptive (Leary et al. 1993), hybridization wastes the reproductive effort approach allowed for the interim evaluation of electrofishing as of Bull Trout and may favor Brook Trout disproportionally when an eradication tool in small- to moderate-sized stream reaches. spawning occurs predominately between Bull Trout females and In this paper, we describe the methods used and the effort ex- Brook Trout males (Allendorf et al. 2001; Kanda et al. 2002). pended to remove Brook Trout from Sun Creek during 1992– Plasticity of life history traits allowing earlier age of maturity 2005, and we provide evidence of Bull Trout recovery between and higher fecundity in Brook Trout also may provide a com- 1992 and 2010. petitive advantage that leads to the displacement of Bull Trout (Moyle 2002; Kennedy et al. 2003). PROJECT AREA An increasingly common management strategy for restoring native stream fishes threatened by introduced competi- Physical Description tors involves isolation of the native species and eradica- The project area included the uppermost 14.6 km of Sun tion of nonnatives (Kruse et al. 2001; Novinger and Rahel Creek, a second-order stream that drains a 72-km2 water- 2003). Restoration projects typically use electrofishing (e.g., shed in the Cascade Mountains of southern Oregon (42◦52N, Thompson and Rahel 1996; Kulp and Moore 2000; Shepard 122◦05W; elevation = 1,340–2,200 m; Figure 1). Sun Creek et al. 2002) or treatment with a piscicide, such as rotenone originates from headwater springs (discharge ∼ 0.05 m3/s), or antimycin-A (hereafter, antimycin; e.g., Rinne et al. 1981; and numerous additional groundwater inputs contribute sub- Hamilton et al. 2009), to remove nonnative fish that are present stantially to the stream through the project area. One major upstream of natural or artificial barriers to immigration. Many perennial tributary enters the creek 2.9 km downstream of the restoration projects have succeeded (e.g., Kulp and Moore 2000; headwaters. Sun Falls, a series of waterfalls located 3.4 km from Lintermans 2000), whereas many others have failed to eliminate the headwaters, prevents upstream movement of fish. Precipi- the target species (Meronek et al. 1996; Meyer et al. 2006). De- tation in the watershed is dominated by heavy winter snowfall, spite a strong need for effective removal techniques to restore and the stream channel is commonly covered by snow from native fish assemblages and the widespread use of piscicide and December through May. Total stream discharge at the down- electrofishing removal in fisheries management, many restora- stream end of the project area varies seasonally from a peak tion projects are not well documented or are described only in of 1.2–2.0 m3/s during early summer snowmelt to a base flow reports that are not readily available (Clancy et al. 1996). of about 0.25 m3/s during early autumn. Water temperatures in Downloaded by [Department Of Fisheries] at 22:29 28 February 2013 We present a 19-year case history of Brook Trout removal the project area range annually from 0◦Cto14◦C, and median and Bull Trout population response in Sun Creek, located in summertime temperatures are less than 8◦C. Water conductivity the Upper Klamath River basin within Crater Lake National ranges from 30 to 46 µS/cm, and pH ranges from 7.1 to 7.8. Park. The National Park Service initiated the project in the early The Sun Creek watershed was inundated with ash and pumice 1990s after a critically imperiled Bull Trout population threat- during the eruption of Mount Mazama roughly 7,700 years ago ened by introduced Brook Trout was detected in Sun Creek (Bacon et al. 2002). Within the project area, the stream’s flood- (Dambacher et al. 1992). Management goals were to eradicate plain lies within a narrow valley that is constrained by steep Brook Trout from the stream, prevent reinvasion of the stream hillslopes of eroded pumice deposits extending tens of meters by nonnative fish, and restore historical Bull Trout abundance deep. Riparian habitat includes subalpine meadow and conif- and distribution in the stream (Buktenica 1997). Methods fol- erous forest, and in-stream habitat consists largely of wood- lowed the recommendations of an advisory panel that was made formed plunge pools, riffles, and rapids. Streambed gradient up of fisheries managers and academic biologists from state and averages approximately 5%, and active channel width ranges federal agencies and universities (see Acknowledgments). After from 1 to 6 m. considering several management strategies, the advisory panel To facilitate restoration and monitoring activities, we divided concluded that Bull Trout could be temporarily removed from Sun Creek longitudinally into 50-m sections beginning at the the stream and held in refuge populations while Brook Trout headwaters (km 0) and ending at an irrigation diversion 22 km BROOK TROUT REMOVAL AND BULL TROUT RESPONSE 119

Management History The uppermost 16 km of Sun Creek have been largely protected from human disturbance since 1886—first by inclusion within a federal forest reserve and then by establishment of Crater Lake National Park in 1902. The portion of the watershed that lies within the national park is managed as a natural area. Except for two small roads crossing the uppermost reaches of the Sun Creek drainage and the introduction of Brook Trout, the upper watershed has remained virtually free from human disturbance, and most of the stream is seldom visited. Downstream of Crater Lake National Park, the stream flows unmodified through Sun Pass State Forest for 6 km, where the landscape is managed by the Oregon Department of Forestry for timber harvest, recreation, and wildlife habitat. At km 22, the entire stream enters an artificial distributary system that supplies irrigation water to private cattle ranches near Fort Klamath, Oregon. Maps prepared by Government Land Office surveyors during the 1870s indicate that Sun Creek was naturally a direct tributary of the Wood River, but irrigation diversions in the lower watershed have greatly reduced Sun Creek’s connectivity with the Wood River for over a century. Timber harvest, livestock grazing, and water withdrawal for irrigation have occurred in the lower portion of the watershed since the late 1800s. To prevent continued Brook Trout invasion and to facilitate Bull Trout recovery, the National Park Service constructed two in-stream fish exclusion barriers in Sun Creek during 1992 (Buktenica 1997). The barrier structures, located at km 11.6 and km 14.6 (Figure 1), are log weirs that create 2-m vertical FIGURE 1. Map of the Sun Creek, Oregon, project area. steps in streambed elevation to prevent upstream movement of trout (Kondratieff and Myrick 2006). Two barriers were built to provide redundant protection against Brook Trout invasion, as downstream (km 22). Sections were landmarked in the field. it was anticipated that one structure could fail over time. Since In this paper, we identify locations by distance (km) from the 1992, the barrier structures have been inspected annually to as- headwaters (Figure 1). sess their physical integrity, and we have made minor repairs to ensure their continued function. Historical Fish Distribution Bull Trout are the only native fish that have been confirmed to occur in Sun Creek, and historically they were distributed METHODS Downloaded by [Department Of Fisheries] at 22:29 28 February 2013 throughout the creek downstream of Sun Falls (Wallis 1948). Wallis (1948) also observed Rainbow Trout Oncorhynchus Brook Trout Eradication mykiss in Sun Creek, but it is unclear whether those Rainbow Electrofishing.—Between 1992 and 2005, Brook Trout were Trout were native or introduced, and no Rainbow Trout have removed by backpack electrofishing throughout Sun Creek up- been observed in the stream since the 1940s. Between 1926 and stream of the lower exclusion barrier (km 0 to km 14.6; Tables 1, 1971, Brook Trout were introduced throughout Sun Creek, both 2). Removal efforts were complicated by the sympatry of Brook within and upstream of the natural distribution of Bull Trout. Trout and Bull Trout, and we employed a variety of electrofish- A 1989 survey found that Brook Trout were present throughout ing techniques at different times and in different parts of the the stream and that Bull Trout distribution had been reduced to a stream to minimize injury or mortality of the few native Bull 1.9-km reach between km 7.5 and km 9.4. Minimum total abun- Trout that remained in the system. Except during 1992, when dance of age-1 and older Bull Trout in 1989 was estimated to AC electrofishing equipment was used, all electrofishing units be 133 fish based on uncalibrated snorkel counts in a stratified, delivered square-waved, pulsed DC with output voltages rang- systematic sample of habitat units (Hankin and Reeves 1988), ing between 700 and 1,000 V, pulse frequencies of 30–60 Hz, with no confidence interval reported. Hybrids of the two species and a duty cycle of 3–12%. Electrofishing methods included also occurred in the stream (Dambacher et al. 1992). (1) upstream single- and multiple-pass protocols, both with and 120 BUKTENICA ET AL.

TABLE 1. Effort expended and number of nonnative Brook Trout removed without block nets (conventional electrofishing); (2) a diver- from Sun Creek upstream of Sun Falls (i.e., between the headwaters [km 0] directed protocol; and (3) a trap-net protocol (Tables 1, 2). and km 3.4) by using conventional electrofishing and antimycin-A treatment, 1992–1998. Between 1992 and 1997, conventional electrofishing was used upstream of Sun Falls (from km 0 to km 3.4 and in the Electrofishing Antimycin-A perennial tributary; Table 1), where Brook Trout occurred in allopatry. During 1992–1994, 50-m sections of the stream were Stream Effort Brook Effort Brook isolated with block nets, and each section was electrofished by reach (person- Trout (person- Trout moving upstream (with one crew member operating the elec- Year (km) days) removed days) removed trofishing unit and two crew members acting as netters) until no 1992 0–3.40 32 176 fish were captured for two out of three consecutive passes (1992) 1.15–3.25 20 44 or until no fish were captured for three consecutive passes (1993 1.15–3.25 22 14 and 1994). Following these intensive multiple-pass protocols, 1993 0–3.40 36 21 additional removal efforts without block nets were conducted 1.10–3.35 6 0 through all or part of the reach in 1992 and 1993 (Table 1). 1994 0–3.40 21 1 During 1995–1997, at least two passes without block nets were 1995 0–3.40 15 0 conducted through the entire reach. 1996 0–3.40 16 3 Downstream of Sun Falls, Brook Trout were removed by 1997 0–3.40 8 1 conventional upstream electrofishing protocols during 1992 0.90–2.00 4 0 and 1993. Because several Bull Trout were injured or killed 1.75–2.25 4 0 during these efforts, however, a diver-directed electrofishing 1998 0–3.40 59 0 protocol was piloted during 1993 and was employed in the Total 184 260 59 0 area containing Bull Trout during subsequent years (Table 2). A snorkel diver moved upstream during the daytime while

TABLE 2. Effort expended and number of nonnative Brook Trout removed from Sun Creek downstream of Sun Falls (between km 3.4 and km 14.6; km 0 = headwaters) by using multiple electroﬁshing techniques (see Methods) and antimycin-A treatment, 1992–2005.

Electroﬁshing Antimycin-A Stream Effort Brook Trout Stream Effort Brook Trout Year Method reach (km) (person-days) removed reach (km) (person-days) removed 1992 Conventional 3.40–4.50; 33 324 8.60–14.60 56 1,906 7.10–9.80 1993 Conventional 3.40–9.80 28 107 Diver directed 5.80–8.35 10 9 1994 Diver directed 3.40–11.60 78 146 1995 Diver directed 3.40–11.60 162 133 1996 Conventional 9.75–11.60 12 7 Diver directed 3.40–10.00 138 279 Downloaded by [Department Of Fisheries] at 22:29 28 February 2013 1997 Conventional 11.60–14.60 16 202 Diver directed 3.40–11.60 224 197 1998 Diver directed 3.40–11.60 184 234 1999 Conventional 7.75–9.60 45 43 Diver directed 3.40–7.80 77 6 Trap net 6.90–10.50 139 259 2000 Trap net 3.40–11.60 310 316 2.65–17.65 144 73 2002 Trap net 6.95–7.95 8 0 2003 Conventional 11.60–14.60 27 6 2004 Conventional 11.60–14.60 15 0 10.50–17.70 32 9 Diver directed 7.10–7.30 5 1 2005 Conventional 11.60–14.60 9 0 11.55–17.70 24 0 Diver directed 8.75–8.80 3 1 Total 1,523 2,270 256 1,988 BROOK TROUT REMOVAL AND BULL TROUT RESPONSE 121

TABLE 3. Stream conditions and details of chemical application for antimycin-A (Fintrol, 11.5% antimycin-A) treatments of Sun Creek, 1992–2005.

Variable 1992 1998 2000 2004 2005 Dates Aug 25–27 Aug 24–28 Aug 21–30 Aug 2–4 Aug 1–2 Discharge (m3/s) 0.20 0.01–0.11 0.08–0.48 0.45 0.33–0.40 Water temperature (◦C) 3.5–11.0 2.2–12.8 4.0–11.0 7.2–10.0 7.2–12.2 pH 7.1–7.7 7.4–7.8 7.6 7.6 – Distance treated (km) 6.00 3.40 15.00 7.20 6.15 Fintrol applied via: Drip stations (L) 1.67 1.40 21.66 8.64 2.50 Backpack sprayer (L) 0.07 0.08 2.21 0.79 0.54 Total Fintrol applied (L) 1.74 1.48 23.87 9.43 3.04 Daily duration of application (h) 5–8 4–8 8–10 6–10 10 Antimycin-A concentration (µg/L) 4–12 9–16 8–12 10 7–11 KMnO4 applied (kg) 76 42 290 117 79

counting Bull Trout. If no Bull Trout were observed in a section genetic analysis indicated that field identification could be un- of stream, an electrofishing crew consisting of one electrofishing reliable (Buktenica 1997). After 1997, identification techniques unit operator and several netters immediately shocked the area were refined, and hybrids that were encountered during elec- to remove Brook Trout. Conventional upstream electrofishing trofishing were removed, but hybrids were not consistently (without snorkel direction) that employed one or two elec- targeted among years. The total number of Brook Trout re- trofishing units continued in downstream portions of the project moved from the stream was recorded and summarized annu- area where Brook Trout were believed to be allopatric (Table 2). ally by stream location. Effort for electrofishing and antimycin To improve Brook Trout removal efficiency and to mini- treatments was calculated in person-days, where 1 person-day mize the risk of fish injury while salvaging Bull Trout prior to equaled one crew member working for 10 h, including travel antimycin treatments, a trap-net electrofishing protocol was de- time to and from field locations. veloped. Short sections of stream (50–75 m) were isolated with Antimycin treatments.—The piscicide antimycin-A (Fintrol conventional block nets upstream and a custom-designed trap Fish Toxicant, 11.5% antimycin-A; Aquabiotics Corp., Bain- net downstream. The trap net was 10.1 m wide × 1.2 m tall bridge Island, Washington) was applied to portions of Sun Creek and was constructed of 0.95-cm, square knotless nylon mesh. on five occasions between 1992 and 2005 (Table 3). The first Two 4.6-m-wide wings directed fish through a fyke tunnel into two treatments (applied in 1992 and 1998) targeted reaches that a net bag (0.9 m wide × 2.4 m long) in the center. The fyke contained allopatric Brook Trout downstream and upstream of entrance was 0.9 m wide × 1.2 m tall and tapered along a 0.6-m the truncated Bull Trout distribution described by Dambacher length to a 0.2-m exit inside the bag. Field crews completed up et al. (1992). The third treatment (in 2000) was applied to the to five consecutive passes through each isolated section; fish entire project area downstream of Sun Falls, with the intent of were herded downstream into the trap net with two backpack eradicating Brook Trout. The final two treatments (in 2004 and electrofishing units (run continuously), and stunned fish were 2005) were used to remove the remaining Brook Trout that were dipnetted along the way. Trap-net electrofishing was used for discovered after the 2000 treatment. Bull Trout, if present, were Downloaded by [Department Of Fisheries] at 22:29 28 February 2013 Brook Trout removal in limited sections of Sun Creek during salvaged from the stream prior to all applications of the pisci- 1999 (Table 2). During 2000, the 8.2-km section of Sun Creek cide (see Refuge populations below). All antimycin treatments between Sun Falls (km 3.4) and the upper exclusion barrier were applied during August and were timed to coincide with (km 11.6) was sampled twice with this protocol, and two small seasonally low discharge but to occur before the onset of Brook segments (totaling 1.2 km) were sampled a third time. The trap- Trout spawning in the fall. net electrofishing effort in 2000 served the dual purpose of In preparation for piscicide treatments, several field and lab- Brook Trout removal and Bull Trout salvage before the stream oratory studies were completed, including inventories of am- was treated with antimycin. phibian and invertebrate fauna, evaluations of water chemistry, Additional targeted electrofishing was conducted between dye release studies to quantify downstream flow rate (i.e., time 2002 and 2005 after Brook Trout were observed by snorkelers of travel) in the stream, and laboratory bioassays to determine who were monitoring the effectiveness of chemical treatment effective dosages of antimycin and potassium permanganate (Table 2). Throughout electrofishing removal efforts, all cap- (KMnO4) in Sun Creek water. Most of the aquatic insects tured Brook Trout were removed and euthanized. Bull Trout × found in the stream were commonly occurring and broadly dis- Brook Trout hybrids were not removed prior to 1997 because tributed taxa with a high dispersal capability—a pattern that 122 BUKTENICA ET AL.

is likely attributable to the relatively recent recovery of the Downstream of each treatment area, KMnO4 was applied invertebrate community after the eruption of Mount Mazama to the stream at a concentration of 4 mg/L to detoxify the an- (7,700 years ago; Bacon et al. 2002)—and therefore these taxa timycin (Table 3). The delivery of KMnO4 started 1 h before were expected to recolonize the stream shortly after treatments the expected arrival of antimycin and continued for 1–2 h after (Wisseman 1992). Long-toed salamanders Ambystoma macro- the last antimycin was expected to pass the station. The KMnO4 dactylum and Cascades frogs Rana cascadae were the only station consisted of a reservoir (190 or 1,500 L, depending on amphibian species identified that were likely to breed within the streamflow) that was attached to a constant-head container with Sun Creek floodplain. These species typically metamorphose a float valve. The KMnO4 outflow was measured at 15-min into terrestrial adults by midsummer and so were unlikely to intervals and was adjusted with a spigot valve. be affected by August piscicide treatments. Bioassays indicated Refuge populations.—To mitigate the risk of accidental Bull that 16 antimycin exposures were lethal to 100% of juvenile Trout extirpation during chemical treatments, Bull Trout were trout in Sun Creek water at pH 7.2 under controlled laboratory removed from Sun Creek and were held in several temporary conditions. One antimycin exposure is equivalent to 1 µgof refuge populations during the course of the project. During the antimycin/L of water for 1 h. 1998 treatment, small populations of Bull Trout were trans- During all treatments, antimycin was added to the stream in planted to an isolated, fishless stream nearby (Lost Creek; n = a manner consistent with the directions for use on the Fintrol 74 fish) and to a portable raceway near the Crater Lake National label and by following the methods later summarized by Moore Park headquarters (n = 123 fish). In September and October et al. (2008). Fintrol was mixed with filtered stream water and 1998 (i.e., after completion of the August 1998 treatment), 78 nonoxynol-9 detergent (62 mL of detergent per 480 mL of Bull Trout from the raceway were introduced into Sun Creek Fintrol), and the mixture was added to Sun Creek using constant- above Sun Falls and the remaining 45 fish from the raceway flow drip stations spaced at intervals based on in-stream time of were added to the Lost Creek population. travel. Each drip station consisted of an 11- or 19-L (3- or 5-gal) Prior to the 2000 treatment, as many Bull Trout as possible bucket connected to a constant-head pet water trough with a were salvaged by trap-net electrofishing and were held either in short piece of garden hose. A small hole was drilled into the bot- the raceway at the national park headquarters (FL > 80 mm; tom of the trough to yield an application rate of approximately n = 486 fish) or at a nearby Oregon Department of Fish and 4.5–4.9 L/h. The quantity of antimycin added to each drip Wildlife hatchery (FL < 80 mm; n = 130 fish). Additional Bull station was adjusted based on stream discharge and delivery rate Trout remained upstream of Sun Falls and in Lost Creek from to obtain target antimycin concentrations of 8–10 µg/L in the the 1998 transplants. A rotary-drum fish screen was installed stream for 6–8 h per application. This dosage, determined by upstream of Sun Falls to prevent transplanted fish from moving monitoring sentinel Brook Trout that were held in the stream, downstream into the treatment area during the 2000 antimycin exceeded the lethal exposure identified in laboratory bioassays treatment. Similarly, block nets prohibited the movement of Bull with juvenile fish to compensate for increased rates of antimycin Trout into the treatment area during the 1992, 2004, and 2005 oxidation in the turbulent lotic environment and greater sizes treatments, and salvaged Bull Trout were placed upstream of of stream-resident Brook Trout compared with trout used in the the block nets. laboratory. A backpack sprayer was used to apply additional One week after the 2000 antimycin treatment was completed, antimycin to springs, side channels, and isolated pools along 480 Bull Trout were returned to Sun Creek from the raceway the creek that could have contained fish but were not connected and were distributed to mimic the spatial and length frequency to the main channel. The total quantity of Fintrol added to the distributions observed during prior Bull Trout salvage efforts. stream between 1992 and 2005 was 39.6 L, or approximately Eighty additional Bull Trout were returned to Sun Creek from Downloaded by [Department Of Fisheries] at 22:29 28 February 2013 4.5 L of the active ingredient antimycin-A (Table 3). the Oregon Department of Fish and Wildlife hatchery in 2001 Treatments in all years were designed with intentional re- after they were large enough to identify reliably to species. dundancy such that an effective dose of piscicide was deliv- Bull Trout were returned to the creek as soon as possible to ered at least twice within the same week. Generally, several minimize any risk associated with captivity. The Bull Trout that segments of stream equivalent to approximately 1.5–3.0 h of were transplanted upstream of Sun Falls subsequently moved in-stream travel time and approximately 50–100 m of elevation downstream over the falls, died, or were moved downstream change were treated per day, starting at the upstream end and by managers in 2002 (n = 8 fish) when it became evident that progressing to the downstream end of the treatment area. An- abundance above the falls was dwindling. To comply with terms timycin application occurred as late in the day as possible to and conditions of a U.S. Fish and Wildlife Service (USFWS) maximize water temperatures during treatment. Sentinel Brook Biological Opinion governing the project (USFWS 1998b), Bull Trout were held in net-pens and block nets were placed within Trout that were introduced into Lost Creek were not repatriated the stream at intervals based on in-stream time of travel to eval- to Sun Creek. uate the effectiveness and neutralization of the piscicide within The raceway that was used to hold Bull Trout during the and downstream of the segments that were being treated. 1998 and 2000 antimycin treatments was a fiberglass-reinforced BROOK TROUT REMOVAL AND BULL TROUT RESPONSE 123

plastic aquaculture tank (Red Ewald Model RW-8; volume = to species, counted, and summarized by 50-m stream section. 8,044 L). The raceway was set up adjacent to a small stream Snorkel surveys were conducted between July and September near the Crater Lake National Park headquarters (∼5kmfrom when stream discharge was generally less than 0.5 m3/s. Dur- Sun Creek) and was filled with water that was gravity fed from ing 1994–2004, when Bull Trout were restricted to the upper the stream through a screened intake. Water flowed constantly portion of the project area, single-pass electrofishing without through the raceway and was returned to the stream at a rate block nets was used to determine the downstream extent of Bull of 0.019 m3/s. An alarm system was installed to automatically Trout distribution; snorkel surveys then proceeded upstream to telephone the project managers if flow into the raceway stopped Sun Falls from the point where Bull Trout were located. The or if the water level in the raceway significantly increased or de- snorkel survey in 2000 was conducted after Bull Trout were creased. A plywood top, artificial grass (AquaMats and Manta- returned to the stream following antimycin treatment. During mats; Meridian Applied Technologies), and clusters of polyvinyl 2005–2008, single-pass electrofishing without block nets was chloride pipe were added to the raceway to increase cover and used to characterize Bull Trout distribution in the reach between reduce stress for fish. Bull Trout were sorted by length into two and immediately upstream of the exclusion barriers (from km raceway compartments (FL = 80–120 mm; and FL > 120 mm) 11.0 to km 14.6), and the snorkel survey began at km 11.2 and and were held for a maximum of 37 d in 1998 or 63 d in 2000. km 11.0. During 2009–2010, the snorkel survey included the en- Fish were fed earthworms and aquatic invertebrates that were tire 11.2-km reach from the lower exclusion barrier (km 14.6) to collected from nearby streams. For transport to and from Sun Sun Falls. Snorkel surveys and electrofishing surveys were not Creek, Bull Trout were sorted by length into 10-L, iced coolers calibrated to estimate total abundance, but they were assumed that were oxygenated with medical-grade oxygen and carried to be consistent methods of assessing Bull Trout abundance and on backpack frames. distribution within the stream. Total abundance of Bull Trout was known in 2000 and was Bull Trout Response estimated in 2009. After the antimycin treatment in 2000, we To determine the effectiveness of Brook Trout removal and knew exactly how many Bull Trout were returned to the stream, to evaluate the Bull Trout population response (abundance and and we assumed that this number was very close to the total num- distribution), single-pass daytime snorkel and electrofishing sur- ber of live Bull Trout present in the stream at that time. In 2009, veys were conducted annually during 1994–2010 (Table 4). A we used a mark–resight approach to estimate total Bull Trout single snorkel diver moved methodically upstream while ex- abundance. Bull Trout (n = 203) were captured by electrofish- amining all available habitat. Observed fish were identified ing throughout the 11.2-km reach between the lower exclusion barrier and Sun Falls. The fish were marked by removing a TABLE 4. Number of Bull Trout observed by single-pass snorkeling and small portion of the upper caudal fin lobe, were tagged with electrofishing surveys of Sun Creek, 1994–2000 (stream reaches are designated PIT tags (23- × 3.4-mm half-duplex tags; Texas Instruments = based on distance from the headwaters; i.e., km 0 headwaters). TI-RFID S-2000), and were released near the site of capture. Electrofishing Snorkeling Only fish larger than 120 mm FL were marked. A stationary PIT antenna at the lower exclusion barrier monitored the emigration Stream Bull Trout Stream Bull Trout of marked fish from the study area to verify population closure. Year reach (km) observed reach (km) observed Approximately 1 month after marking, we conducted a single- pass daytime snorkel survey of the entire reach and counted 1994 10.5–11.6 0 3.4–10.5 116 both marked and unmarked fish. Total abundance of Bull Trout 1995 10.2–11.6 0 3.4–10.2 97 exceeding 120 mm FL was estimated using the joint hypergeo-

Downloaded by [Department Of Fisheries] at 22:29 28 February 2013 1996 10.0–11.6 5 3.4–9.7 142 metric maximum likelihood estimator in program NOREMARK 1997 9.6–14.6 4 3.4–9.6 161 (White 1996). 1998 9.7–11.6 5 3.4–9.7 110 1999 9.1–9.6 3 3.4 –9.1 134 2000 4.9–9.4 128 RESULTS 2001 9.8–11.6 1 3.4–9.8 241 Brook Trout Eradication 2002 3.4–9.4 183 Brook Trout were successfully eradicated from the 3.4-km 2003 9.0–14.6 2 3.4–9.0 365 reach upstream of Sun Falls by electroﬁshing that required 184 2004 9.8–14.6 2 3.4–9.8 471 person-days (54 person-days/km) applied over 6 years. In total, 2005 11.6–14.6 9 3.4–11.2 699 260 Brook Trout were removed from this reach. An additional 2006 11.0–14.6 18 3.4–11.0 507 59 person-days of effort (17 person-days/km) were spent treat- 2007 11.0–14.6 31 3.4–11.0 651 ing this reach with antimycin during 1998 to ensure that no 2008 11.0–14.6 35 3.4–11.0 613 Brook Trout remained in the stream (Table 1). The 1998 an- 2009 3.4–14.6 455 timycin treatment killed sentinel Brook Trout, but no additional 2010 3.4–14.6 692 wild Brook Trout were found dead in the reach, suggesting that 124 BUKTENICA ET AL.

previous electrofishing removal had been successful. Antimycin was effectively neutralized by KMnO4 addition during the 1998 treatment, and no mortality occurred below the detoxification area. Downstream of Sun Falls, 2,270 Brook Trout were removed during 1992–2005 by a combination of conventional upstream electrofishing (n = 689 fish), diver-directed electrofishing (n = 1,006 fish), and trap-net electrofishing (n = 575 fish; Table 2). Antimycin treatments in 1992, 2000, 2004, and 2005 killed at least 1,988 Brook Trout downstream of Sun Falls (Table 2). Of these, 1,906 fish were removed by the 1992 treatment. The count of Brook Trout removed by antimycin included Brook Trout that were captured in block nets or found in the stream channel during treatments; it is possible that additional fish were killed by antimycin treatments but not observed. The antimycin treatment in 2000 did not eradicate Brook Trout from the treatment area. Surviving Brook Trout were observed between the exclusion barriers in 2003, prompting the 2004 and 2005 antimycin treatments. Single Brook Trout were observed and removed upstream of the upper exclusion barrier during 2004 (at km 7.2) and 2005 (at km 8.75; Table 2). Since 2005, no Brook Trout have been observed in Sun Creek upstream of the lower artificial barrier (km 14.6). Antimycin treatments killed at least 42 Bull Trout that were missed by pretreatment electrofishing salvage efforts, including 1 Bull Trout during the 1992 treatment, 27 fish during 2000, 7 fish during 2004, and 7 fish during 2005. No Bull Trout mortality occurred during the 1998 treatment. Six Bull Trout died while being held in the raceway, and two were found dead immediately after being returned to the stream in 2000. No Bull Trout mortality occurred during transportation.

Bull Trout Response Annual snorkel and electroﬁshing surveys indicated that Bull Trout increased in abundance and expanded in distribution during the course of the project (Table 4; Figure 2). The Bull Trout count remained relatively low between 1994 and 2000 during efforts to remove Brook Trout by electroﬁshing. Abundance increased from 2001 to 2005 after the 2000 antimycin treatments, Downloaded by [Department Of Fisheries] at 22:29 28 February 2013 and since 2005 the counts of Bull Trout have remained relatively high but variable (Table 4). After the 2000 antimycin treatment, 480 Bull Trout (≥age 1) were returned to the stream, and this was presumably the total

←−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−− FIGURE 2. Mean number of Bull Trout observed per 250-m segment of Sun Creek by snorkel diving (solid line) and electroﬁshing (dotted line) during 1994–2000 (i.e., prior to full-stream antimycin treatment) and during each 2- year interval thereafter. Box-and-whisker plots represent spatial distribution of observed Bull Trout during each time period (vertical band within box = median; ends of box = interquartile range; whiskers = maximum observed distribution). Vertical dashed lines on each panel indicate the locations of Sun Falls (km 3.4) and two artiﬁcial barriers (km 11.6 and km 14.6). BROOK TROUT REMOVAL AND BULL TROUT RESPONSE 125

number of Bull Trout present. During the 2009 mark–resight channels of downstream reaches, several years of electrofish- survey, 2 of 203 fin-clipped and PIT-tagged Bull Trout were ing were insufficient to eradicate Brook Trout. Our results are detected leaving the study reach prior to the completion of the consistent with literature reports that eradication attempts with resight effort. Divers observed 47 of the remaining 201 fin- electrofishing require considerable effort (e.g., Kulp and Moore clipped Bull Trout and positively identified the mark status of 2000; Shepard et al. 2002) and often prove unsuccessful (e.g., 381 total Bull Trout (>120 mm FL). Because the number of em- Thompson and Rahel 1996; Meyer et al. 2006). igrants was low (∼1% of tagged fish), we considered population Two alternative techniques that were developed for closure during the resight effort to be a valid assumption, and we the project—diver-directed electrofishing and trap-net estimated the total abundance of Bull Trout larger than 120 mm electrofishing—allowed us to greatly reduce Brook Trout FL to be 1,606 fish (95% confidence interval = 1,311–2,097 abundance prior to eradication attempts by chemical treatment. fish). Diver-directed electrofishing removed Brook Trout while The length of stream occupied by Bull Trout during summer minimizing harm to Bull Trout, but the technique was labor snorkel and electrofishing surveys increased during the project intensive, with low potential as a tool for eradication. Trap-net period, and by 2010 their distribution extended at least 11.2 km electrofishing was very effective for removing Brook Trout from Sun Falls to the lower exclusion barrier (Figure 2). The and salvaging Bull Trout prior to chemical treatments, as was upstream extent of Bull Trout distribution is constrained by Sun evident from the relatively small number of fish that were found Falls. The lower extent of the distribution is not constrained by dead during the 2000 antimycin treatment. Moreover, down- barriers within the project area, and PIT tag detections suggested stream herding of fish into a trap net appeared to pose a lower that some Bull Trout moved downstream over the lower exclu- risk of fish injury than conventional electrofishing, although we sion barrier. Bull Trout were observed between the exclusion did not explicitly test this hypothesis. Due to the high capture barriers as early as 2004 and 2005. By 2010, however, the Bull efficiency of trap-net electrofishing, the technique may have Trout density within the lowest 4 km of the project area remained utility as an eradication tool for streams when toxicants are not substantially lower than the density in areas of peak abundance an acceptable alternative or when sensitive species co-occur upstream. Larger fish appeared to lead the downstream expan- with the target species. We are unaware of diver-directed sion. For example, in 2007, the 31 Bull Trout that were captured electrofishing or trap-net electrofishing techniques being used downstream of km 11.0 ranged from 166 to 282 mm FL, whereas elsewhere for fish removal in streams prior to their use in this the largest Bull Trout captured elsewhere in the stream was project. 241 mm FL. As the downstream extent of distribution expanded, The portable raceway that was used to hold native Bull Trout peak Bull Trout abundance remained centered near the area during antimycin treatments was another novel technique imple- where Bull Trout were located in 1989. In all years, the major- mented for the Sun Creek project, and it shows great potential for ity of Bull Trout were observed in a 2–3-km section of stream use in other native fish restoration projects. The raceway system (i.e., km 6 to km 9; Figure 2). was relatively inexpensive, required no mechanical pumps, and could be set up and maintained in a somewhat remote location. DISCUSSION We were concerned that holding wild Bull Trout in the raceway would cause them to be stressed, aggressive toward one another, Brook Trout Eradication or cannibalistic, with consequent injuries or deaths. However, The long-term, adaptive nature of the Sun Creek project al- raceway-related mortalities, including those during transporta- lowed us to qualitatively evaluate the effectiveness of multiple tion and captivity, were minimal (1%). Bull Trout were observed electrofishing methods and antimycin treatment as means of in and around the artificial cover that was placed in the raceway, Downloaded by [Department Of Fisheries] at 22:29 28 February 2013 eradicating nonnative Brook Trout. The strengths, weaknesses, and fish also “schooled” in open areas of the tank. Aggressive and efficacy of the removal techniques varied with stream size behavior was observed primarily when the fish were fed. and complexity, presence or absence of sympatric Bull Trout, We found that antimycin was effective for removing Brook and available resources. The combination of techniques allowed Trout from Sun Creek, but more than one treatment was for removal of Brook Trout while minimizing risk to a Bull Trout required. Brook Trout were discovered upstream of exclusion population that was likely near extirpation prior to implementa- barriers between 2003 and 2005, and we suspect that these tion of the project. fish likely survived the antimycin treatment in 2000 (perhaps Conventional backpack electrofishing was effective for erad- as young of the year) by occupying springs or poorly mixed icating allopatric Brook Trout in the smallest headwater reach side-channel habitats that were missed by backpack sprayers. of Sun Creek upstream of Sun Falls, and this approach may The alternative possibility—that Brook Trout reinvaded the be applicable for eradicating nonnative trout in other, similar project area after the 2000 treatment by passing upstream over montane streams. The headwater reach was characterized by a the exclusion barriers—seems less likely because jump heights narrow channel (<2 m) with few undercut banks or complex at the barriers are more than twice the maximum jump height pools; nevertheless, eradication of Brook Trout from this reach identified for Brook Trout in laboratory trials (Kondratieff and required 6 years of repeated effort. In the larger, more-complex Myrick 2006), and more importantly, no additional Brook Trout 126 BUKTENICA ET AL.

have been found upstream of the barriers during 5 years of throughout the 11.2-km restored reach. The substantial increase monitoring since 2005. Based on this experience, we conclude in Bull Trout abundance and distribution during the years af- that repeat application of antimycin during two consecutive ter the 2000 antimycin treatment provides strong evidence that years is more reliable for removing trout than the application the native species was limited by interspecific interaction with of two treatments during a single year. Small fish that survive Brook Trout. Despite removal of thousands of Brook Trout from an initial treatment in off-channel habitat may grow, move into the stream during the 1990s and commensurate reductions in the main channel, and become more susceptible to the toxicant Brook Trout density, a strong increase in Bull Trout abundance during a second treatment. The final Sun Creek treatments did not occur until the entire stream was treated with antimycin between the barriers (i.e., 2004 and 2005) were planned to take in 2000. Disproportionate loss of reproductive success for Bull place in two consecutive years for this reason. Trout through hybridization (Kanda 2002) may explain the ini- We believe that there is no longer a viable population of tial slow response of Bull Trout while Brook Trout remained in Brook Trout from the headwaters to the lower migration barrier the stream. in Sun Creek and that Brook Trout were eradicated from this It is unclear why Bull Trout have not expanded more rapidly reach. Although eradication is difficult to demonstrate conclu- or evenly into vacant habitats after Brook Trout removal. By sively, each successive monitoring year without Brook Trout contrast, Brook Trout are known to invade vacant stream habitat detections increases our confidence that no Brook Trout remain. relatively quickly. For example, Brook Trout repeatedly rein- Ultimately, the success of the Sun Creek project is attributable to vaded a stream segment in Colorado within 8 months after they the long-term effectiveness monitoring and targeted Brook Trout had been removed (Peterson and Fausch 2003); Brook Trout removal efforts that were implemented after the streamwide also recolonized a defaunated stream reach in Virginia within chemical treatment in 2000. 3 years of a catastrophic debris flow (Roghair et al. 2002). In Electrofishing eradication of Brook Trout upstream of Sun Sun Creek, we found that Brook Trout from the upstream reach Falls required approximately threefold more effort (54 person- recolonized the area between artificial barriers and became well days/km or 540 h/km of stream over 6 years) than antimycin established there within 5 years of the 1992 antimycin treatment treatments of the same stream reach (17 person-days/km or (Table 2). 170 h/km, applied over a period of several weeks during Because resident Bull Trout and Brook Trout are both cold- 1 year). Similarly, Moore et al. (2005) reported that the eradica- water char species that occupy headwater streams, we assumed tion of nonnative trout from small streams of the Great Smoky early in the project that if Bull Trout became allopatric, they Mountains required approximately 1,400 h/km of stream when would recolonize and inhabit Sun Creek in a manner similar to using multiple-pass electrofishing protocols versus 400 h/km that exhibited by the invasive Brook Trout. Unlike Brook Trout, when using antimycin application. Piscicide treatments require however, Bull Trout did not persist when transplanted to the more planning, environmental regulatory compliance, and state small headwater reaches above Sun Falls. The presence of Bull and federal permitting than electrofishing and may have greater Trout between the exclusion barriers during 2005 suggests that effects on nontarget organisms, such as aquatic invertebrates Bull Trout quickly colonized the area after the 2004 antimycin (Vinson et al. 2010). When environmental effects are accept- treatment, yet Bull Trout density remains much lower than the able or mitigated, however, the use of antimycin requires less density previously attained by Brook Trout in the same stream long-term commitment of time. Due to the increasing need for reaches. Restricted movement or poor spawning success may restoration of native fish assemblages in combination with lim- limit Bull Trout dispersal in Sun Creek. Relatively few Bull ited financial and human resources, the application of toxicants Trout generations have elapsed since removal, however, and we such as antimycin has broad utility as an efficient and cost- expect that Bull Trout will continue to increase in abundance Downloaded by [Department Of Fisheries] at 22:29 28 February 2013 effective approach to nonnative fish removal provided that the and colonize vacant habitat over time. native fish populations can be isolated from reinvasion. Bull Trout persisted in a small reach of upper Sun Creek The resources dedicated to restoring Sun Creek Bull Trout decades after Brook Trout invasion and this core area has re- cannot be expended for restoration of every population of Bull mained the area of highest abundance during Bull Trout re- Trout or other native fish threatened by nonnative competitors. covery, suggesting that the habitat in this reach provided some Such intensive effort may be warranted only for critical popula- benefit to Bull Trout that conferred “environmental resistance” tion segments with special status or local management interest to invasion (Fausch 2008). A study of habitat in the stream found (Clancy et al. 1996). that peak Bull Trout density both before and after Brook Trout removal was located between two spring areas (∼km 5.3 and Bull Trout Response km 8.4), and Bull Trout distribution was positively associated Removal of Brook Trout from Sun Creek has resulted in with low turbidity, upstream springs, deeper pools, and higher Bull Trout abundance increasing approximately tenfold in the stream temperatures (average daily temperature between 7◦C stream since 1989. Distribution also has expanded beyond the and 8◦C; Renner 2005). Because summer temperatures through- 1.9-km section where Bull Trout persisted in sympatry with out the stream remain well below the thermal tolerances of Bull Brook Trout prior to restoration, and Bull Trout are now present Trout (Selong et al. 2001), we propose a yet-untested hypothesis BROOK TROUT REMOVAL AND BULL TROUT RESPONSE 127

that an important effect of groundwater spring habitats in Sun impacts on migration will have to be mitigated. Migratory Bull Creek might be to improve spawning success (Baxter and Trout may hold a competitive advantage over sympatric Brook McPhail 1999) or overwinter survival by ameliorating winter Trout due to larger body size and higher fecundity (Fraley and ice conditions rather than to provide coldwater refuge in the Shepard 1989; Rieman and McIntyre 1993), and thus Bull Trout summer. Presumably, habitat throughout Sun Creek was suitable may be able to persist even if complete removal or exclusion of for Bull Trout prior to Brook Trout invasion, as Wallis (1948) Brook Trout from the lower watershed is not feasible. reported that Bull Trout were present throughout the drainage. Efforts to restore local fish populations are often lengthy Therefore, the presence of groundwater springs or some other and are never easily accomplished, and the Sun Creek project favorable habitat characteristic may help to explain why Bull is no exception. After nearly two decades of work have been Trout persisted in the core area when Brook Trout were present; invested, Bull Trout are recovering in upper Sun Creek, but however, habitat quality alone should not prevent the Bull Trout continued removal of introduced fish from lower Sun Creek will distribution from expanding after Brook Trout removal. be necessary to promote Bull Trout recovery throughout the watershed. Such long-term sustained effort, although difficult Prospect for the Future and costly, yields immeasurable value by preserving the native The USFWS recovery plan for Bull Trout in the Klamath biodiversity of fishes. River basin (USFWS 2002) calls for stabilization of existing Bull Trout populations (i.e., prevention of further declines in ACKNOWLEDGMENTS abundance and distribution), downstream expansion of headwa- Special thanks are due to Edwin (Phil) Pister, chair of the ter resident populations, and re-establishment of migratory life Crater Lake Bull Trout Advisory Panel, and to panel mem- history expression to promote Bull Trout colonization of previ- bers Fred Allendorf, Norm Anderson, Tom Felando (deceased), ously occupied habitat and genetic exchange among currently Doug Markle, and Steve Moore. Gary Larson (U.S. Geologi- isolated populations. Removal of introduced Brook Trout from cal Survey, emeritus) provided invaluable advice and participa- the project area stabilized the headwater Bull Trout population tion throughout the project. Credit for program success rests on in upper Sun Creek. The population is increasing in abundance the collective commitment and dedication of Crater Lake Na- and distribution, and the short-term prospect for Sun Creek Bull tional Park field crews for 19 years—too many crew members Trout is favorable. To promote long-term resilience of this Bull to acknowledge by name. Park managers and biologists Scott Trout population, however, the Bull Trout distribution must ex- Stonum, Ken Stahlnecker, Mac Brock, and Craig Ackerman pand downstream of the exclusion barriers into the remaining championed the program over the years. Funding was provided unrestored portion of Sun Creek—a reach that is currently occu- by the National Park Service, the National Park Foundation in pied by introduced Brook Trout and Brown Trout Salmo trutta. partnership with Target Stores, the Crater Lake Natural His- Furthermore, connectivity with other upper basin streams con- tory Association, and the USFWS Klamath Basin Ecosystem taining imperiled Bull Trout populations must be re-established Restoration Office. Philip Howell, Kevin Meyer, Seth Wenger, to promote the recovery of Bull Trout in the region. Jason Dunham, Clint Muhlfeld, Carter Kruse, and one anony- State, federal, and private partners have initiated a new mous reviewer provided helpful comments on earlier drafts of project to apply the restoration approach used in the Crater the manuscript. Reference to trade names does not imply en- Lake National Park reach to the lower portion of Sun Creek (km dorsement by the U.S. Government. 14.6 to km 22.0). Goals of the project include removing introduced salmonids, re-establishing a native fish assemblage, and preventing nonnative fish from reinvading this lower reach. This REFERENCES Downloaded by [Department Of Fisheries] at 22:29 28 February 2013 strategy necessitates construction of additional artificial barriers Allan, J. D., and A. S. Flecker. 1993. Biodiversity conservation in running to prevent Brook Trout reinvasion, thus highlighting a tradeoff waters: identifying the major factors that threaten destruction of riverine species and ecosystems. BioScience 43:32–43. for the management of native fish. The conservation benefits of Allendorf, F. W., R. F. Leary, P. Spruell, and J. K. Wenburg. 2001. The prob- artificial barriers, including protection against competition, dis- lems with hybrids: setting conservation guidelines. Trends in Ecology and placement, and hybridization, must be weighed against the risks Evolution 16:613–622. of genetic isolation and disruption of migration (Peterson et al. Bacon, C. R., J. V. Gardner, L. A. Mayer, M. W. Buktenica, P. Dartnell, D. 2008; Fausch et al. 2009). In Sun Creek, there is little evidence of W. Ramsey, and J. E. Robinson. 2002. Morphology, volcanism, and mass wasting in Crater Lake, Oregon. Geological Society of America Bulletin extant migratory Bull Trout, and land use practices in the basin 114:675–692. have likely selected against migration for many generations. Baxter, J. S., and J. D. McPhail. 1999. The influence of redd site selection, Exclusion barriers are warranted because the population is evo- groundwater upwelling, and over-winter incubation temperature on survival lutionarily significant, is one of the few Bull Trout populations of Bull Trout (Salvelinus confluentus) from egg to alevin. Canadian Journal remaining in the region, and has faced imminent displacement of Zoology 77:1233–1239. Buchanan, D. V.,M. L. Hanson, and R. M. Hooton. 1997. Status of Oregon’s Bull by invasive species (Fausch et al. 2009). We anticipate that as Trout: distribution, life history, limiting factors, management considerations, the density and distribution of Sun Creek Bull Trout increase, and status. Technical Report to the Bonneville Power Administration, Report migratory behavior may be expressed, and the artificial barriers’ DOE/BP-34342–5, Oregon Department of Fish and Wildlife, Portland. 128 BUKTENICA ET AL.

Buktenica, M. W. 1997. Bull Trout restoration and Brook Trout eradication Markle, D. F. 1992. Evidence of Bull × Brook Trout hybrids in Oregon. at Crater Lake National Park, Oregon. Pages 127–135 in W. C. Mackay, Pages 58–67 in P. J. Howell and D. V. Buchanan, editors. Proceedings of M. K. Brewin, and M. Monita, editors. Friends of the Bull Trout confer- the Gearhart Mountain Bull Trout workshop. American Fisheries Society, ence proceedings. Bull Trout Task Force (Alberta), Trout Unlimited Canada, Oregon Chapter, Corvallis. Calgary. Meronek, T. G., P. M. Bouchard, E. R. Buckner, T. M. Burri, K. K. Demmerly, Clancy, C., C. Frissell, and T. Weaver. 1996. Assessment of methods for removal D. C. Hatleli, R. A. Klumb, S. H. Schmidt, and D. W. Coble. 1996. A review or suppression of introduced fish to aid in Bull Trout recovery. Montana Bull of fish control projects. North American Journal of Fisheries Management Trout Scientific Group, Governor’s Bull Trout Restoration Team, Technical 16:63–74. Report, Montana Fish, Wildlife, and Parks, Helena. Meyer, K. A., J. A. Lamansky Jr., and D. J. Schill. 2006. Evaluation of an Cucherousset, J., and J. D. Olden. 2011. Ecological impacts of non-native fresh- unsuccessful Brook Trout electrofishing removal project in a small Rocky water fishes. Fisheries 36:215–230. Mountain stream. North American Journal of Fisheries Management 26:849– Dambacher, J. M., M. W. Buktenica, and G. L. Larson. 1992. Distribution, 860. abundance, and habitat utilization of Bull Trout and Brook Trout in Sun Moore, S. E., M. A. Kulp, J. Hammonds, and B. Rosenlund. 2005. Restora- Creek, Crater Lake National Park, Oregon. Pages 30–36 in P. J. Howell and tion of Sams Creek and an assessment of Brook Trout restoration meth- D. V. Buchanan, editors. Proceedings of the Gearhart Mountain Bull Trout ods: Great Smoky Mountains National Park. National Park Service, Water workshop. American Fisheries Society, Oregon Chapter, Corvallis. Resources Division, Technical Report NPS/NRWRD/NRTR-2005/342, Fort Fausch, K. D. 2008. A paradox of trout invasions in North America. Biological Collins, Colorado. Invasions 10:685–701. Moore, S. E., M. A. Kulp, B. Rosenlund, J. Brooks, and D. Propst. 2008. Fausch, K. D., B. E. Rieman, J. B. Dunham, M. K. Young, and D. P. A field manual for the use of antimycin A for restoration of native fish Peterson. 2009. Invasion versus isolation: trade-offs in managing native populations. National Park Service, Natural Resource Program Center, Report salmonids with barriers to upstream movement. Conservation Biology 23: NPS/NRPC/NRR-2008/033, Fort Collins, Colorado. 859–870. Moyle, P. B. 2002. Inland fishes of California. University of California Press, Fraley, J. J., and B. B. Shepard. 1989. Life history, ecology and population Berkeley. status of migratory Bull Trout (Salvelinus confluentus) in the Flathead Lake Nelson, J. S. 1965. Effects of fish introductions and hydroelectric development river system, Montana. Northwest Science 63:133–143. on fishes in the Kananaskis River system, Alberta. Journal of the Fisheries Gunckel, S. L., A. R. Hemmingsen, and J. L. Li. 2002. Effect of Bull Research Board of Canada 22:721–753. Trout and Brook Trout interactions on foraging habitat, feeding behav- Novinger, D. C., and F. J. Rahel. 2003. Isolation management with artificial ior, and growth. Transactions of the American Fisheries Society 131:1119– barriers as a conservation strategy for Cutthroat Trout in headwater streams. 1130. Conservation Biology 17:772–781. Hamilton, B. T., S. E. Moore, T. B. Williams, N. Darby, and M. R. Vinson. 2009. Peterson, D. P., and K. D. Fausch. 2003. Upstream movement by nonnative Comparative effects of rotenone and antimycin on macroinvertebrate diversity Brook Trout (Salvelinus fontinalis) promotes invasion of native Cutthroat in two streams in Great Basin National Park, Nevada. North American Journal Trout (Oncorhynchus clarki) habitat. Canadian Journal of Fisheries and of Fisheries Management 29:1620–1635. Aquatic Sciences 60:1502–1516. Hankin, D. G., and G. H. Reeves. 1988. Estimating total fish abundance and total Peterson, D. P., B. E. Rieman, J. B. Dunham, K. D. Fausch, and M. K. Young. habitat area in small streams based on visual estimation methods. Canadian 2008. Analysis of trade-offs between threats of invasion by nonnative Brook Journal of Fisheries and Aquatic Sciences 45:834–844. Trout (Salvelinus fontinalis) and intentional isolation for native Westslope Kanda, N., R. F. Leary, and F. W. Allendorf. 2002. Evidence of introgres- Cutthroat Trout (Oncorhynchus clarkii lewisi). Canadian Journal of Fisheries sive hybridization between Bull Trout and Brook Trout. Transactions of the and Aquatic Sciences 65:557–573. American Fisheries Society 131:772–782. Rahel, F. J. 2000. Homogenization of fish faunas across the United States. Kennedy, B. M., D. P. Peterson, and K. D. Fausch. 2003. Different life histo- Science 288:854–856. ries of Brook Trout populations invading mid-elevation and high-elevation Ratliff, D. E., and P. J. Howell. 1992. The status of Bull Trout populations in Cutthroat Trout streams in Colorado. Western North American Naturalist Oregon. Pages 10–17 in P.J. Howell and D. V.Buchanan, editors. Proceedings 63:215–223. of the Gearhart Mountain Bull Trout workshop. American Fisheries Society, Kondratieff, M. C., and C. A. Myrick. 2006. How high can Brook Trout jump? Oregon Chapter, Corvallis. a laboratory evaluation of Brook Trout jumping performance. Transactions Renner, D. A. 2005. Distribution and habitat use of Bull Trout following the of the American Fisheries Society 135:361–370. removal of nonnative Brook Trout. Master’s thesis. Oregon State University, Krueger, C. C., and B. May. 1991. Ecological and genetic effects of salmonid Corvallis. Downloaded by [Department Of Fisheries] at 22:29 28 February 2013 introductions in North America. Canadian Journal of Fisheries and Aquatic Rieman, B. E., and J. D. McIntyre. 1993. Demographic and habitat require- Sciences 48(Supplement 1):66–77. ments for conservation of Bull Trout. U.S. Department of Agriculture, Forest Kruse, C. G., W. A. Hubert, and F. J. Rahel. 2001. An assessment of headwa- Service, Intermountain Research Station, General Technical Report INT-302, ter isolation as a conservation strategy for Cutthroat Trout in the Absaroka Ogden, Utah. Mountains of Wyoming. Northwest Science 75:1–11. Rinne, J. N., W. L. Minckley, and J. N. Hanson. 1981. Chemical treatment of Kulp, M. A., and S. E. Moore. 2000. Multiple electrofishing removals for Ord Creek, Apache County, Arizona, to re-establish Arizona trout. Journal of eliminating Rainbow Trout in a small southern Appalachian stream. North the Arizona-Nevada Academy of Science 16:74–78. American Journal of Fisheries Management 20:259–266. Roghair, C. N., C. A. Dolloff, and M. K. Underwood. 2002. Response of Leary, R. F., F. W. Allendorf, and S. H. Forbes. 1993. Conservation genetics a Brook Trout population and instream habitat to a catastrophic flood of Bull Trout in the Columbia and Klamath river drainages. Conservation and debris flow. Transactions of the American Fisheries Society 131:718– Biology 7:856–865. 730. Leary, R. F., F. W. Allendorf, and K. L. Knudsen. 1983. Consistently high Selong, J. H., T. E. McMahon, A. V. Zale, and F. T. Barrows. 2001. Effect meristic counts in natural hybrids between Brook Trout and Bull Trout. of temperature on growth and survival of Bull Trout, with application of an Systematic Zoology 32:369–376. improved method for determining thermal tolerance in fishes. Transactions Lintermans, M. 2000. Recolonization by the Mountain Galaxias Galaxias olidus of the American Fisheries Society 130:1026–1037. of a montane stream after the eradication of Rainbow Trout Oncorhynchus Shepard, B. B., R. Spoon, and L. Nelson. 2002. A native Wests- mykiss. Marine and Freshwater Research 51:799–804. lope Cutthroat Trout population responds positively after Brook Trout BROOK TROUT REMOVAL AND BULL TROUT RESPONSE 129

removal and habitat restoration. Intermountain Journal of Sciences 8: coterminous United States; final Rule. Federal Register 64:210(1 November 191–211. 1999):58910–58933. Thompson, P. D., and F. J. Rahel. 1996. Evaluation of depletion–removal elec- USFWS (U.S. Fish and Wildlife Service). 2002. Klamath River recovery unit, trofishing of Brook Trout in small Rocky Mountain streams. North American Oregon. Chapter 2 in Bull Trout (Salvelinus confluentus) draft recovery plan. Journal of Fisheries Management 16:332–339. USFWS, Portland, Oregon. USFWS (U.S. Fish and Wildlife Service). 1998a. Endangered and threatened Wallis, O. L. 1948. Trout studies and a stream survey of Crater Lake National wildlife and plants; determination of threatened status for the Klamath River Park. Master’s thesis. Oregon State University, Corvallis. and Columbia River distinct population segments of Bull Trout. Federal White, G. C. 1996. NOREMARK: population estimation from mark–resighting Register 63:111(10 June 1998):31647–31674. surveys. Wildlife Society Bulletin 24:50–52. USFWS (U.S. Fish and Wildlife Service). 1998b. Biological opinion on Crater Wisseman, R. W. 1992. A survey of the aquatic macroinvertebrates in Annie, Lake National Park’s Bull Trout recovery activities in Sun Creek. USFWS, Lost, Sun and Sand creeks, Crater Lake National Park, Oregon: with special Klamath Falls, Oregon. emphasis on potential impacts to the fauna from the Bull Trout population USFWS (U.S. Fish and Wildlife Service). 1999. Endangered and threatened enhancement project. Report to Crater Lake National Park, Aquatic Biology wildlife and plants; determination of threatened status for Bull Trout in the Associates, Corvallis, Oregon. Downloaded by [Department Of Fisheries] at 22:29 28 February 2013 This article was downloaded by: [Department Of Fisheries] On: 28 February 2013, At: 22:29 Publisher: Taylor & Francis Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

North American Journal of Fisheries Management Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/ujfm20 Factors Limiting Brook Trout Biomass in Northeastern Vermont Streams Jud F. Kratzer a & Dana R. Warren b a Vermont Fish and Wildlife Department, 1229 Portland Street, Suite 201, Saint Johnsbury, Vermont, 05819, USA b Department of Fisheries and Wildlife, Oregon State University, Nash Hall, Corvallis, Oregon, 97331, USA Version of record first published: 25 Jan 2013.

To cite this article: Jud F. Kratzer & Dana R. Warren (2013): Factors Limiting Brook Trout Biomass in Northeastern Vermont Streams, North American Journal of Fisheries Management, 33:1, 130-139 To link to this article: http://dx.doi.org/10.1080/02755947.2012.743934

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material. North American Journal of Fisheries Management 33:130–139, 2013 C American Fisheries Society 2013 ISSN: 0275-5947 print / 1548-8675 online DOI: 10.1080/02755947.2012.743934

ARTICLE

Factors Limiting Brook Trout Biomass in Northeastern Vermont Streams

Jud F. Kratzer* Vermont Fish and Wildlife Department, 1229 Portland Street, Suite 201, Saint Johnsbury, Vermont 05819, USA Dana R. Warren Department of Fisheries and Wildlife, Oregon State University, Nash Hall, Corvallis, Oregon 97331, USA

Abstract Habitat, water chemistry, and water temperature requirements and preferences have been well documented for stream-dwelling Brook Trout Salvelinus fontinalis, but fisheries managers rarely know what factors limit Brook Trout abundance in their jurisdictions. We measured various habitat (width, depth,% pool, and wood abundance), water chemistry (pH and conductivity), and water temperature metrics and estimated Brook Trout biomass at 33 stream reaches in northeastern Vermont to determine what factors were most strongly related to Brook Trout biomass, with the ultimate goal of predicting whether adding wood to these streams could be expected to increase Brook Trout abundance. We fit generalized additive models to investigate potential linear and nonlinear relationships between Brook Trout biomass and the various habitat, chemistry, and temperature metrics. Akaike’s information criterion was used to rank candidate models. The top-ranked model included the duration of water temperatures exceeding 20◦C, total wood density, and maximum riffle depth, and it predicted that Brook Trout biomass could be expected to increase with increasing woody habitat as long as water temperature does not exceed 20◦C for 200 h or more per summer. The model also predicted that the benefits of adding woody habitat would be more pronounced in streams with deeper riffles. The absence of stream pH and pool area from the top-ranked models was surprising. Our results highlight the importance of evaluating the local relationships between fish biomass and stream habitat as well as the important influence of study design on the results and conclusions.

The concept of limitation is fundamental to ecology habitat complexity, cover, and thermal conditions), and deter-

Downloaded by [Department Of Fisheries] at 22:29 28 February 2013 (Shelford 1913), and much of our effort in the fields of fish ecol- mining which features are important within given regions is ogy and fisheries management has been devoted to identifying necessary for an overall understanding of the relationship be- the factors that limit fish production. In streams, the availability tween fish production and the stream environments in which of suitable habitat is widely acknowledged as a common limit- they live. ing factor for fish production and abundance (Poff and Huryn We were particularly interested in the degree to which in- 1998; Rosenfeld 2003; Rosenfeld and Taylor 2009; Minns et al. stream woody habitat relates to fish abundance in these systems. 2011), but the term “habitat” can encompass many elements Wood in streams can benefit fish by increasing pool frequency, of the stream environment. The specific habitat features limit- pool size, and habitat complexity (Montgomery et al. 1995; ing fish production and abundance can vary within and across Sundbaum and Naslund¨ 1998; Crook and Robertson 1999; Gur- streams, which creates the need for regionally focused habitat nell et al. 2005). Large wood and large wood jams are key struc- assessments. There are, however, a number of metrics which tural elements in forested streams (Gurnell et al. 2002, 2005; consistently arise as important features in a stream (e.g., pools, Gregory et al. 2003), and wood input to streams is expected to

*Corresponding author: [email protected] Received August 10, 2012; accepted October 23, 2012

130 BROOK TROUT BIOMASS IN VERMONT STREAMS 131

increase across northeastern North America as a result of both and whether a focus on the addition of wood in habitat restora- natural processes and anthropogenic activities. As riparian tion is appropriate in these streams given the numerous other forests mature and recover from historic land use and forests factors that can affect fish abundance. Although the relationship shift from the stem-exclusion phase to a more complex gap- between fish abundance and stream habitat has been evaluated dynamic phase of stand development, wood loading to streams elsewhere in this region (van Zyll De Jong et al. 1997; Kocov- across the eastern USA will increase (Keeton et al. 2007; sky and Carline 2005; Waco and Taylor 2010; Warren et al. Warren et al. 2009). Warming climate conditions, which allow 2010; McKenna and Johnson 2011), this study is unique in two invasive species such as the hemlock wooly adelgid Adelges primary ways. First, we quantified fish abundance and stream tsugae and the emerald ash borer Agrilus planiplennis to move habitat across a high number of sites within a relatively small north, will also promote wood loading to streams across the area that encompassed a range of gradients and substrate types. region through the mortality of host trees (Evans et al. 2011). In We assessed low-gradient tributaries in addition to the boulder- addition to natural increases in stream wood, restoration efforts dominated headwater streams that are often considered “typical” often add wood to increase habitat complexity and enhance for the region. This high-resolution field analysis controlled for stream recovery (Roni et al. 2002; Kail and Hering 2005; regional variability in climate, stocking histories, fish commu- Nagayama et al. 2008; Nagayama and Nakamura 2010). While nities, and underlying geological conditions, which allowed us a number of studies—both observational and experimental— to focus specifically on local habitat and its influence on stream have found significant positive relationships between in-stream fish. Second, we used a statistical approach that allowed us to wood and fish abundance (Burgess and Bider 1980; Flebbe and assess both linear and nonlinear relationships between habitat Dolloff 1995; Solazzi et al. 2000; Roni and Quinn 2001; White features and fish abundance. et al. 2011), increasing wood does not necessarily lead to an increase in fish, and in many streams the benefits of wood for METHODS stream fish remain equivocal (Thompson 2006; Stewart et al. All of the 33 stream reaches were within or just outside of 2009; Anton´ et al. 2011). This uncertainty in wood function is Essex County, Vermont, which is the least populated county in particularly relevant in the northeastern USA, where glaciation the state (Figure 1). Stream reaches ranged in length from 21 to has left an abundance of coarse material that can create large 94 m (Table 1). The land cover is largely dominated by northern pools even when wood is absent (Warren and Kraft 2003). hardwood forests, with pockets of spruce–fir forest. The phys- Beyond physical structures in the stream environment, the iographic region is known as the Northeastern Highlands, and term “habitat” can also encompass the chemical and thermal the underlying bedrock is mostly granite. The area has a long conditions of a stream (Baker et al. 1996; Todd et al. 2008; history of logging, including heavy clear-cutting and log driv- Chu et al. 2010; Waco and Taylor 2010; McKenna and John- ing on the major rivers. Logging remains a major land use in son 2011). Limits imposed by these two factors—temperature the area, but large portions of land have been conserved by the and chemistry—can reduce fish populations independent of the Vermont Agency of Natural Resources and the U.S. Fish and availability of structural elements (Baker et al. 1996; Siitari et al. Wildlife Service. 2011). For example, low pH conditions (chronic or episodic Assessments of fish abundance, water temperature, water acidification) in many headwater streams in the eastern USA chemistry, and habitat occurred in the summer and fall of 2011. can substantially reduce fish diversity and abundance (Baker Brook Trout abundance was estimated during July and August et al. 1996; Baldigo and Lawrence 2000; Warren et al. 2010). with multiple-pass depletion and a maximum likelihood estima- Temperature regimes may affect fish at both low and high ex- tor (Carle and Strub 1978). Electrofishing was conducted using tremes. Stress associated with elevated stream temperatures can either battery-powered backpack units or gasoline-powered gen- Downloaded by [Department Of Fisheries] at 22:29 28 February 2013 decrease abundance via direct mortality and emigration in search erators stationed on shore or in a canoe. Captured Brook Trout of thermal refugia (Baird and Krueger 2003; Xu et al. 2010a, were measured and weighed before being returned to the stream. 2010b; Wenger et al. 2011). In contrast, particularly low temper- Water temperatures were monitored using Onset HOBO atures can decrease fish abundance directly via reduced growth temperature loggers. The loggers were deployed in May and rates and fecundity of individual fish and indirectly through a re- retrieved in late September, with water temperatures being duction in overall ecosystem productivity (Mullner and Hubert recorded every 30 min. Temperature data were used to calculate 2005; Coleman and Fausch 2007a, 2007b). Given the potential two metrics taken from Butryn (2010): duration over 20◦C and for chemical and thermal conditions to limit fish production in 22◦C, which were simply the amount of time (in hours) that streams, our habitat assessment explicitly included these factors water temperatures exceeded the threshold values of 20◦Cor along with the physical structure of the stream environment. 22◦C. These two temperature thresholds were chosen because Our goal was to determine which physical habitat (width, Brook Trout prefer temperatures less than 20◦C (Jobling 1981) depth,% pool, and wood abundance), water chemistry (pH and and exhibit a heat-shock response at 22◦C and above (Lund et al. conductivity), and water temperature metrics were most strongly 2003). related to Brook Trout Salvelinus fontinalis abundance in north- Water chemistry metrics were measured during low-flow eastern USA streams. We were particularly interested in wood conditions in July and August and included conductivity, pH, 132 KRATZER AND WARREN Downloaded by [Department Of Fisheries] at 22:29 28 February 2013

FIGURE 1. Locations of the 33 stream reaches sampled to evaluate Brook Trout biomass and habitat relationships in northeastern Vermont. [Figure available in color online.] BROOK TROUT BIOMASS IN VERMONT STREAMS 133

TABLE 1. Summary of ﬁsh, temperature, chemistry, and physical habitat metrics measured for 33 reaches in northeastern Vermont. Not all of the metrics were utilized in the generalized additive models.

Metric Description Mean SD Minimum Maximum BKT Brook Trout biomass (kg/ha) 22 22 0 123 dur20 Duration of water temperatures over 20◦C (h) 75 124 0 586 dur22 Duration of water temperatures over 22◦C (h) 16 37 0 184 cond Conductivity (µS) 47 26 16 148 pH 7.3 0.5 5.8 8.2 alk Alkalinity (mg/L CaCO3)1711240 L Reach length (m) 49 16 21 94 wetW Mean wetted width (m) 5 2 2 10 BFW Mean bankfull width (m) 10 4 4 23 % Pool Percent of reach area classiﬁed as pool 30 23 0 100 maxpool Maximum pool depth for the reach (cm) 70 25 37 125 maxriff Maximum rifﬂe depth for the reach (cm) 34 15 6 67 maxrun Maximum run depth for the reach (cm) 44 20 21 110 woodtot Total wood within the bankfull area (no./ha) 391 363 24 1,710 woodfun Wood having some perceived function (no./ha) 298 348 0 1,710 slope Water surface slope (%) 1.9 1.2 0.3 4.5 elev Elevation (m) 416 56 301 546 drain Drainage area (km2)1511250

and alkalinity. Conductivity and pH were measured in the field Trout biomass (kg/ha). The GAMs were fit using a generalized with a digital pH–conductivity meter. Alkalinity was measured cross-validation procedure using the gam function in the mgcv in the laboratory by titration (APHA 2005). Most of the alka- package in R (version 2.10.1; Wood 2004, 2006, 2008, 2010). linity values used in the study came from 2011 data, but water Before running the models, we used histograms to assess the samples from some stations were lost when the laboratory was normality of our data and log10 transformed all data that were flooded by tropical storm Irene in late August 2011, so alkalinity not normally distributed. It would be statistically inappropri- values measured in 2008 were used for those stations. ate and time-consuming to investigate all the models that could Habitat was assessed during low-flow conditions in Septem- be developed from all combinations of habitat metrics (Burn- ber using standard procedures developed by the New Hampshire ham and Anderson 2002), so we developed 24 potential models Fish and Game Department (John Magee, personal communi- based on a priori assumptions informed by observations made cation). The first step was to delineate the riffle, run, and pool in the field. We assumed that temperature and wood would be habitat units within the study reach and measure the length, the two factors most strongly related to Brook Trout biomass, typical wetted width, and typical bankfull width for each unit. so all one- and two-predictor GAMs included a water tempera- Typical widths were measured at the point that appeared to be ture metric and most three- and four-predictor GAMs included most representative of the average width of the habitat unit. one temperature metric and one wood metric. We limited the Downloaded by [Department Of Fisheries] at 22:29 28 February 2013 We used the length and width data to calculate average wet- degrees of freedom used to smooth each independent variable ted width, average bankfull width, and total wetted area for the to a maximum of three. Models were compared using Akaike’s study reach and the percent pool habitat. We also recorded the information criterion corrected for small sample sizes (AICc; maximum depths for each riffle, run, and pool. Each piece of Burnham and Anderson 2002). wood within the bankfull area having a diameter of at least 10 cm was recorded along with the habitat type (riffle, run, or pool); wood type (log, rootwad, or wood jam); and the expected function of the wood (none, pool formation, sediment retention, RESULTS fish cover, or some combination). Finally, we estimated slope Brook Trout biomass ranged from 0 to 123 kg/ha, with a mean using a clinometer, elevation using Google Earth, and drainage of 22 kg/ha (Table 1). The only station where no Brook Trout area using the U.S. Geological Survey’s StreamStats Web site were collected was the warmest one, where water temperatures (USGS 2012). exceeded 20◦C for 586 h. Water temperature never exceeded We fit multiple generalized additive models (GAMs) with 20◦C at nine stations. Alkalinity and conductivity were low, thin-plate regression splines to investigate the potential for linear with minimums of 2 mg/L CaCO3 and 16 µS, respectively. and nonlinear relationships between the various water temper- Wood densities ranged widely, from a minimum of 24 to a ature, water chemistry, and physical habitat metrics and Brook maximum of 1,710 pieces/ha. 134 KRATZER AND WARREN

TABLE 2. The generalized additive models for Brook Trout biomass ranked described the relationship between maximum rifﬂe depth and 2 by AICc, along with the R , log likelihood (logLik), and number of parameters Brook Trout biomass. (k). Included in k are the parameters in the GAM and the smoothing parameters used in the splines.

2 Model AICc R logLik k DISCUSSION The top-ranked model suggested that water temperature, dur20 + woodtot 27.19 0.766 0.97 9.8 wood density, and maximum riffle depth were all related to + maxriff Brook Trout biomass in headwater streams of northeastern Ver- dur20 + woodtot 31.36 0.762 1.13 10.8 mont. Water temperature was the controlling factor in stream + maxriff + wetW ◦ reaches where it exceeded 20 C for 200 h or more. Relatively dur20 + woodtot + cond 33.87 0.663 –6.48 7.7 low Brook Trout biomass was predicted for those reaches re- dur20 + woodtot 33.88 0.639 –8.2 6.7 gardless of wood density or riffle depth (Figure 3). For stations dur20 + woodfun 34.10 0.655 –6.95 7.5 where water temperatures were suitably cool, total wood density dur20 + woodtot + % pool 34.15 0.703 –3.14 9.5 was an important factor in accounting for Brook Trout biomass, dur20 + maxriff + wetW 34.57 0.682 –4.77 8.8 with a strong positive relationship between wood density and dur20 + woodtot 37.04 0.625 –8.24 7.6 the biomass of Brook Trout. The relationship between maxi- + maxpool mum riffle depth and Brook Trout biomass was not as strong, dur20 + woodtot + pH 37.49 0.639 –7.19 8.3 but this factor was an important component of our best model, dur22 + woodfun 39.77 0.587 –9.96 7.4 with Brook Trout biomass tending to increase with increases in dur20 + maxriff 40.31 0.589 –9.7 7.7 maximum riffle depth. dur20 40.96 0.498 –14.8 4.7 Brook trout require cold water. The incipient lethal tempera- dur20 + pool 42.03 0.558 –11.09 7.4 ture, or the temperature at which at least 50% of the test subjects dur22 + maxriff 42.16 0.535 –12.34 6.7 ◦ are expected to die, is 24.5 C (McCormick et al. 1972). Brook dur20 + cond 42.32 0.504 –14.01 5.7 ◦ trout exhibit a heat-shock response at 22 C and above (Lund dur22 + % pool 42.92 0.544 –11.71 7.3 et al. 2003). Various studies have found preferred temperatures dur22 + woodtot 43.00 0.518 –13.09 6.5 ◦ of 14–19 C (Jobling 1981). Of the two temperature metrics dur20 + pH 43.16 0.504 –13.81 6.1 ◦ ◦ (duration over 20 C and 22 C), the amount of time that water dur20 + maxpool 43.38 0.488 –14.54 5.7 ◦ temperatures exceeded 20 C was most closely related to Brook dur22 + maxriff + wetW 44.37 0.571 –9.67 8.8 Trout biomass, suggesting that preferred temperatures are more dur22 46.50 0.401 –17.99 4.4 deterministic of Brook Trout biomass in our study area than are dur22 + cond 48.46 0.400 –17.38 5.5 stressful or lethal temperatures. Water temperatures exceeded dur22 + maxpool 48.60 0.393 –17.45 5.5 ◦ 24 C at five study sites. We collected Brook Trout at the four dur22 + pH 49.06 0.445 –15.12 7.1 ◦ sites where water temperatures exceeded 24 C for 2–17 h, where it is likely that cold spring seeps allowed trout to survive periods of lethal water temperatures (Baird and Krueger 2003). No The models that included temperature and wood metrics were Brook Trout were present at the warmest site, where tempera- the best predictors of Brook Trout biomass (Table 2). The dura- tures exceeded 24◦C for 184 h. tion of water temperatures over 20◦C (dur20) was a better predic- We demonstrated that wood was indeed an important fea- tor than the duration of water temperatures over 22◦C (dur22). ture of Brook Trout habitat in this region. Rather than there Downloaded by [Department Of Fisheries] at 22:29 28 February 2013 Both wood metrics performed similarly when paired with dur20. being a strictly linear relationship between wood abundance Maximum riffle depth was included in the top-ranked model, and and fish biomass, wood’s influence on the Brook Trout popula- the increase in R2 from the two-predictor model that included tions in these streams appeared to be tied to a threshold wood only dur20 and woodtot to the three-predictor model that also density. Wood was positively and significantly related to Brook included maxriff was 0.13. Trout biomass once wood reached or surpassed a density of Brook trout biomass generally decreased with increasing wa- 100 pieces/ha. At low wood densities, the minor contributions ter temperature and increased with increasing wood density, but to pool formation or habitat complexity were not sufficient to these relationships were not linear (Figure 2). The top-ranked increase biomass above and beyond that in nonwood conditions. model predicted relatively high Brook Trout biomass for streams This may be due to influences on total fish abundance or to the in which water temperatures exceeded 20◦C for less than ap- presence or absence of a few larger fish that hold in and defend proximately 20 h but decreasing Brook Trout biomass with in- only the highly complex habitat resulting from the interaction of creasing dur20. Brook Trout biomass increased with increasing multiple pieces of wood rather than individual pieces. The po- wood density only when woodtot exceeded approximately 100 tential importance of interaction between multiple wood pieces pieces/ha, with the strongest relationship being observed when was also highlighted by the AICc analysis, which indicated that woodtot exceeded 200 pieces/ha. A straight line sufficiently a simple measure of total wood count was a better predictor of BROOK TROUT BIOMASS IN VERMONT STREAMS 135 Downloaded by [Department Of Fisheries] at 22:29 28 February 2013

FIGURE 2. Partial residual plots for the three predictors included in the top-ranked model: (a) the duration of water temperatures exceeding 20◦C(h),(b) total wood density (pieces/ha), and (c) maximum riffle depth (cm). The shaded areas represent two standard errors. The y-axis represents the effect of each metric on Brook Trout biomass, where s is the smoother term and the number in parentheses is the equivalent degrees of freedom (edf). An edf of 1.0 corresponds to a linear relationship, while larger edfs correspond to increasingly nonlinear relationships. The merging of confidence limits in panel (c), where the line passes through zero on the x-axis, is the result of the identifiability constraints applied to the smoothed terms (see Wood 2006:222). 136 KRATZER AND WARREN

limited number of wood pieces may not necessarily affect total fish densities, a large number of wood pieces may indeed lead to measureable increases in fish biomass (or abundance). White et al. (2011) found that wood additions to streams in Colorado led to a long-term increase in fish production, but those logs were explicitly placed and anchored to promote pool formation. Only a small proportion of the logs that naturally recruit to a stream may create pools, or pool creation may require wood jams rather than individual wood pieces (especially in higher-gradient systems with more stable substrates). In these cases, it makes sense that a minimum threshold of wood abundance may be needed before a clear influence of wood on fish biomass can be quantified. The greater descriptive ability of the model including total wood count rather than just the number of functional wood pieces suggests that wood function during the nonsurvey periods (e.g., during high-flow events) may persist. While much less important than water temperature and wood, riffle depth was also included as a key factor in our top model relating habitat characteristics to Brook Trout biomass. The riffle is the shallowest part of the stream, so a stream with deeper riffles should generally be deeper overall. Trout often select deepwater habitat (Berg et al. 1998), and several studies have demonstrated a relationship between the width : depth ratio and salmonid abundance (Chisholm and Hubert 1986; Lanka et al. 1987; Scarnecchia and Bergersen 1987; Kozel and Hubert 1989). While width : depth ratio was not explicitly included in our analysis, both width and depth measures were and the model with the second lowest AICc included both mean wetted width and maximum riffle depth. Depth may be a more important habitat component in larger streams, where wood is less abundant. In streams that lack wood, Brook Trout may seek other sources of cover, such as large boulders, deep water, or surface turbulence (Berg et al. 1998; Warren and Kraft 2003). In addition to providing the cover of surface turbulence, deep FIGURE 3. Brook Trout biomass predicted by the top-ranked model at various riffles may also provide preferred feeding locations for Brook ◦ durations of water temperatures exceeding 20 C (dur20 [h]) and total wood Trout (Fausch 1984; Hughes and Dill 1990; Hayes and Jowett densities at (a)–(c) three different maximum riffle depths. 1994; Baker and Coon 1997). During electrofishing surveys on the larger streams that generally lacked woody habitat, we rou- total trout biomass than the number of wood pieces that clearly tinely observed concentrations of Brook Trout in the deepest Downloaded by [Department Of Fisheries] at 22:29 28 February 2013 performed some function, such as pool formation or cover. So, riffles, and the largest Brook Trout at these study sites were despite perceived function during summer surveys, the relation- often found in these deep-riffle areas. ship between wood and trout biomass was limited at low wood Our results contrast somewhat with those of Warren et al. densities. When wood occurs at high densities, it is more likely (2010), who found pool area, cover, and stream pH to be the to interact with other wood to create a higher degree of habi- primary habitat factors accounting for the variability in total tat complexity and accumulate into a wood jam, which may be fish biomass in streams in northern New York and northern Ver- more likely to influence pool formation or provide cover than mont. As neither wood frequency nor temperature were included an individual piece of wood. in their best model, however, these results are not as contradic- This result was consistent with expectations regarding the tory as they at first appear. As with any study evaluating biotic role of wood in these glaciated streams. Wood was important responses across a gradient, the range of values within each but it was not the only feature of importance, and wood only factor can influence its importance in accounting for the vari- truly mattered when there was a lot of it. Berg et al. (1998) ability in the response metric. The pH values that we observed found that wood was not the primary preferred habitat type in had a smaller range and were generally higher than the val- streams of the Sierra Nevada Mountains, but wood was utilized ues observed in the Warren et al. (2010) study. Conversely, the to a greater degree than its availability. Therefore, although a Warren et al. (2010) study had a limited range in temperature, BROOK TROUT BIOMASS IN VERMONT STREAMS 137

with almost all streams occurring in cool, well-forested environmental Conservation. John Magee, New Hampshire Fish vironments. The lack of a strong relationship between trout and Game, developed the physical habitat assessment proto- abundance and pool area was somewhat surprising. Inconsis- col that we used. Thanks to Tyler Wagner at the Pennsylvania tency with the Warren et al. (2010) result on this point may be Cooperative Fish and Wildlife Research Unit for invaluable sta- due to the focus here on Brook Trout alone rather than Brook tistical advice. Finally, thanks to two anonymous reviewers and Trout and Slimy Sculpin Cottus cognatus together, though it is the journal editors for helpful comments during the publication more likely that within this region temperature and the com- process. Funding for this study was provided by the Federal Aid bined habitat functions of large wood have a stronger influence in Sport Fish Restoration Act and Vermont fishing license sales. on trout than pool area alone. The contrast between these two studies highlights the importance of study design and regional context when considering factors related to fish abundance or REFERENCES biomass in streams. Collecting data across a large and strong Anton,´ A., A. Elosegi, L. Garc´ıa-Arberas, J. D´ıez, and A. Rallo. 2011. Restora- habitat gradient is more likely to yield a significant result for a tion of dead wood in Basque stream channels: effects on Brown Trout popu- factor in terms of accounting for the variability in biomass or lation. Ecology of Freshwater Fish 20:461–471. abundance. However, in most practical field studies a compara- APHA (American Public Health Association). 2005. Standard methods for the ble gradient among factors is unlikely to be achieved. This is examination of water and wastewater, 21st edition. APHA, Washington, D.C. often due to regional conditions which restrict the values within Baird, O. E., and C. C. Krueger. 2003. Behavioral thermoregulation of Brook and Rainbow Trout: comparison of summer habitat use in an Adirondack river, a region (e.g., pH, which had a narrower range in our study New York. Transactions of the American Fisheries Society 132:1194–1206. area than it would across the entire northeastern USA). This is Baker, E. A., and T. G. Coon. 1997. Development and evaluation of alternative not necessarily detrimental to studies of fish ecology and fish habitat suitability criteria for Brook Trout. Transactions of the American management, but it does highlight the need to consider regional Fisheries Society 126:65–76. conditions when evaluating the relative importance of various Baker, J. P., J. Van Sickle, C. J. Gagen, D. R. DeWalle, W. E. Sharpe, R. F. Carline, B. P. Baldigo, P. S. Murdoch, D. W. Bath, W. A. Kretser, habitat characteristics. Fisheries managers in other regions are H. A. Simonin, and P. J. Wigington Jr. 1996. Episodic acidification of small likely to observe that Brook Trout abundance is most strongly streams in the northeastern United States: effects on fish populations. Eco- related to the habitat features in their region that have the largest logical Applications 6:422–437. variability. Baldigo, B. P., and G. B. Lawrence. 2000. Composition of fish communities in We drew two primary conclusions. First, trout biomass in relation to stream acidification and habitat in the Neversink River, New York. Transactions of the American Fisheries Society 129:60–76. northeastern Vermont’s headwater streams was strongly associ- Berg, N., A. Carlson, and D. Azuma. 1998. Function and dynamics of woody ated with stream temperature, high wood density, and deepwater debris in stream reaches in the central Sierra Nevada, California. Canadian habitat. Management activities that promote cooler temperatures Journal of Fisheries and Aquatic Sciences 55:1807–1820. (e.g., planting riparian shade trees), increase wood abundance Burgess, S. A., and J. R. Bider. 1980. Effects of stream habitat improvements (e.g., chop-and-drop wood additions), or create deeper habitat on invertebrates, trout populations, and mink activity. Journal of Wildlife Management 44:871–880. (e.g., wing deflectors and V-weirs) would probably increase to- Burnham, K. P., and D. R. Anderson. 2002. Model selection and multimodel tal Brook Trout biomass in these streams. The abiotic stream inference: a practical information-theoretic approach, 2nd edition. Springer- characteristics accounting for the greatest variability in trout Verlag, New York. biomass differed somewhat from the most important factors Butryn, R. S. 2010. Summer stream temperature as an indicator of coldwater identified in an earlier study in this region (Warren et al. 2010) fish distribution. Master’s thesis. University of Vermont, Burlington. Carle, F. L., and M. R. Strub. 1978. A new method for estimating population but are broadly consistent with the results of other studies (Berg size from removal data. Biometrics 34:621–630. et al. 1998; Poff and Huryn 1998; Rosenfeld 2003; Rosenfeld Chisholm, I. M., and W. A. Hubert. 1986. Influence of stream gradient on Downloaded by [Department Of Fisheries] at 22:29 28 February 2013 and Taylor 2009; Minns et al. 2011). This led us to our sec- standing stock of Brook Trout in the Snowy Range, Wyoming. Northwest ond conclusion: the habitat factors that limit fish abundance or Science 60:137–139. biomass may be highly variable between regions, and conclu- Chu, C., N. E. Jones, and L. Allin. 2010. Linking the thermal regimes of streams in the Great Lakes basin, Ontario, to landscape and climate variables. River sions about which factors are important may also be influenced Research and Applications 26:221–241. by study design or, more specifically, the range of chemical and Coleman, M. A., and K. D. Fausch. 2007a. Cold summer temperature limits physical attributes included in the study. recruitment of age-0 Cutthroat Trout in high-elevation Colorado streams. Transactions of the American Fisheries Society 136:1231–1244. Coleman, M. A., and K. D. Fausch. 2007b. Cold summer temperature regimes cause a recruitment bottleneck in age-0 Colorado River Cutthroat Trout ACKNOWLEDGMENTS reared in laboratory streams. Transactions of the American Fisheries Society We thank the many people who helped collect data for this 136:639–654. study. Vermont Fish and Wildlife staff members Pete Emerson, Crook, D. A., and A. I. Robertson. 1999. Relationships between riverine fish Kandis Barrett, and Ann Hare and Trout Unlimited staff mem- and woody debris: implications for lowland rivers. Marine and Freshwater Research 50:941–953. bers Joe Norton, Joe Martell, and Ben Matthews assisted with Evans, D. M., W. M. Aust, C. A. Dolloff, B. S. Templeton, and J. A. Peterson. the collection of the fish and habitat data. Alkalinity titrations 2011. Eastern hemlock decline in riparian areas from Maine to Alabama. were performed by Jim Kellogg, Vermont Department of En- Northern Journal of Applied Forestry 28:97–104. 138 KRATZER AND WARREN

Fausch, K. D. 1984. Profitable stream positions for salmonids: relating spe- Nagayama, S., and F. Nakamura. 2010. Fish habitat rehabilitation using wood cific growth rate to net energy gain. Canadian Journal of Zoology 62: in the world. Landscape and Ecological Engineering 6:289–305. 441–451. Poff, N. L., and A. D. Huryn. 1998. Multi-scale determinants of secondary Flebbe, P. A., and C. A. Dolloff. 1995. Trout use of woody debris and habitat in production in Atlantic Salmon (Salmo salar) streams. Canadian Journal of Appalachian wilderness streams of North Carolina. North American Journal Fisheries and Aquatic Sciences 55(Supplement 1):201–217. of Fisheries Management 15:579–590. Roni, P., T. J. Beechie, R. E. Bilby, F. E. Leonetti, M. M. Pollock, and G. R. Pess. Gregory, S. V., K. L. Boyer, and A. M. Gurnell, editors. 2003. The ecology and 2002. A review of stream restoration techniques and a hierarchical strategy management of wood in world rivers. American Fisheries Society, Sympo- for prioritizing restoration in Pacific Northwest watersheds. North American sium 37, Bethesda, Maryland. Journal of Fisheries Management 22:1–20. Gurnell, A. M., H. Piegay,´ F. J. Swanson, and S. V. Gregory. 2002. Large wood Roni, P., and T. P. Quinn. 2001. Density and size of juvenile salmonids in re- and fluvial processes. Freshwater Biology 47:601–619. sponse to placement of large woody debris in western Oregon and Washington Gurnell, A. M., K. Tockner, P. Edwards, and G. Petts. 2005. Effects of de- streams. Canadian Journal of Fisheries and Aquatic Sciences 58:282–292. posited wood on biocomplexity of river corridors. Frontiers in Ecology and Rosenfeld, J. 2003. Assessing the habitat requirements of stream fishes: an the Environment 3:377–382. overview and evaluation of different approaches. Transactions of the Ameri- Hayes, J. W., and I. G. Jowett. 1994. Microhabitat models of large drift-feeding can Fisheries Society 132:953–968. Brown Trout in three New Zealand rivers. North American Journal of Fish- Rosenfeld, J., and J. Taylor. 2009. Prey abundance, channel structure and the eries Management 14:710–725. allometry of growth rate potential for juvenile trout. Fisheries Management Hughes, N. F., and L. M. Dill. 1990. Position choice by drift-feeding salmonids: and Ecology 16:202–218. model and test for Arctic Grayling (Thymallus arcticus) in subarctic mountain Scarnecchia, D. L., and E. P. Bergersen. 1987. Trout production and standing streams, interior Alaska. Canadian Journal of Fisheries and Aquatic Sciences crop in Colorado’s small streams, as related to environmental features. North 47:2039–2048. American Journal of Fisheries Management 7:315–330. Jobling, M. 1981. Temperature tolerance and the final preferendum—rapid Shelford, V. E. 1913. Animal communities in temperate America. University of methods for the assessment of optimum growth temperatures. Journal of Chicago Press, Chicago. Fish Biology 19:439–455. Siitari, K. J., W. W. Taylor, S. A. C. Nelson, and K. E. Weaver. 2011. The Kail, J., and D. Hering. 2005. Using large wood to restore streams in Central influence of land cover composition and groundwater on thermal habitat Europe: potential use and likely effects. Landscape Ecology 20:755–772. availability for Brook Charr (Salvelinus fontinalis) populations in the United Keeton, W. S., C. E. Kraft, and D. R. Warren. 2007. Mature and old-growth ripar- States of America. Ecology of Freshwater Fish 20:431–437. ian forests: structure, dynamics, and effects on Adirondack stream habitats. Solazzi, M. F., T. E. Nickelson, S. L. Johnson, and J. D. Rodgers. 2000. Effects Ecological Applications 17:852–868. of increasing winter rearing habitat on abundance of salmonids in two coastal Kocovsky, P. M., and R. F. Carline. 2005. Stream pH as an abiotic gradient Oregon streams. Canadian Journal of Fisheries and Aquatic Sciences 57:906– influencing distributions of trout in Pennsylvania streams. Transactions of 914. the American Fisheries Society 134:1299–1312. Stewart, G. B., H. R. Bayliss, D. A. Showler, W. J. Sutherland, and A. S. Pullin. Kozel, S. J., and W. A. Hubert. 1989. Testing of habitat assessment models for 2009. Effectiveness of engineered in-stream structure mitigation measures to small trout streams in the Medicine Bow National Forest, Wyoming. North increase salmonid abundance: a systematic review. Ecological Applications American Journal of Fisheries Management 9:458–464. 19:931–941. Lanka, R. P.,W. A. Hubert, and T. A. Wesche. 1987. Relations of geomorphology Sundbaum, K., and I. Naslund.¨ 1998. Effects of woody debris on the growth to stream habitat and trout standing stock in small Rocky Mountain streams. and behaviour of Brown Trout in experimental stream channels. Canadian Transactions of the American Fisheries Society 116:21–28. Journal of Zoology 76:56–61. Lund, S. G., M. E. A. Lund, and B. L. Tufts. 2003. Red blood cell Hsp 70 Thompson, D. M. 2006. Did the pre-1980 use of in-stream structures improve mRNA and protein as bio-indicators of temperature stress in the Brook Trout streams? a reanalysis of historical data. Ecological Applications 16:784–796. (Salvelinus fontinalis). Canadian Journal of Fisheries and Aquatic Sciences Todd, A. S., M. A. Coleman, A. M. Konowal, M. K. May, S. Johnson, N. K. 60:460–470. M. Vieira, and J. F. Saunders. 2008. Development of new water temperature McCormick, J. H., K. E. F. Hokanson, and B. R. Jones. 1972. Effects of tem- criteria to protect Colorado’s fisheries. Fisheries 33:433–443. perature on growth and survival of young Brook Trout, Salvelinus fontinalis. USGS (U.S. Geological Survey). 2012. StreamStats for Vermont. USGS, Journal of the Fisheries Research Board of Canada 29:1107–1112. Reston, Virginia. Available: water.usgs.gov/osw/streamstats/Vermont.html. McKenna, J. E., Jr., and J. H. Johnson. 2011. Landscape models of Brook (July 2012). Trout abundance and distribution in lotic habitat with field validation. North van Zyll De Jong, M. C., I. B. Cowx, and D. A. Scruton. 1997. An evaluation Downloaded by [Department Of Fisheries] at 22:29 28 February 2013 American Journal of Fisheries Management 31:742–756. of instream habitat restoration techniques on salmonid populations in a New- Minns, C. K., R. G. Randall, K. E. Smokorowski, K. D. Clarke, A. Velez-´ foundland stream. Regulated Rivers: Research and Management 13:603–614. Espino, R. S. Gregory, S. Courtenay, and P.LeBlanc. 2011. Direct and indirect Waco, K. E., and W. W. Taylor. 2010. The influence of groundwater withdrawal estimates of the productive capacity of fish habitat under Canada’s policy for and land use changes on Brook Charr (Salvelinus fontinalis) thermal habitat the management of fish habitat: where have we been, where are we now, and in two coldwater tributaries in Michigan, USA. Hydrobiologia 650:101–116. where are we going? Canadian Journal of Fisheries and Aquatic Sciences Warren, D. R., and C. E. Kraft. 2003. Brook Trout (Salvelinus fontinalis)re- 68:2204–2227. sponse to wood removal from high-gradient streams of the Adirondack Moun- Montgomery, D. R., J. M. Buffington, R. D. Smith, K. M. Schmidt, and G. Pess. tains (N.Y., U.S.A.). Canadian Journal of Fisheries and Aquatic Sciences 1995. Pool spacing in forest channels. Water Resources Research 31:1097– 60:379–389. 1105. Warren, D. R., C. E. Kraft, W. S. Keeton, J. S. Nunery, and G. E. Likens. 2009. Mullner, S. A., and W. A. Hubert. 2005. Low summer water temperatures Dynamics of wood recruitment in streams of the northeastern US. Forest influence occurrence of naturalized salmonids across a mountain watershed. Ecology and Management 258:804–813. North American Journal of Fisheries Management 25:1034–1040. Warren, D. R., M. M. Mineau, E. J. Ward, and C. E. Kraft. 2010. Relating fish Nagayama, S., Y. Kawaguchi, D. Nakano, and F. Nakamura. 2008. Methods for biomass to habitat and chemistry in headwater streams of the northeastern and fish responses to channel remeandering and large wood structure place- United States. Environmental Biology of Fishes 88:51–62. ment in the Shibetsu River restoration project in northern Japan. Landscape Wenger, S. J., D. J. Isaak, C. H. Luce, H. M. Neville, K. D. Fausch, J. B. Dunham, and Ecological Engineering 4:69–74. D. C. Dauwalter, M. K. Young, M. M. Elsner, B. E. Rieman, A. F. Hamlet, BROOK TROUT BIOMASS IN VERMONT STREAMS 139

and J. E. Williams. 2011. Flow regime, temperature, and biotic interactions Wood, S. N. 2008. Fast stable direct fitting and smoothness selection for gen- drive differential declines of trout species under climate change. Proceedings eralized additive models. Journal of the Royal Statistical Society B 70: of the National Academy of Sciences of the USA 108:14175–14180. 495–518. White, S. L., C. Gowan, K. D. Fausch, J. G. Harris, and W. C. Saunders. 2011. Wood, S. N. 2010. GAMs with GCV/AIC/REML smoothness estimation and Response of trout populations in five Colorado streams two decades after GAMMs by PQL: package ‘mgcv’. R Foundation, Vienna. Available: cran.r- habitat manipulation. Canadian Journal of Fisheries and Aquatic Sciences project.org/web/packages/mgcv/mgcv.pdf. (May 2010). 68:2057–2063. Xu, C. L., B. H. Letcher, and K. H. Nislow. 2010a. Context-specific influence Wood, S. N. 2004. Stable and efficient multiple smoothing parameter estimation of water temperature on Brook Trout growth rates in the field. Freshwater for generalized additive models. Journal of the American Statistical Associ- Biology 55:2253–2264. ation 99:673–686. Xu, C. L., B. H. Letcher, and K. H. Nislow. 2010b. Size-dependent survival Wood, S. N. 2006. Generalized additive models: an introduction with R. Chap- of Brook Trout Salvelinus fontinalis in summer: effects of water temperature man and Hall/CRC Press, Boca Raton, Florida. and stream flow. Journal of Fish Biology 76:2342–2369. Downloaded by [Department Of Fisheries] at 22:29 28 February 2013 This article was downloaded by: [Department Of Fisheries] On: 28 February 2013, At: 22:30 Publisher: Taylor & Francis Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

North American Journal of Fisheries Management Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/ujfm20 Growth per Molt of Snow Crab in the Eastern Bering Sea David Somerton a , Scott Goodman b , Robert Foy c , Lou Rugolo c & Laura Slater d a Groundfish Assessment Program, Alaska Fisheries Science Center, 7600 Sand Point Way Northeast, Seattle, Washington, 98125, USA b Natural Resources Consultants, Incorporated, 4039 21st Avenue West, Seattle, Washington, 98199, USA c Kodiak Laboratory, Alaska Fisheries Science Center, 301 Research Court, Kodiak, Alaska, 99615, USA d Alaska Department of Fish and Game, Commercial Fisheries Division, 211 Mission Road, Kodiak, Alaska, 99615, USA Version of record first published: 25 Jan 2013.

To cite this article: David Somerton , Scott Goodman , Robert Foy , Lou Rugolo & Laura Slater (2013): Growth per Molt of Snow Crab in the Eastern Bering Sea, North American Journal of Fisheries Management, 33:1, 140-147 To link to this article: http://dx.doi.org/10.1080/02755947.2012.732671

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material. North American Journal of Fisheries Management 33:140–147, 2013 C American Fisheries Society 2013 ISSN: 0275-5947 print / 1548-8675 online DOI: 10.1080/02755947.2012.732671

ARTICLE

Growth per Molt of Snow Crab in the Eastern Bering Sea

David Somerton* Groundﬁsh Assessment Program, Alaska Fisheries Science Center, 7600 Sand Point Way Northeast, Seattle, Washington 98125, USA Scott Goodman Natural Resources Consultants, Incorporated, 4039 21st Avenue West, Seattle, Washington 98199, USA Robert Foy and Lou Rugolo Kodiak Laboratory, Alaska Fisheries Science Center, 301 Research Court, Kodiak, Alaska 99615, USA Laura Slater Alaska Department of Fish and Game, Commercial Fisheries Division, 211 Mission Road, Kodiak, Alaska 99615, USA

Abstract The growth per molt of snow crab Chionoecetes opilio in the eastern Bering Sea is poorly known, primarily because snow crabs are difﬁcult to sample during their spring molting period due to the presence of sea ice. To provide better estimates of growth per molt for the stock assessment model used to manage the snow crab ﬁshery, a study was conducted to collect juvenile crabs with a bottom trawler and return them to holding facilities at Dutch Harbor and Kodiak, Alaska. Since relatively few of these crabs subsequently molted, these data were augmented with molting data collected by two previous, unpublished studies. Based on a total of 35 observations, snow crabs grew according to the following premolt (PRMW) and postmolt (POMW) carapace width relationships: POMW =−4.00 + 1.46(PRMW), where PRMW < 36.1 mm, and POMW = 48.83 + 1.17(PRMW − 36.1), where PRMW ≥ 36.1 mm.

Snow crabs Chionoecetes opilio range from the northwest At- poorly known because snow crabs occupy areas that are remote lantic Ocean, through the Arctic to both the eastern and western from research facilities and often covered by sea ice during the

Downloaded by [Department Of Fisheries] at 22:30 28 February 2013 North Pacific Ocean (Elner 1982; Rand and Logerwell 2011). In spring molting period. Alaskan waters, snow crabs supports a commercial fishery in the Prior studies of snow crab growth and that of its Alaskan con- eastern Bering Sea that produced landings peaking at 149,000 gener the Tanner crab C. bairdi have used two approaches to es- metric tons in 1991 but in recent years (2008–2010) has aver- timate growth per molt. The first is based on the identification of aged about 25,000 metric tons (Turnock and Rugolo 2010). The molt classes or instars in width-frequency distributions of crabs stock assessment model used in the management of this fish- collected with bottom trawls (Sainte-Marie et al. 1995; Alunno- ery (Turnock and Rugolo 2010) considers the growth of snow Bruscia and Sainte-Marie 1998). Although the eastern Bering crabs to be a function of the growth in carapace width gained Sea snow crab population is sampled annually each summer by at each molt and the frequency of molting. While the growth the National Marine Fisheries Service bottom trawl survey, the per molt has been well documented for Atlantic populations catchability of the trawl for small crabs is so low (Somerton and (Sainte-Marie et al. 1995; Alunno-Bruscia and Sainte-Marie Otto 1999) that width-frequency modes are not clearly defined 1998; Comeau et al. 1998), in the eastern Bering Sea it is still over much of the juvenile width range. The second approach is

*Corresponding author: [email protected] Received July 23, 2012; accepted September 13, 2012

140 SNOW CRAB GROWTH IN THE BERING SEA 141

based on the collection of crabs just prior to their molt and the (BSFRF) to collect juvenile snow crabs during their spring molt- maintenance of the crabs in captivity until molting is completed ing period and transport them from the collecting areas in the (Donaldson et al. 1981; Sainte-Marie et al. 1995; Stone et al. northwest to holding facilities at both Dutch Harbor and Ko- 2003), which is usually applied to molting populations situated diak, Alaska, to be maintained there until they molted. The data fairly close to holding facilities. from this field collection were then compared with and, in some Indication that the second approach might be possible for cases, combined with unpublished data from prior studies. The eastern Bering Sea snow crabs, despite the sampling complica- combined data were then analyzed to produce a growth-per-molt tions posed by sea ice, was provided by a spring trawl sampling function that allows prediction of postmolt carapace width from study for mature females that captured some premolt juvenile premolt width and that is more suitable for use in the snow crab males (Rugolo et al. 2005), 14 of which subsequently molted. stock assessment model. Although not all of these crabs were allowed to harden sufficiently to obtain good postmolt width measurements, these were the first direct snow crab growth observations in the east- METHODS ern Bering Sea and were subsequently used to estimate the Sample collection.—Immature snow crabs were collected growth-per-molt function in the snow crab stock assessment north of the Pribilof Islands (Figure 1) on 11–13 April 2011 model (Turnock and Rugolo 2010). aboard a BSFRF-chartered commercial stern trawler, Half Moon Because this function is so important for determining the an- Bay, by probing the seaward edge of the ice cover for available nual catch limits of snow crabs but was so weakly supported by trawling sites and using short (about 5-min) tows of a Nephrops the available data, a 2011 cooperative project was developed by trawl (Conan et al. 1994) modified by the addition of weight the Alaska Fisheries Science Center of the National Marine Fish- on the footrope to increase the catchability of small individuals. eries Service (NMFS), the Alaska Department of Fish and Game The snow crabs caught in each tow were sorted by sex and (ADF&G), and the Bering Sea Fisheries Research Foundation 10-mm carapace width categories and then placed into mesh Downloaded by [Department Of Fisheries] at 22:30 28 February 2013

FIGURE 1. Collection locations of the immature snow crabs used in the transit study, Dutch Harbor holding study, and National Marine Fisheries Service Kodiak holding study. 142 SOMERTON ET AL.

TABLE 1. Number of immature snow crabs collected from the eastern Bering 1999), then the carapace width (excluding lateral spines) of the Sea for growth observations during the April 2011 sampling trip, by sex and each crab and its cast exoskeleton were measured to the nearest width. 0.1 mm using digital calipers. In addition, the sex and morpho- Number collected metric maturity of the crab were then determined. For females, maturity was recognized by a large increase in abdominal width Carapace width (mm) Males Females relative to carapace width, while for males, maturity was rec- 20–29 154 204 ognized by a large increase in chela height relative to carapace 30–39 119 157 width (Jadamec et al. 1999). The data from the transit study 40–49 25 79 consisted of 14 crabs, including the three pseudo-individuals 50–59 60 20 representing the means of multiple observations. 60–69 52 55 Dutch Harbor holding study.—After reaching the holding 70–79 18 0 facility in Dutch Harbor on April 15, the snow crabs were trans- 80–89 66 0 ferred to lantern nets and maintained without feeding for 30 d 90–99 36 0 or until molting or death. Each lantern net consisted of a stack 100–109 4 0 of 10 circular compartments 48 cm in diameter and 20 cm high. Eight of the nets were shortened from 2.1 to 1.2 m in length, and Total 534 515 then placed horizontally in a stainless steel tank (2.6 × 1.5 × 1.0 m), with flow-through seawater. The tanks were located outside on a dock. In addition, because of space limitations in the bags maintained in deck tanks with circulating seawater. The tank, the two remaining lantern nets were hung vertically off the objective of this sampling was to collect crabs that were in a dock and allowed to expand to their full length. Each of the 100 stage just prior to their molt, which can be recognized by a compartments was loaded with two crabs sufficiently different greenish tint to the carapace (Hoenig et al. 1994). However, in width to allow an unambiguous match of a crab with its cast crabs in some categories of sex and width did not display this exoskeleton. Both sexes and the full width range collected at characteristic but were still retained in the hope that molting sea were included in the holding experiment, even though not might occur during the holding experiment (Table 1). all of the crabs displayed the greenish color indicative of an im- The intended design of the molting experiment was to trans- minent molt. The nets were monitored almost daily by ADF&G port the crabs to Dutch Harbor, where half would be kept in staff and when a molting crab was identified it was allowed 3 outdoor live tanks for up to 1 month without feeding and half d to harden, then measured as above. The data from the Dutch would be shipped to the NMFS Kodiak Laboratory, where they Harbor holding study included nine crabs. would be kept with daily feeding in indoor tanks under more NMFS Kodiak holding study.—In addition to the Dutch Har- controlled conditions; we were unsure which treatment would bor holding experiment, 200 snow crabs with approximately be more effective. For a variety of reasons, the handling and the same carapace width and sex proportions as those kept in stress that the crabs experienced differed, so we considered the Dutch Harbor were shipped (air freight) to Kodiak, Alaska, in growth observations as if they came from three separate studies: wet burlap and maintained in tanks with flow-through seawater the transit study, the Dutch Harbor holding study, and the NMFS at the Kodiak Fisheries Research Center. Unlike the Dutch Har- Kodiak holding study. bor experiment, the crabs were fed daily and maintained longer The transit study.—During transport from the collection ar- than the intended 30-d duration of the experiment. As with the eas to the holding facilities, some snow crabs began molting. Dutch Harbor crabs, Kodiak crabs were allowed 3 d to harden Downloaded by [Department Of Fisheries] at 22:30 28 February 2013 Those recognized to be in the early stages of the molt were before their carapace widths were measured identically to those placed in individual cages to complete the molt in isolation, but in the Dutch Harbor holding experiment. These data from the others completed the molt while in the mesh bags with other NMFS Kodiak holding study included 25 crabs. crabs. When only a single crab or several easily identified crabs Because the total number of snow crabs molting in the above (e.g., differences in size, missing appendages or sex) had com- experiments was quite small, we also considered data from the pleted molting the newly molted crabs were matched with their following two previous unpublished growth-per-molt studies on cast exoskeletons. However, when the crabs in a mesh bag were eastern Bering Sea snow crabs. so similar in appearance that they could not be unambiguously Cooperative seasonal study.—As part of an ADF&G and matched to their cast exoskeletons, the premolt and postmolt NMFS cooperative seasonality study of snow crab reproductive widths of all unmatched crabs were averaged and subsequently biology in the eastern Bering Sea (Rugolo et al. 2005), premolt used to represent the premolt and postmolt widths of a single crabs were captured with a bottom trawl during March 2003 crab. This happened in three mesh bags, with sample sizes of and brought back to Dutch Harbor, where they were maintained three, eight, and three crabs, and in all cases the crabs were the briefly (<5 d), without feeding, in tanks supplied with flow- same sex and almost identical in size. Molted crabs were allowed through seawater. During both transit and holding, 14 of these at least 3 d to fully expand and partially harden (Jadamec et al. crabs molted and were subsequently measured, but not with a SNOW CRAB GROWTH IN THE BERING SEA 143

standardized postmolt waiting period. For our study, only crabs to as the Hiatt model (Hiatt 1948), which has been used in previ- that were allowed to harden for 3 d were considered for analysis. ous growth studies of Chionoecetes (Somerton 1980; Donaldson The data from the cooperative seasonality study included seven et al. 1981; Sainte-Marie et al. 1995; Stone et al. 2003). This crabs. model, which consists of two straight lines with an intersection ADF&G Kodiak holding study.—Beginning in 2007, point near the premolt size where immature crabs develop their ADF&G scientists collected immature snow crabs during the gonads and transition to adolescents (Sainte-Marie et al. 1995), summer NMFS eastern Bering Sea bottom trawl survey (Stauf- can be expressed as fer 2004). Crabs were held in the vessel’s seawater hold or in ∗ live tanks on the deck until the vessel returned to Dutch Harbor, Yi = A + BX i +∈i Xi ≤ X (1) where crabs were then shipped via air freight to Kodiak. These ∗ ∗ ∗ Yi = Y + D(Xi − X )+∈i Xi > X , (2) crabs were subsequently maintained in individual cages within tanks with flow-through seawater chilled to about 3◦C and fed where X is the premolt width (mm), Y is the postmolt width twice weekly. Crabs were initially measured for carapace width, i i (mm), ∈ is a normal random error, and X*istheX value and those surviving to molt 6–9 months later were remeasured i and Y*= A + BX*istheY value at the intersection point. after a waiting period of at least 3 d after the molt. The data Because we did not have prior knowledge of X*, we used an from the ADF&G holding study included 41 crabs. iterative model-fitting approach described in Somerton (1980). Choosing the best data.—Before beginning the analysis, we Briefly this method consists of first choosing a range of pre- excluded the data for some snow crabs to create a more ho- molt carapace widths likely to include the intersection point, mogenous data set. First, we excluded observations from fe- then iteratively fitting a segmented straight line model (linear males molting to maturity (terminal molt) because previous regression; function lm; R Development Core Team 2008) with studies have shown that the growth increment gained at the a constrained intersection point progressively increased by 0.1- maturity molt is significantly less than an immature molt at the mm increments across this range. The best parameter estimates same premolt width (Moriyasu et al. 1987; Alunno-Bruscia and (A, B, D, and X*) are then chosen as those producing the lowest Sainte-Marie 1998; L. Slater, unpublished data) and we had value of the residual sum of squares. too few terminal molting females to adequately define these in- After the model was fitted to the highest-ranked data set, crements. However, we retained the data for males molting to determining the consistency of lower-ranked data and pooling maturity because previous studies indicated either no difference of consistent data were done iteratively as follows. First, the data (Sainte-Marie et al. 1995) or only a slight reduction in growth at from the next lower ranked study were pooled with the original the terminal molt (Moriyasu et al. 1987). Second, we excluded data and then the model was fit to the combined data. Second, 11 observations for crabs missing more than two limbs because, the residual sum of squares about the model fit to the combined for Tanner crabs, limb loss has been found to have a negative data (RSS ) and the original data (RSS ) were compared by effect on growth increment (Stone et al. 2003). Additionally, 2 1 calculating the quotient male and female growth data were combined because, except for the female maturity molt, previous studies were not able to − RSS2 RSS1 / − detect a sexual difference in growth per molt (Moriyasu et al. (N2 N1) , RSS1/ 1987). (N1) − 4 Since none of the above studies produced sufficient data to adequately describe a growth-per-molt function by itself, we which is distributed as an F-statistic with N2 − N1 and N1 − examined the potential for pooling data sets by examining their 4 degrees of freedom and where N1 and N2 are the numbers of Downloaded by [Department Of Fisheries] at 22:30 28 February 2013 consistency to each other. The data sets differ in the length of crabs in the two data sets (Montgomery and Peck 1982). Third, time that occurred between capture and molting and the amount if any test was not significant, the data were pooled; otherwise of handling each crab received. Previous studies on Tanner crabs the data were skipped, and the comparison progressed to the indicated that such factors could have a negative effect on growth next lesser ranked data. For studies where the data were signif- per molt (Stone et al. 2003); we therefore ranked the data sets icantly different, the deviations of the test data from the model based on the premise that the crabs receiving the least handling predictions, expressed as percent bias, were plotted against the should produce growth-per-molt values most representative of elapsed time since the start of the experiment. natural conditions. From this perspective, the data collected in The error about the final model predicting the expected mean the transit study was considered the best, followed in sequence postmolt width for each premolt width was estimated using by the cooperative seasonal study, the Dutch Harbor holding bootstrapping (Efron and Tibshirani 1993). This was done by study, the NMFS Kodiak holding study, and the ADF&G holding randomly subsampling individual crabs, with replacement, until study. the sample size was the same as the original, then fitting the Fitting a statistical model to the data.—The statistical model segmented-line model to the sampled observations. Repeating we used to describe the relationship between premolt and post- this process 500 times, produced 500 estimates of the predicted molt carapace width was a segmented straight line, often referred mean postmolt width for each premolt width. The error in 144 SOMERTON ET AL.

the predicted mean was then estimated as the variance of the Kodiak holding data were significantly different from the pooled replicates at each premolt width. In addition, for graphical data (P < 0.001; Table 2; Figure 2D) and were excluded from purposes, the empirical 95% confidence intervals (CIs) were the model. calculated by sorting the replicates at each premolt width and The best-fitting model to the selected data (35 values) had then choosing the 12th and 488th elements of the sorted array an intersection point at a premolt width (POMW) of 36.1 mm as the lower and upper CI bounds. (Figure 2D) and produced an R2 of 99.9%. The fitted model and In addition to the expected variance of future postmolt width, estimated parameter values when the premolt width (PRMW) < we also estimated the variance of the predicted postmolt width 36.1 mm was observations at each premolt width. This was done by adding the variance of the observations about the fitted model to the vari- POMW =−4.00 + 1.46 × PRMW ; ance of the mean calculated above (Draper and Smith, 1981). Before this addition, we linearly regressed the squared devia- when PRMW ≥ 36.1 mm the resulting equation was tions of the observations against premolt width and tested the slope for significance, to determine whether the variance about POMW = 48.83 + 1.17(PRMW − 36.1). the model changed with premolt width. Since the slope was not significant (P = 0.40), this component of variance was consid- The uncertainty in the predicted mean postmolt width as a ered a constant. The resulting error, expressed as the standard function of premolt width, expressed as the 95% CI, is shown deviation, is needed in the snow crab stock assessment model in Figure 3. The uncertainty in the predicted postmolt width to calculate the postmolt width probability distribution for each observations as a function of premolt width, expressed as the premolt width. standard deviation, is shown in Figure 4. To examine the effect of holding time on the growth per molt, the percent bias of the NMFS Kodiak holding data was RESULTS plotted against the elapsed time between capture and molting The data selection process proceeded as follows: The coop- (Figure 5). The bias is initially fairly stable but soon increases erative seasonal data were not significantly different from the with time, and after about 40 d, reaches an apparent asymptote transit data (P = 0.115; Table 2; Figure 2A) and were pooled. (excluding the one extreme value) of about −9%. Percent bias The Dutch Harbor holding data were not significantly different was similarly calculated for the ADF&G Kodiak holding data; from the previously pooled data (P = 0.264; Table 2; Figure 2B) however, because the holding times were so much longer, only and were also pooled. The NMFS Kodiak holding data were sig- the mean bias was calculated. This bias was −8.2%. nificantly different from the pooled data (P < 0.001; Table 2) when all of the data were included, but were not significantly different (P = 0.192; Table 2; Figure 2C) when the maximum DISCUSSION holding time was limited to 30 d (i.e., the maximum holding The five sets of snow crab growth data we examined dif- time of the Dutch Harbor holding study). Finally, the ADF&G fered in the length of time crabs were held before molting and whether they were fed during this time. Growth holding studies conducted for Tanner crabs, which provided the model for this TABLE 2. Test for inclusion of snow crab molting data sets in a common study, also differed in these respects. While Donaldson et al. pool. The data sets are ranked from best to worst based on the relative amount (1981) had a maximum holding duration of 7 d and did not sup- of handling (length of time in captivity, shipping stress, provision of food, etc.) ply food, Stone et al. (2003) had a maximum holding duration that crabs received before molting. Included are the numbers of observations

Downloaded by [Department Of Fisheries] at 22:30 28 February 2013 (N) and the significance of the test of similarity to the pooled observations. of 33 d and did supply food, although all crabs subsequently When the significance was ≥0.05, the data set was pooled with the (previously molting did not eat. Stone et al. (2003) found that the growth accepted) data sets above it in the list. When the significance was <0.05, the data increment was negatively affected by increasing holding time set was excluded and the next study on the list was tested. The NMFS Kodiak but considered the effect to be small for holding times of less holding study was first tested including all data on molted crabs regardless of than 14 d. This reduction in growth increment with increasing the length of time in captivity, then again by only considering crabs in captivity forlessthan30d. holding time was also apparent in the data we examined. The NMFS holding study data were significantly different from the Data set N Significance (P) previously pooled data (Table 2), but not when the maximum holding time was restricted to 30 d, the same as the maximum Transit study 14 holding time used in the Dutch Harbor holding study. When Cooperative seasonality study 6 0.115 the percent bias of all NMFS holding study data, relative to the Dutch Harbor holding study 9 0.264 previously fitted model, was plotted against holding time (Fig- NMFS Kodiak holding study 25 <0.001 ure 5), it increased with time to an apparent asymptote of about NMFS Kodiak holding study (t < 30 d) 6 0.192 −9% at about 40 d. Likewise, percent bias of the ADF&G hold- ADF&G Kodiak holding study 41 <0.001 ing study data, which had considerably longer holding times, SNOW CRAB GROWTH IN THE BERING SEA 145

FIGURE 2. Comparison of snow crab premolt versus postmolt carapace widths from the different studies. Each panel shows an accepted set of data (open circles),

Downloaded by [Department Of Fisheries] at 22:30 28 February 2013 the fit of a segmented model to the accepted set of data, and a set of data being tested for consistency (filled circles). In panel (A) the accepted set is the transit data (TS) and the test set is the cooperative seasonality data (CSS); in (B) the accepted set is the pooled TS and CSS data and the test set is the Dutch Harbor holding data (DHHS); in (C) the accepted set is the pooled TS, CSS, and DHHS data and the test set is the NMFS Kodiak holding data (NKHS) restricted to holding times of ≤30 d; and In (D) the accepted set is the pooled TS, CSS, DHHS, and NKHS data and the test set is the AFD&G holding data. The fitted model in panel (D) is the final eastern Bering Sea growth-per-molt model.

was nearly the same at −8.2%. This indicates that long-term ern Bering Sea snow crabs so that collection for growth studies captivity of these crabs resulted in a reduced growth increment, could be done under less rigorous conditions than is currently and that this holding effect requires a month or more to fully possible. develop. In contrast to this, Sainte-Marie et al. (1995) found no Besides the summary estimates from tagging studies reported apparent reduction in growth per molt for Canadian snow crabs in McBride (1982) and the few observations reported in Rugolo held in captivity with feeding until they molted. Although our et al. (2005) that are used in this study, there are no previous esti- analysis supports the notion that long-term holding suppresses mates of snow crab growth per molt in Alaskan waters. However, growth increment, the effect may not be universal and perhaps there are two prior studies of snow crab growth per molt for east- a more benign holding protocol could be developed for eastern Canadian stocks, one based on modes in width-frequency 146 SOMERTON ET AL.

FIGURE 3. Growth-per-molt model for snow crabs in the eastern Bering FIGURE 5. Percent bias in the snow crab growth-per-molt data from the Sea (solid line), the empirical 95% conﬁdence intervals (dotted lines), and the National Marine Fisheries Service holding study relative to that predicted by growth-per-molt models of Comeau et al. (1998) and Sainte-Marie et al. (1995; the accepted model. This bias is shown plotted against the time in days from dashed lines) for snow crabs in eastern Canada. collection. Note that the bias is initially small, then increases asymptotically with time. Except for the one extreme value, the bias stabilizes at about −9% data (Comeau et al. 1998) and the second based on the direct after 40 d. The vertical line indicates the maximum holding time (i.e., 30 d) observation of molting crabs (Sainte-Marie et al. 1995), both of used as one of the criteria for data acceptance. which reported growth-per-molt functions that can be directly compared with the function presented here. In our study, the segmented straight-line model had an intersection point at a premolt width of 36.1 mm, which is remarkably similar to the intersection points in the studies of Comeau et al. (1998; 36.7 mm) and Sainte-Marie et al. (1995; 36.6 mm). Since the intersection point is considered to be the carapace width associated with a transition between immature and adolescent life history phases (Sainte-Marie et al. 1995), this similarity indicates that the size dependency of this life history event is preserved between the distantly separated Alaskan and Cana- dian populations. However, for premolt widths greater than the intersection points, postmolt widths predicted by the Canadian models are below the lower 95% CI bound of the Alaskan Downloaded by [Department Of Fisheries] at 22:30 28 February 2013 model, indicating that Alaska snow crabs may have a slightly greater growth per molt over this width range (Figure 3). In addition to the mean itself, we estimated the standard deviation of future postmolt width observations about the expected mean because of the way growth is considered in the snow crab stock assessment model (Turnock and Rugolo 2010). The model considers growth to be a discrete event in which molting crabs in each premolt width interval are increased in width according to a postmolt-width probability distribution function. Our estimates of the mean and standard deviation of postmolt width at FIGURE 4. Standard deviations of the predicted postmolt width observations each premolt width are intended to parameterize this probability as a function of premolt width. The standard deviation increases at the ends of distribution function; however, our small sample size may have the two lines and at their joint point (36.1 mm). Values of standard deviations at premolt widths >36 are used in the snow crab stock assessment model to resulted in an underestimate of the standard deviation. compute the postmolt width probability distribution for each premolt width To increase the sample size and width range of our growth- class. per-molt data, we pooled the observations collected during three SNOW CRAB GROWTH IN THE BERING SEA 147

phases of our study with those from a previous study, all of which Draper, N., and H. Smith. 1981. Applied regression analysis, 2nd edition. Wiley, were potentially influenced by different experimental condi- New York. tions. To minimize the differences in experimental effects we Efron, B., and R. J. Tibshirani. 1993. An introduction to the bootstrap. Chapman and Hall, New York. eliminated crabs that were held for longer than 30 d or had lost Elner, R. W. 1982. Overview of the snow crab Chionoecetes opilio fishery more than two legs because previous work indicated that these in Atlantic Canada. Pages 3–19 in Proceedings of the international sympo- variables were influential on the growth increment of Chionoe- sium on the genus Chionoecetes: Lowell Wakefield fisheries symposia series. cetes (Stone et al. 2003). Other variables, such as water temper- University of Alaska, Sea Grant College Program, Report AK-SG-82-10, ature, provision of food, or other aspects that were not consistent Fairbanks. Hiatt, R. W. 1948. The biology of the lined shore crab, Pachygrapsus crassipes, across these studies, could have influenced growth per molt and Randall. Pacific Science 2:135–213. led to a bias in the estimated growth function. However, the seg- Hoenig, J. M., E. G. Dawe, and P. G. O’Keefe. 1994. Molt indicators and growth mented line model that we used fits our growth-per-molt data per molt for male snow crabs (Chionoecetes opilio). Journal of Crustacean extremely well, as is evidenced by a value of R2 approaching Biology 14:273–279. 1.0. This indicates that any variability in the experimental con- Jadamec, L. S., W. E. Donaldson, and P. Cullenberg. 1999. Biological field techniques for Chionoecetes crabs. University of Alaska, Sea Grant College ditions experienced by the crabs in this study did not result in Program, Report AK-SG-99-02, Fairbanks. variability in growth per molt. The low residual error coupled McBride, J. 1982. Tanner crab tag development and tagging experiments 1978– with the clear linearity in each of the growth phase lines of 1982. Pages 383–402 in Proceedings of the international symposium on the the segmented model indicate that our model, although based genus Chionoecetes: Lowell Wakefield fisheries symposia series. University on relatively few crabs, probably describes the growth-per-molt of Alaska, Sea Grant College Program, Report AK-SG-82-10, Fairbanks. Montgomery, D. C., and E. A. Peck. 1982. Introduction to linear regression function of eastern Bering Sea snow crabs quite well. analysis. Wiley, New York. Moriyasu, M., G. Y. Conan, P. Mallet, Y. J. Chiasson, and H. Lacroix. 1987. Growth at molt, molting season and mating of snow crab (Chionoecetes ACKNOWLEDGMENTS opilio) in relation to functional and morphometric maturity. International We thank the staff of the Dutch Harbor office of the Alaska Council for the Exploration of the Sea, C.M. 1987/K:21, Copenhagen. Department of Fish and Game (Heather Fitch, Rachel Alin- R Development Core Team. 2008. R: a language and environment for statisti- sunurin, Britta Baechler, Ian Fo, Loren St Amand, and Trent cal computing. R Foundation for Statistical Computing, Vienna. Available: www.R-project.org. (July 2012). Hartill) for help with the snow crab holding study, the staff of Rand, K. M., and E. A. Logerwell. 2011. The first demersal trawl survey of Unisea Inc. (Todd Shoup, Don Graves, and Guy Collins) for benthic fish and invertebrates in the Beaufort Sea since the late 1970s. Polar providing the dock and live tank facilities in Dutch Harbor, Biology 34:475–488. and the captain and crew of the F/V Half Moon Bay for their Rugolo, L. J., D. Pengilly, R. MacIntosh, and K. Gravel. 2005. Reproductive support through the rigors of snow crab sampling. This study dynamics and life history of snow crab (Chionoecetes opilio) in the eastern Bering Sea. Final Completion Report to the Alaska Department of Fish and was partially funded by the North Pacific Fisheries Research Game, Bering Sea Snow Crab Fishery Restoration Research, National Marine Foundation. Fisheries Service, Alaska Fisheries Science Center, Seattle. Sainte-Marie, B., S. Raymond, and J. C. Brethes.ˆ 1995. Growth and maturation of the benthic stages of male snow crab, Chionoecetes opilio (Brachyura: REFERENCES Majidae). Canadian Journal of Fisheries and Aquatic Sciences 52:903–924. Alunno-Bruscia, M., and B. Sainte-Marie. 1998. Abdomen allometry, ovary de- Somerton, D. A. 1980. Fitting straight lines to Hiatt growth diagrams: a re- velopment, and growth of female snow crab, Chionoecetes opilio (Brachyura, evaluation. ICES Journal of Marine Science 39:15–19. Majidae), in the northwestern Gulf of St. Lawrence. Canadian Journal of Fish- Somerton, D. A., and R. S. Otto. 1999. Net efficiency of a survey trawl for snow eries and Aquatic Sciences 55:459–477. crab, Chionoecetes opilio, and Tanner crab, C. bairdi. U.S. National Marine Comeau, M., G. Y. Conan, F. Maynou, G. Robichaud, J. C. Therriault, and M. Fisheries Service Fishery Bulletin 97:617–625. Starr. 1998. Growth, spatial distribution, and abundance of benthic stages of Stauffer, G. 2004. NOAA protocols for groundfish bottom trawl surveys of the Downloaded by [Department Of Fisheries] at 22:30 28 February 2013 the snow crab (Chionoecetes opilio) in Bonne Bay, Newfoundland, Canada. nation’s fishery resources. NOAA Technical Memorandum NMFS-SPO-65. Canadian Journal of Fisheries and Aquatic Sciences 55:262–279. Stone, R. P., M. M. Masuda, and J. E. Clark. 2003. Growth of male Tanner crabs Conan, G. Y., M. Comeau, C. Gosset, G. Robichaud, and C. Gara¨ıcoechea. Chionoecetes bairdi in a southeast Alaska estuary. Alaska Fishery Research 1994. The Bigouden Nephrops trawl, and the Devismes trawl, two otter Bulletin 10:137–148. trawls efficiently catching benthic stages of snow crab (Chionoecetes opilio), Turnock, B. J., and L. J. Rugolo. 2010. Stock assessment of eastern Bering Sea and American lobster (Homarus americanus). Canadian Technical Report of snow crab. Pages 27–135 in Crab Plan Team, editors. Stock assessment and Fisheries and Aquatic Sciences 1992. fishery evaluation report for the king and Tanner crab fisheries of the Bering Donaldson, W. E., R. T. Cooney, and J. R. Hilsinger. 1981. Growth, age Sea and Aleutian Islands regions. North Pacific Fishery Management Coun- and size at maturity of Tanner crab, Chionoecetes bairdi M. J. Rathbun, cil, Anchorage, Alaska. Available: https://alaskafisheries.noaa.gov/npfmc/ in the northern Gulf of Alaska (Decapoda, Brachyura). Crustaceana 40: PDFdocuments/resources/SAFE/CrabSAFE/CrabSAFE2011.pdf. (January 286–302. 2013). This article was downloaded by: [Department Of Fisheries] On: 28 February 2013, At: 22:30 Publisher: Taylor & Francis Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

North American Journal of Fisheries Management Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/ujfm20 Can Management Reduce Harvest Inequality in Recreational Fisheries? David A. Seekell a , Timothy J. Cline b d & Raymond J. Winchcombe c a Department of Environmental Sciences, University of Virginia, Clark Hall, 291 McCormick Road, Charlottesville, Virginia, 22904, USA b Center for Limnology, University of Wisconsin, 680 North Park Street, Madison, Wisconsin, 53706, USA c Cary Institute of Ecosystem Studies, Plant Science Building, Box AB, Millbrook, New York, 12545, USA d School of Aquatic and Fishery Sciences, University of Washington, 1122 Northeast Boat Street, Seattle, Washington, 98195, USA Version of record first published: 25 Jan 2013.

To cite this article: David A. Seekell , Timothy J. Cline & Raymond J. Winchcombe (2013): Can Management Reduce Harvest Inequality in Recreational Fisheries?, North American Journal of Fisheries Management, 33:1, 148-152 To link to this article: http://dx.doi.org/10.1080/02755947.2012.746246

PLEASE SCROLL DOWN FOR ARTICLE

The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material. North American Journal of Fisheries Management 33:148–152, 2013 C American Fisheries Society 2013 ISSN: 0275-5947 print / 1548-8675 online DOI: 10.1080/02755947.2012.746246

MANAGEMENT BRIEF

Can Management Reduce Harvest Inequality in Recreational Fisheries?

David A. Seekell* Department of Environmental Sciences, University of Virginia, Clark Hall, 291 McCormick Road, Charlottesville, Virginia 22904, USA Timothy J. Cline1 Center for Limnology, University of Wisconsin, 680 North Park Street, Madison, Wisconsin 53706, USA Raymond J. Winchcombe Cary Institute of Ecosystem Studies, Plant Science Building, Box AB, Millbrook, New York 12545, USA

recreational fisheries management has been to promote a rela- Abstract tively equal distribution of harvest with the goal of improving Harvest inequality, a situation in which most of the fish are angler satisfaction with fisheries (cf. Hudgins and Davies 1984; harvested by a disproportionately small number of anglers, is a Cook et al. 2001; Radomski 2003; Paukert et al. 2007). How- characteristic of most recreational fisheries. Harvest inequality develops when a few anglers harvest a very large number of fish ever, a fundamental understanding of the patterns and dynamics or when many anglers harvest few fish. Identifying the cause of of harvest inequality is lacking, and the efficacy of management harvest inequality is critical to understanding the potential for in influencing harvest inequality is unknown (Cook et al. 2001; management to reduce the inequality. Management efforts aimed Radomski et al. 2001; Paukert et al. 2007). at reducing the top anglers’ take will have only a modest impact Harvest inequality occurs when most of the fish are har- if the harvest inequality is actually caused by many anglers harvesting no fish. We measured harvest inequality in 20 years of creel vested by a disproportionately small number of anglers (Cook census data from a trout stream in southeastern New York. We et al. 2001). Knowing the cause of harvest inequality is critical calculated Lorenz curve asymmetry coefficients (S) to test whether to understanding the potential effects of management because harvest inequality was attributable to small harvests by many an- most harvest regulations only influence the top harvesters. Har- glers or large harvests by a few anglers. Harvest inequality in the vest inequality develops if a few anglers take a large number of fishery was consistently high and the S-value was always less than 1.0, indicating that harvest inequality was caused by many anglers fish relative to the average harvest (Cook et al. 2001). However, harvesting no fish rather than by few anglers harvesting many fish. harvest inequality can also develop if many anglers take no fish. This influence becomes stronger with increased harvest. We con- The few anglers that do take fish then appear to harvest a dispro- Downloaded by [Department Of Fisheries] at 22:30 28 February 2013 clude that management is unlikely to influence the magnitudes of portionately large number of fish, even if their harvest consists of harvest inequality in recreational fisheries because regulations do only a few fish (i.e., the harvest is below a creel limit). This dis- not target the principal cause of harvest inequality. tinction in the cause of harvest inequality is important because creel limits (or reduced creel limits) are often implemented to Angler satisfaction with fisheries is related to the evenness of reduce harvest inequality (see Cook et al. 2001; Radomski et al. the distribution of catch among anglers (cf. Hudgins and Davies 2001; Radomski 2003; Paukert et al. 2007), but the limits will 1984). However, the distribution of catch is heavily influenced have only a modest impact in reducing harvest inequality if it is by angler effort, which is difficult to regulate (Hilborn 1985; caused by a situation of many anglers harvesting no fish. Radomski 2003; Sullivan 2003; Seekell 2011). Harvest is more Reports of the magnitude of harvest inequality are rare. Cook manageable than catch, and consequently a principal focus of et al. (2001) measured harvest inequality in Minnesota lakes by

*Corresponding author: [email protected] 1Present address: School of Aquatic and Fishery Sciences, University of Washington, 1122 Northeast Boat Street, Seattle, Washington 98195, USA. Received May 10, 2012; accepted October 29, 2012

148 MANAGEMENT BRIEF 149

using complete-trip creel surveys. Those authors found that the magnitude of harvest inequality varied between six commonly targeted species, but all of the species were characterized by an abundance of anglers with no harvest. This result is likely attributable to the fact that harvest is not the principal motivation of all recreational freshwater anglers, and many anglers strictly participate in voluntary catch-and-release fishing (e.g., Chipman and Helfrich 1988; Sutton 2003; Schramm and Gerard 2004; Arlinghaus 2006). The results of Cook et al. (2001) suggest that harvest inequality is due to low harvest by many anglers rather than to large harvest by a few anglers. However, there has not been a rigorous test to discriminate between the two potential causes of harvest inequality. In this study, we analyzed long-term creel census data for Brown Trout Salmo trutta from a delimited section of Wappinger Creek near Millbrook, New York. We applied statistical measures of inequality to test whether small harvests by many anglers or large harvests by a few anglers were causing harvest inequality; our goal was to infer the potential efficacy of management in influencing the magnitude of harvest inequality.

METHODS FIGURE 1. Hypothetical Lorenz curves (modified from Damgaard and Weiner = 2000 and Seekell et al. 2011b). The dark-gray bold line is the line of equality Data.—We retrieved angler logs (n 1,499) archived by the (e.g., 10% of fishing effort results in 10% of harvest, 30% of effort results in Cary Institute of Ecosystem Studies (hereafter, Cary Institute) 30% of harvest, and so on). The light-gray solid line is the axis of symmetry. The from 1988 to 2007 for the Brown Trout fishery in the East Branch black and gray dashed lines are Lorenz curves representing the same amount of Wappinger Creek where the creek flows through the Cary of inequality (same Gini coefficient) but with different shapes described by the Institute’s property (41◦478.9628N, 73◦4429.2086W). asymmetry coefficient S (i.e., S < 1.0 for the gray dashed curve; S > 1.0 for the black dashed curve). When S is less than 1.0, the distribution of harvest is Seekell et al. (2011a) provided a detailed description of the dominated by a large number of anglers that harvest few or no fish. When S survey methodology. Briefly, these data represent a complete is greater than 1.0, the distribution of harvest is dominated by a few anglers census of recreational fishing in this fishery. All anglers recorded that harvest a very large number of fish (cf. Chen et al. 2010). The black effort (h), catch, harvest, and party size for each trip as a con- Lorenz curve (solid curve) has greater inequality (larger Gini coefficient) than dition of receiving a fishing permit from the Cary Institute. the dashed curves, but it is symmetrical (S = 1.0). Records were submitted after each trip. We used records with a party size of one angler to minimize potential confounding fac- this methodology is that differently shaped harvest distributions tors associated with variable group sizes (Seekell et al. 2011a). can display the same level of inequality (i.e., they can have the We believe that the potential for confounding variability from same Gini coefficient value). For example, the dashed black and substitutability or targeting of different species in influencing gray Lorenz curves in Figure 1 represent distributions of harvest catching power or harvesting decisions was minimal because with the same level of inequality. However, the distributions are Downloaded by [Department Of Fisheries] at 22:30 28 February 2013 99% of the fish caught were Brown Trout (Cook et al. 2001; different because one is dominated by a large number of anglers Arlinghaus et al. 2007). that harvest few fish (the gray dashed Lorenz curve, describ- Statistical analysis.—Harvest inequality was visualized by ing a distribution in which about 50% of the anglers harvest plotting Lorenz curves and was quantified by calculating Gini about 10% of the fish), whereas the other is dominated by a coefficients (Smith 1990; Baccante 1995; Cook et al. 2001). few anglers that harvest large numbers of fish (the black dashed These methods are described in detail by Seekell et al. (2011a). Lorenz curve, describing a distribution in which about 10% of Briefly, Lorenz curves and Gini coefficients assess the relation- the anglers harvest about 50% of the fish). Hence, the shape of ship between the cumulative proportion of fish harvested and the Lorenz curve can be used to infer whether harvest inequality the cumulative proportion of effort. If there is harvest inequal- is caused by (1) few anglers harvesting a relatively large num- ity, Lorenz curves become more concave and Gini coefficients ber of fish (relative to the mean harvest) or (2) many anglers increase from 0 to near 1.0. We calculated Gini coefficients for harvesting a small number of fish (cf. Chen et al. 2010). To each year according to the method of Damgaard and Weiner quantify these different Lorenz curve shapes, we calculated the (2000). Lorenz curve asymmetry coefficient (S; Damgaard and Weiner The Gini coefficient is 2 times the area between the Lorenz 2000). An S less than 1.0 describes a Lorenz curve in which the curve and a hypothetical curve describing a scenario in which point with a slope equal to 1.0 is below the axis of symmetry anglers harvest an equal proportion of fish. A consequence of (solid gray diagonal line in Figure 1), indicating low harvest 150 SEEKELL ET AL.

by many anglers (cf. Chen et al. 2010). An S greater than 1.0 describes a Lorenz curve in which the point with a slope equal to 1.0 is above the axis of symmetry, indicating relatively large harvests by a few anglers (cf. Chen et al. 2010). An S equal to 1.0 represents a symmetrical Lorenz curve like the solid black curve shown in Figure 1. We plotted Lorenz curves from catch distributions reported by Seekell et al. (2011a) for this fishery as a basis for comparison with our harvest distribution Lorenz curves. Pearson’s product- moment correlation coefficients between harvest inequality (as represented by Gini coefficients), harvest per trip, S, and year were calculated as basic tests for trend; we calculated correlations for these variables to understand how they co-vary. We controlled for confounding trends by using partial correlation coefficients; this method reduces the chance of identifying a spurious correlation between two variables.

RESULTS AND DISCUSSION Harvest inequality in the Wappinger Creek Brown Trout fishery was extreme (mean Gini coefficient = 0.85; SD = 0.08; range = 0.68–0.95). The harvest Gini coefficients were considerably greater than the values reported by Seekell et al. (2011a) FIGURE 2. Lorenz curves from the 1989 Brown Trout fishing season in = Wappinger Creek. The shapes of these curves are typical for catch and harvest for catch distributions in this fishery (mean Gini coefficient distributions (e.g., Seekell et al. 2011a and Cook et al. 2001, respectively). 0.51; SD = 0.07; range = 0.43–0.71). This result means that The catch Lorenz curve (dashed curve) is relatively symmetrical, with only a there is nothing inherent in the process of catching fish that moderate level of inequality; the harvest curve (solid curve) is asymmetrical causes extreme harvest inequality—rather, the extreme inequal- (asymmetry coefficient S < 1.0), with a high level of inequality. ity develops from the decisions made by anglers after the fish are caught. The harvest Lorenz curve displays a large number of comes more unequal as the number of nonharvesting anglers zero-harvest trips relative to the number of trips with zero catch, increases (Figure 3). This is because when many anglers do whereas the curves are close to each other at the upper ends not harvest fish, the point where the cumulative proportion of of their distributions (Figure 2). This suggests that the greater harvest becomes greater than zero moves closer to the axis of harvest inequality relative to catch inequality is due to the abun- symmetry (Figure 3). This increases the value of S but also in- dance of nonharvesting anglers rather than to large harvests by a creases the Gini coefficient by moving the Lorenz curve near the few anglers. In the interest of brevity, we do not present figures point furthest from the line of equality (Figure 3). For instance, for all 20 years, but the shapes of the catch and harvest Lorenz the insets in Figure 3 show harvest Lorenz curves for 2 years curves in Figure 2 are representative of all catch and harvest (1988 and 2007) with different levels of inequality (Gini coeffi- curves from the Wappinger Creek data set. Furthermore, the cients = 0.65 and 0.81, respectively). The increased inequality Lorenz curves in Figure 2 are also very similar to Lorenz curves in 2007 (the black shaded area between the line of equality and Downloaded by [Department Of Fisheries] at 22:30 28 February 2013 that have been reported for other fisheries. The symmetrical the Lorenz curve) relative to 1988 is due to a greater relative catch Lorenz curve in Figure 2 is similar in shape and mag- abundance of zero harvests. The S increases because there are nitude to the catch Lorenz curves presented by Seekell (2011) so many zero-harvest trips with increased inequality that the and Mosindy and Duffy (2007) for other fisheries. The harvest point where the Lorenz curve departs from zero harvest is very Lorenz curve in Figure 2 is similar in shape and magnitude to close to the axis of symmetry. However, the S is always less harvest Lorenz curves presented by Cook et al. (2001) for six than 1.0 (mean = 0.82; SD = 0.1; range = 0.60–0.94), indicat- species in Minnesota lakes. These similarities suggest that the ing that harvest inequality is always caused by an abundance of patterns identified in our analysis can be generalized to other nonharvesting anglers. inland recreational fisheries. The mean harvest per trip declined during the study (r = Harvest inequality and S-values increased significantly dur- −0.71, P < 0.001). This decline caused the S and the Gini ing the study period (Gini coefficient: r = 0.613, P < 0.001; coefficient to increase because these values are both inversely S-value: r = 0.689, P < 0.001). The Gini coefficients and S- related to mean harvest per trip (S-value: partial r =−0.924, values were strongly correlated after we controlled for poten- P < 0.001; Gini coefficient: partial r =−0.756, P < 0.001; tially confounding trends (partial r = 0.916, P < 0.001). The Figure 4). This result indicates that higher harvest is associated positive correlation indicates that the distribution of harvest be- with decreased inequality as well as an increased influence of MANAGEMENT BRIEF 151

these limits are generally arbitrary and affect only a few anglers because few anglers harvest the limit (Cook et al. 2001). Creel limits may be ineffective for transferring harvest from top harvesters to nonharvesting anglers because many anglers strictly participate in catch-and-release angling and are unlikely to change their motivation based on regulations aimed at harvest- oriented anglers (Beardmore et al. 2011). When we recalculated Gini coefficients and Lorenz curve S-values while excluding anglers that strictly participated in voluntary catch-and-release fishing, the Gini coefficients were slightly lower (by 0.1 on average), indicating less inequality (mean = 0.75; SD = 0.1; range = 0.54–0.88) among harvesting anglers relative to the total population of anglers. However, the values of S also decreased (by 0.08 on average), indicating that harvest inequality was still caused by an abundance of trips with little or no harvest (mean = 0.74; SD = 0.14; range = 0.44–1.03). There was one exception to this: the S-value for 1988 was 1.03, indicat- FIGURE 3. Relationship between the Lorenz curve asymmetry coefficient (S) ing that the contribution to inequality was approximately equal and the Gini coefficient. The insets show Lorenz curves representing the first between anglers harvesting no fish and those harvesting many and last years (1988 and 2007) of the data set for the Wappinger Creek Brown fish. Trout fishery. In each inset, the area between the Lorenz curve and the line of equality is shaded black. The level of inequality was higher in 2007 than in 1988, as indicated by the larger shaded region for 2007. However, the S is MANAGEMENT IMPLICATIONS higher for 2007 because the overabundance of zero harvests pushes the point of Although it is a common characteristic of recreational fresh- the Lorenz curve where the slope is 1.0 in the direction of the axis of symmetry. water fisheries, harvest inequality is likely unmanageable. Har- vest inequality is caused by many anglers harvesting no fish nonharvesting anglers in creating the inequality. Hence, as the rather than by few anglers harvesting many fish. We suggest harvest increases, the harvest inequality in these fisheries will that minimization of harvest inequality should not be a primary become difficult or impossible to manage through restrictions goal of management because harvest inequality does not con- aimed at harvest-oriented anglers. flict with the motivations of recreational anglers and, at least Creel limits are usually implemented in part to reduce har- in this fishery, it is not caused by a few anglers obtaining large vest inequality (Cook et al. 2001; Paukert et al. 2007). However, harvests.

ACKNOWLEDGMENTS This study was supported by the National Science Founda- tion’s Graduate Research Fellowship Program, the University of Virginia, the University of Wisconsin–Madison, and the Cary Institute of Ecosystem Studies. We thank Jim Kitchell for helpful comments on an earlier version of this manuscript. Downloaded by [Department Of Fisheries] at 22:30 28 February 2013

REFERENCES Arlinghaus, R. 2006. On the apparently striking disconnect between motivation and satisfaction in recreational fishing: the case of catch orientation of German anglers. North American Journal of Fisheries Management 26: 592–605. Arlinghaus, R., S. J. Cooke, J. Lyman, D. Policansky, A. Schwab, C. Suski, S. G. Sutton, and E. B. Thorstad. 2007. Understanding the complexity of catch-and-release in recreational fishing: an integrative synthesis of global knowledge from historical, ethical, social, and biological perspectives. Re- views in Fisheries Science 15:75–167. Baccante, D. 1995. Assessing catch inequality in Walleye angling fisheries. North American Journal of Fisheries Management 15:661–665. Beardmore, B., W. Haider, L. M. Hunt, and R. Arlinghaus. 2011. The importance FIGURE 4. Relationship between Brown Trout harvest per trip and either the of trip context for determining primary angler motivations: are more special- Lorenz curve asymmetry coefficient or the Gini coefficient. Linear regression ized anglers more catch-oriented than previously believed? North American lines are drawn to emphasize the relationships. Journal of Fisheries Management 31:861–879. 152 SEEKELL ET AL.

Chen, D., X. Ma, H. Mu, and P. Li. 2010. The inequality of natural resources Radomski, P. 2003. Initial attempts to actively manage recreational fishery consumption and its relationship with the social development level based on harvest in Minnesota. North American Journal of Fisheries Management the ecological footprint and the HDI. Journal of Environmental Assessment 23:1329–1342. Policy and Management 12:69–86. Radomski, P., G. C. Grant, P. C. Jacobson, and M. F. Cook. 2001. Visions for Chipman, B. D., and L. A. Helfrich. 1988. Recreational specializations and recreational fishing regulations. Fisheries 26(5):7–18. motivations of Virginia river anglers. North American Journal of Fisheries Schramm, H. L., Jr., and P. D. Gerard. 2004. Temporal changes in fishing moti- Management 8:390–398. vation among fishing club anglers in the United States. Fisheries Management Cook, M. F., T. J. Goeman, P. J. Radomski, J. A. Younk, and P. C. Jacobson. and Ecology 11:313–321. 2001. Creel limits in Minnesota: a proposal for change. Fisheries 26(5): Seekell, D. A. 2011. Recreational freshwater angler success is not significantly 19–26. different from a random catch model. North American Journal of Fisheries Damgaard, C., and J. Weiner. 2000. Describing inequality in plant size or Management 31:203–208. fecundity. Ecology 81:1139–1142. Seekell, D. A., C. J. Brosseau, T. J. Cline, R. J. Winchcombe, and L. J. Zinn. Hilborn, R. 1985. Fleet dynamics and individual variation: why some people 2011a. Long-term changes in recreational catch inequality in a trout stream. catch more fish than others. Canadian Journal of Fisheries and Aquatic Sci- North American Journal of Fisheries Management 31:1110–1115. ences 42:2–13. Seekell, D. A., P. D’Odorico, and M. L. Pace. 2011b. Virtual water transfers Hudgins, M. D., and W. D. Davies. 1984. Probability angling: a recreational fish- unlikely to redress inequality in global water use. Environmental Research ery management strategy. North American Journal of Fisheries Management Letterss [online serial] 6:024017. 4:431–439. Smith, C. L. 1990. Resource scarcity and inequality in the distribution of catch. Mosindy, T. E., and M. J. Duffy. 2007. The use of angler diary surveys to eval- North American Journal of Fisheries Management 10:269–278. uate long-term changes in Muskellunge populations on Lake of the Woods, Sullivan, M. G. 2003. Active management of Walleye fisheries in Alberta: dilem- Ontario. Environmental Biology of Fishes 79:71–83. mas of managing recovering fisheries. North American Journal of Fisheries Paukert, C., M. McInerny, and R. Schultz. 2007. Historical trends in creel Management 23:1343–1358. limits, length-based limits, and season restrictions for black basses in the Sutton, S. G. 2003. Personal and situational determinants of catch-and-release United States and Canada. Fisheries 32:62–72. choice of freshwater anglers. Human Dimensions of Wildlife 8:109–126. Downloaded by [Department Of Fisheries] at 22:30 28 February 2013 This article was downloaded by: [Department Of Fisheries] On: 28 February 2013, At: 22:31 Publisher: Taylor & Francis Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

North American Journal of Fisheries Management Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/ujfm20 Application of Stock Density Indices as a Tool for Broad- Scale Population Assessment for Four Cyprinid Species in Central Italy Giovanni Pedicillo a , Massimo Lorenzoni a , Antonella Carosi b & Lucia Ghetti c a Dipartimento di Biologia Cellulare e Ambientale, Università di Perugia, Via Elce di Sotto, Perugia, 06123, Italy b Provincia di Terni, Servizio Programmazione Ittico-Faunistica, Via Plinio il Giovane 21, Terni, 05100, Italy c Regione dell'Umbria, Servizio Programmazione Forestale, Faunistico-Venatoria ed Economia Montana, Via Mario Angeloni 61, 06100, Perugia, Italy Version of record first published: 30 Jan 2013.

To cite this article: Giovanni Pedicillo , Massimo Lorenzoni , Antonella Carosi & Lucia Ghetti (2013): Application of Stock Density Indices as a Tool for Broad-Scale Population Assessment for Four Cyprinid Species in Central Italy, North American Journal of Fisheries Management, 33:1, 153-162 To link to this article: http://dx.doi.org/10.1080/02755947.2012.747453

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material. North American Journal of Fisheries Management 33:153–162, 2013 C American Fisheries Society 2013 ISSN: 0275-5947 print / 1548-8675 online DOI: 10.1080/02755947.2012.747453

ARTICLE

Application of Stock Density Indices as a Tool for Broad-Scale Population Assessment for Four Cyprinid Species in Central Italy

Giovanni Pedicillo* and Massimo Lorenzoni Dipartimento di Biologia Cellulare e Ambientale, Universita` di Perugia, Via Elce di Sotto, 06123 Perugia, Italy Antonella Carosi Provincia di Terni, Servizio Programmazione Ittico-Faunistica, Via Plinio il Giovane 21, 05100 Terni, Italy Lucia Ghetti Regione dell’Umbria, Servizio Programmazione Forestale, Faunistico-Venatoria ed Economia Montana, Via Mario Angeloni 61, 06100 Perugia, Italy

Abstract Stock-density indices are standardized methods for analyzing length-frequency data that quantify the length structure of a fish population into length categories that are of interest to recreational anglers. We adapted North American relative stock density (RSD) and proportional stock density (PSD) indices to four Italian endemic or native cyprinid species by means of two methods. In accordance with the traditional approach, the benchmarks of length categories were established according to Gabelhouse’s percentage classification and calculated on the basis of the largest individual in the data set for each species. In the second method, asymptotic length and size at maturity were used to define the length categories for index calculation. Both methods were tested on length-frequency data from fish collected from sampling sites in the Tiber River basin. The results showed that the traditional approach displayed better applicability, required less sampling effort, and provided a better insight into the length structure of the fish populations studied. Downloaded by [Department Of Fisheries] at 22:31 28 February 2013 The length structure of a fish population is one of the most numerically and provide appropriate numerical estimations of commonly used assessment tools in fisheries, as it reflects in- deviations from natural length structures in fish populations. teractions among the dynamic rates of recruitment, growth, and Such indices are commonly applied in North American fish com- mortality. Length-frequency distributions are among the oldest munities, and yield results that are easy to communicate to fish- methods used to assess length structure in fish populations (Ev- ery managers (Gassner et al. 2003). Proportional stock density erhart and Youngs 1981). However, these can often be difficult is the percentage of “quality-sized” individuals within the total to interpret because there are few standards by which to assess number of “stock-sized” individuals. Stock and quality size des- whether the length frequency is optimal or expected for a given ignations are based on the response and expectation of anglers; situation. Proportional stock density (PSD) and relative stock fish of stock length have little recreational value, while quality density (RSD) enable length-frequency data to be estimated length is the minimum size that most anglers like to catch.

*Corresponding author: [email protected] Received January 17, 2012; accepted November 4, 2012 153 154 PEDICILLO ET AL.

Proportional stock density compresses the entire length- METHODS frequency distribution into a single number, thereby engen- The study area consisted of the watercourses in the upper and dering a probable loss of information (Gabelhouse 1984). To middle portions of Tiber River from its source to its confluence analyze the length structure of a fish population in greater with the Treja River (Figure 1). The Tiber River, the third-longest detail, Gabelhouse (1984) added three additional categories: river in Italy, originates on Mount Fumaiolo (about 1,270 m the proportion of “preferred-length” fish, the proportion of above sea level) and is 406 km long. Its basin, the second-largest “memorable-length” fish, and the proportion of “trophy-length” Italian river catchment, covers more than 17,000 km2 and has an fish in the population. Gablehouse (1984) called this length cate- average elevation of 524 m. The area investigated, correspond- gorization system the RSD. The reference values for each length ing to a surface area of 9,832 km2 (57% of the total drainage area class vary by species and are defined as percentages of the length of the Tiber River), included numerous tributaries, the most im- of the all-tackle world-record fish. Because PSD and RSD do not portant ones being the Nestore River (watershed = 1,033 km2), always reflect population density (Willis et al. 1993), Guy et al. the Paglia River (1,338 km2), the Chiascio River (5,963 km2), (2007) suggested that proportional stock density be changed to and the Nera River (4,280 km2) (Mearelli et al. 1994). More de- proportional size distribution. tailed information on the characteristics of the Tiber River basin Although both PSD and RSD are frequently used in North and its fish populations is available in Lorenzoni et al. (2006) America, they are rarely used in Europe. The first attempt to and Pedicillo et al. (2010). The field studies were carried out apply North American PSD and RSD indices in Europe was between 1992 and 2008; a total of 70 streams and rivers were made by Gassner et al. (2003) to assess the ecological in- included in the study for a total of 113–253 sampling stations, tegrity of two Austrian prealpine lakes according to the Eu- depending on the species. Sampling was performed by means of ropean Union’s Water Framework Directive (WFD) (European Parliament and Council 2000). In their study, the authors used a different method to determine the thresholds for index calculation, based no longer on the percentage of all-tackle world- record fish, but on two biological parameters: the maximum length recorded for the species examined in the study area and the length at maturity. Zick et al. (2007) applied this method to Arctic Char Salvelinus alpinus in Austrian lakes, and Pedi- cillo et al. (2010) applied it to Brown Trout Salmo trutta in central Italian streams, using the asymptotic length (L∞) of the von Bertalanffy growth model and length at maturity. Volta and Oggioni (2010) did the same in order to develop a multimetric index (Lake Fish Index) to assess the ecological status of Italian lakes in accordance with the WFD. In this case, those authors reported reference values for the stock and quality lengths of the key fish species in Italian lakes. As yet, however, reference values for the length classes of fish species in running waters in Italy are lacking. As these indices may be useful indicators in future assessments of the quality of Ital- ian rivers according to the WFD, the goal of our research was Downloaded by [Department Of Fisheries] at 22:31 28 February 2013 to determine, with regard to the principle native fish species present in the Tiber River basin, the reference values delimiting the length classes used to calculate structure indices by means of the two methods available in the literature: the traditional method of Gabelhouse (1984) adapted to the local conditions of fish species in the Tiber River basin (traditional approach), and that of Gassner et al. (2003) modified by Zick et al. (2007) (biological approach). The fish species considered were the cyprinids Cavedano Chub Squalius squalus, Rovella Rutilus rubilio, Horse Barbel Barbus tyberinus, and Italian Riffle Dace Telestes muticellus. The additional aim was to compare the length categories obtained by the two methods and structural indices computed from the length categories by the two methods for sample sites FIGURE 1. Study area (in grey) shown in the River Tiber basin in central across the Tiber River basin. Italy. STOCK DENSITY INDICES FOR FOUR FISH SPECIES IN CENTRAL ITALY 155

electrofishing with electric stunning devices of different pow- The length at maturity was calculated from the mean L∞ by ers, according to the features of the watercourse involved. All applying the empirical equation log10Lm = 0.8979 log10L∞ − specimens caught were identified by species, counted, and mea- 0.0782 (Froese and Binohlan 2000). sured (total length [TL] ± 1 mm) (Anderson and Neumann On the basis of these thresholds, in accordance with the 1996), and scale samples for age determination were taken from biological approach, the length classes were defined as follows the body area, as described by DeVries and Frie (1996). When (Gassner et al. 2003; Zick et al. 2007; Pedicillo et al. 2010; large numbers of specimens were sampled, scales were only Volta 2010; Volta and Oggioni 2010): Stock (S) = Q − [(T − collected from a subsample for each 10-mm length increment. Q)/3], Quality (Q) = Lm, Preferred (P) = Q + [(T − Q)/3], Age determination by means of scale analysis was confirmed Memorable (M) = Q{[(T − Q)/3]2} and Trophy (T) = 80% of and integrated by applying Petersen’s length-frequency method the L∞. (Bagenal and Tesch 1978). Both methods were tested on the length classes for the PSD To assess the minimum lengths for the size categories for the [(number of fish ≥ minimum quality length / number of fish ≥ RSD index calculation, two different methods were used. For minimum stock length) × 100] by analyzing their applicability, a given species, length classes are traditionally defined as per- the number of specimens needed in order to obtain reliable centage lengths of the all-tackle world-record fish (Gabelhouse estimates of PSD, and the efficacy of this index in picking out 1984). In accordance with this method, length-categorization the differences in the length structure of different samples. standards for stock, quality, preferred, memorable, and trophy The applicability of the two methods was evaluated with length were established by adopting the percentages suggested regard to each species by comparing the percentages of sampling by Gabelhouse (1984) but adapted to the local conditions. In- sites for which it was not possible to determine PSD because deed, because there are not official world records for the fish the length of all specimens was below stock length. species considered in this study, we used the largest fish in our According to Willis and Scalet (1989) a fish population with data set (TLmax) for each species (traditional approach). The a high PSD consists mainly of large individuals, whereas a pop- length of largest fish in the sample is typically used for species ulation with a low PSD consists mainly of small individuals. The lacking an official world record (Milewski and Brown 1994; ability of PSD to vary as a function of the fish mean length (Mi- Zick et al. 2007; Pedicillo et al. 2010). The minimum lengths randa 2007) was used to assess the efficacy of the two methods of the categories were established by using the arithmetic mean proposed in order to highlight the differences in length com- of each length class, rounded to the nearest 5-mm increment. position of the samples analyzed. For each species, the PSD For example, to select the minimum stock length for Horse Bar- values calculated by the two methods were plotted as a function bel, a length-range equivalent to 20–26% of the largest fish in of the mean length of specimens from each sampling site, and our data set (TLmax = 502 mm) was chosen; the corresponding the regression parameters were examined (Pedicillo et al. 2010). length class was 101–131 mm and the proposed stock length Our basic assumption was the relationship between percentage was 120 mm. of large fish in a sample and the corresponding mean length: the With regard to the second method (biological approach), in higher the percentage of large fish (and hence the PSD) is, the accordance with Gassner et al. (2003), two thresholds were higher the mean length of fish in the sample will be. An AN- defined for each species considered: the asymptotic length (L∞) COVA was used to compare the two relationships (traditional of the von Bertalanffy (1938) growth function and the length at approach versus biological approach) for each species using maturity (Lm). mean TL as continuous predictor and the methods (traditional The asymptotic length was expressed as the mean value esti- and biological approaches) as factor. The validity conditions for mated in the study area; the von Bertalanffy growth parameters the ANCOVA have been checked by means of the Levene’s test Downloaded by [Department Of Fisheries] at 22:31 28 February 2013 were calculated by using the mean length at age, for homogeneity of variances. To demonstrate the utility of stock density indices for assess- [−k(t−t )] ment of fish-size structure and to examine the sensitivity and L = L∞ 1 − e 0 , t applicability of the methods to variations in length structure, the stock and quality lengths for Cavedano Chub were applied to where Lt is the theoretical length at age t, k is the rate at which the sampling data from a Chiascio River site. This population the asymptotic length is approached, and t0 is the theoretical was sampled during the autumn in two different years, 1999 and age (in years) at which the length of the specimens is zero. 2005. Indeed, about this time changes in the quality habitat of Some populations do not have asymptotic growth trajectories; the sample site occurred and at the same time the population therefore, the L∞ and k parameters calculated by means of this underwent changes in size structure. equation may be unrealistic (Zivkovˇ et al. 1999). To ensure Confidence intervals (CIs) for PSD were calculated on the that our analyses were not affected by such problems, we ex- basis of sample size of stock-length fish using normal approx- cluded populations with L∞ more than 50% greater than the imation (Gustafson 1988). The mean values of PSD calculated maximum length observed in each population (Pedicillo et al. for each species by means of the two methods were com- 2010). pared using the Mann-Whitney U-test. The chi-square test for 156 PEDICILLO ET AL.

TABLE 1. Total length (mm) descriptive statistics for ﬁsh species collected in overall sample.

TL values (mm) Sampling Specimens 25th 75th Species sites (n)(N) Mean Median Minimum Maximum percentile percentile SD Horse Barbel 211 13,708 157.9 162.0 20.0 502.0 100.0 205.0 69.3 Cavedano Chub 253 24,873 140.3 120.0 15.0 470.0 75.0 191.0 79.3 Rovella 207 28,008 77.8 75.0 15.0 208.0 60.0 94.0 26.9 Italian Rifﬂe Dace 113 12,996 76.9 75.0 20.0 190.0 52.0 98.0 28.1

differences in probabilities according to Conover (1980) was (Table 2). Based on the biological approach, the von Berta- used to analyze the differences in PSD values from two sam- lanffy growth function parameters of the populations used to pling sites. The thresholds calculated for preferred, memorable, generate the mean asymptotic length and the length at matu- and trophy classes are not discussed here. They have, neverthe- rity for each species can be found in Appendix 1 (available less, been provided, as they may contribute to the development online at https://bio.unipg.it/download/PSD/Appendix1), while and dessemination of these indices. the length-categorization system is summarized in Table 3. The two methods produced different results. Specifically, the proposed stock length according to the traditional approach RESULTS was considerably lower than that calculated by the biological The length composition of the total sample for each species approach in all species selected. This aspect influences the pos- considered is reported in Table 1. The maximum TL observed sibility of applying these structure indices. Indeed, the higher for each species was used to define the thresholds for the stock- the value of the stock length is in relation to the lengths typ- density index calculation according to the traditional approach ical of the species, the greater the probability will be that the

TABLE 2. Traditional approach: classiﬁcation of the length classes and minimum thresholds for index calculation. The thresholds were established by adopting the percentages suggested by Gabelhouse (1984) and calculated on the basis of the largest ﬁsh in the data set (TLmax). Percent (%) of the Length classes Minimum length maximum length (mm) based on (mm) for size Species Size category (Gabelhouse 1984) maximum length category

Horse Barbel (TLmax = 502 mm) Stock (S) 20–26 100–131 120 Quality (Q) 36–41 181–206 190 Preferred (P) 45–55 226–276 250 Memorable (M) 59–64 296–321 310 Trophy (T) 74–80 371–402 390

Cavedano Chub (TLmax = 470 mm) Stock (S) 20–26 94–122 110

Downloaded by [Department Of Fisheries] at 22:31 28 February 2013 Quality (Q) 36–41 169–193 180 Preferred (P) 45–55 211–258 230 Memorable (M) 59–64 273–301 290 Trophy (T) 74–80 348–376 360

Rovella (TLmax = 208 mm) Stock (S) 20–26 42–54 50 Quality (Q) 36–41 75–85 80 Preferred (P) 45–55 94–114 100 Memorable (M) 59–64 123–133 130 Trophy (T) 74–80 154–166 160

Italian Rifﬂe Dace (TLmax = 190 mm) Stock (S) 20–26 38–4.9 40 Quality (Q) 36–41 68–78 70 Preferred (P) 45–55 86–105 90 Memorable (M) 59–64 112–122 120 Trophy (T) 74–80 141–152 150 STOCK DENSITY INDICES FOR FOUR FISH SPECIES IN CENTRAL ITALY 157

TABLE 3. Biological approach: classiﬁcation of the length classes and minimum thresholds for index calculation. The thresholds were established on the basis of the asymptotic length (L∞) of the von Bertalanffy (1938) growth function and the length at maturity (Lm). Equation for the size Minimum length (mm) Species Size category category calculation for size category Horse Barbel Stock (S) S = Q − [(T − Q)/3] 220 Lm = 257.3 mm Quality (Q) Q = 257.3 260 L∞ = 454.8 mm Preferred (P) P = Q + [(T − Q)/3] 290 Memorable (M) M = Q + {[(T − Q)/3] × 2} 330 Trophy (T) T = 454.8 × 0.8 360 Cavedano Chub Stock (S) S = Q − [(T − Q)/3] 250 Lm = 290.8 mm Quality (Q) Q = 290.8 290 L∞ = 521.3 mm Preferred (P) P = Q + [(T − Q)/3] 330 Memorable (M) M = Q + {[(T − Q)/3] × 2} 370 Trophy (T) T = 521.3 × 0.8 420 Rovella Stock (S) S = Q − [(T − Q)/3] 120 Lm = 137.6 Quality (Q) Q = 137.6 140 L∞ = 226.6 Preferred (P) P = Q + [(T – Q)/3] 150 Memorable (M) M = Q + {[(T − Q)/3] × 2} 170 Trophy (T) T = 226.6 × 0.8 180 Italian Rifﬂe Dace Stock (S) S = Q − [(T − Q)/3] 110 Lm = 137.6 Quality (Q) Q = 137.6 130 L∞ = 226.6 Preferred (P) P = Q + [(T − Q)/3] 140 Memorable (M) M = Q + {[(T − Q)/3] × 2} 150 Trophy (T) T = 226.6 × 0.8 160

indices cannot be calculated for a given population, since all The stock and quality length values calculated by means of the specimens caught may be smaller than the stock length. On the two methods were applied to the remaining sampling sites the basis of stock lengths calculated by the traditional approach in order to determine the corresponding value of PSD. The for Horse Barbel, only in four instances (2%) out of a total of comparison of mean PSD values for each species highlights 211 sampling sites investigated were all specimens below this the statistical differences between the two methods (Table 5). threshold (Table 4); for the other species investigated, none of Moreover, the percentage of sampling sites with significant dif- the specimens caught at any of the sites were below the stock ferences between PSD values obtained from the two methods length. By contrast, when the biological approach was applied, varied from 47.4% for Horse Barbel and Cavedano Chub, to the number of sampling sites at which all specimens were below 79.0%, for Italian Riffle Dace. In these sampling sites in which the stock length was far higher, varying from 8% (9 out of 113) the PSD determined by the traditional approach was signifi- for Italian Riffle Dace to 14% for Horse Barbel (30 out of 211) cantly higher than that determined by the biological approach

Downloaded by [Department Of Fisheries] at 22:31 28 February 2013 and Rovella (30 out of 207). the percentage was very high for each species, ranging from 89.0%, for Horse Barbel up to 100.0%, for Rovella and Italian Riffle Dace. The relationship between the PSD and mean fish length from TABLE 4. Percentage of sampling sites in which it was not possible to cal- each sampling site is shown in Figure 2. The Traditional ap- culate the proportional stock density (PSD) because all specimens were smaller proach displayed a highly significant positive relationship be- than stock length. The number of sampling sites (x) out of the total (y) is reported as x/y in parentheses. tween PSD and fish length in all selected species (Table 6). When the biological approach was applied, however, no rela- Traditional Biological tionship emerged between these two parameters for any of the Species approach approach species analyzed, except for Cavedano Chub. For this species, both methods displayed a highly significant relationship be- Horse Barbel 2% (4/211) 14% (30/211) tween PSD and length, but the traditional approach yielded a Cavedano Chub 0% (0/253) 12% (30/253) higher correlation coefficient (r), coefficient of determination Rovella 0% (0/207) 14% (30/207) (r2), and slope than did the biological approach (Table 6). In Italian Riffle Dace 0% (0/113) 8% (9/113) each of the species considered, the differences between the two 158 PEDICILLO ET AL.

TABLE 5. Comparison between mean proportional stock density (PSD) values determined by a Mann–Whitney U-test. PSDT = PSD calculated by the traditional approach, PSDB = PSD calculated by the biological approach, min = minimum value, max = maximum value. PSDT = PSDB: percentage of sampling sites with signiﬁcant differences in PSD values obtained from the two methods (comparison performed by chi-square test in sampling sites with both reliable PSD values); PSDT > PSDB: percentage of sampling sites in which PSDT is signiﬁcantly higher than PSDB. The number of sampling sites (x) out of the total (y) is reported as x/y in parentheses.

PSDT = PSDB (%) Mean PSDT ± Mean PSDB ± Mann–Whitney (chi-square test: PSDT > Species 95% CI (min–max) 95% CI (min–max) U-test P < 0.05) PSDB (%) Horse Barbel 52.55 ± 3.57 (14–93) 40.09 ± 4.57 (12–77) Z = 2.337 47.4 (18/38) 89.0 (16/18) P = 0.019 Cavedano Chub 49.96 ± 3.33 (2–97) 51.46 ± 4.00 (21–79) Z =−0.364 44.7 (17/38) 94.1 (16/17) P = 0.716 Rovella 51.33 ± 3.29 (4–93) 35.97 ± 3.58 (18–63) Z = 4.551 68.4 (26/38) 100.0 (26/26) P < 0.001 Italian Rifﬂe Dace 58.67 ± 4.06 (10–98) 34.32 ± 5.23 (9–60) Z = 5.096 79.0 (15/19) 100.0 (15/15) P < 0.001

relationships proved to be highly signiﬁcant as determined from reported in Figure 3. In 1999 the biological water quality was ANCOVA. very poor. Indeed the extended biotic index (EBI; Ghetti 1986), The age structure of the Cavedano Chub population sampled an index used to evaluate the overall water quality based on the in the years 1999 and 2005 in a stretch of the Chiascio River is composition of the macrobenthic fauna, described the sampling Downloaded by [Department Of Fisheries] at 22:31 28 February 2013

FIGURE 2. Relationship between mean TL of ﬁsh samples analyzed and related proportional stock density (PSD) for each species analyzed. Dashed lines denote 0.95 conﬁdence intervals. STOCK DENSITY INDICES FOR FOUR FISH SPECIES IN CENTRAL ITALY 159

TABLE 6. Linear regression analysis (Proportional stock density [PSD] versus mean TL); comparison between the traditional approach and the biological approach was performed by ANCOVA.

Linear regression results Species Traditional approach Biological approach ANCOVA result Horse Barbel PSD =−0.07 + 3.10TL PSD = 41.46 + 0.13TL F = 23.97, P < 0.001 r2 = 0.337, r = 0.581, P < 0.001 r2 = 0.001, r = 0.031, P = 0.839 Cavedano Chub PSD =−7.61 + 3.66TL PSD = 25.14 + 1.32TL F = 17.71, P < 0.001 r2 = 0.521, r = 0.722, P < 0.001 r2 = 0.203, r = 0.451, P = 0.002 Rovella PSD =−54.42 + 13.31TL PSD = 25.43 + 1.31TL F = 153.41, P < 0.001 r2 = 0.654, r = 0.809, P < 0.001 r2 = 0.029, r = 0.172, P = 0.277 Italian Rifﬂe Dace PSD =−44.22 + 13.49TL PSD = 27.29 + 0.81TL F = 104.98, P < 0.001 r2 = 0.730, r = 0.855, P < 0.001 r2 = 0.016, r = 0.125, P = 0.589

site as an “environment very polluted” (Lorenzoni et al. 2010). as almost 54% of the specimens were between age 1 + and A total of 95 Cavedano Chub specimens were caught and the age 4 + . Moreover, fish density increased from 0.162 to 0.261 age structure of the population appeared to be unbalanced; fish specimens/m2 and the number age-0 + specimens was high, varied from age 1 + to age 5 + , and the population was dom- demonstrating the suitability of the site for reproduction of the inated by age-2 + fish (70% of the specimens); moreover, the species. Applying the length categorization calculated using the absence of age-0 + fish in the sample indicated that the site traditional approach the PSD value was 41 (80% CI =±6.1); was not suitable for the reproduction of the species. Applying n = 126 and 51 for stock- and quality-length fish, respectively, the length categorization calculated by means of the traditional out of a total of 208 fish caught. For the biological approach, approach to the sample, n = 73 for the stock-length fish and n = the sample sizes were insufficient to reliably calculate the PSD 20 for quality-length fish; the related PSD value was 27 (80% (stock-length fish, n = 4; quality-length fish, n = 1). CI =±7.7), demonstrating that there was a shortage of large fish in the population. Based on the biological approach, the DISCUSSION = stock- and quality-length fish were n 9 and 3, respectively, The assessment of fish populations provides fundamental and consequently the sample size was too small for a reliable information for the management of fish species. Indeed, the PSD estimate (Gustafson 1988). Over time the water quality species considered in this study are endemic in Italy and, except at the site improved and in 2005 the EBI value described the for Cavedano Chub, classified as “Near Threatened” (Rovella, sampling site as an “environment where signs of pollution can Italian Riffle Dace, and Horse Barbel) according to the Interna- be detected” (Lorenzoni et al. 2012). Moreover the Cavedano tional Union for Conservation of Nature criteria (IUCN 2001). Chub population appeared more balanced than it had been in Moreover, Cavedano Chub and Horse Barbel are important sport the past: the age structure comprised seven age-classes, from fishes. Readily obtainable indices that reflect the length structure + + age 0 to age 6 , with a better distribution of the specimens of fish populations and assist in making inferences about population parameters are needed by fisheries managers (Carline et al. 1984) from the point of view of both conservation and sus- Downloaded by [Department Of Fisheries] at 22:31 28 February 2013 tainable exploitation. Moreover, the establishment of the WFD in 2000 has raised an urgent need to expand our knowledge of aquatic biological communities (including fish) in Italian water- ways. The WFD mandates assessment of the ecological status of all European rivers, lakes, and transitional waters on the basis of a wide array of biotic variables, including the composition, abundance, and population structure of fish communities. North American stock density PSD and RSD indices may be useful proxies for determining the population structure of fish, a labor- intensive process that is demanded by the WFD. The two methods used in our study to calculate the reference values for the length categories in stock-density indices yielded fairly dissimilar thresholds. The values calculated by means of FIGURE 3. Age structures of Cavedano Chub population in Chiascio River the traditional approach were considerably lower than those cal- for the years 1999 and 2005. culated by the biological approach in all selected species. The 160 PEDICILLO ET AL.

definition of stock length is the minimum length of fish that pro- among those reported in the literature that might be adopted as a vides recreational value, while quality length is the minimum model and, if possible, also applied to other fish species, regard- size of a fish most anglers would like to catch (Anderson and less of whether their length at maturity is known. The reference Neumann 1996). Though Italian Riffle Dace and Rovella are values for the stock length calculated by means of the biological fished in Tiber River basin, no minimum legal limit exists for approach also seem to be too high. The value of the stock length these species. According to the traditional approach, the min- is a fundamental aspect in determining the possibility of apply- imum stock length for Italian Riffle Dace and Rovella would ing these structure indices. Indeed, the higher the value of the seem to be of little value when assessing the quality of fishery. stock length is, the greater the probability will be that the index In fact fish of this species are small, and in the Tiber River basin cannot be calculated, since all the specimens in the sample may they reach 70 mm on average. Hence, the length of 40 mm for be below this threshold. Analysis of the data reported in this Rovella, and 50 mm for Italian Riffle Dace are considered ac- study reveals that, for all the species investigated, the thresholds ceptable by local anglers. Moreover, because these species are calculated for this length class by means of the biological ap- “Near Threatened,” our aim is the establishment of minimum proach proved to be markedly higher than those determined by stock and quality lengths that reflect the biology of the species, means of the traditional approach. Thus, if the reference values and which will give the fisheries managers a tool for evaluating calculated by means of the biological approach are used, it is the quality of a population structure, not only from the point of highly likely that, for a given population, the indices cannot be view of the fishery but also for the conservation and sustainable calculated, as the lengths of all the specimens are below the exploitation of the species. stock length. According to the biological approach, the length at maturity When stock and quality lengths calculated by the two meth- and the asymptotic length were used to define the specific length ods were used to compute the PSD, they produced different classes for PSD and RSD calculation. Generally, the length at results. The PSD values yielded by the traditional approach which fish of a given population become mature is an impor- were higher than those yielded by the biological approach. tant biological parameter for the management (Jennings et al. According to Willis and Scalet (1989), low-PSD populations 1998). Froese and Binohlan (2000) observed that the age at first are dominated by small, slow-growing fish with low condition maturity is a function of size, and they gave an empirical for- factors. By contrast, high-PSD populations may be less dense, mula to calculate the length at maturity of a given population have better growth, and be in better condition. However, other based on the related asymptotic length, L∞. Therefore, defining factors can disrupt this relationship (e.g., exploitation) (Willis accurate L∞ values is critical because they are central to the et al. 1993; Pedicillo et al. 2010). Miranda (2007), in simu- computation of length categories using the biological approach. lated length distributions for three reference species (i.e., Large- Indeed, when L∞ increases the stock and quality lengths also mouth Bass Micropterus salmoides, crappies Pomoxis spp. and increase. To ensure that our analyses were not biased by un- Bluegill Lepomis macrochirus) at three levels of interval annual realistic L∞ values, we excluded populations with L∞ greater mortality, found that as mean length decreased, PSD values than 50% higher than the maximum length observed in each dropped. Guy and Willis (1990) noted that Largemouth Bass population (Pedicillo et al. 2010). Moreover, the coefficient of PSD was correlated with mean length of Largemouth Bass in determination (r2) may help to evaluate how well the model fits South Dakota ponds. Pedicillo et al. (2010) found a significant the data. In fact the growth curves used in this study have high relationship between PSD determined by the traditional method r2 values varying from 0.89 to 1.00, indicating that the models and mean length in Brown Trout, while no relation existed be- fit the data quite well. tween PSD calculated by the biological method and length. According to the biological approach, minimum quality Similarly in our study the results of the linear regression be- Downloaded by [Department Of Fisheries] at 22:31 28 February 2013 length is the size at maturity (Pedicillo et al. 2010; Volta 2010; tween the values of PSD and the mean length of the specimens Volta and Oggioni 2010), but the reference values calculated from the various populations revealed that the traditional ap- by this method seem to be too high with regard to the local proach yielded a significant positive relationship between the conditions for all species analyzed (Bianco and Santoro 2004; sizes of the specimens from the single populations and the re- Pompei et al. 2011). These results are in agreement with those lated PSD values. However, the r2 values suggested there is of Pedicillo et al. (2010) for Brown Trout. Moreover, for Rov- considerable variation in PSD not explained by its relationship ella and Italian Riffle Dace, the length at maturity calculated with TL. The regression model explaind only about one-third by means of the empirical formula provided by Froese and Bi- of the variation in PSD for Horse Barbel, while it explained nohlan (2000) proved to be close to the maximum length of nearly three-quarters of the variation for Italian Riffle Dace. Be- the specimens caught in large numbers from the watercourses cause Cavedano Chub and Horse Barbel showed greater sizes investigated. Actually the lengths at maturity for some species than Italian Riffle Dace and Rovella, they have a higher size considered in this study are available in the literature; however, range and consequently higher variance in length distribution. to determine length at maturity, we chose to use the more gen- For this reason some barbel or chub populations could have the eral formula proposed by Froese and Binohlan (2000). Indeed, same PSD value even if they have different mean lengths. This the aim of the present study was to select a general method from could explain the lower explained percentage by the model for STOCK DENSITY INDICES FOR FOUR FISH SPECIES IN CENTRAL ITALY 161

barbel and chub in the PSD–mean length relationship. When the At present the applicability of the length-category values ob- biological approach was used, however, no significant relation- tained in this study is limited to the Tiber River basin because ship emergeed between these two parameters. The two methods they are computed from fish sampled from this area. Impor- therefore displayed different behaviors according to the lengths tant future tasks will be to (1) test the application of indices of the specimens from the various samples analyzed. Indeed, calculated using the traditional approach on the largest possible regression analysis seemed to indicate that the traditional number of populations from the Tiber River basin, (2) define the method was more sensitive to changes in the average size of target values for balanced populations, and (3) correlate these specimens in the samples, and, hence, was better able to high- indices with other parameters, for example, body condition. light the differences in the length-frequency distributions of the Moreover, it will be important to increase our investigations by populations analyzed (Pedicillo et al. 2010). extending the application of these indices to other fish species. Generally, the first attempt to adapt stock-density indices to the local conditions of the Tiber River basin showed that these REFERENCES indices can be useful tools for characterizing fish populations. Anderson, R. O., and R. M. Neumann. 1996. Length, weight, and associated Indeed, the traditional approach applied to the Cavedano Chub structural indices. Pages 447–482 in B. R. Murphy and D. W. Willis, editors. population in the Chiascio River appeared to work quite well, Fisheries techniques, 2nd edition. American Fisheries Society, Bethesda, reflecting the temporal variations in age structure of tested pop- Maryland. Bagenal, T. B., and F. W. Tesch. 1978. Age and growth. Pages 101–136 in T. B. ulation. The objective range for PSD values is typically 30–60 Bagenal, editor. Methods for assessment of fish production in fresh waters, (Penczak et al. 1998) and indicates a balanced fish population. 3rd edition. Blackwell Science Publications, Oxford, UK. In 1999 the water quality of the sampling site was very poor Bianco, P. G., and E. Santoro. 2004. Biologia della rovella, del cavedano e del and the Cavedano Chub population was highly unbalanced. The barbo nei bacini del Parco Nazionale del Cilento e Vallo di Diano (Pisces; related PSD value confirmed this condition, being below the Cyprinidae). Biologia Ambientale 18:85–91. (In Italian.) Carline, R. F., B. L. Johnson, and T. J. Hall. 1984. Estimation and inter- objective range, and indicated a population dominated by small pretation of proportional stock density for fish populations in Ohio im- fish. Over time the water quality of the sampling site improved poundments. North American Journal of Fisheries Management 4:139– and the Cavedano Chub population appeared more balanced 154. with a better distribution of the specimens in the age-classes. Conover, W. J. 1980. Practical nonparametric statistics, 2nd edition. Wiley, New Again the PSD value agreed with this change and fells in the York. DeVries, D. R., and R. V. Frie. 1996. Determination of age and growth. Pages objective range, hence indicating a well-structured population. 483–512 in B. R. Murphy and D. W. Willis, editors. Fisheries techniques, 2nd On the contrary, when the stock and quality length calculated edition. American Fisheries Society, Bethesda, Maryland. by means the biological approach was applied, the sample size European Parliament and Council. 2000. Water framework directive 2000/60/EC became too small for reliable PSD estimates, even if the number establishing a framework for community action in the field of water policy. of fish caught was quite high in both years. We achieved similar Official Journal of the European Communities L327:1–73. Everhart, W. H., and W. D. Youngs. 1981. Principles of fishery science, 2nd results in a previous study (Pedicillo et al. 2010) in which we edition. Cornell University Press, Ithaca, New York. tested the methods on Brown Trout length-frequency data from Froese, R., and C. Binohlan. 2000. Empirical relationships to estimate asymp- 263 sampling sites in the Tiber River basin. In that study, when totic length, length at first maturity and length at maximum yield per recruit the sampling sites were disaggregated according to the various in fishes, with a simple method to evaluate length frequency data. Journal of fishing regulations (catch and release, unfished, and fished), the Fish Biology 56:758–773. Gabelhouse, D. W., Jr. 1984. A length-categorization system to assess traditional method was better able than the biological approach fish stocks. North American Journal of Fisheries Management 4:273– to reveal the differences in fish-length distribution among the 285. various types of management practices. Gassner, H., G. Tischler, and J. Wanzenbock.¨ 2003. Ecological integrity as- Downloaded by [Department Of Fisheries] at 22:31 28 February 2013 Therefore, all analyses seem to show that the traditional sessment of lakes using fish communities: suggestions of new metrics devel- method has greater sensitivity and efficacy to highlight the dif- oped in two Austrian prealpine lakes. International Review of Hydrobiology 88:635–652. ferences in the length-frequency data than does the biological Ghetti, P. F. 1986. I macroinvertebrati nell’analisi di qualita` dei corsi d’acqua: approach and has a greater practical applicability requiring less manuale di applicazione. Indice Biotico: E.B.I. Modificato, Provincia Au- sampling effort. Hence, this application could contribute to es- tonoma di Trento, Italy. (In Italian.) tablishing the ecological status of rivers, in accordance with Gustafson, K. A. 1988. Approximating confidence intervals for indices of fish the European Union’s WFD (WFD 2000). Moreover the stan- population size structure. North American Journal of Fisheries Management 8:139–141. dards provided in this study offer resource managers another Guy, C. S., R. M. Neumann, D. W. Willis, and R. O. Anderson. 2007. Propor- tool for identifying and establishing management goals, evalu- tional size distribution: a further refinement of population size structure index ating management objectives, and providing insight into envi- terminology. Fisheries 32:348. ronmental conditions. Nevertheless fisheries managers should Guy, C. S., and D. W. Willis. 1990. Structural relationships of Largemouth Bass use stock-density indices as only one of their assessment tools. and Bluegill populations in South Dakota ponds. North American Journal of Fisheries Management 10:338–343. By combining size-structure indices with other tools such as IUCN (International Union for Conservation of Nature). 2001. IUCN red catch per unit effort, fish condition, and growth assessment, list categories and criteria. IUCN Species Survival Commission, Gland, population and community analyses will be more reliable. Switzerland. 162 PEDICILLO ET AL.

Jennings, S., J. D. Reynolds, and S. C. Mills. 1998. Life history correlates the Assino Creek (Umbria, Italy). Knowledge and Management of Aquatic of responses to fisheries exploitation. Proceedings of the Royal Society of Ecosystems [online serial] 400:article 09. DOI: 10.1051/kmae/2011011. London B 265:333–339. Volta, P. 2010. Analisi della struttura di popolazione di coregone lavarello Lorenzoni, M., V.Angeli, A. Bicchi, G. Pedicillo, A. Carosi, P. Viali, G. Baldini, (Coregonus lavaretus Linneus 1758) in tre laghi profondi italiani medi- L. Ghetti, A. Zeetti, M. Natali, A. Biscaro Parrini, R. Dolciami, A. Mezzetti, ante indici stock density. Studi Trentini di Scienze Naturali 87:257–260. L. Tancioni, M. C. Bocale, S. Manfredini, and G. Tozzi. 2012. La carta ittica (In Italian.) di II livello del bacino dei fiumi Chiascio e Topino, aggiornamenti. Regione Volta, P., and A. Oggioni. 2010. Specie ittiche chiave e tipo-specifiche nei laghi Umbria, Perugia, Italy. (In Italian.) naturali dell’ecoregione Alpina: approccio storico e proposta di metriche per Lorenzoni, M., L. Ghetti, A. Carosi, and R. Dolciami. 2010. La fauna ittica ed l’analisi dello stato di qualita` della fauna ittica ai sensi della direttiva sulle i corsi d’acqua dell’Umbria. Regione Umbria, Perugia, Italy. (In Italian.) acque 2000/60/CE. [Key- and type-specific fish species in natural lakes of Lorenzoni, M., M. Mearelli, and L. Ghetti. 2006. Native and exotic fish species Italian Alpine ecoregions reconstructed from historical data: a preliminary in the Tiber River watershed (Umbria - Italy) and their relationship to the index to assess the quality status of fish fauna according to WFD 2000/60/CE.] longitudinal gradient. Bulletin Français de la Pecheˆ et de la Pisciculture Studi Trentini di Scienze Naturali 87:97–104. (In Italian.) 382:19–44. von Bertalanffy, L. 1938. A quantitative theory of organic growth. Human Mearelli, M., M. Lorenzoni, G. Giovinazzo, and M. L. Petesse. 1994. Carta ittica Biology 10:181–213. della regione Umbria: metodologie adottate e risultati. Rivista di Idrobiologia Willis, D. W., C. S. Guy, and B. R. Murphy. 1991. Development and evaluation 33:129–149. of a standard weight (Ws) equation for Yellow Perch. North American Journal Milewski, C. L., and M. L. Brown. 1994. Proposed standard weight (Ws) equa- of Fisheries Management 11:374–380. tion and length-categorization standards for stream-dwelling Brown Trout Willis, D. W., B. R. Murphy, and C. S. Guy. 1993. Stock density indices: (Salmo trutta). Journal of Freshwater Ecology 9:111–116. development, use, and limitations. Reviews in Fisheries Science 1:203– Miranda, L. E. 2007. Approximate sample sizes required to estimate length 222. distributions. Transactions of the American Fisheries Society 136:409–415. Willis, D. W., and C. G. Scalet. 1989. Relations between proportional stock den- Pedicillo, G., A. Carosi, L. Ghetti, and M. Lorenzoni. 2010. Population size sity and growth and condition of Northern Pike populations. North American structure indices and growth standards for Salmo (trutta) trutta Linnaeus, Journal of Fisheries Management 9:488–492. 1758 in central Italy. Knowledge and Management of Aquatic Ecosystems Zick, D., H. Gassner, M. Rinnerthaler, P. Jager,¨ and R. A. Patzner. 2007. [online serial] 399:article 02. DOI: 10.1051/kmae/2010030. Application of population size structure indices to Arctic Charr Salvelinus Penczak, T., Ł. Głowacki, W. Galicka, and H. Koszal´ınski. 1998. A long-term alpinus (L.) in Alpine lakes in Austria. Ecology of Freshwater Fish 16:54– study (1985–1995) of fish populations in the impounded Warta River, Poland. 63. Hydrobiologia 368:157–173. Zivkov,ˇ M. T., T. A. Trichkova, and G. N. Raikova-Petrova. 1999. Biological Pompei, L., A. Carosi, G. Pedicillo, E. Rocchini, and M. Lorenzoni. 2011. reasons for the unsuitability of growth parameters and indices for comparing Age and growth analysis of the chub, Squalius squalus (Bonaparte, 1837), in fish growth. Environmental Biology of Fishes 54:67–76. Downloaded by [Department Of Fisheries] at 22:31 28 February 2013 This article was downloaded by: [Department Of Fisheries] On: 28 February 2013, At: 22:31 Publisher: Taylor & Francis Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

North American Journal of Fisheries Management Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/ujfm20 Experimental Evaluation of Size-Dependent Predation by Adult Post-Spawned Rainbow Smelt on Larval Lake Whitefish Dimitry Gorsky a c & Joseph Zydlewski b a Department of Wildlife Ecology, University of Maine, Orono, Maine, 04469, USA b U.S. Geological Survey Maine Cooperative Fish and Wildlife Research Unit, Department of Wildlife Ecology, University of Maine, Orono, Maine, 04469, USA c U.S. Fish and Wildlife Service, Lower Great Lakes Fish and Wildlife Conservation Office, 1101 Casey Road, Basom, New York, 14013, USA Version of record first published: 30 Jan 2013.

To cite this article: Dimitry Gorsky & Joseph Zydlewski (2013): Experimental Evaluation of Size-Dependent Predation by Adult Post-Spawned Rainbow Smelt on Larval Lake Whitefish, North American Journal of Fisheries Management, 33:1, 163-169 To link to this article: http://dx.doi.org/10.1080/02755947.2012.750632

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material. North American Journal of Fisheries Management 33:163–169, 2013 C American Fisheries Society 2013 ISSN: 0275-5947 print / 1548-8675 online DOI: 10.1080/02755947.2012.750632

ARTICLE

Experimental Evaluation of Size-Dependent Predation by Adult Post-Spawned Rainbow Smelt on Larval Lake Whiteﬁsh

Dimitry Gorsky*1 Department of Wildlife Ecology, University of Maine, Orono, Maine 04469, USA Joseph Zydlewski U.S. Geological Survey Maine Cooperative Fish and Wildlife Research Unit, Department of Wildlife Ecology, University of Maine, Orono, Maine 04469, USA

Abstract Introduced landlocked Rainbow Smelt Osmerus mordax are hypothesized to be a major factor in the decline of Lake Whitefish Coregonus clupeaformis populations in many lakes. We sought to identify the size of Lake Whitefish preyed upon by adult Rainbow Smelt and how the efficiency of Rainbow Smelt predation changes Lake Whitefish ontogeny. In a laboratory setting, we exposed larval Lake Whitefish of increasing sizes to groups of seven Rainbow Smelt (>100 mm) for a 1-h period and observed predation behaviors and efficiencies. In each trial, the group of Rainbow Smelt consumed at least one larval Lake Whitefish, which were up to 45 mm in length (up to 89% of Rainbow Smelt gape width). Predation efficiency, the total number of Lake Whitefish consumed by Rainbow Smelt during trials, was 100% after Lake Whitefish hatching and followed a decreasing sigmoidal response to increasing lengths of Lake Whitefish. The apparent predatory window of Rainbow Smelt on Lake Whitefish is from Lake Whitefish size at hatch (∼12 mm) to approximately 34 mm. Rainbow Smelt generally required multiple attacks to capture a single Lake Whitefish. The capture efficiencies of Rainbow Smelt decreased from 30% as Lake Whitefish length increased and were highly variable within each Lake Whitefish size-group. The overall impact that Rainbow Smelt predation will have on Lake Whitefish populations is dependent on the growth rate of Lake Whitefish, environmental conditions that cause the Lake Whitefish hatching period to coincide with Rainbow Smelt spawning events, and the degree of overlap in habitat use between spawning Rainbow Smelt, nonspawning subadult Rainbow Smelt, and hatching Lake Whitefish. Downloaded by [Department Of Fisheries] at 22:31 28 February 2013 Populations of smelts of the genus Osmerus have been iden- Myers et al. 2009). In both the Laurentian Great Lakes and tified as major larval predators in freshwater lakes of both North smaller inland lakes, Rainbow Smelt can affect recruitment of America and Eurasia (Foltz and Norden 1977; Sterligova 1979; native coregonids, specifically Cisco Coregonus artedi and Lake Loftus and Hulsman 1986; Myers et al. 2009). In inland lakes Whitefish C. clupeaformis (O’Gorman 1974; Loftus and Huls- throughout North America, Rainbow Smelt O. mordax popu- man 1986; Hrabik et al. 1998; Myers et al. 2009). Although lations are often nonnative as a result of both legal and illegal Rainbow Smelt predation has been implicated in causing re- stocking and subsequent expansions (Evans and Waring 1987). cruitment declines for Lake Whitefish and Ciscoes in the Great These introduced populations are often implicated in the decline Lakes, other factors certainly have contributed, such as over- of native fish populations (Anderson and Smith 1971; Loftus fishing, habitat alteration, invasive species, and changes in the and Hulsman 1986; Evans and Waring 1987; Evans et al. 1988; lower food webs (Hartman 1973; Christie 1974; Hoff 2004).

*Corresponding author: dimitry [email protected] 1Present address: U.S. Fish and Wildlife Service, Lower Great Lakes Fish and Wildlife Conservation Ofﬁce, 1101 Casey Road, Basom, New York 14013, USA. Received August 29, 2011; accepted November 7, 2012 163 164 GORSKY AND ZYDLEWSKI

In order to be efficient predators of larval fish Rainbow Smelt occurred by the third week of November. Fish were removed predation on larval coregonids is gape-limited at a threshold size from the trap nets and transported to shore for processing. Eggs of greater than 100 mm (Selgeby et al. 1978; Loftus and Huls- were manually expelled and fertilized. They were then water- man 1986; Haakana and Huuskonen 2009; Myers et al. 2009). hardened at Clear Lake and transported to the Enfield State Fish Loftus and Hulsman (1986) found strong evidence of Rainbow Hatchery, Enfield, Maine, where they were incubated in McDon- Smelt predation on Lake Herring (i.e., Cisco) and Lake White- ald jars. Before hatch, eggs were transported to the Aquaculture fish, where Rainbow Smelt ranged from 85 to 215 mm. Mac- Research Center located at the University of Maine, Orono. The Crimmon and Pugsley (1979) found fish remains in Rainbow eggs remained in McDonald jars until hatching. After hatch- Smelt over 160 mm, and Stedman and Argyle (1985) found core- ing, Lake Whitefish larvae were held in dark blue, cylindrical, gonid remains in Rainbow Smelt over 100 mm. This size trans- plastic tanks measuring 30 cm in diameter and 60 cm tall. Lake lates to age-2 and older Rainbow Smelt in the Great Lakes and Whitefish began consuming 2-d-old Artemia nauplii (Great Salt similar or older fish for other lakes with these species present. Lake strain, ∼450 µm; Brine Shrimp Direct, Ogden, Utah) at The emergence of Cisco and Lake Whitefish larvae from 2 d posthatch. Water temperature was maintained at 10◦Cin spawning substrates coincides temporally with Rainbow Smelt each tank using a flow-through system with a water bath, and spawning in the early spring, closely following ice-out. Spawn- overhead fluorescent lights provided light at a 12 h light : 12 h ing locations for these coregonids vary from lake inlets and dark cycle at a level of 350 lx. streams to windswept shores, but commonly comprise clean, Larval whitefish were separated into 20 size-groups of 15 fish rocky substrate with sufficient interstitial spaces for egg incu- each, and group mean fork lengths ranged from 9.5 to 44.5 mm bation (Hart 1931; Nester and Poe 1984; Begout´ Anras et al. (Table 1). These groups were each placed in 37-L glass aquaria, 1999). Eggs are incubated over the winter and hatch when wa- the sides of which were blackened. Water temperature was main- ter temperatures approach 10◦C (Faber 1969; Nester and Poe tained at 10◦C and 50% water exchanges were performed daily. 1984; Begout´ Anras et al. 1999). Similarly, Rainbow Smelt tend Fish were fed 600 Artemia/L, which were rinsed with freshwater, to use windswept shores, shoals, and streams for spawning. five times a day throughout the experiment. Adult landlocked After spawning, adult Rainbow Smelt are voracious feeders, Rainbow Smelt were obtained after spawning from bait dealers as feeding is often suspended during spawning and followed who legally collected them through the ice from local systems. by intense feeding behaviors (MacCrimmon and Pugsley 1979; All Rainbow Smelt were of sexual maturity and therefore con- Loftus and Hulsman 1986). If larval coregonid spawning and sidered adult. Twenty-one Rainbow Smelt, 113–128 mm TL, nursery habitat occurs in proximity to Rainbow Smelt spawning were used in this experiment (Table 2). Seven adult Rainbow areas, Rainbow Smelt predation could have devastating effects Smelt were placed in each of three tanks containing 100 L of on coregonid recruitment. flow-through freshwater at ∼10◦C. Rainbow Smelt were not fed Applying the hypothesis that predation by adult Rainbow outside the actual predation trials and were acclimated for 3 d Smelt may lower recruitment of Lake Whitefish, we sought to before the first trial. characterize the predation efficiency of adult Rainbow Smelt on Predation trials.—Before each predation trial, digital im- larval Lake Whitefish in a laboratory setting and in so doing ages were taken of larval Lake Whitefish, grouped by tank, establish how and when predation may be responsible for low for size measurements. All larvae in a single trial were anes- recruitment rates of coregonids in general. We hypothesized that thetized using 5 mL of tricaine methanesulfonate (MS-222; overall Rainbow Smelt predation rates and predation efficiency 100 mg/L, buffered with 0.20 mM NaCO3,pH= 7.0) in 250 mL would decrease as larval Lake Whitefish length increases, up to of water. All larvae were photographed together on a 20–mm- a point where Rainbow Smelt are no longer efficient predators grid petri dish using a digital camera (Canon Powershot S3 IS, Downloaded by [Department Of Fisheries] at 22:31 28 February 2013 on Lake Whitefish larvae. Our objective was to determine the 6.0 mp) mounted 30 cm above the petri dish. Size measurements length range of larval Lake Whitefish that are susceptible to were determined using digital imaging techniques and ImageJ Rainbow Smelt predation, which is an important component software (D. Gorsky and J. Zydlewski, unpublished; Rasband, of assessing risk from spatiotemporal overlap with predatory ImageJ: http://rsb.info.nih.gov/ij/). Larval fish were revived in Rainbow Smelt. aerated freshwater and then allowed to recover for an hour before experimentation. Photographic measurements of Rainbow Smelt were again taken at the conclusion of all of the predation METHODS trials. Study species.—Sexually mature Lake Whitefish were col- Every 4 d, three predation trials were performed using each lected during spawning periods using 1.5-m-deep trap nets with group of Rainbow Smelt once for a total of 20 trials taking 30.5-m leads. Trap nets were set in 1–2 m of water along place on seven separate days between June 26 and July 13, known spawning movement corridors in Clear Lake (township 2009. We used the same three groups of Rainbow Smelt for T10 R11 WELS, Piscataquis County, Maine; 46◦3116.02N, each set of predation trials without exchanging individuals. Tri- 69◦733.97W). Trap nets were checked twice a week starting als were performed in the same tanks that the Rainbow Smelt October 15 until ice restricted access to the nets, which typically were held in. These tanks were circular white tanks, 1 m across RAINBOW SMELT PREDATION ON LARVAL LAKE WHITEFISH 165

TABLE 1. Mean Lake Whiteﬁsh length and measurements of documented predation behaviors for predation trials using Rainbow Smelt. Predicted body height of larval ﬁsh was calculated using the regression of length to body height.

Predicted Larval Mean larval body height whitefish Trial efficiency Capture efficiency Predator behavior length (mm) (mm) used (n) Pursuits Attacks Captures (Captures/fish) (Captures/Attack) (Attacks/Pursuit) 9.5 2.1 15 1.00 12.8 2.5 15 1.00 26.7 4.4 15 26 90 13 0.87 0.14 3.46 28.9 4.6 15 33 60 10 0.67 0.17 1.82 29.5 4.7 15 33 56 8 0.53 0.14 1.70 30.0 4.8 15 36 66 13 0.87 0.20 1.83 30.2 4.8 15 33 62 9 0.60 0.15 1.88 30.5 4.9 14 87 124 8 0.57 0.06 1.43 30.6 4.9 15 17 37 10 0.67 0.27 2.18 31.3 5.0 15 43 63 4 0.27 0.06 1.47 33.2 5.2 15 57 58 6 0.40 0.10 1.02 33.6 5.3 13 24 16 3 0.23 0.19 0.67 35.3 5.5 15 26 21 3 0.20 0.14 0.81 35.5 5.5 15 30 6 1 0.07 0.17 0.20 35.7 5.5 15 38 54 4 0.27 0.07 1.42 37.1 5.7 17 26 48 5 0.29 0.10 1.85 38.9 5.9 14 17 13 3 0.21 0.23 0.76 39.4 6.0 15 36 12 1 0.07 0.08 0.33 42.8 6.5 14 41 33 3 0.21 0.09 0.80 44.5 6.7 15 32 27 2 0.13 0.07 0.84

and 1.5 m deep. During predation trials, the water level was Predation trials were videotaped for 60 min using a high- lowered to improve visibility from the surface to approximately definition, digital, video recorder and wide-angle lens (Canon 25 cm depth. For every set of predation trials, we rotated the Vixia HV30) fixed to a frame at a distance of 1 m from the tanks of Rainbow Smelt that would receive the smallest Lake water surface, directly over the tank. Videos were transferred Whitefish group. Trials were performed at 1000 hours. A total to a computer and reviewed using Microsoft Media Player (Mi- of 15 Lake Whitefish larvae were introduced to each tank of crosoft, Redmond, Washington). During video review, we docu- Rainbow Smelt. Two Lake Whitefish larvae were introduced at mented “pursuits,” “attacks,” “captures,” and, at the end of each a time at 2-min intervals using a PVC tube that placed the fish trial, the number of actively swimming larval fish remaining. at the side of the tank. The final individual larva was introduced We identified “pursuits” as movements consisting of multiple singly. thrusts and sudden direction change towards larvae. “Attacks” Downloaded by [Department Of Fisheries] at 22:31 28 February 2013

TABLE 2. Total length (TL), weight, and gape measurements of adult Rainbow Smelt for each tank used in predation trials.

Predation trial Measurement A B C Length range (mm) 113–125 119–128 118–128 Mean length (mm) ± SE 119.7 ± 1.8 122.3 ± 1.5 121.9 ± 1.7 Weight range (g) 7.4–10.4 8.3–11.8 8.4–12.4 Mean weight (g) ± SE 8.9 ± 0.4 9.8 ± 0.5 9.7 ± 0.5 Gape height range (mm) 11.7–15.2 10.8–12.8 11.5–13.7 Mean gape height (mm) ± SE 13.3 ± 0.5 11.7 ± 0.3 12.7 ± 0.3 Gape width range (mm) 6.6–8.3 7.3–8.0 7.5–8.2 Mean gape width (mm) ± SE 7.3 ± 0.2 7.5 ± 0.1 7.8 ± 0.1 166 GORSKY AND ZYDLEWSKI

were aggressive pursuits sometimes involving a C-start type of attack, where an individual curls its body followed by a quick flex back to straight to provide high burst-swimming movement. “Pursuits” differed from “attacks” in that Rainbow Smelt broke off the pursuits before larvae were in striking distance. Finally “captures” included attacks that caused larval Lake Whitefish mortality either by injury or consumption. The first two predation trials were performed in a square tank without video review so only the trial efficiency was measured. To determine whether Rainbow Smelt were gape-limited during any predation trial performed, we compared the limiting dimensions of each species, gape width for Rainbow Smelt, and head height of Lake Whitefish. Gape measurements for Rainbow Smelt were performed using a caliper, and Lake Whitefish head height was measured using the same procedure as measuring the fish’s TL. Data analysis.—Three summary statistics of predation were analyzed across all size trials: (1) overall predation rates, FIGURE 1. Overall predation rate of adult Rainbow Smelt preying on lar- (2) capture efficiency, and (3) pursuits per attack. Overall pre- val Lake Whitefish of increasing mean total lengths. Predation rate represents dation rates were measured based on the total number of prey the number of captured prey per total number of prey introduced. Error bars represent the SE in larval lengths used in each predation trial. captured relative to the total number of larval fish introduced during the 60-min period. Variation in predation rates relative to fish size was modeled using a logistic functional response model tween the number of attacks needed for capture and larval Lake in which Lake Whitefish larval length was used as the indepen- Whitefish length was not significant (F = 1.68, P = 0.214). dent variable and overall Rainbow Smelt predation rate was used 1, 16 Variation in capture efficiency tended to vary with larval size, as the dependent variable. Capture efficiency was calculated by and the highest variability in capture success occurred at smaller observing the frequency of attacks resulting in successful cap- lengths. Adult Rainbow Smelt were never able to consume larval ture during the trial period. The capture efficiency was then fish with high capture efficiencies (>30% capture efficiency). regressed against larval length using linear regression. Finally, For larval fish with lengths less than 35 mm, Rainbow Smelt the number of pursuits per attack was determined to identify capture efficiency ranged between 6% and 31%, while efficien- predatory behavior changes across larval lengths. These data cies for larvae above 35 mm ranged between 7% and 23%. were likewise analyzed using linear regression. All regressions Across size-groups, the maximum capture efficiency generally and coefficients were tested using an α-value of 0.05. decreased as larval lengths increased (Figure 2).

RESULTS Predation by adult Rainbow Smelt on larval Lake Whitefish occurred in every trial performed (Figure 1). During the first two trials, when larvae were less than 15 mm, all larval Lake Whitefish were consumed by adult Rainbow Smelt. Predation rates were high (>50%) for larval Lake Whitefish with lengths Downloaded by [Department Of Fisheries] at 22:31 28 February 2013 less than 30 mm (Figure 1). Overall predation dropped quickly from approximately 50% to below 10% as larval Lake Whitefish lengths increased from 25 to 40 mm. Once larval fish lengths exceeded 40 mm, predation was low; typically only one larval Lake Whitefish was consumed per trial. The change in predation rates followed a logistic functional response with larval size as follows:

e(6.4−0.2x) π(y) = , 1 + e(6.4−0.2x)

where x is the mean length of the larval lake whitefish and π(y) 2 = = < FIGURE 2. Capture efficiency in individual tanks of Rainbow Smelt on larval is the predation rate (G 37.2, df 19, P 0.001). Lake Whitefish of different mean total lengths. Capture efficiency represents Adult Rainbow Smelt generally required multiple attacks to the number of attacks made per capture. Linear regression of data was not capture larval Lake Whitefish (Figure 2). The relationship be- significant. RAINBOW SMELT PREDATION ON LARVAL LAKE WHITEFISH 167

3.0

2.0 y = 83.93·e-0.13x r2 = 0.80, p<0.001

Pursuits/Attack 1.0

0.0 030354045 Mean lake whitefish length (mm)

FIGURE 3. Observed attacks per pursuits of Rainbow Smelt on larval Lake Whiteﬁsh of increasing mean total lengths. Error bars represent the SE in larval lengths used in each predation trial. Linear regression of data are signiﬁcant following the equation: y =−0.11x + 5.05, with r2 = 0.47 and P = 0.002.

Rainbow Smelt predatory behavior changed as larval Lake Whitefish lengths increased (Figure 3). The ratio of pursuits to 2 attacks followed a negative-sloped linear function (r = 0.80, FIGURE 4. Rainbow Smelt gape width and Lake Whitefish head height in F1, 16 = 27.5, P < 0.001). When larval fish were small (<35 mm) relation to the fork length (FL) of both species. Circles represent the gape width adult Rainbow Smelt were more likely to proceed from pursuit to of the 21 Rainbow Smelt used in predation trials in relation to FL. Triangles actual attack once a pursuit was initiated, as opposed to making represent the head height of Lake Whitefish in relation to FL. Length and head height of Lake Whitefish follow a significant linear regression: y = 0.13x − numerous pursuits before actually attacking larger larval fish 0.86, with r2 = 0.71 and P < 0.001. (>35 mm). At the conclusion of the study Rainbow Smelt had fork larval Lake Whitefish was from 12 mm, the typical size at lengths between 113 and 128 mm and gape widths between 6.6 hatch, to 43 mm, the approximate length when overall predation and 8.3 mm and had an overall mean of 7.6 mm (Table 2). Larval rate was below 10%. However, in systems like the Great Lakes Lake Whitefish head height increased with length and followed where smelt have been observed to feed on fish after reaching a significant linear relationship (r2 = 0.71, F = 121.6, P < 1, 50 lengths of 130 mm (MacCrimmon and Pugsley 1979; Evans and 0.001; Figure 4). The largest measured head height was 7.3 mm Waring 1987) and where age-3 and older adult Rainbow Smelt or 96% of the mean gape width of Rainbow Smelt and larger exceed 150 mm TL (Gorman 2007), the limit of the predatory than three of the smallest Rainbow Smelt’s gape width. Larval window for Lake Whitefish may exceed 43 mm. As Lake fish over 40 mm had a mean head height of 6.7 mm, which was Whitefish larvae grew, Rainbow Smelt became less efficient 89% of the mean gape width of Rainbow Smelt. Larval Lake at capturing Lake Whitefish and behaviorally were more likely Whitefish less than 40 mm in length had a mean head height of to pursue fish rather than attack them. The decline in trial and 5.3 mm or 70% of mean Rainbow Smelt gape width. capture efficiencies as Lake Whitefish length increased suggests Downloaded by [Department Of Fisheries] at 22:31 28 February 2013 that predator avoidance behaviors by Lake Whitefish and gape DISCUSSION limitation of Rainbow Smelt contributes to lower predation Rainbow Smelt interactions with Cisco and Lake Whitefish rates. Some of the larger Lake Whitefish groups reached head have been extensively studied (Anderson and Smith 1971; heights near the limit of the Rainbow Smelt gape width, and it Loftus and Hulsman 1986; Evans and Waring 1987; Evans et al. is possible some Rainbow Smelt were gape-limited and could 1988; Hrabik et al. 1998; Myers et al. 2009). The results from not prey on the larger Lake Whitefish. Before reaching this size, these studies have documented predatory ability by Rainbow avoidance behavior by Lake Whitefish may have already been Smelt on Lake Whitefish. This predation has been suggested driving predation efficiency down. Behaviors such as schooling, to be the cause for observed declines in whitefish recruitment swimming deeper, and burst swimming may reduce predation (Loftus and Hulsman 1986; Myers et al. 2009). Although our (Eklov¨ and Diehl 1994); all of these behaviors, including swim- experiment tested Rainbow Smelt predation on Lake Whitefish ming along the bottom, were observed in our study. Christensen in a laboratory setting, the results of our study demonstrateed (1996) suggested gape limitation often may not be realized since the size-dependent relationship between Lake Whitefish prey avoidance behaviors alone can control the maximum size of and Rainbow Smelt predation. In our experiment, the predatory prey ingested. In nature, any of the observed avoidance behav- window for Rainbow Smelt of 113–128 mm TL that consumed iors in our experiment would most probably increase survival 168 GORSKY AND ZYDLEWSKI

sharply, but in the confines of our experimental tanks, one or two At a majority of the sizes tested in this study, Rainbow Smelt avoidance maneuvers did not sufficiently remove the individual never had high capture efficiencies, but were nonetheless eventu- from the threat of predation, and thus our predation estimates ally successful in consuming a proportion of the larvae presented are most probably high. Natural environments would allow to them within the 1-h trial. Laboratory studies of predation have larvae to escape and evade the fish beyond the limits imposed shown similar or higher capture efficiencies for their respective by our 1-m-diameter experimental tanks. prey compared with our observations of Rainbow Smelt preda- The effectiveness of Rainbow Smelt predation on the smaller tion on Lake Whitefish (Christensen 1996; Brooking et al. 1998). larvae suggests that in lakes that contain both species, the highest Some of these studies showed capture efficiencies of above 50% potential impact of predation by Rainbow Smelt may occur soon for the smallest prey items. It is possible that Rainbow Smelt after Lake Whitefish larvae hatch and are occupying surface wa- capture efficiencies of larval Lake Whitefish may be higher ters near spawning areas. After hatching, larval Lake Whitefish when Lake Whitefish are closer to the size at which they hatch. occupy shallow surface waters for the first 3 weeks (Hart 1931). The first two trials of our study were performed using a slightly The sharp reduction in predation efficiency by Rainbow Smelt different design, and we were not able to quantify capture effi- on Lake Whitefish between 24 and 36 mm indicates that it is ciency when larval lengths were less than 20 mm. The larvae of essential for Lake Whitefish to grow quickly through this crit- the size we tested may have already begun to perform behaviors ical period to avoid predation. Growth rates of Lake Whitefish that limited predation efficiency. Brooking et al. (1998) first ob- are variable and depend on temperature and prey availability. served avoidance behavior in Walleye Sander vitreus larvae after Laboratory studies of larval Lake Whitefish growth rates show 24 d posthatch at lengths of 16.3 mm. This early observation that in typical spring temperatures (∼10◦C) and with abundant of avoidance behaviors may be occurring with Lake Whitefish. prey availability (>600 zooplanton/L), larval Lake Whitefish The fact that Rainbow Smelt capture efficiencies were always may take over 74 d from time of hatch to pass through the de- low may be an indication that the Lake Whitefish larvae used in fined predatory window (Gorsky and Zydlewski, unpublished). our experiments could already perform some level of predator As demonstrated by Myers et al. (2009), increases in the length avoidance. of time that larval fish are exposed to predation will lead to larval In conclusion, this study showed that if Rainbow Smelt mortality approaching 100%. For higher survival in the face of greater than 100 mm have high overlap in habitat use with Rainbow Smelt predation pressure, coregonid populations will Lake Whitefish larvae, they can be predators within a critical require abundant prey and appropriate temperatures to allow for size window, which could lead to larval whitefish recruitment faster growth, but additionally larval fish will also have to use failure. We determined that for Rainbow Smelt of 113–128 mm predatory avoidance techniques to ensure survival. TL, the predatory window for Lake Whitefish occurs from size To fully understand the magnitude of predation pressure of at hatch to about 43 mm. In systems such as the Great Lakes Rainbow Smelt on Lake Whitefish populations, information where adult Rainbow Smelt reach sizes >150 mm TL, the limit about individual populations and spawning habitat use in each of the predatory window for Lake Whitefish is probably greater. lake will be required. Past studies have been inconsistent in doc- Larvae may be able to increase survival if temperature and food umenting the influence of Rainbow Smelt predation on Lake conditions allow for higher growth, and if larvae use predator Whitefish or Cisco in general (Selgeby et al. 1978). Rainbow avoidance behaviors to minimize predation. In order to effec- Smelt are considered opportunistic feeders, so high densities and tively assess the impact of Rainbow Smelt predation on Lake habitat overlap may be required for this interaction to take place. Whitefish recruitment, knowledge about the abundance and size Specific examples show that when spawning habitats overlap, structure of the Rainbow Smelt population, and timing and habi- the effect of predation can be extremely high (Loftus and Huls- tat use by spawning and larval fish for both species in the lake Downloaded by [Department Of Fisheries] at 22:31 28 February 2013 man 1986; Myers et al. 2009). Additionally, when there are or area of the lake of interest is required. high densities of Rainbow Smelt, few prey items consumed per individual can still have major effects on recruitment. In Lake Ontario, cyclical patterns in recruitment were observed ACKNOWLEDGMENTS for Rainbow Smelt due to cannibalism in which low numbers of We thank Peter Bourque and John Boland of the Maine De- prey items per individual were documented (Lantry and Stewart partment of Inland Fisheries and Wildlife for their cooperation 2000). This suggests that even when low numbers of prey items and financial support for this project. We also thank Dave Basley per predator are observed, the effect of high predator density from the Ashland headquarters of the Maine Department of In- will cause the predatory influence to be more substantial. Past land Fisheries and Wildlife and Henry Hartley from the Enfield observations of low predation levels by Rainbow Smelt on larval State Fish Hatchery for their help in egg collection and rearing. fish suggest there is a weak effect due to predation. This may be We thank Ellen Marsden, Linda Kling, Yong Chen, and Michael an underrepresentation of the predatory impact simply because Kinnison for reviewing and commenting on the manuscript. Fi- of the high numbers of Rainbow Smelt that are able to prey on nally, we extend our gratitude to Neil Greenberg, Linda Kling, the individuals. Jackie Hunter, and all the undergraduate technicians at the RAINBOW SMELT PREDATION ON LARVAL LAKE WHITEFISH 169

University of Maine Aquaculture Research Center for providing Gorman, O. T. 2007. Changes in a population of exotic Rainbow Smelt in laboratory space and help with rearing live food and larval fish. Lake Superior: boom to bust, 1974–2005. Journal of Great Lakes Research This research was supported by the Maine Department of Inland 33(Supplement 1):75–90. Haakana, H., and H. Huuskonen. 2009. Predation of smelt on vendace larvae: Fisheries and Wildlife, U.S. Geological Survey Maine Cooper- experimental and field studies. Ecology of Freshwater Fish 18:226–233. ative Fish and Wildlife Research Unit, and the Department of Hart, J. L. 1931. The spawning and early life history of the whitefish, Core- Wildlife Ecology, University of Maine. Mention of trade names gonus clupeaformis (Mitchill), in the Bay of Quinte, Ontario. Contributions does not constitute endorsement by the U.S. Government. to Canadian Biology and Fisheries 6:165–214. Hartman, W. L. 1973. Effects of exploitation, environmental changes, and new species on the fish habitats and resources of Lake Erie. Great Lakes Fishery Commission Technical Report 22. REFERENCES Hoff, M. H. 2004. Biotic and abiotic factors related to Lake Herring recruitment Anderson, E. D., and L. L. Smith Jr. 1971. Factors affecting abundance of Lake in the Wisconsin waters of Lake Superior, 1984–1998. Journal of Great Lakes Herring (Coregonus artedii Lesueur) in western Lake Superior. Transactions Research 30(Supplement 1):423–433. of the American Fisheries Society 100:691–707. Hrabik, T. R., J. J. Magnuson, and A. S. McLain. 1998. Predicting the effects Begout´ Anras, M. L., P. M. Cooley, R. A. Bodaly, L. Anras, and R. J. P. of Rainbow Smelt on native fish in small lakes: evidence from long-term Fudge. 1999. Movement and habitat use by Lake Whitefish during spawn- research on two lakes. Canadian Journal of Fisheries and Aquatic Sciences ing in a boreal lake: integrating acoustic telemetry and geographic infor- 55:1364–1371. mation systems. Transactions of the American Fisheries Society 128:939– Lantry, B. F., and D. J. Stewart. 2000. Population dynamics of Rainbow Smelt 952. (Osmerus mordax) in Lakes Ontario and Erie: a modeling analysis of canni- Brooking, T. E., L. G. Rudstam, M. H. Olson, and A. J. VanDeValk. 1998. Size- balism effects. Canadian Journal of Fisheries and Aquatic Sciences 57:1594– dependent Alewife predation on larval Walleyes in laboratory experiments. 1606. North American Journal of Fisheries Management 18:960–965. Loftus, D. H., and P. F. Hulsman. 1986. Predation on larval Lake Whitefish Christensen, B. 1996. Predator foraging capabilities and prey antipredator be- (Coregonus clupeaformis) and Lake Herring (C. artedii) by adult Rainbow haviours: pre- versus postcapture constraints on size-dependent predator–prey Smelt (Osmerus mordax). Canadian Journal of Fisheries and Aquatic Sciences interactions. Oikos 76:368–380. 43:812–818. Christie, W. J. 1974. Changes in the fish species composition of the Great Lakes. MacCrimmon, H. R., and R. W. Pugsley. 1979. Food and feeding of the Rainbow Journal of the Fisheries Research Board of Canada 31:827–854. Smelt (Osmerus mordax) in Lake Simcoe, Ontario. Canadian Field-Naturalist Eklov,¨ P., and S. Diehl. 1994. Piscivore efficiency and refuging prey: the impor- 93:266–271. tance of predator search mode. Oecologia 98:344–353. Myers, J. T., M. L. Jones, J. D. Stockwell, and D. L. Yule. 2009. Reassessment Evans, D. O., J. J. Houston, and G. N. Meredith. 1988. Status of the Lake Simcoe of the predatory effects of Rainbow Smelt on ciscoes in Lake Superior. whitefish, Coregonus clupeaformis, in Canada. Canadian Field-Naturalist Transactions of the American Fisheries Society 138:1352–1368. 102:103–113. Nester, R. T., and T. P.Poe. 1984. Predation on Lake Whitefish eggs by Longnose Evans, D. O., and P. Waring. 1987. Changes in the multispecies, winter angling Suckers. Journal of Great Lakes Research 10:327–328. fishery of Lake Simcoe, Ontario, 1961–83: invasion by Rainbow Smelt, Os- O’Gorman, R. 1974. Predation by Rainbow Smelt (Osmerus mordax) on young- merus mordax, and the roles of intra- and interspecific interactions. Cana- of-the-year Alewives (Alosa pseudoharengus) in the Great Lakes. Progressive dian Journal of Fisheries and Aquatic Sciences 44(Supplement 2):182– Fish-Culturist 36:223–224. 197. Selgeby, J. H., W. R. MacCallum, and D. V. Swedberg. 1978. Predation by Faber, D. J. 1969. Ecological observations on newly hatched Lake Whitefish in Rainbow Smelt (Osmerus mordax) on Lake Herring (Coregonus artedii)in South Bay, Lake Huron. Pages 481–500 in C. C. Lindsey and C. S. Woods, western Lake Superior. Journal of the Fisheries Research Board of Canada editors. Biology of coregonid fishes: papers presented at the international 35:1457–1463. symposium on biology of coregonid fishes. University of Manitoba Press, Stedman, R. M., and R. L. Argyle. 1985. Rainbow Smelt (Osmerus mordax)as Winnipeg. predators on young Bloaters (Coregonus hoyi) in Lake Michigan. Journal of Foltz, J. W., and C. R. Norden. 1977. Food habits and feeding chronology of Great Lakes Research 11:40–42. Rainbow Smelt, Osmerus mordax, in Lake Michigan. U.S. National Marine Sterligova, O. D. 1979. Smelt Osmerus eperlanus (L.) and its role in ichthy- Fisheries Service Fishery Bulletin 75:637–640. ofauna of Lake Saäm¨ aj¨ arvi.¨ Journal of Ichthyology 19:793–800. Downloaded by [Department Of Fisheries] at 22:31 28 February 2013 This article was downloaded by: [Department Of Fisheries] On: 28 February 2013, At: 22:32 Publisher: Taylor & Francis Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

North American Journal of Fisheries Management Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/ujfm20 Growth and Survival of Largemouth Bass following Supplemental Feeding of Bluegills in Small Impoundments Stephen R. Woodard a , Russell A. Wright a & Dennis R. DeVries a a Department of Fisheries and Allied Aquacultures, Auburn University, 203 Swingle Hall, Auburn, Alabama, 36849, USA Version of record first published: 30 Jan 2013.

To cite this article: Stephen R. Woodard , Russell A. Wright & Dennis R. DeVries (2013): Growth and Survival of Largemouth Bass following Supplemental Feeding of Bluegills in Small Impoundments, North American Journal of Fisheries Management, 33:1, 170-177 To link to this article: http://dx.doi.org/10.1080/02755947.2012.754802

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material. North American Journal of Fisheries Management 33:170–177, 2013 C American Fisheries Society 2013 ISSN: 0275-5947 print / 1548-8675 online DOI: 10.1080/02755947.2012.754802

ARTICLE

Growth and Survival of Largemouth Bass following Supplemental Feeding of Bluegills in Small Impoundments

Stephen R. Woodard, Russell A. Wright, and Dennis R. DeVries* Department of Fisheries and Allied Aquacultures, Auburn University, 203 Swingle Hall, Auburn, Alabama 36849, USA

Abstract We investigated whether providing supplemental pellet feed to Bluegills Lepomis macrochirus in small impoundments increased their growth and reproductive rates after stocking and whether the supplemental feeding of Bluegills had positive indirect effects on Largemouth Bass Micropterus salmoides during the initial stocking year. In 2010, we stocked adult Bluegills and Largemouth Bass into nine 0.1-ha ponds; three replicate ponds were each provided with no pelleted food (control), a low ration (1.52 kg·ha−1·d−1), or a high ration (2.68 kg·ha−1·d−1). Supplemental feeding yielded increases in the growth of newly stocked Bluegills, the gonadosomatic index of female Bluegills, and the density of larval Bluegills within 4 months of stocking. In contrast, the sizes of adult and age-0 Largemouth Bass and the density and biomass of age-0 Bluegills during fall were not affected by feeding within the time frame of our study. The results demonstrate that supplemental pellet feeding is useful when the management goal is to increase Bluegill size and reproductive output and that these effects develop soon after stocking. We conclude that supplemental feeding is a beneﬁcial enhancement technique for application to Bluegills in recently stocked impoundments.

Small impoundments or ponds are manmade water features, Threadfin Shad Dorosoma petenense (Noble 1981), installing typically with a surface area of less than 40 ha, that are abundant fish attractors, and using supplemental pellet feed (Schmittou across the USA, with an estimated 2.6 million small impound- 1969; Berger 1982; Porath et al. 2003). Because fish growth ments located on private land (Smith et al. 2002; Dauwalter often is limited by food availability (Roff 1984), supplemental and Jackson 2005). These systems are used for a variety of feeding with prepared pellet feeds from feeders may increase purposes, such as storing water for irrigation, watering live- fish productivity and the condition of individuals in stunted fish stock, or swimming, but they are most commonly stocked with populations (Berger 1982; Murnyak et al. 1984).

Downloaded by [Department Of Fisheries] at 22:32 28 February 2013 Largemouth Bass Micropterus salmoides and Bluegills Lep- Supplemental feeding can act both as an attractant (i.e., to omis macrochirus for recreational fishing purposes (Swingle make Bluegills more available to anglers) and as a method of and Smith 1940; Swingle 1949; Modde 1980; Dauwalter and improving growth and condition of Bluegills and other prey Jackson 2005; Haley 2009). Bluegills are important prey for fish (Schmittou 1969; Berger 1982; Murnyak et al. 1984). This Largemouth Bass, and both of these species can experience increase in food availability should indirectly improve growth reduced growth in small impoundments due to competition, and condition of the top piscivore in the pond—Largemouth lack of food resources, or improper harvest (Swingle and Smith Bass. Although pelleted feed is unlikely to be directly con- 1940; Swingle 1949). Thus, owners of small impoundments em- sumed by Largemouth Bass themselves, large adult Bluegills ploy a wide variety of enhancement techniques to increase the benefit Largemouth Bass through the production of offspring, size of Bluegills and Largemouth Bass, maximize catch, and and an increase in adult Bluegill condition would likely re- maintain long-term harvest (Haley 2009). Pond enhancements sult in more or better-condition age-0 Bluegills to serve as can include the stocking of supplemental prey species such as prey items for Largemouth Bass. However, this potential link

*Corresponding author: [email protected] Received May 13, 2012; accepted November 20, 2012

170 LARGEMOUTH BASS GROWTH AND SURVIVAL 171

between supplemental pelleted feed and improved prey re- immediately after Bluegills spawned, adult Largemouth Bass sources for Largemouth Bass has not been quantified. (210–260 mm TL) were stocked into each pond at a rate of 250 Although the feeding of Bluegills can increase their size and fish/ha. condition (Schmittou 1969; Berger 1982; Murnyak et al. 1984; Supplemental pellet feeding occurred from April to August Hoagland et al. 2003; Twibell et al. 2003), the effects of supple- according to three experimental treatments: a no-feed treatment mental feeding on their reproductive rates is unknown. Mature (control), a low ration of pelleted feed (1.52 kg·ha−1·d−1), and female Bluegills in unfertilized ponds had lower gonad weights a high ration of pelleted feed (2.68 kg·ha−1·d−1). These ra- than mature females in fertilized ponds, suggesting preferential tions were chosen because pond owners rarely use feeding rates allocation of energy to somatic tissue when food is scarce (Aday greater than 1.12 kg·ha−1·d−1 (N. Haley, Alabama Coopera- et al. 2006). Thus, resource availability may influence Bluegill tive Extension System, personal communication), and so by total reproductive output, and increasing the resources available increasing the feeding rate above 1.12 kg·ha−1·d−1, we hoped to to Bluegills through supplemental feeding may increase their maximize the effects of feeding. The three treatments were ran- egg production and spawning frequency during each season. domly assigned to the nine ponds (3 ponds/treatment). Feed was This, in turn, may increase the food supply for and the condition dispersed by automatic directional fish feeders (23-L capacity; of Largemouth Bass during that season. To address these top- Moultrie Feeders, LLC), which were installed beside each treat- ics, we conducted a single-year study that included nine ponds ment pond. Timers on the feeders were calibrated to disperse during 2010. We examined whether supplemental feeding in the defined amount of food once daily at 0800 h. The supplied initially stocked impoundments produced changes in Bluegill feed was a 36% crude protein, 3.18-mm-diameter, commercial growth, condition, survival, density, and reproductive rates, as floating catfish pellet (F-R-M Catfish Fingerling Grower; Flint increases in these measures could potentially benefit Large- River Mills, Inc., Bainbridge, Georgia). mouth Bass populations. Further, we examined the lengths and To ensure that fish in all ponds experienced similar environ- weights of Largemouth Bass to determine whether the size of ments and thus to confirm that any observed changes in fish these fish was enhanced during that initial year in ponds where condition were due to feeding treatments, we assessed a variety Bluegills were provided with supplemental feed. of abiotic and biotic measures in all ponds. Water temperature and dissolved oxygen were measured at 1-m depth in the deep end of each pond; temperature was measured at 2-h intervals METHODS with Hobo pendant data loggers (Model UA-002-64), and dis- To determine whether a relationship existed between feeding solved oxygen (mg/L) was measured at 0900 hours every 3 d rates and Bluegill and Largemouth Bass condition during the with a Yellow Springs Instruments Model 51B meter. Every 14 initial stocking year, we conducted a whole-pond experiment at d, water samples were collected in 500-mL, dark polyethylene Auburn University’s E.W. Shell Fisheries Center, South Auburn bottles and were placed directly on ice. From these samples, tur- Unit, Auburn, Alabama. In November 2009, nine 0.1-ha ponds bidity (NTU) was measured with a nephelometer (Micro TPW; (2 m deep at one end and <0.5 m deep in the opposite end) were HF Scientific, Inc.); samples were then filtered onto a 47-mm- drained and allowed to dry. The ponds were filled in January diameter, glass-fiber disk (Millipore), and chlorophyll a was 2010, and the alkalinity in all ponds was 26–30 mg/L. All ponds extracted in cold 95% ethanol for 24 h. Chlorophyll-a concen- were fertilized with 10-34-0 (N-P-K) liquid fertilizer at a rate of trations (µg/L) were determined with a fluorometer (Turner De- 9.4 L/ha to stimulate a phytoplankton bloom (Wright and Masser signs Aquafluor; Welschmeyer 1994). Every 14 d, zooplankton 2004). Additional fertilizer was added periodically during the were also collected (at 0800–1000 hours) in the deep end of each experiment to maintain a uniform Secchi depth of 80 cm in all pond by conducting a vertical tow from 1 m to the surface with a Downloaded by [Department Of Fisheries] at 22:32 28 February 2013 ponds and to prevent excessive growth of weeds. During May, zooplankton net (30-cm diameter; 50-µm mesh). Samples were pondweeds Potamogeton spp. started to grow in the ponds; the preserved in 90% ethanol. In the laboratory, zooplankton were herbicide Reward (active ingredient: diquat dibromide) with a examined under a dissecting microscope until the entire sample sinking drift control adjuvant (Pro-Mate Sinker) was therefore was counted or until at least 200 individuals of the most abun- applied once to all ponds at a rate of 9.4 L/ha. Dissolved oxygen dant taxa were counted (Dettmers and Stein 1992; Welker et al. levels were monitored closely for the next 72 h; when the levels 1994). Cladocerans were identified to genus; copepods were dropped to less than 4.0 mg/L, water was flowed through all identified as cyclopoids, calanoids, harpacticoids, or nauplii. ponds to circulate and aerate the water. Dead weeds were raked Once per week during the Bluegill spawning season (April– out of the ponds after the herbicide treatment. October; Swingle and Smith 1942), larval fish were sampled In February, we stocked reproductively mature, small adult with a 0.5-m-diameter, 500-µm-mesh ichthyoplankton net that Bluegills (75–120 mm TL; Belk 1995) at the recommended was pulled across the length of each pond at approximately Bluegill fingerling (40–70 mm TL) rate of 2,500 fish/ha (Modde 1.0 m/s; two ichthyoplankton samples were collected from each 1980; Wright and Masser 2004; Dauwalter and Jackson 2005); pond on each sampling date. The net was fitted with a flowmeter the fingerling rate was used because there are no recommen- to allow determination of the towing speed and water volume dations for the stocking of Bluegills of this size. In April, sampled, which were used for estimation of larval fish density. 172 WOODARD ET AL.

To determine whether there were differences in growth rates tent (Glover et al. 2010). To determine whether Largemouth among Bluegill cohorts across feeding treatments (e.g., whether Bass were growing as a result of eating pelleted feed directly, adult Bluegills in the high-ration treatment could produce faster- we examined the stomachs of all adult Largemouth Bass and growing larvae after a repeated number of spawns), sagittal 10 age-0 Largemouth Bass collected from each pond at the end otoliths were removed from larval Bluegills (n = 5 fish from each of the experiment to verify whether feed was present. Finally, pond on each sample date) to estimate their age in days (Taubert to determine whether Largemouth Bass were in better condi- and Coble 1977). Otoliths were mounted on microscope slides tion from eating supplementally fed Bluegills, we measured the using thermoplastic cement, ground with 1,000-grit sandpaper caloric densities of adult Largemouth Bass by using the tech- on the sagittal plane, and polished. Two independent readers niques outlined by Glover et al. (2010). then counted daily rings of each otolith under oil immersion by Statistical analyses.—All data were analyzed with the Sta- use of a compound microscope (400 × magnification). If the tistical Analysis System (SAS; SAS Institute, Inc., Cary, North two age estimates were within 10% of each other, they were Carolina). Analysis of variance (ANOVA) tests were all mixed averaged; if not, they were recounted until a difference of less model (MIXED procedure in SAS; SAS 2008), wherein pond than 10% was reached. means (random factor) for each sampling date were nested To collect random samples of adult Bluegills and Large- within the three feeding treatments. Normality and homogene- mouth Bass, we sampled a 15-min electrofishing transect in ity of variance were assessed to ensure that the assumptions each pond every 30 d by using a boat-mounted, pulsed-DC of ANOVA were met. When significant treatment effects were electrofisher (5.0 GPP, 1,000 W; Smith-Root, Inc.). Collected detected, comparisons of least-squares means were used to ex- fish were weighed (g) and measured (mm TL) and then were amine where the differences existed. returned to the pond. At the end of the experiment (i.e., in Au- Values of Secchi depth, temperature, dissolved oxygen, tur- gust), ponds were drained and all fish were harvested. Adult and bidity, chlorophyll a, and zooplankton density collected during age-0 fish were randomly collected and counted, weighed (g), the field season were log-transformed to meet the assumptions measured (mm TL), and frozen for laboratory analyses. For age- of ANOVA; each variable was then analyzed with repeated- 0 Bluegills, a randomly drawn subsample of approximately 500 measures (RM) ANOVA. A first-order autoregressive covari- individuals from each pond was weighed to the nearest gram. ance structure was used in all of the RM-ANOVAs to account This weight was then extrapolated to estimate the total number for correlations among observations within the fixed feeding of age-0 Bluegills in the pond based on the entire weight of all treatments. For the fish that were obtained by electrofishing and collected age-0 Bluegills. measured for TL and weight, we tested for differences in mean For male and female Bluegills in frozen samples, we calcu- adult Bluegill weight, mean adult Largemouth Bass weight, rela- lated the gonadosomatic index (GSI) as an estimate of reproductive weight (Wr; used to estimate condition; Wege and Anderson tive potential (Crim and Glebe 1990). We used bomb calorimetry 1978), and TL among treatments by using RM-ANOVAs. We to determine whether there were differences in energy density also used RM-ANOVAs to test for among-treatment differences (i.e., condition) of both adult and age-0 Bluegills among feeding in mean density and growth (mm/d; from otolith measures) treatments. Adult Bluegills (>130 mm TL; n = 10 randomly of larval Bluegills that were collected during ichthyoplankton selected individuals from each pond) were dried at 70◦C until a sampling. final, constant mass (<0.01 g difference in mass between days) At harvest, percent survival of adult Bluegills or Largemouth was achieved and recorded. Each dried adult was then blended Bass was calculated as nf/ni, where ni is the initial number of to a homogeneous mixture by using a coffee grinder, and from fish stocked into a given pond and nf is the number of fish this mixture we made two 0.1–0.2-g pellets per fish. A ran- recovered from the pond (Sager and Winkelman 2006). Adult Downloaded by [Department Of Fisheries] at 22:32 28 February 2013 domly selected sample of age-0 Bluegills from each pond was Bluegill GSIs were compared among treatments and between placed into seven 10-mm size-classes (15–24 mm, 25–34 mm, sexes with ANOVA. Density and biomass of age-0 Bluegills etc., up to 75–84 mm; n = 5 fish per size-class) and dried to a and the density, biomass, mean weight, and mean TL of age-0 constant weight (<0.01 g difference in weight between days). Largemouth Bass were each compared among treatments by use These dried age-0 Bluegills were each homogenized with mor- of ANOVA. The caloric density of adult Largemouth Bass, adult tar and pestle; they were then dried to a constant weight once Bluegills, and age-0 Bluegill size-classes were also compared more to remove any moisture that accumulated during the mix- among treatments by using ANOVA. ing process (Glover et al. 2010). We made two 0.1–0.2-g pellets per age-0 Bluegill from each size-class (n = 10 pellets per size- class) for each pond. For the smallest size-class (15–24 mm), multiple age-0 Bluegills were dried and combined to create pel- RESULTS lets that were large enough for analysis; thus, more than five age-0 Bluegills of this size-class were used in the pellets. Pel- Abiotic and Zooplankton Measures lets were ignited in a semi-microbomb calorimeter (Model 1425 Secchi depth, temperature, dissolved oxygen, turbidity, and Model 6725; Parr Instrument Co.) to measure caloric con- chlorophyll a, and zooplankton density did not differ among LARGEMOUTH BASS GROWTH AND SURVIVAL 173

TABLE 1. Mean (±SE) temperature, Secchi depth, dissolved oxygen, turbidity, chlorophyll a, and zooplankton density measured in experimental ponds at the E.W. Shell Fisheries Center, Auburn, Alabama, during January–August 2010. None of the variables differed signiﬁcantly among feeding treatments.

Feeding treatment

Variable Control Low ration High ration F2, 6 P Temperature (◦C) 25.3 ± 0.16 25.3 ± 0.16 25.0 ± 0.16 0.85 0.47 Secchi depth (cm) 91.4 ± 19.9 87.4 ± 17.3 80.5 ± 15.9 0.27 0.78 Dissolved oxygen (mg/L) 5.50 ± 0.86 5.71 ± 0.49 5.07 ± 0.88 2.86 0.13 Turbidity (NTU) 13.0 ± 5.31 15.2 ± 4.36 15.9 ± 5.58 0.32 0.74 Chlorophyll a (µg/L) 15.3 ± 4.68 17.7 ± 9.47 18.1 ± 3.46 0.08 0.92 Zooplankton density 9.69 × 104 ± 1.16 × 104 1.13 × 105 ± 2.61 × 104 1.13 × 105 ± 2.83 × 104 0.37 0.71 (number/m3)

treatments (Table 1). As such, any differences in fish variables not differ among the treatments (F34, 100 = 0.95, P = 0.56). There among ponds were due to feeding treatments. were no differences in larval Bluegill growth among treatments based on analysis of otolith daily rings (F2, 6 = 0.24, P = 0.79). Bluegills Percent survival of adult Bluegills was similarly high among During February through the end of the experiment (i.e., all ponds (71–75%). At the end of the experiment, the whole- August), adult Bluegill weight (F2, 6 = 11.1, P = 0.0096; body caloric density of adult Bluegills was greater for the = = Figure 1) and TL (F2, 6 = 25.2; P = 0.0012) varied; the weight of high-ration treatment than for the control (F2, 6 4.3, P Bluegills in the high-ration treatment was greater than weights 0.05; Figure 3); there was no difference in adult whole-body for the low-ration treatment and the control (least-significant- caloric density between the high- and low-ration treatments. difference post hoc test: P < 0.05). The feeding treatment Adult GSI did not differ among treatments (F2, 6 = 2.93, P × time interactions were significant for adult Bluegill weight = 0.13), but the TL × treatment interaction was significant = = (F12, 36 = 5.65, P < 0.0001; Figure 1) and TL (F12, 36 = 3.23, (F2, 436 3.00, P 0.05). Thus, there was a nonlinear ef- P = 0.0032) because there were no differences among treat- fect of TL on GSI. Female GSI values were greater in the ments in March or April, whereas the values for the high-ration high-ration treatment than in the low-ration treatment and the treatment were greater than those for the low-ration treatment control, which were similar to one another (F2, 436 = 21.4, P < and the control during May–August. Adult Bluegill Wr did not 0.0001; Figure 4). There were no differences in male GSI differ between the high-ration treatment and the control (F2, 6 = among treatments. The mean density or number of Bluegills per 4.97, P = 0.053). Larval Bluegill density increased significantly pond (F2, 6 = 0.52, P = 0.62; Table 2) and biomass (F2, 6 = 0.43, = with feeding level (F2, 6 = 5.17, P = 0.049; Figure 2). The feed- P 0.67) of fall age-0 Bluegills did not differ among treatments. ing treatment × time interaction for larval Bluegill density did The whole-body caloric density of age-0 Bluegills also did not differ among treatments (F2, 6 = 3.72, P = 0.089; Figure 5). Downloaded by [Department Of Fisheries] at 22:32 28 February 2013

FIGURE 1. Mean (±SE) weight (g) of adult Bluegills in experimental ponds 3 within each of three feeding treatments (3 ponds/treatment): unfed control (cir- FIGURE 2. Mean (±SE) densities (number of ﬁsh/m ) of larval Bluegills cles, solid line), low-ration feeding (inverted triangles, dotted line), and high- in experimental ponds within the three feeding treatments (unfed control, low ration feeding (squares, dashed line). ration, and high ration) from April to August 2010. 174 WOODARD ET AL.

TABLE 2. Total count and biomass (±SE) of age-0 Bluegills collected from the experimental ponds at the E.W. Shell Fisheries Center during August 2010. There were no differences among feeding treatments.

Feeding treatment Total count (n)Biomass(g) Control 1,191 ± 1,110 5,231 ± 3,968 Low ration 371 ± 194 3,144 ± 790 High ration 362 ± 163 2,128 ± 1,048

FIGURE 5. Mean (±SE) caloric density (cal/g wet weight [wwt]) at harvest in relation to total length at harvest for age-0 Bluegills from three feeding treatments (unfed control, low ration, and high ration) in 2010.

Largemouth Bass During April–August, there were no among-treatment differences in weight (F2, 6 = 1.35, P = 0.33; Figure 6), Wr (F2, 6 = 0.18, P = 0.84), or TL (F2, 6 = 0.77, P = 0.50) of the originally stocked adult Largemouth Bass. Largemouth Bass survival was low at 20–38% per pond and did not differ among treatments (F2, 6 = 1.07, P = 0.40). Additionally, there were no differences FIGURE 3. Mean (±SE) caloric density (cal/g wet weight [wwt]) of adult among treatments in the whole-body caloric density of adult Bluegills from the three feeding treatments (unfed control, low ration, and high Largemouth Bass (F2, 6 = 0.02, P = 0.98). Age-0 Largemouth ration) in 2010. Means with the same letter are not signiﬁcantly different. Bass, which were the offspring of the stocked adults, exhibited no signiﬁcant differences in density (F2, 6 = 0.10, P = 0.91) or total biomass (F2, 6 = 1.52, P = 0.29) at harvest. Additionally, neither the weight (F2, 6 = 1.75, P = 0.25; Figure 7) nor the TL (F2, 6 = 1.34, P = 0.33) of age-0 Largemouth Bass differed Downloaded by [Department Of Fisheries] at 22:32 28 February 2013

FIGURE 4. Mean (±SE; error calculated from least-squares means) gonadosomatic index (%) of adult Bluegills in experimental ponds within three feeding FIGURE 6. Mean (±SE) weight (g) of adult Largemouth Bass in experimental treatments (unfed control, low ration, and high ration) during 2010. For a given ponds subjected to three Bluegill feeding treatments (unfed control, low ration, sex, treatment means with the same letter are not signiﬁcantly different. and high ration) during 2010. LARGEMOUTH BASS GROWTH AND SURVIVAL 175

feeding on the fish might have been even more pronounced if feeding treatment ponds had been compared with control ponds that received no fertilization. Overall, the positive effects of supplemental feeding in terms of increased weight, length, and caloric density showed a clear benefit to adult Bluegills after initial stocking. The way fish allocate growth to ovaries is complex, as it depends on body condition, feeding rate, body size, time of year, and a variety of other factors. Thus, we used a conservative GSI (Crim and Glebe 1990), which is an adequate index for measuring energy allocation of fish that are in good to very good condition. The TL × treatment interaction in the GSI analysis was significant, indicating an effect of size on Bluegill GSI. Therefore, the significant difference in female Bluegill GSI between the feeding treatments is due to the Bluegills attaining a larger size in the high-ration treatment. Increased adult Bluegill GSIs should correspond with an increase in the number of eggs spawned and the number of larvae produced (Wootton 1979; Roff 1984), and we found that larval Bluegill densities were greater in the feeding treatments than in the control. Given the greater larval densities in the feeding treatments, it is surprising that we did not observe any differences in age-0 Bluegill numbers or biomass among treatments. A high density of age-0 Bluegills in a single control pond increased the control mean value, but differences among treatments were not statistically significant (Table 2). Because a complex set of interacting environmental factors contributes to larval fish recruitment (Houde 1987; Cargnelli and Gross 1996; Tomcko and Pierce 2005), there may have been limitations (e.g., density- dependent competition) that prevented the greater number of FIGURE 7. Mean (±SE) total length (mm) and weight (g) of age-0 Large- mouth Bass in experimental ponds subjected to three Bluegill feeding treatments Bluegill larvae produced in the feeding treatments from trans- (unfed control, low ration, and high ration) during 2010. For a given variable, lating into greater numbers of age-0 individuals. Furthermore, treatment means with the same letter did not differ significantly. the increased numbers of Bluegill larvae in the feeding treatments might not have had sufficient time to grow and mature into age-0 Bluegills by the time harvest occurred; a longer pe- among treatments. No pelleted feed was found in the stomachs riod before harvest would have provided more prey items for of any age-0 or adult Largemouth Bass. the adult and age-0 Largemouth Bass. Finally, it is possible that we missed some age-0 Bluegills in the mud at the pond bottoms when the experiment ended, but we believe that this did not Downloaded by [Department Of Fisheries] at 22:32 28 February 2013 DISCUSSION differ among treatments. The growth rate of stocked Bluegills responded positively The low survival rate (20–38%) of adult Largemouth Bass to supplemental feeding, as has been observed in other studies among all ponds resulted in small sample sizes and reduced (Schmittou 1969; Berger 1982). We were also able to show that statistical power, likely preventing the detection of differences. the growth of Bluegills can vary depending on the feeding rate. The food supply for the adult Largemouth Bass was not limiting None of the measured abiotic or zooplankton variables differed at the beginning of the experiment given that all individuals (in among the treatments; this is likely attributable to the fertilizer all treatments) had access to the same prey resources; later in the we added to ensure that plankton blooms and Secchi depth were experiment, Largemouth Bass in the feeding treatments would consistent among the ponds (Swingle and Smith 1939). Because have had access to Bluegills with higher energetic content rela- these variables did not differ among treatments, the differences tive to those in control ponds. However, temperatures are high in fish growth and condition were due to the feeding treatments. during summer, and fish energetics suggest that Largemouth Additionally, fertilization alone (in the absence of pellet feed- Bass do not exhibit substantial growth when water temperatures ing) can enhance Bluegill condition through the stimulation of exceed 27◦C (Niimi and Beamish 1974). Temperature might primary and secondary production (Swingle and Smith 1939; have masked growth effects in the present study, but the average Haley 2009); therefore, the observed effects of supplemental water temperature at 1-m depth was 25◦C over the course of 176 WOODARD ET AL.

the experiment, and the Largemouth Bass also had deeper water Madeline Wedge, Emily DeVries, Kirk Mears, Zachary De- (2 m deep) available as a refuge. Vries, and Kristina Woodard. We also thank Tammy DeVries The effects of temperature would have been lower for age-0 for the bomb calorimetry and zooplankton counts and David Largemouth Bass than for adults because of the smaller body Glover for help with statistical analyses. Alan Wilson provided size of age-0 fish (Niimi and Beamish 1974); therefore, because excellent and insightful comments on previous versions of this larval Bluegill density differed among feeding treatments, we work as well as supplying insight on the importance of lower expected to see an increased growth rate of age-0 Largemouth trophic levels. Lindsay Huebner provided useful editorial advice Bass. However, age-0 Largemouth Bass had to compete with the on drafts of this manuscript. We thank three anonymous review- adult Largemouth Bass, perhaps limiting their growth potential. ers for comments that helped to improve this manuscript. Gener- Over a longer duration, more pronounced effects of Bluegill re- ous donations from William Ireland and feeders from Moultrie production could result from stored lipid carryover from previ- Feeders, LLC, supported this research. ous years (Fischer et al. 1998); thus, a study of longer duration might have resulted in detectable differences in Largemouth Bass growth and condition among feeding treatments. If the experiment had been run for another year, it is possible that REFERENCES more differences in age-0 Largemouth Bass would have been Aday, D. D., D. P. Philipp, and D. H. Wahl. 2006. Sex-specific life history Lepomis macrochirus detected; for example, age-0 Largemouth Bass would have had patterns in Bluegill ( ): interacting mechanisms influence individual body size. Oecologia 147:31–38. larger prey items with increased energetic content available to Belk, M. C. 1995. Variation in growth and age at maturity in Bluegill sunfish: them if the Bluegills that were larvae at the end of summer had genetic or environmental effects? Journal of Fish Biology 47:237–247. been given time to mature during fall. Largemouth Bass also Berger, T. A. 1982. Supplemental feeding of a wild Bluegill population. North typically gain more weight in the fall than in the summer (Gar- American Journal of Fisheries Management 2:158–163. vey and Marschall 2003), so a continuation of the experiment Cargnelli, L. M., and M. R. Gross. 1996. The temporal dimension in fish recruitment: birth date, body size, and size-dependent survival in a sunfish (Bluegill: might also have yielded detectable differences in Largemouth Lepomis macrochirus). Canadian Journal of Fisheries and Aquatic Sciences Bass size. Because there was no evidence of pelleted feed in 53:360–367. the stomachs of any sampled Largemouth Bass, we conclude Crim, L. W., and B. D. Glebe. 1990. Reproduction. Pages 529–553 in C. B. that any benefits received by Largemouth Bass from the feed- Schreck and P. B. Moyle, editors. Methods for fish biology. American Fish- ing treatments resulted indirectly from the feed-related benefits eries Society, Bethesda, Maryland. Dauwalter, D. C., and J. R. Jackson. 2005. A re-evaluation of U.S. state fish- provided to the Bluegills. stocking recommendations for small, private, warmwater impoundments. We have shown that supplemental pellet feed provides bene- Fisheries 30(8):18–28. fits to Bluegills that are stocked in small impoundments. Thus, Dettmers, J. M., and R. A. Stein. 1992. Food consumption by larval Gizzard if the management goal is to increase adult Largemouth Bass Shad: zooplankton effects and implications for reservoir communities. Trans- condition during the season immediately after stocking, then actions of the American Fisheries Society 121:494–507. Fischer, R. U., J. D. Congdon, C. D. Difonzo, and J. M. Aho. 1998. Lipid and other management techniques, such as stocking a supplemen- reproductive cycles of Bluegills (Lepomis macrochirus) exposed to 35 years tal prey species for the Largemouth Bass (e.g., Threadfin Shad; of elevated environmental temperatures. Journal of Thermal Biology 23:359– Noble 1981), may be more cost effective than providing sup- 367. plemental feed to Bluegills. However, our results demonstrate Garvey, J. E., and E. A. Marschall. 2003. Understanding latitudinal trends in fish that supplemental pellet feeding is clearly beneficial when the body size through models of optimal seasonal energy allocation. Canadian Journal of Fisheries and Aquatic Sciences 60:938–948. pond management goal is to directly increase Bluegill size and Glover, D. C., D. R. DeVries, R. A. Wright, and D. A. Davis. 2010. Sam- reproductive output, and these effects on Bluegills will likely ple preparation techniques for determination of fish energy density via bomb Downloaded by [Department Of Fisheries] at 22:32 28 February 2013 result in carryover benefits to Largemouth Bass during the next calorimetry: an evaluation using Largemouth Bass. Transactions of the Amer- season. Therefore, for pond owners wishing to promote quick ican Fisheries Society 139:671–675. growth and reproduction of Bluegills after initial stocking, we Haley, N. V. 2009. Privately owned small impoundments of central Alabama: a survey and evaluation of management techniques and enhancements. Master’s · −1· −1 suggest feeding at a high ration (e.g., 2.68 kg ha d )forthe thesis. Auburn University, Auburn, Alabama. first year. If pond owners choose to feed at such a rate, how- Hoagland, R. H., III, D. A. Davis, N. A. Tuan, and W. J. McGraw. 2003. ever, they should be aware that they are artificially increasing Evaluation of practical Bluegill diets with varying protein and energy levels. the carrying capacity of the pond. A cessation in feeding could North American Journal of Aquaculture 65:147–150. then have damaging effects on the ecosystem; therefore, feeding Houde, E. D. 1987. Fish early life dynamics and recruitment variability. Pages 17–29 in R. D. Hoyt, editor. 10th annual larval fish conference. American should be undertaken as a long-term management activity. Fisheries Society, Symposium 2, Bethesda, Maryland. Modde, T. 1980. State stocking policies for small warmwater impoundments. Fisheries 5(5):13–17. ACKNOWLEDGMENTS Murnyak, D. F., M. O. Murnyak, and L. J. Wolgast. 1984. Growth of stunted and nonstunted Bluegill sunfish in ponds. Progressive Fish-Culturist 46:133–138. We are extremely grateful to the many graduate students Niimi, A. J., and F. W. H. Beamish. 1974. Bioenergetics and growth of Large- and technicians who helped with this project in the field and mouth Bass (Micropterus salmoides) in relation to body weight and temper- laboratory: Tommy Purcell, Craig Roberts, Brandon Simcox, ature. Canadian Journal of Zoology 52:447–456. LARGEMOUTH BASS GROWTH AND SURVIVAL 177

Noble, R. L. 1981. Management of forage fishes in impoundments of the south- Swingle, H. S., and E. V.Smith. 1942. Management of farm fish ponds. Alabama ern United States. Transactions of the American Fisheries Society 110:738– Polytechnic Institute, Agricultural Experiment Station Bulletin 254. 750. Taubert, B. D., and D. W. Coble. 1977. Daily rings in otoliths of three species Porath, M. T., L. D. Pape, and J. J. Jackson. 2003. Evaluating supplemen- of Lepomis and Tilapia mossambica. Journal of the Fisheries Research Board tal feeding of a Bluegill population and validating species utilization by of Canada 34:332–340. application of a toxic fish food. Journal of Freshwater Ecology 18:593– Tomcko, C. M., and R. B. Pierce. 2005. Bluegill recruitment, growth, population 597. size structure, and associated factors in Minnesota lakes. North American Roff, D. A. 1984. The evolution of life history parameters in teleosts. Canadian Journal of Fisheries Management 25:171–179. Journal of Fisheries and Aquatic Sciences 41:989–1000. Twibell, R. G., K. A. Wilson, S. Sanders, and P. B. Brown. 2003. Evaluation of Sager, C. R., and D. L. Winkelman. 2006. Effects of increased feeding fre- experimental and practical diets for Bluegill Lepomis macrochirus and hybrid quency on growth of hybrid Bluegill in ponds. North American Journal of Bluegill L. cyanellus × L. macrochirus. Journal of the World Aquaculture Aquaculture 68:313–316. Society 34:487–495. SAS (Statistical Analysis Systems). 2008. SAS version 9.1.3. SAS Institute, Wege, G. J., and R. O. Anderson. 1978. Relative weight (Wr): a new index of Cary, North Carolina. condition for Largemouth Bass. Pages 79–91 in G. D. Novinger and J. G. Schmittou, H. R. 1969. Some effects of supplemental feeding and controlled Dillard, editors. New approaches to the management of small impoundments. fishing in Largemouth Bass–Bluegill populations. Proceedings of the An- American Fisheries Society, North Central Division, Special Publication 5, nual Conference Southeastern Association of Game and Fish Commissioners Bethesda, Maryland. 22(1968):311–320. Welker, M. T., C. L. Pierce, and D. H. Wahl. 1994. Growth and survival of larval Smith, S. V., W. H. Renwick, J. D. Bartley, and R. W. Buddemeier. 2002. fishes: roles of competition and zooplankton abundance. Transactions of the Distribution and significance of small, artificial water bodies across the United American Fisheries Society 123:703–717. States landscape. Science of the Total Environment 299:21–36. Welschmeyer, N. A. 1994. Fluorometric analysis of chlorophyll a in the presence Swingle, H. S. 1949. Experiments with combinations of Largemouth Black Bass, of chlorophyll b and pheopigments. Limnology and Oceanography 39:1985– Bluegills, and minnows in ponds. Transactions of the American Fisheries 1992. Society 76:46–62. Wootton, R. J. 1979. Energy cost of egg production and environmental deter- Swingle, H. S., and E. V. Smith. 1939. Fertilizers for increasing the natural minants of fecundity in teleost fishes. Symposia of the Zoological Society of food for fish in ponds. Transactions of the American Fisheries Society 68: London 44:133–159. 126–135. Wright, R., and M. Masser. 2004. Management of recreational fish ponds in Al- Swingle, H. S., and E. V. Smith. 1940. Experiments on the stocking of fish abama. Alabama Cooperative Extension Service, Circular ANR-577, Auburn ponds. Progressive Fish-Culturist 7(49):19–20. University, Auburn. Downloaded by [Department Of Fisheries] at 22:32 28 February 2013 This article was downloaded by: [Department Of Fisheries] On: 28 February 2013, At: 22:32 Publisher: Taylor & Francis Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

North American Journal of Fisheries Management Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/ujfm20 Frequency and Severity of Trauma in Fishes Subjected to Multiple-Pass Depletion Electrofishing Frank M. Panek a & Christine L. Densmore a a U.S. Geological Survey, Leetown Science Center, National Fish Health Research Laboratory, 11649 Leetown Road, Kearneysville, West Virginia, 25430, USA Version of record first published: 29 Jan 2013.

To cite this article: Frank M. Panek & Christine L. Densmore (2013): Frequency and Severity of Trauma in Fishes Subjected to Multiple-Pass Depletion Electrofishing, North American Journal of Fisheries Management, 33:1, 178-185 To link to this article: http://dx.doi.org/10.1080/02755947.2012.754803

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material. North American Journal of Fisheries Management 33:178–185, 2013 American Fisheries Society 2013 ISSN: 0275-5947 print / 1548-8675 online DOI: 10.1080/02755947.2012.754803

ARTICLE

Frequency and Severity of Trauma in Fishes Subjected to Multiple-Pass Depletion Electroﬁshing

Frank M. Panek* and Christine L. Densmore U.S. Geological Survey, Leetown Science Center, National Fish Health Research Laboratory, 11649 Leetown Road, Kearneysville, West Virginia 25430, USA

Abstract The incidence and severity of trauma associated with multiple-pass electrofishing and the effects on short-term (30-d) survival and growth of Rainbow Trout Oncorhynchus mykiss, Brook Trout Salvelinus fontinalis, and five representative co-inhabiting nontarget or bycatch species were examined. Fish were held in four rectangular fiberglass tanks (190 × 66 cm) equipped with electrodes, a gravel–cobble stream substrate, and continuous water flow. Fish were exposed to one, two, or three electroshocks (100-V, 60-Hz pulsed DC) spaced 1 h apart or were held as a control. The heterogeneous field produced a mean (±SD) voltage gradient of 0.23 ± 0.024 V/cm (range = 0.20–0.30 V/cm) with a duty cycle of 30% and a 5-s exposure. Radiographs of 355 fish were examined for evidence of spinal injuries, and necropsies were performed on 303 fish to assess hemorrhagic trauma in soft tissue. Using linear regression, we demonstrated significant relationships between the number of electrical shocks and the frequency and severity of hemorrhagic and spinal trauma in each of the nontarget species (Potomac Sculpin Cottus girardi, Channel Catfish Ictalurus punctatus, Fathead Minnow Pimephales promelas, Green Sunfish Lepomis cyanellus, and Largemouth Bass Micropterus salmoides). Most of the injuries in these species were either minor or moderate. Rainbow Trout and Brook Trout generally sustained the highest incidence and severity of injuries, but those injuries were generally independent of the number of treatments. The 30-d postshock survival for the trout species was greater than 94%; survival for the bycatch species ranged from 80% (Fathead Minnow) to 100% (Green Sunfish and Channel Catfish). There were no significant differences in 30-d postshock condition factors despite observations of altered feeding behavior lasting several days to 1 week posttreatment in several of the study species.

Depletion sampling by multiple-pass electrofishing is an im- increasing stream width and to increase the precision of the pop- portant fisheries management tool for use in wadeable streams ulation estimate. During electrofishing, many individual fish are

Downloaded by [Department Of Fisheries] at 22:32 28 February 2013 (Hilborn and Walters 1992). This technique has been used rou- exposed to one electrofishing pass, removed from the popula- tinely by federal and state fishery managers for several decades tion, and processed. However, multiple passes and exposures to as a reliable means to obtain quantitative data on fish populations electrical fields are required to obtain progressive depletions in (Van Deventer and Platts 1983). In addition to its usefulness as catch. Those fish that escape the initial electrofishing pass are a tool for estimating population size, multiple-pass electrofish- subjected to a second electrofishing pass or additional passes ing coupled with seining in wadeable streams can provide ac- before being removed from the population. curate and precise estimates of species composition, richness, In striving to obtain reliable population estimates for trout or and relative abundance (Kennard et al. 2006). These techniques other game species, there are often unintended effects of elec- generally employ maximum-likelihood functions that estimate trofishing on nontarget species (Miranda and Kidwell 2010), population size by extrapolating the rate of decline in catch and these effects could be compounded when multiple-pass per unit effort over multiple sampling units of effort. Gener- electrofishing is used. There have been many studies on the ally, biologists increase the number of sampling units to address potentially harmful effects of AC, DC, and pulsed-DC (PDC)

*Corresponding author: [email protected] Received June 27, 2012; accepted November 20, 2012

178 ELECTROFISHING HEMORRHAGIC AND SPINAL TRAUMA 179

electrofishing on salmonids (Snyder 2003), but we know much To determine the effects of multiple-pass depletion sampling less about the effects on nongame species (Kocovsky et al. on both the target trout species and the nontarget or co-inhabiting 1997; Holliman et al. 2003; Dolan and Miranda 2004; Clement´ species, we selected Rainbow Trout as our primary species and and Cunjak 2010; Miranda and Kidwell 2010) or the poten- established our PDC electrofishing field to provide for an op- tial effects of exposure to multiple shocks (Panek and Dens- timal capture response in the experimental tanks. We selected more 2011). Collectively, previous studies have demonstrated a 100-V, 60-Hz PDC as the electrofishing treatment because it re- pattern of injury, primarily musculoskeletal tissue damage, as sulted in an ideal field for producing the desired effect (narcosis well as a variety of other deleterious effects, including reduced but minimal tetany) on Rainbow Trout. Fish were exposed to stamina, lethargy, behavioral changes, reduced cardiac function, this field for 5 s in 190- × 66-cm, rectangular fiberglass tanks and changes in fertility. equipped with a head box that provided a constant directional Our objective was to evaluate the effects of multiple-pass water flow of 26 L/min. This regulated flow was maintained at electrofishing on Rainbow Trout Oncorhynchus mykiss, Brook a depth of 20 cm over a gravel–cobble substrate to simulate a Trout Salvelinus fontinalis, and representative co-inhabiting natural stream environment. The continuous flow also provided nontarget or bycatch species. We compared the incidence and a constant water temperature (12.5◦C) and ambient conductivity severity of trauma among fishes in identical electrofishing fields (490 µS/cm) for the duration of the experiment. Specific con- under controlled laboratory conditions, and we also examined ductivity and ambient water temperature were recorded by using the effects on short-term (30-d postshock) changes in fish sur- an Oakton 10 Series pH/conductivity meter. One tank served as vival and growth. a control (no shock), and three tanks served as experimental tanks receiving treatments of one, two, or three shocks (i.e., 1 tank/treatment). Tanks were equipped with 7- × 30-cm alu- METHODS minum electrodes at the tank ends; electrodes were connected We evaluated three-pass PDC electrofishing effects on seven by Number-10 AWG (American wire gauge) braided copper freshwater fish species: two commonly targeted species (Brook wire to a Smith-Root Model LR-24 backpack electrofisher. The Trout and Rainbow Trout) and five bycatch species (Potomac characteristics of the electrofishing field were measured with a Sculpin Cottus girardi, Channel Catfish Ictalurus punctatus, Fluke Model 196B-003 oscilloscope. We used a voltage gradi- Fathead Minnow Pimephales promelas, Green Sunfish Lepomis ent (VG) probe and oscilloscope to measure the VG (V/cm) in cyanellus, and Largemouth Bass Micropterus salmoides). Chan- each cell of a 24-cell field grid; these empirical values were used nel Catfish, Fathead Minnow, and Green Sunfish were harvested to obtain average and peak VGs. This treatment produced a het- from production ponds at the U.S. Geological Survey (USGS) erogeneous field with a mean (±SD)VGof0.23 ± 0.024 V/cm Leetown Science Center (Kearneysville, West Virginia). Rain- (range = 0.20–0.30 V/cm) delivered with a duty cycle of 30% bow Trout were provided by the Center for Cold and Cool Water in a 5-s exposure. Aquaculture (U.S. Department of Agriculture, Kearneysville, Each species was evaluated in a separate electrofishing chal- West Virginia), and the Brook Trout were provided by Ridge lenge using the control tank and the three experimental tanks. State Fish Hatchery (West Virginia Department of Natural Re- The sizes, weights, and number of individuals of each species sources, Berkeley Springs). Largemouth Bass were purchased utilized in these trials are summarized in Table 1. Condition fac- from a commercial grower, and Potomac Sculpin were collected tors (KTL) were determined from TL and weight (g) measure- by hand nets in Hopewell Run on the grounds of the Leetown ments in accordance with the method of Anderson and Neumann Science Center. All fish were acclimated to tank conditions for (1996). In general, we attempted to minimize the effects of size 30 d prior to their use in experiments. by using individuals less than 200 mm TL as test fish. Potential Downloaded by [Department Of Fisheries] at 22:32 28 February 2013

TABLE 1. Number of ﬁsh of each species used in multiple-pass electroﬁshing evaluations, presented with minimum (min), maximum (max), and mean TLs (mm) and weights (g).

TL (mm) Weight (g) Species n Min Max Mean Min Max Mean Rainbow Trout 131 75 138 107 4.4 25.8 13.5 Brook Trout 100 89 163 131 6.1 35.1 21.0 Potomac Sculpin 75 50 89 68 1.4 9.6 4.0 Channel Catﬁsh 100 107 180 137 8.2 40.9 19.5 Fathead Minnow 111 55 85 65 2.0 8.5 3.6 Green Sunﬁsh 60 98 186 152 22.7 143.7 76.2 Largemouth Bass 98 70 125 89 5.6 24.7 9.6 180 PANEK AND DENSMORE

short-term effects of electrofishing on growth were assessed by At the conclusion of the 30-d period, all fish in both the ex- calculating 30-d postexposure KTL and comparing these values perimental and control tanks were euthanized with MS-222 and to those of the control fish in each experimental lot via ANOVA frozen for radiographic analysis. Rainbow Trout were radio- and a Tukey pairwise comparison test (SYSTAT Software, Inc., graphed with a Xerox-126 Processor at the Virginia–Maryland Richmond, California). Regional College of Veterinary Medicine (Blacksburg, Virginia) Multiple electrofishing exposures simulated three-pass de- by following the procedures of Smith and Smith (1994). Ra- pletion sampling, with fish shocked at time 0, 1 h after the diographs of the other species were taken at the USGS Na- initial shock, and 2 h after the initial shock. A subsample of no tional Wildlife Health Center (Madison, Wisconsin) by using fewer than 10 fish was removed for evaluation of acute hem- a TechAmerica MT8030 x-ray machine at 18–24 kV for 120– orrhagic trauma 30 min after the final electrofishing treatment, 180 s depending upon species and resolution of the radiograph. or, in the case of the controls, at the beginning of the treatment All fish were radiographed in lateral view, and a subsample period. Fish that were collected from the control and exper- (25%) of each species was radiographed in dorsoventral view. imental tanks were euthanized with a lethal aqueous dose of We examined all fish for spinal injuries (compressions, mis- tricaine methanesulfonate (MS-222; Argent Chemical Labora- alignments, and fractures) by following standardized criteria tories, Inc., Redmond, Washington) and were then frozen for for electrofishing-induced spinal injuries (Reynolds 1996). subsequent necropsy. Hemorrhagic trauma was scored by us- The scoring criteria for both hemorrhagic trauma and spinal ing the criteria modified from Reynolds (1996; Table 2). The trauma (Reynolds 1996) provide a range of scores from 0 to remaining groups in each tank were monitored for moribund 3, where 0 represents no injury and 3 represents severe injury and dead fish at 0.5, 1, 2, 4, 8, 16, and 24 h and then daily for (Table 2). We used SYSTAT II (SYSTAT Software, Inc.) to the next 30 d after the third electrofishing pass. Dead fish were obtain mean and SE estimates for hemorrhagic and spinal trauma removed and frozen for necropsy and subsequent radiography by treatment effects. We also developed an additive trauma score by following the recommendations of Reynolds (1996). Due to (ATS) that combined the means and SEs within each species to constraints of tank availability, space, and logistics, we were un- examine overall differences in electrofishing effects. The ATS is able to conduct simultaneous replications of each electrofishing classified as follows: 0 represents no injury; >0 to 2 represents treatment. slight or minor injury; >2 to 4 represents moderately severe injury; and >4 to 6 represents severe injury. The SEs for these TABLE 2. Criteria for assigning trauma scores to hemorrhages and spinal combined means were calculated by the method of Baker and damage in fish that have been exposed to electrofishing fields (modified from Nissim (1963). The 30-d posttreatment changes in K were Reynolds 1996). TL compared using ANOVA, and significance was tested (at α = Type of injury Score Criteria 0.05) with the Durbin–Watson D-statistic in SYSTAT II. We used linear regression techniques (McDonald 2009) to assess Internal 0 No hemorrhage present. relationships between the ATS or the frequency of hemorrhagic hemorrhage 1 Mild hemorrhage with one or more and spinal injury and the number of electrical shocks for each wounds in the muscle, separate species. from the spinal column. 2 Moderate hemorrhage with one or more wounds located on the spinal RESULTS column and each wound involving Frequency of Hemorrhagic and Spinal Trauma fewer than two vertebrae. In total, 675 fish of seven species were held as a control or

Downloaded by [Department Of Fisheries] at 22:32 28 February 2013 3 Severe hemorrhage with one or more received one-shock, two-shock, or three-shock treatments at a large wounds over two vertebrae in maximum peak VG of 0.30 V/cm and a peak power density of size on the spinal column. 44 µW/cm3. This field produced narcosis but minimal tetany in Spinal damage 0 No spinal damage evident. Rainbow Trout and was considered optimal for electrofishing 1 Vertebral compression involving more in our experimental tanks. Of the species that were exposed than two vertebrae in one location to this electrofishing field, Rainbow Trout and Brook Trout along spine. were clearly the most susceptible to hemorrhagic and spinal 2 Vertebral compression, misalignment injury. Rainbow Trout experienced hemorrhagic injury rates of and/or dislocation involving more 20% after one shock and up to 60% after two shocks or three than two vertebrae in single or shocks (Figure 1). Likewise, exposures in Brook Trout resulted multiple locations along spine. in hemorrhagic injury rates of 50–80%, but these injuries did 3 Vertebral fracture of one vertebra or not appear to be dose dependent. Spinal injuries were variable complete separation of more than in both species of trout, with the incidence of trauma being as two vertebrae. high as 69.2% in Rainbow Trout after two shocks and up to ELECTROFISHING HEMORRHAGIC AND SPINAL TRAUMA 181 Downloaded by [Department Of Fisheries] at 22:32 28 February 2013

FIGURE 1. (a) Frequency of hemorrhagic trauma, (b) frequency of spinal trauma, and (c) mean (±SE) additive trauma scores for seven fish species that were exposed to one, two, or three electrical shocks at 100-V, 60-Hz pulsed DC under controlled laboratory conditions (RBT = Rainbow Trout; BKT = Brook Trout; PTS = Potomac Sculpin; FHM = Fathead Minnow; CHC = Channel Catfish; GRS = Green Sunfish; LMB = Largemouth Bass). 182 PANEK AND DENSMORE

TABLE 3. Linear regression tests of significance for the relationships be- generally more susceptible to hemorrhagic trauma than to spinal tween percent frequency of hemorrhagic trauma (HT), percent frequency of trauma (Figure 1). After three shocks, the average percentage spinal/vertebral trauma (ST), or the additive trauma score (ATS) for seven fish of hemorrhagic trauma in the nontarget species was 48.3 ± species and exposure to three consecutive treatments in a heterogeneous, 100-V, ± 60-Hz pulsed-DC electrofishing field under controlled laboratory conditions at a 7.98% (mean SD) and the average for spinal injury was mean (±SD) voltage gradient of 0.23 ± 0.024 V/cm (range = 0.20–0.30 V/cm; 32.6 ± 20.88%. significance: *P ≤ 0.10, **P ≤ 0.05). Of the nontarget species examined, the Fathead Minnow was most susceptible to electrofishing-related trauma, with 60% of Species Variable df R2 P the tested fish sustaining hemorrhagic injuries and 56.2% sus- Rainbow Trout HT 2 0.90 0.05** taining spinal injuries after three shocks. There was a statistically ST 2 0.24 0.51 significant (P ≤ 0.05) increase in hemorrhagic trauma with an ATS 2 0.78 0.12 increasing number of electrical exposures for all species except Brook Trout HT 2 0.34 0.42 Largemouth Bass (Table 3). A similar relationship was evident ST 2 0.63 0.21 for spinal trauma in Fathead Minnow (r2 = 0.91, P ≤ 0.05), ATS 2 0.64 0.20 Channel Catfish (r2 = 0.98, P ≤ 0.01), and Green Sunfish (r2 = Potomac Sculpin HT 2 0.90 0.05** 0.99, P ≤ 0.01). ST 2 0.80 0.11 ATS 2 0.89 0.06* Severity of Hemorrhagic and Spinal Trauma Fathead Minnow HT 2 0.97 0.02** As measured by the ATS, the severity of injury due to hem- ST 2 0.91 0.04** orrhage and spinal compression, dislocation, or fracture across ATS 2 0.95 0.02** the three shock treatments was most evident in Rainbow Trout Channel Catfish HT 2 0.98 0.01** and Brook Trout (Figure 1). Among the species tested, Brook ST 2 0.98 0.01** Trout sustained the highest ATS(1.63) observed after one shock. ATS 2 0.92 0.04** However, additional shocks resulted in few additional injuries Green Sunfish HT 2 0.90 0.05** to Brook Trout, and after three shocks the severity of observed ST 2 0.99 0.01** injuries in both Rainbow Trout and Brook Trout were similar. ATS 2 0.86 0.07* Rainbow Trout experienced an ATS of 1.8 ± 0.14 (mean ± Largemouth Bass HT 2 0.73 0.14 SE) after three shocks, and Brook Trout under the same condi- ST 2 0.21 0.54 tions experienced comparable levels of trauma (ATS = 1.7 ± ATS 2 0.80 0.11 0.14; Figure 1). Linear regression tests of significance failed to demonstrate any relationship between the number of shocks and the ATS for either Brook Trout or Rainbow Trout (Table 3). We also noted that up to 30% of the Rainbow Trout and 7% of the 60% in Brook Trout after three shocks (Figure 1). The average Brook Trout that were exposed to the electric field exhibited ex- percentage of hemorrhagic and spinal trauma in both species ternal, predorsal fin branding or banding (Panek and Densmore after the three-shock treatment was 55.0 ± 7.07% (mean ± 2011). SD) and 43.3 ± 7.98%, respectively. For Rainbow Trout, there In the nontarget species, the severity of hemorrhagic and was a significant linear relationship between the number of spinal injury was generally less than that observed in either of shocks and the incidence of hemorrhagic trauma (r2 = 90, P < the salmonids, and the ATS for Fathead Minnow and Channel 0.05), but such a relationship was not observed for the incidence Catfish increased significantly (P ≤ 0.05) with an increasing Downloaded by [Department Of Fisheries] at 22:32 28 February 2013 of spinal trauma (r2 = 0.24, P = 0.51; Table 3). Brook Trout number of electrical shocks (Table 3). This same relationship did not exhibit a dose-related response for either hemorrhagic was also significant (P ≤ 0.10) for Potomac Sculpin, Green trauma (r2 = 0.34, P = 0.42) or spinal/vertebral trauma (r2 = Sunfish, and Largemouth Bass (Table 3). Of particular note was 0.63, P = 0.21). Hemorrhagic trauma and spinal trauma in the level of trauma sustained by Fathead Minnow after the three- Brook Trout were initially high at 70% and 53.3%, respectively, shock treatment, as the injuries would be considered moderately after one shock. These measures remained high after additional severe (ATS = 2.65). Overall, after three shocks, the average electrofishing exposures (Figure 1). ATS for the nontarget species was 1.60 ± 0.69 (mean ± SD), In general, nontarget species exhibited fewer hemorrhagic which is considered minor or slight on the ATS injury scale. and spinal injuries than either of the salmonids, but the incidence of these injuries increased as the number of shocks Thirty-Day Postshock Survival and Growth increased—a pattern of injury that was only marginally appar- The effects of multiple-pass electrofishing were monitored ent in the Rainbow Trout and Brook Trout. There were con- for 30 d posttreatment to assess interspecies differences in sur- siderable interspecies differences in the responses of nontarget vival and growth. In no case was morbidity or mortality observed fish to increasing numbers of electrical shocks, and the fish were in the control group for any species (Table 4). Rainbow Trout, ELECTROFISHING HEMORRHAGIC AND SPINAL TRAUMA 183

TABLE 4. Thirty-day postshock survival (%) of fish that were subjected to TABLE 5. Thirty-day postshock condition factors (i.e., KTL) of fish that were one, two, or three electrical exposures under controlled laboratory conditions in subjected to one, two, or three electrical exposures in a heterogeneous, 100-V, a heterogeneous, 100-V, 60-Hz pulsed-DC electrofishing field at a mean (±SD) 60-Hz pulsed-DC electrofishing field at a mean (±SD) voltage gradient of 0.23 voltage gradient of 0.23 ± 0.024 V/cm (range = 0.20–0.30 V/cm). ± 0.024 V/cm (range = 0.20–0.30 V/cm).

One Two Three Species Treatment Mean SD n Species Control shock shocks shocks Rainbow Trout Control 1.064 0.101 10 Rainbow Trout 100 100 94 98 One shock 1.006 0.128 21 Brook Trout 100 100 100 100 Two shocks 1.104 0.161 20 Potomac Sculpin 100 92 100 100 Three shocks 1.009 0.125 20 Fathead Minnow 100 100 80 80 Brook Trout Control 0.894 0.058 15 Channel Catfish 100 100 100 100 One shock 0.909 0.070 15 Green Sunfish 100 100 100 100 Two shocks 0.978 0.062 15 Largemouth Bass 100 96 96 92 Three shocks 0.947 0.197 15 Potomac Sculpin Control 1.186 0.103 5 One shock 1.202 0.142 10 Two shocks 1.108 0.246 10 the target or baseline species for our interspecies comparisons, Three shocks 1.062 0.197 10 demonstrated reduced survival with multiple-shock treatments, Channel Catfish Control 0.733 0.062 15 but these reductions were minimal and never exceeded 6% One shock 0.684 0.050 15 (Table 4). The same electrofishing treatments had no effect on Two shocks 0.684 0.062 15 Brook Trout, which demonstrated 100% survival in multiple- Three shocks 0.731 0.062 15 shock treatments. Fathead Minnow Control 1.214 0.126 19 The effect of electrofishing treatments on the 30-d survival of One shock 1.203 0.143 20 the nontarget species varied and was most pronounced for Fat- Two shocks 1.293 0.088 16 head Minnow, in which mortality was 20% after the two-shock Three shocks 1.308 0.160 16 and three-shock treatments. We noted mortality of less than 8% Green Sunfish Control 2.050 0.246 10 in Potomac Sculpin and Largemouth Bass, and multiple-shock One shock 1.988 0.171 10 electrofishing had no effect on the 30-d survival of either Chan- Two shocks 2.080 0.237 10 nel Catfish or Green Sunfish (Table 4). Three shocks 1.965 0.095 10 During this 30-d postshock evaluation, we qualitatively ob- Largemouth Bass Control 1.228 0.141 13 served changes in feeding behavior of some species. For in- One shock 1.238 0.150 14 stance, we observed a reduction in feeding by Rainbow Trout Two shocks 1.252 0.170 15 and Brook Trout for up to 1 d posttreatment. Similarly, both Three shocks 1.287 0.108 13 Potomac Sculpin and Green Sunfish refused to feed for up to 3 d after exposure to the same PDC electrical field. Largemouth Bass, which sustained the fewest electrofishing injuries, did not this hypothesis, those authors electrofished twice within a 1–3-d resume normal feeding until 8 d postexposure. These periods period on a monthly basis for up to 7 months and found that of reduced feeding corresponded with obvious signs of lethargy the percentage of Rainbow Trout and Brown Trout Salmo trutta and reduced swimming activity. At the conclusion of the 30- with instantaneous growth rates less than average was signif- Downloaded by [Department Of Fisheries] at 22:32 28 February 2013 d period, we failed to document any significant changes (P ≤ icantly greater for fish that were electrofished four months or 0.05) in KTL for any species or treatment (Table 5). more. Similar observations were made in field studies involving the multiple-pass depletion sampling of Cutthroat Trout Oncorhynchus clarkii (Mesa and Schreck 1989). However, in DISCUSSION his review of the literature, Snyder (2003) concluded that elec- Our results show that the frequency and severity of muscu- trofishing caused no overall long-term effects on growth. Our loskeletal trauma in both Rainbow Trout and Brook Trout were findings agree with Snyder’s (2003) interpretation of the liter- not clearly related to the number of electrofishing exposures. ature. Although we noted altered feeding behavior in Rainbow Furthermore, multiple exposures to electrofishing, as would be Trout and Brook Trout, this behavioral change appeared to be experienced during three-pass depletion sampling, had no sig- short term and had no significant effect on 30-d postshock KTL. nificant effects on survival or growth of Rainbow Trout or Brook The 30-d postshock monitoring of fish indicated that elec- Trout. There are several studies that suggest a relationship be- trofishing had a minimal influence on survival. For Rainbow tween multiple-pass electrofishing and fish growth or survival. Trout and Brook Trout that were subjected to three shocks, Gatz et al. (1986) hypothesized that repeated electrofishing ex- the observed 30-d posttreatment survival was 98% and 100%, posures might result in measurable effects on growth. To test respectively. Similarly, for the nontarget species, the 30-d 184 PANEK AND DENSMORE

postshock survival ranged from 80% to 100%. Of the seven evaluate the rates of injury in nontarget fishes. Thirty-day post- species evaluated, three species (Brook Trout, Channel Catfish, shock survival was 100% for Brook Trout and, at the worst, was and Green Sunfish) experienced no mortality. These observa- 94% for Rainbow Trout, suggesting minimal life-threatening tions are consistent with those of Miranda and Kidwell (2010), effects. For the non-trout species, the observed mortality ranged who reported that electrofishing mortality in several species un- from 0% in Green Sunfish and Channel Catfish to as much as der controlled laboratory conditions averaged 16%. 20% in Fathead Minnow. These mortality rates are well within This is not to imply that electrofishing effects on Rainbow the range of natural mortality observed in other species (Pardue Trout and Brook Trout were benign. In fact, some of the most 1993; Quist and Guy 2001; Sammons and Maceina 2009). How- serious musculoskeletal injuries observed during this study were ever, despite the apparent acceptable electrofishing field and low documented in these two species. These observations are con- mortality in our study, these fishes experienced high frequencies sistent with other studies that have demonstrated a high suscep- of musculoskeletal trauma. The effects of these stressors on the tibility of salmonids to electrofishing injuries (Snyder 2003). fish’s blood chemistry and stress response have not been fully For instance, Sharber and Carothers (1988) detected spinal and described or evaluated (Panek and Densmore 2011). hemorrhagic injuries in 40–67% of Rainbow Trout that were Our results clearly suggest that multiple-pass PDC elec- collected by PDC electrofishing in the Colorado River. In a trofishing for target populations of trout results in unintended 3-year study evaluating the effects of Rainbow Trout stock- trauma to co-inhabiting bycatch species under conditions that ing on native fish populations in an Ozark stream, Walsh et al. are optimal for the capture of Rainbow Trout—that is, effec- (2004) found hemorrhages in up to 90% of Rainbow Trout that tive capture without inducing tetany. We demonstrated statisti- were shocked with 60-Hz AC. Overall, we found that although cally significant relationships (P ≤ 0.05) between an increasing the frequency of both hemorrhagic trauma and spinal trauma number of electrical shocks and the frequency of hemorrhagic was high in Rainbow Trout, the severity of these injuries after and spinal trauma in four of the five nontarget species, and we three-pass sampling was, at worse, slight to moderate on the demonstrated that these injuries will become progressively more ATS scale. In Brook Trout, the high incidence of hemorrhagic severe as the number of electrical shocks increases. For instance, injury (70%) after the one-shock treatment was essentially sus- the frequency of hemorrhagic trauma in Potomac Sculpin, Fat- tained in multiple-shock scenarios without incurring additional head Minnow, Channel Catfish, Green Sunfish, and Largemouth trauma. This was also observed for the incidence of spinal injury, Bass collectively averaged 16.0 ± 11.4% (mean ± SD) after wherein the average rate of injury after three shocks was 51 ± single-pass electrofishing, 38.0 ± 13.0% after two passes, and 10.2% (mean ± SD) and the severity of these injuries remained 48.3 ± 7.9% after three passes. Likewise, spinal trauma rates relatively constant across the treatments (ATS = 1.76 ± 0.12). increased from 11.9 ± 9.3% after single-pass electrofishing to Clearly, both of the trout species were highly susceptible to in- 21.7 ± 6.2% after two shocks and finally to 32.5 ± 20.8% after jury, supporting the commonly held view that salmonids in PDC three shocks. These rates of injury were considerably higher than fields are prone to high rates of injury (Snyder 2003). those documented by Miranda and Kidwell (2010) for nongame The banding or branding observed in Rainbow Trout and species that were exposed to single-pass DC and PDC elec- Brook Trout consistently occurred in fish that exhibited mid- trofishing, with the incidence of hemorrhagic trauma averaging dorsal hemorrhagic trauma, either with or without spinal frac- 2% and the incidence of spinal trauma averaging 6%. In our tures. This condition has been noted by others and attributed to experiments, the ATS values for the bycatch species increased electrofishing-induced hemorrhagic trauma (Ainslie et al. 1998). from 0.74 after one shock to 1.64 after three shocks. However, However, not all fish with hemorrhagic trauma exhibit the brand- despite the progressive increase in the frequency of trauma with ing effect; this was clearly the case in Brook Trout, as 7% of increasing electrofishing exposures, the severity of these injuries Downloaded by [Department Of Fisheries] at 22:32 28 February 2013 the fish demonstrated branding, whereas up to 70% experienced for most of the species we tested would be considered slight to hemorrhagic trauma. Our observations agree with previous stud- minor within the range of the ATS (0–6). ies (Sharber and Carothers 1988; Sharber and Sharber Black Fishery biologists need to be cognizant of the potential harm 1999) indicating that the presence of branding in salmonids, to nontarget aquatic life that may result from electrofishing, par- while indicative of hemorrhagic injury, should not be consid- ticularly when exposing fish to multiple electrical shocks. The ered a reliable measure of the degree or severity of the injury. target species of the electrofishing should be collected by using Our results for Brook Trout suggest that this commonly held the lowest possible power that results in efficient capture, and assumption could be seriously misleading in assessments of the crews should work to minimize the number of electrofishing degree of musculoskeletal trauma. passes that are required to accomplish the sampling objective. The results also suggest the need for a cautionary approach At times, this may be difficult since depletion sampling requires by field biologists when conducting salmonid population sur- three or more passes for effective population estimation (Ri- veys. In our work, we deliberately set our test electrofishing ley and Fausch 1992; Bacon and Youngson 2007). Differential field to a condition that produced narcosis but minimal tetany in rates of trauma experienced by fish exposed to multiple-pass Rainbow Trout. The purpose was to simulate reasonable opera- electrofishing may affect the assumption of equal capture vul- tional conditions for collecting fish in population surveys and to nerability, which is inherent to estimates of population size. ELECTROFISHING HEMORRHAGIC AND SPINAL TRAUMA 185

From an ecological perspective, little is known about the Kocovsky, P. M., C. Gowan, K. D. Fausch, and S. C. Riley. 1997. Spinal injury sublethal effects of electrofishing on either the individual or rates in three wild trout populations in Colorado after eight years of backpack the affected populations (Panek and Densmore 2011). Based electrofishing. North American Journal of Fisheries Management 17:308– 313. on the present results, we know that injuries to nontarget or McDonald, J. H. 2009. Handbook of biological statistics, 2nd edition. Sparky bycatch species can be significant and, most often, dose depen- House Publishing, Baltimore, Maryland. dent. In some species, this sublethal stress has been shown to Mesa, M. G., and C. B. Schreck. 1989. Electrofishing mark–recapture and deple- reduce swimming stamina (Mitton and McDonald 1994), in- tion methodologies evoke behavioral and physiological changes in Cutthroat duce lethargy and elicit behavioral changes (Sigismondi and Trout. Transactions of the American Fisheries Society 118:644–658. Miranda, L. E., and R. H. Kidwell. 2010. Unintended effects of electrofishing on Weber 1988), reduce fertility (Muth and Ruppert 1996), or im- nongame fishes. Transactions of the American Fisheries Society 139:1315– pair cardiac function (Schreer et al. 2004). The population-level 1321. effects of sublethal exposures from multiple-pass electrofishing Mitton, C. J. A., and D. G. McDonald. 1994. Consequences of pulsed DC warrant further investigation, particularly for fish communities electrofishing and air exposure to Rainbow Trout (Oncorhynchus mykiss). supporting rare or threatened species that may be placed in fur- Canadian Journal of Fisheries and Aquatic Sciences 51:1791–1798. Muth, R. T., and J. B. Ruppert. 1996. Effects of two electrofishing currents on ther jeopardy by the effects of multiple-pass electrofishing. captive ripe Razorback Suckers and subsequent egg-hatching success. North American Journal of Fisheries Management 16:473–476. ACKNOWLEDGMENTS Panek, F. M., and C. L. Densmore. 2011. Electrofishing and the effects of depletion sampling on fish health: a review and recommendations for addi- We thank Christine Bunck, Nathan Ramsay, Dottie Johnson, tional study. Pages 299–308 in R. C. Cipriano, A. W. Bruckner, and I. S. Allison Klein, and Stephanie Steinfeldt at the USGS National Shchelkunov, editors. Bridging America and Russia with shared perspec- Wildlife Health Center (Madison, Wisconsin) for their assis- tives on aquatic animal health: proceedings of the third bilateral conference tance in sample preparation and radiography. We are also grate- between the United States and Russia. Khaled bin Sultan Living Oceans ful to Alan Temple, Steven Moore, and Bruce Taggart for provid- Foundation, Landover, Maryland. Pardue, G. B. 1993. Life history and ecology of the Mud Sunfish (Acantharchus ing constructive suggestions on earlier drafts of the manuscript. pomotis). Copeia 1993:533–540. Use of trade, product, or firm names does not imply endorsement Quist, M. C., and C. S. Guy. 2001. Growth and mortality of prairie stream fishes: by the U.S. Government. relations with fish community and instream habitat characteristics. Ecology of Freshwater Fish 10:88–96. Reynolds, J. B. 1996. Electrofishing. Pages 221–253 in B. R. Murphy and D. REFERENCES W. Willis, editors. Fisheries techniques, second edition. American Fisheries Ainslie, B. J., J. R. Post, and A. J. Paul. 1998. Effects of pulsed and continuous Society, Bethesda, Maryland. DC electrofishing on juvenile Rainbow Trout. North American Journal of Riley, S. C., and K. D. Fausch. 1992. Underestimation of trout population size Fisheries Management 18:905–918. by maximum-likelihood removal estimates in small streams. North American Anderson, R. O., and R. M. Neumann. 1996. Length, weight and associated Journal of Fisheries Management 12:768–776. structural indices. Pages 447–482 in B. R. Murphy and D. W. Willis, editors. Sammons, S. M., and M. J. Maceina. 2009. Variation in growth and survival of Fisheries techniques, 2nd edition. American Fisheries Society, Bethesda, Bluegills and Redbreast Sunfish in Georgia rivers. North American Journal Maryland. of Fisheries Management 29:101–108. Bacon, P. J., and A. F. Youngson. 2007. The application of electro-fishing to Schreer, J. F., S. J. Cooke, and K. B. Connors. 2004. Electrofishing-induced produce census estimates of juvenile salmonid populations within defined cardiac disturbance and injury in Rainbow Trout. Journal of Fish Biology areas: taking stock of the options. Scottish Fisheries Research Report 67. 64:996–1014. Baker, R. W. R., and J. A. Nissim. 1963. Expressions for combining standard Sharber, N. G., and S. W. Carothers. 1988. Influence of electrofishing pulse errors of two groups and for sequential standard error. Nature 198:1020. shape on spinal injuries in adult Rainbow Trout. North American Journal of Clement,´ M., and R. A. Cunjak. 2010. Physical injuries in juvenile Atlantic Fisheries Management 8:117–122. Salmon, Slimy Sculpin, and Blacknose Dace attributable to electrofishing. Sharber, N. G., and J. Sharber Black. 1999. Epilepsy as a unifying principle North American Journal of Fisheries Management 30:840–850. in electrofishing theory: a proposal. Transactions of the American Fisheries Downloaded by [Department Of Fisheries] at 22:32 28 February 2013 Dolan, C. R., and L. E. Miranda. 2004. Injury and mortality of warmwater Society 128:666–671. fishes immobilized by electrofishing. North American Journal of Fisheries Sigismondi, L. A., and L. J. Weber. 1988. Changes in avoidance response time Management 24:118–127. of juvenile Chinook Salmon exposed to multiple acute handling stresses. Gatz, A. J., Jr., J. M. Loar, and G. F. Cada. 1986. Effects of repeated elec- Transactions of the American Fisheries Society 117:196–201. troshocking on instantaneous growth of trout. North American Journal of Smith, S. A., and B. J. Smith. 1994. Xeroradiographic and radiographic anatomy Fisheries Management 6:176–182. of the Channel Catfish, Ictalurus punctatus. Veterinary Radiology and Ultra- Hilborn, R., and C. J. Walters. 1992. Quantitative fisheries stock assessment: sound 35:384–390. choice, dynamics and uncertainty. Routledge, Chapman and Hall, New York. Snyder, D. E. 2003. Electrofishing and its harmful effects on fish. U.S. Geologi- Holliman, F. M., J. B. Reynolds, and T. J. Kwak. 2003. A predic- cal Survey, Information and Technology Report USGS/BRD/ITR-2003-0002, tive risk model for electroshock-induced mortality of the endangered Denver, Colorado. Cape Fear Shiner. North American Journal of Fisheries Management 23: Van Deventer, J. S., and W. S. Platts. 1983. Sampling and estimating fish popula- 905–912. tions from streams. Transactions of the North American Wildlife and Natural Kennard, M. J., B. J. Pusey, B. D. Harch, E. Dore, and A. H. Arthington. Resources Conference 48:349–354. 2006. Estimating local stream fish assemblage attributes: sampling effort Walsh, M. G., D. L. Winkelman, and R. J. Bahr. 2004. Electrofishing injury and and efficiency at two spatial scales. Marine and Freshwater Research 57: short-term mortality in hatchery-reared Rainbow Trout stocked into an Ozark 635–653. stream. North American Journal of Fisheries Management 24:316–321. This article was downloaded by: [Department Of Fisheries] On: 28 February 2013, At: 22:33 Publisher: Taylor & Francis Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

North American Journal of Fisheries Management Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/ujfm20 Evidence of Wild Juvenile Lake Trout Recruitment in Western Lake Michigan S. Dale Hanson a , Mark E. Holey a , Ted J. Treska a , Charles R. Bronte a & Ted H. Eggebraaten a a U.S. Fish and Wildlife Service, Green Bay Fish and Wildlife Conservation Office, 2661 Scott Tower Drive, New Franken, Wisconsin, 54229, USA Version of record first published: 29 Jan 2013.

To cite this article: S. Dale Hanson , Mark E. Holey , Ted J. Treska , Charles R. Bronte & Ted H. Eggebraaten (2013): Evidence of Wild Juvenile Lake Trout Recruitment in Western Lake Michigan, North American Journal of Fisheries Management, 33:1, 186-191 To link to this article: http://dx.doi.org/10.1080/02755947.2012.754804

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material. North American Journal of Fisheries Management 33:186–191, 2013 American Fisheries Society 2013 ISSN: 0275-5947 print / 1548-8675 online DOI: 10.1080/02755947.2012.754804

MANAGEMENT BRIEF

Evidence of Wild Juvenile Lake Trout Recruitment in Western Lake Michigan

S. Dale Hanson,* Mark E. Holey, Ted J. Treska, Charles R. Bronte, and Ted H. Eggebraaten U.S. Fish and Wildlife Service, Green Bay Fish and Wildlife Conservation Ofﬁce, 2661 Scott Tower Drive, New Franken, Wisconsin 54229, USA

rotated every 6 years to help differentiate year-classes at re- Abstract capture. Also, a proportion of the 1984, 1985, 1988–2004, and Lake Trout Salvelinus namaycush were extirpated from Lake 2009 year-classes, and all fish in the 2010–2011 year-classes Michigan by the early 1950s, and as part of an effort to restore were tagged with coded wire tags (CWT) and marked with an naturally reproducing populations, hatchery-reared fish have been stocked since the early 1960s. Stocked fish are marked with a fin clip adipose fin clip. These marks and tags allowed stocked fish to be to differentiate them from wild, lake-produced Lake Trout; mark- differentiated from wild fish. Fin clip marking error, measured ing error for the 2007–2010 year-classes of Lake Trout stocked by as the percentage of fish that were unintentionally released with- federal hatcheries averaged 3.0%. Egg deposition, emergent fry, out fin clips, averaged about 6% in federal hatcheries between and wild juvenile Lake Trout have previously been observed, but no 1990 and 2001 (Bronte et al. 2007). Despite nearly five decades sustained wild recruitment has been measured in assessment surveys or in sport and commercial fishery catches. In 2011 and 2012, of stocking, there has been no consistent wild juvenile Lake we caught juvenile Lake Trout in gill-net and bottom trawl catches Trout recruitment in Lake Michigan as evidenced by the recov- that were targeting Bloater Coregonus hoyi in water depths greater ery of unclipped fish (Holey et al. 1995; Bronte et al. 2007, than 80 m. Unclipped, wild Lake Trout represented 20% of all Lake 2008; Madenjian 2012). Trout caught in a southern offshore region of Lake Michigan. In Stocked Lake Trout survive to spawn in Lake Michigan, northwestern Lake Michigan wild recruits represented from 10% to 27% of the 2007–2009 year-classes, and we recovered a small and egg deposition (Marsden and Janssen 1997; Marsden et al. number of wild Lake Trout from the 2010 year-class. This is the 2005), fry emergence (Jude et al. 1981; Wagner 1981; Janssen first evidence for consecutive year-classes of naturally produced 2006), and recruitment of wild age-1 and older Lake Trout Lake Trout surviving beyond the fry stage in Lake Michigan. (Rybicki 1991) have been reported. However, recruitment of wild age-1 and older Lake Trout has been not been consistent. Population assessments performed between 1983 and 1989 in Lake Michigan once supported the world’s largest commer- Grand Traverse Bay and nearby Platte Bay in northeastern Lake

Downloaded by [Department Of Fisheries] at 22:33 28 February 2013 cial fishery for Lake Trout Salvelinus namaycush, but in the Michigan documented that 15% of the 1976 year-class, 8.9% of 1950s all populations in Lake Michigan and in most other the 1981 year-class, and 5.7% of the 1983 year-class comprised Great Lakes were extirpated due to predation by Sea Lamprey unclipped, presumably wild Lake Trout (Rybicki 1991), but Petromyzon marinus and overfishing (Wells and McLain 1972; none thereafter. Holey et al. 1995; Hansen 1999). In 1960, the U.S. Fish and Since 2010, the USFWS has collected Bloater Coregonus Wildlife Service (USFWS) began stocking Lake Trout finger- hoyi gametes during winter at multiple deepwater (>80 m) lings and yearlings to restore self-sustaining populations in Lake locations in western Lake Michigan to support reintroduction Michigan; Illinois, Michigan, and Wisconsin state agencies also efforts into Lake Ontario. As part of these collections, many stocked Lake Trout in later years. Between 2000 and 2010, an juvenile Lake Trout were captured, and here we summarize the average of 2.8 million fall-fingerling and yearling Lake Trout demographics of this bycatch that provide evidence of consecu- were stocked annually in Lake Michigan. Stocked fish have tive and consistent natural reproduction by Lake Trout in Lake been marked by the removal of one or more fins; fin clips were Michigan.

*Corresponding author: dale [email protected] Received July 13, 2012; accepted November 26, 2012

186 MANAGEMENT BRIEF 187

METHODS and was defined as the percentage of fish sampled without a Commercial fishers were contracted in February 2011 and fin clip or CWT. We summarized marking error as a lake-wide January and February 2012 to fish with 65-mm-stretch-mesh average and, because adult Lake Trout generally reside within monofilament gill nets at depths between 82 and 135 m, 60 km 68–100 km of their stocking location (Rybicki and Keller 1978; southeast of Milwaukee, Wisconsin, in 2011, and 60 km north- Schmalz et al. 2002; Bronte et al. 2007), at a finer spatial scale east of Sturgeon Bay off the northern Door Peninsula, Wiscon- for fish stocked in nearshore and offshore areas near Milwaukee, sin, in 2012 (Figure 1; Table 1). Nets were fished multiple nights nearshore areas near Manitowoc, and nearshore areas near the prior to lifting. Also, a commercial bottom trawl (net opening, northern Door Peninsula (Figure 1). The marking error for fish about 2.1 m high × 15.2 m wide) was used to sample fish on stocked near the stocking area will provide the best reference February 24, 2012, 22 km east of Manitowoc, Wisconsin, in for comparison of local recovery data when using the proportion depths between 87 and 89 m. Lake Trout were removed from of unclipped fish as evidence of natural reproduction. the total catch and brought to the laboratory for processing. All To ensure that the “wild” classification of unclipped Lake Lake Trout were measured for total length (TL, mm) and fin clips Trout were not false positives, we scanned all unclipped Lake were recorded. Year-class was assigned to individual unclipped Trout captured in both 2011 and 2012 for the presence of CWTs. Lake Trout caught in 2012 by aging the sagittal otolith. Otoliths All fish were also closely inspected for signs of fin regeneration were mounted and thin-sectioned, and the annuli were counted (false positive) as indicated by uneven lengths of paired fins or to assign year-class. In 2011, otoliths were not collected from wavy deformations of fin rays. We then reported the proportion unclipped fish. Year-class representation among the unclipped of unclipped Lake Trout from the total number of Lake Trout Lake Trout caught in 2011 was inferred by comparing the length captured at each survey area. frequency of the unclipped fish to the age–length frequencies of fin-clipped fish less than 500 mm, making the assumption that growth rates of wild and stocked fish were the same. Fin-clipped RESULTS fish larger than 500 mm were not used in this comparison be- For the 2007–2010 year-classes of Lake Trout stocked in cause more than 90% of the unclipped fish in 2011 were less Lake Michigan, mean weighted lake-wide marking error av- than 500 mm, and because of variable growth the same fin clip eraged 3.0%, and was 2.5% for fish stocked near Milwaukee for hatchery fish larger than 500 mm can represent two or more and <1% near Manitowoc and the northern Door Peninsula. year-classes. Stocked Lake Trout from the 2010 year-class were marked with Unclipped Lake Trout can occur from three possibilities: an adipose fin clip and tagged with a CWT. Loss of CWTs av- (1) they are of wild origin (true positive), (2) they were released eraged 2.5% in the hatchery for the 2010 year-class; hence, the from the hatchery without a fin clip unintentionally (false pos- probability that this year-class was stocked with no CWT and itive), or (3) they were released from the hatchery with a fin no fin clip was <0.09%. We inferred wild recruitment when clip but the fin (or fins) regenerated before capture and went the proportion of unclipped Lake Trout exceeded 3.0% of all unrecognized (false positive). Fin clip marking error and, when Lake Trout caught, which we noted was a conservative estimate applicable, CWT tagging error were determined from a sam- given that marking error in Wisconsin waters was lower than ple of fish within each raceway prior to their being stocked, the lake-wide mean.

TABLE 1. Summary of the Bloater survey gill-net and trawl effort and the associated Lake Trout bycatch for a southern offshore area of Lake Michigan in 2011 and offshore areas of northwestern Lake Michigan in 2012.

Downloaded by [Department Of Fisheries] at 22:33 28 February 2013 Depth Percentage Sample date Nights Latitude Longitude range CPUE·km−1· Number Number unclipped (month/day/year) ﬁshed Location (◦N) (◦W) (m) night−1 caught unclipped (%) 2/10/2011 4 Milwaukee 42.808 86.662 122–135 24.2 295 59 20.0 1/21/2012 10 Door Peninsula 45.162 86.776 95–130 0.5 14 6 42.8 1/25/2012 11 Door Peninsula 45.143 86.799 96–128 0.4 15 2 13.3 2/2/2012 8 Door Peninsula 45.180 86.754 94–132 1.1 27 0 0.0 2/3/2012 7 Door Peninsula 45.132 86.835 84–124 1.1 23 2 8.7 2/9/2012 9 Door Peninsula 45.158 86.731 93–130 0.7 20 3 15.0 2/14/2012 12 Door Peninsula 45.181 86.761 82–132 1.2 44 10 22.7 2/16/2012 13 Door Peninsula 45.121 86.764 82–125 0.6 23 2 8.7 2/23/2012 9 Door Peninsula 45.179 86.712 84–132 0.5 15 1 6.7 2/28/2012a Manitowoc 44.091 87.405 87–89 4 3 75.0

aTrawl tow, all other data describe gill-net lifts. 188 HANSON ET AL. Downloaded by [Department Of Fisheries] at 22:33 28 February 2013

FIGURE 1. Lake Michigan showing gill-net and trawl survey locations in 2011 and 2012 where Lake Trout were captured in Bloater spawner surveys. MANAGEMENT BRIEF 189

0.20 3.7% were from the 2005 year-class, 42.6% from the 2006 year- Unclipped class, 45.8% from the 2007 year-class, and 6.0% were from the 0.15 Clipped 2008 year-class. No Lake Trout less than 500 mm possessed an 0.10 2011 total catch adipose fin clip or a CWT. 59 of 295 ( 20.0 %) unclipped In 2012, eight gill-net lifts off the northern Door Penin- 0.05 sula were made between January 21 and February 23. A to- 0.00 tal of 180 Lake Trout were caught and the mean CPUE was 0.20 0.8 (SD = 0.3). Lake Trout mean TL was 353 mm (SD = 0.15 RP clip, 2008 year-class 86 mm); 14.4% (26 of 180) of all Lake Trout were unclipped and none of these fish contained CWTs. Six fin-clipped fish 0.10 n = 13 contained CWTs: two fish of the 2009 year-class and one fish 0.05 of the 2010 year-class all stocked in the Northern Refuge, and 0.00 one each from the 1999, 2000, and 2003 year-classes stocked in

0.20 the Southern Refuge. The single bottom trawl tow near Mani- towoc caught four Lake Trout. Three ﬁsh were unclipped (75%) 0.15 RV clip, 2007 year-class and had a mean length of 190 mm (SD = 13 mm), while a 0.10 n = 99 224-mm ﬁn-clipped Lake Trout with a CWT code of the 2010

0.05 year-class stocked at multiple locations as fall fingerlings was also captured. The percentages of unclipped fish caught in 2012 0.00 were 22.7, 27.0, 10.0, and 60.0% for the 2007, 2008, 2009, and 0.20 2010 year-classes, respectively (Figure 3). Size comparisons 0.15 RPLV clip, 2006 year-class within year-classes indicate unclipped Lake Trout were smaller than fin-clipped fish at ages 2 and 3, but modal size was similar 0.10 n = 92 for fish captured at ages 4 and 5 (Figure 3). 0.05

Proportion of fish in each length bin within0.00 the total catch 0.20 DISCUSSION We have provided the first evidence of consistent recruitment LP clip, 2005 year-class 0.15 of wild age-1 and older Lake Trout in Lake Michigan based on 0.10 n = 8 the percent of unclipped Lake Trout that exceeded the 3.0% 0.05 lakewide marking error for Lake Trout stocked between 2007 and 2010. In northwestern Lake Michigan we observed contri- 0.00 200 260 320 380 440 500 560 620 680 740 800 bution of wild recruits for the 2007 (22.7%), 2008 (27.0%), 2009 (10.0%), and 2010 (60.0%) year-classes (though the sample size Length bin (mm) for the 2010 year-class was small). The length frequencies of FIGURE 2. Stacked barplot showing the length frequencies for fin-clipped these fish at ages 4 and 5 were similar between wild and stocked and unclipped Lake Trout sampled near the Southern Refuge, 2011. The top fish, which suggests we can be reasonably confident in our panel shows the proportion of unclipped and fin-clipped Lake Trout in each year-class assignments where no otolith was available. Twenty = length bin relative to the total catch (n 295), and the lower four panels show percent of the Lake Trout we sampled in the southern offshore the proportion of fin-clipped fish in each length bin, by year-class, from the total Downloaded by [Department Of Fisheries] at 22:33 28 February 2013 number of fin-clipped fish caught (n = 216). The year-class-specific fin clips location in Lake Michigan were wild recruits, and length fre- were right pectoral (RP) for 2008, right ventral (RV) for 2007, right pectoral quency comparisons suggested these were most probably age-5 and left ventral (RPLV) for 2006, and left pectoral (LP) for 2005. fish from the 2006 year-class and age-4 fish from the 2007 year- class. The occurrence of these wild fish is much higher than can be attributed to marking error and provides evidence of some In 2011, three gill-net lifts were made in southern Lake successful Lake Trout natural reproduction during 2007–2010 Michigan, but Lake Trout bycatch was only retained from the in northwestern Lake Michigan, in addition to the 2006 and last lift on February 10 (Table 1). A total of 295 Lake Trout 2007 year-classes in the southern portion of the lake. were caught resulting in a CPUE of 24.2 fish/km of net per Our findings are corroborated by a small yet notable increase night fished. Lake Trout mean TL was 385 mm (SD = 101 mm) in the mean percentage of unclipped Lake Trout caught in the and 93% of the fish were less than 500 mm. The percentage of standardized, multiagency, spring, graded-mesh gill net survey. unclipped fish was 20.0% (59 of 295) and no CWTs or fin re- The spring survey targets Lake Trout at depths between 15 and generation were detected in these unclipped fish. The majority 61 m in nine nearshore areas and offshore in the Northern and of the unclipped fish were probably from the 2006 and 2007 Southern Refuges (Schneeberger et al. 1998). Generally Lake year-classes based on the age–length frequencies of stocked Trout recruit to this survey’s graded mesh (63.5, 76.2, 88.9, fish (Figure 2). For fin-clipped fish less than 500 mm (n = 216), 101.6, 114.3, 127, 139.7, and 152.4 mm stretch mesh) beginning 190 HANSON ET AL.

0.2 and later year-classes become vulnerable to the spring survey, Unclipped but the percentage of unclipped recoveries lakewide reflects Clipped 2012 total catch both the proportion of wild recruits relative to stocked fish 0.1 and the spatial scale over which natural recruits are pro- 29 of 184 ( 15.8 %) unclipped duced. In 2011, 16 of 154 (10.4%) Lake Trout sampled in the spring survey near Waukegan, Illinois, were unclipped (Steve 0.0 Robillard, Illinois Department of Natural Resources, personal 0.3 communication), which provides evidence that wild recruits are also being produced in Illinois waters in addition to the southern 0.2 offshore and nearshore waters near Manitowoc and the northern 2010 year-class Door Peninsula where we detected wild fish. 0.1 3 of 5 ( 60.0 %) unclipped Impediments to natural reproduction of Lake Trout in Lake 0.0 Michigan have been evaluated (Bronte 2003; Bronte et al. 2007, 2008) and are summarized as having poor survival of early life 0.3 stages, a lakewide population too low to overcome bottlenecks, 0.2 and inappropriate stocking practices. The management strat- 2009 year-class egy to restore Lake Trout in Lake Michigan has been revised 0.1 11 of 110 ( 10.0 %) unclipped and implemented by the Lake Michigan Committee (Dexter et al. 2010) to address stocking practices and low abundance 0.0 in selected areas of the lake, but the changes enacted will take 0.3 years to evaluate their effectiveness. Currently Lake Trout total and adult abundance remains below target levels (Madenjian 0.2 2012). Thiamine deficiency complex has been implicated as

Proportion of fish within each length bin 2008 year-class the primary cause for poor survival of early life stages. This 0.1 10 of 37 ( 27.0 %) unclipped complex is linked to a maternal diet of Alewives Alosa pseudo- 0.0 harengus that causes a thiamine deficiency in eggs that result 0.3 in poor hatching success and high posthatch mortality among fry (Fisher et al. 1996). In Lake Michigan, mean egg thiamine 0.2 levels in Lake Trout have increased in recent years and are corre- 2007 year-class lated with a concomitant decrease in Alewife abundance (Riley 0.1 5 of 22 ( 22.7 %) unclipped et al. 2011). Lake Trout eggs now exceed the 4-nmol/g thiamine concentration threshold level and meet the definition of viable 0.0 eggs for restoration efforts (Bronte et al. 2008) at most sites; 160 240 320 400 480 560 640 720 800 this may partially explain the recent detection of wild recruits. Length (mm) Based on our results, it appears that some reproduction of Lake Trout is now occurring in western Lake Michigan, albeit FIGURE 3. Stacked barplot showing the length frequencies for fin-clipped and unclipped Lake Trout caught in gill-net and trawl sampling, 2012. The top at a low level. As progress toward the abundance and egg quality panel shows the total catch while the lower four panels show the proportion targets for rehabilitation continues, we would expect increasing within each length bin by year-class. levels of wild recruitment to occur. Downloaded by [Department Of Fisheries] at 22:33 28 February 2013

at age 3 with a modal age of 5 (Lake Michigan Lake Trout ACKNOWLEDGMENTS Working Group, unpublished data). Between 1999 and 2010, The authors thank commercial fishers Charlie Henrikson, <3% of the Lake Trout catch from all sites combined were un- Dan Anderson, Michael LeClair, and the crews of the Alicia Rae clipped. In 2011, 4.3% of all Lake Trout caught in the spring and Susie Q for their assistance completing the field surveys. survey were unclipped and these fish were smaller (mean = Rachel Van Dam prepared the otoliths for aging. The findings 563 mm) than fin-clipped fish (638 mm) (Madenjian 2012). and conclusions in this article are those of the authors and do Our Bloater surveys employed 65-mm stretch mesh in deep- not necessarily represent the views of the U.S. Fish and Wildlife water (>80 m) habitats preferred by juvenile Lake Trout (El- Service. rod and Schneider 1987) and we caught mostly small Lake Trout < 500 mm. The 2006 year-class of wild Lake Trout captured in our survey probably contributed to the 2011 spring REFERENCES Bronte, C. R., M. E. Holey, C. P. Madenjian, J. L. Jonas, R. M. Claramunt, survey catch (as 5-year-old fish); however, later year-classes P. C. McKee, M. L. Toneys, M. P. Ebener, B. Breidert, G. W. Fleischer, R. were not yet fully recruited. We predict the percentage of wild Hess, A. W. Martell Jr., and E. J. Olsen. 2007. Relative abundance, site fi- fish will continue to increase as natural recruits from 2007 delity, and survival of adult Lake Trout in Lake Michigan from 1999 to 2001: MANAGEMENT BRIEF 191

implications for future restoration strategies. North American Journal of Fish- Jude, D. J., S. A. Klinger, and M. D. Enk. 1981. Evidence of natural reproduction eries Management 27:137–155. by planted Lake Trout in Lake Michigan. Journal of Great Lakes Research Bronte, C. R., J. L. Jonas, M. E. Holey, R. L. Eschenroder, M. L. Toneys, P. 7:57–61. McKee, B. Breidert, R. M. Claramunt, M. P.Ebener, C. C. Krueger, G. Wright, Madenjian, C. P. 2012. Lake Trout working group report. Report to the Lake and R. Hess. 2003. Possible impediments to Lake Trout restoration in Lake Michigan Committee, Windsor, Ontario, March 19, 2012, Great Lakes Fish- Michigan. Great Lakes Fishery Commission, Lake Michigan Committee, Ann ery Commission, Ann Arbor, Michigan. Arbor, Michigan. Marsden, J. E., and J. Janssen. 1997. Evidence of Lake Trout spawning on a Bronte,C.R.,C.C.Krueger,M.E.Holey,M.L.Toneys,R.L.Eshenroder,andJ. deep reef in Lake Michigan using an ROV-based egg collector. Journal of L. Jonas. 2008. A guide for the rehabilitation of Lake Trout in Lake Michigan. Great Lakes Research 23:450–457. Great Lakes Fishery Commission, Miscellaneous Publication 2008-01, Ann Riley, S. C., J. Rinchard, D. C. Honeyfield, A. N. Evans, and L. Begnoche. 2011. Arbor, Michigan. Increasing thiamine concentrations in Lake Trout eggs from Lakes Huron and Dexter, J. L., Jr., B. T. Eggold, T. K. Gorenflo, W. H. Horns, S. R. Robillard, Michigan coincide with low Alewife abundance. North American Journal of and S. T. Shipman. 2010. A fisheries management implementation strategy Fisheries Management 31:1052–1064. for the rehabilitation of Lake Trout in Lake Michigan. Great Lakes Fishery Rybicki, R. W. 1991. Growth, mortality, recruitment and management of Lake Commission, Lake Michigan Committee, Ann Arbor, Michigan. Trout in eastern Lake Michigan. Michigan Department of Natural Resources, Elrod, J. H., and C. P. Schneider. 1987. Seasonal bathythermal distribution Fisheries Research Report 1863, Ann Arbor. of juvenile Lake Trout in Lake Ontario. Journal of Great Lakes Research Rybicki, R. W., and M. Keller. 1978. The Lake Trout resource in Michigan 13:121–134. waters of Lake Michigan, 1970–1976. Michigan Department of Natural Re- Fisher, J. P., J. D. Fitzsimons, G. F. Combs Jr., and J. M. Spitsbergen. 1996. sources, Fisheries Research Report 1863, Lansing. Naturally occurring thiamine deficiency causing reproductive failure in Finger Schmalz, P. J., M. J. Hansen, M. E. Holey, P. C. McKee, and M. L. Toneys. 2002. Lakes Atlantic Salmon and Great Lakes Lake Trout. Transactions of the Lake Trout movements in northwestern Lake Michigan. North American American Fisheries Society 125:167–178. Journal of Fisheries Management 22:737–749. Hansen, M. J. 1999. Lake Trout in the Great Lakes: basinwide stock collapse Schneeberger, P., M. Toneys, R. Elliott, J. Jonas, D. Clapp, R. Hess, and and binational restoration. Pages 417–454 in W. W. Taylor and C. P. Fer- D. Passino-Reader. 1998. Lakewide assessment plan for Lake Michigan reri, editors. Great Lakes fisheries policy and management. Michigan State fish communities. Great Lakes Fishery Commission, Lake Michigan Tech- University Press, East Lansing. nical Committee, Ann Arbor, Michigan. Available: www.glfc.org/pubs/ Holey,M.E.,R.W.Rybicki,G.W.Eck,E.H.BrownJr.,J.E.Marsden,D.S. SpecialPubs/lwasses01.pdf. (July 2012). Lavis, M. L. Toneys, T. N. Trudeau, and R. M. Horrall. 1995. Progress toward Wagner, W. C. 1981. Reproduction of planted Lake Trout in Lake Lake Trout restoration in Lake Michigan. Journal of Great Lakes Research Michigan. North American Journal of Fisheries Management 1:159– 21(Supplement 1):128–151. 164. Janssen, J., D. J. Jude, T. A. Edsall, R. W. Paddock, N. Wattrus, M. Toneys, and Wells, L., and A. L. McLain. 1972. Lake Michigan: effects of exploitation, P. McKee. 2006. Evidence of Lake Trout reproduction at Lake Michigan’s introductions, and eutrophication on the salmonid community. Journal of the mid-lake reef complex. Journal of Great Lakes Research 32:749–763. Fisheries Research Board of Canada 29:889–898. Downloaded by [Department Of Fisheries] at 22:33 28 February 2013 This article was downloaded by: [Department Of Fisheries] On: 28 February 2013, At: 22:33 Publisher: Taylor & Francis Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

North American Journal of Fisheries Management Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/ujfm20 A Probabilistic Model for Assessing Passage Performance of Coastal Cutthroat Trout through Corrugated Metal Culverts N. Phil Peterson a , Ryan K. Simmons a , Tamre Cardoso b & Jeffrey T. Light c a West Fork Environmental, Post Office Box 4455, Olympia, Washington, 98501, USA b Terrastat Consulting Group, 6322 14th Avenue Northeast, Seattle, Washington, 98115, USA c Plum Creek Timber Company, 380 Northwest 1st Street, Toledo, Oregon, 97191, USA Version of record first published: 01 Feb 2013.

To cite this article: N. Phil Peterson , Ryan K. Simmons , Tamre Cardoso & Jeffrey T. Light (2013): A Probabilistic Model for Assessing Passage Performance of Coastal Cutthroat Trout through Corrugated Metal Culverts, North American Journal of Fisheries Management, 33:1, 192-199 To link to this article: http://dx.doi.org/10.1080/02755947.2012.750633

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material. North American Journal of Fisheries Management 33:192–199, 2013 C American Fisheries Society 2013 ISSN: 0275-5947 print / 1548-8675 online DOI: 10.1080/02755947.2012.750633

ARTICLE

A Probabilistic Model for Assessing Passage Performance of Coastal Cutthroat Trout through Corrugated Metal Culverts

N. Phil Peterson* and Ryan K. Simmons West Fork Environmental, Post Ofﬁce Box 4455, Olympia, Washington 98501, USA Tamre Cardoso Terrastat Consulting Group, 6322 14th Avenue Northeast, Seattle, Washington 98115, USA Jeffrey T. Light Plum Creek Timber Company, 380 Northwest 1st Street, Toledo, Oregon 97191, USA

Abstract We conducted a series of volitional trials with wild-caught resident Coastal Cutthroat Trout Oncorhynchus clarkii clarkii in a 12.2-m-long, 1.8-m-diameter culvert test facility to develop a probabilistic model for predicting rates of upstream passage over a wide range of average velocities. The results of the passage trials indicated that the percentage of fish attempting passage and the percentage of fish successfully passing decreased as the trial target average velocity increased. At our highest trial average velocity of 2.4 m/s, 31% of test fish that chose to attempt passage passed after two nights of observation. Passage performance was generally better for larger fish, but this pattern was only statistically significant for a single trial (1.9 m/s). Fish ascended through the pipe more quickly as velocity increased. At higher test velocities fish favored the left side of the pipe (looking downstream), which contained a reduced-velocity zone created by the slightly oblique orientation of culvert corrugations. Our data provide the basis for a logistic model describing the probability of passage for Coastal Cutthroat Trout through bare corrugated metal culverts with no outlet drop. Empirical studies testing fish passage, such as this one, can inform culvert assessment protocols currently in use.

Corrugated metal culverts are a relatively inexpensive a significant and ongoing financial investment and obligation Downloaded by [Department Of Fisheries] at 22:33 28 February 2013 stream crossing structure, but they can inhibit fish movement for private and public land managers alike. Increasingly the due to high velocities and physical drops at the outfall (Park expectation and regulatory requirement is that stream crossings et al. 2008; Rolls 2011). Where unimproved forest roads in provide for upstream fish movement. the Pacific Northwest intersect small to midsized streams, the For stream-dwelling fish, movement is often necessary for crossing structure of choice has long been the corrugated metal gaining access to spawning and rearing habitat, food resources, pipe (CMP); thousands of these structures currently exist in and thermal or velocity refugia. Barriers to upstream movement steep headwater streams. Forest roads are critical for moving can fragment habitats and isolate populations, reducing gene logs and other forest commodities to market and maintaining flow and the genetic diversity important for maintaining long- access for timber harvest, forest management activities, term population viability (Morita and Yamamoto 2002; Wofford and public recreation. Maintenance of these extensive road et al. 2005). Factors that may impede upstream movement at networks, including the stream crossing structures, represents culverts include a high velocity or an inadequate depth (both

*Corresponding author: [email protected] Received April 9, 2012; accepted November 7, 2012

192 ASSESSING PASSAGE OF CUTTHROAT TROUT 193

of which are related to steep slope), an excessive drop at the data could lay the foundation for an empirically derived function outlet, or an accumulation of debris at the inlet. It is common for Cutthroat Trout passability through CMPs, providing a tool for culverts to create partial barriers (i.e., they are barriers on for quantitatively assessing the risk to nonpassage and inform some flows, to some species, or for some life stages of a partic- decisions on culvert replacement. We tested the volitional pas- ular species; Kemp and O’Hanley 2010). For the unimproved sage performance of Coastal Cutthroat Trout in an experimental forest road infrastructure in the Pacific Northwest, these partial culvert test facility. These tests spanned a range of average ve- barriers often occur in relatively steep headwater streams within locities typical of those found at road crossings in headwater reaches populated only by resident Coastal Cutthroat Trout On- stream systems and were used to develop a probabilistic model corhynchus clarkii clarkii, a species whose habitat is naturally for assessing passage success. fragmented by permanent and transient barriers such as bedrock waterfalls and slides, boulder and cobble cascades, and log jams (Gresswell et al. 2004). The additional habitat fragmentation METHODS and loss of connectivity caused by culverts representing partial Study site.—Fish passage trials were performed at an experi- or complete barriers has been cited as a concern for Coastal mental culvert test bed (CTB) facility located at the Washington Cutthroat Trout (Latterell et al. 2003; Gresswell et al. 2004). Department of Fish and Wildlife Skookumchuck Fish Hatchery Identification, prioritization, and remediation of fish passage (for a detailed description of facility construction see Pearson barriers have become key elements of stream and salmonid et al. 2005). The facility provides a platform for testing fish restoration efforts throughout western North America. An ar- passage through culverts of different diameters, with the ca- ray of methods for assessing potential barriers at road crossing pacity for discharges of up to 0.57 m3/s and adjustable slopes structures has been developed to date. While some informa- of up to 10%. Its water source originates from the Skookum- tion is available documenting fish passage through culverts, chuck Reservoir immediately upstream from the hatchery, and these data were not collected in a systematic way to capture the an in-line electromagnetic flowmeter (McCrometer) allows for range of passage conditions fish are likely to face for the given the instantaneous measurement of discharge. This study used a crossing structure. Thus, the most common barrier assessment single helically wound corrugated steel culvert with a circular approaches use other methods for determining whether fish are diameter of 1.83 m and a length of 12.2 m. Corrugations had capable of passage. For example, simulation-based assessments an amplitude of 2.54 cm and a wavelength of 7.62 cm, with a such as FishXing (Furniss et al. 2008) incorporate hydrological right-hand spiraled pitch of 5.3◦. The pipe was left bare for all and fish swimming capability data. Recent studies have docu- trials, free of baffle structures or natural substrate. Discharge and mented cases where simulations provide conclusions incongru- slope were varied to achieve a range of target water velocities, ent with field observations, overstating the level of passage risk and tailwater pool surface elevation was controlled by an ad- to fish (Cahoon et al. 2005; Karle 2005; Burford et al. 2009). justable stop-log weir to maintain similar pipe outlet conditions Other assessment approaches rely on expert systems that apply across all trials. An enclosure constructed of a wooden frame a passage value based on the judgment of a field crew (e.g., covered by 6.4-mm wire mesh was used to contain fish within WDFW 2009), though these are rarely validated with empiri- the tailwater pool at the entrance to the culvert. cal datasets (Kemp and O’Hanley 2010). In the case of both Detection system.—Detection of fish movement during pas- simulation and expert-based approaches, empirical fish passage sage trials relied on the development of a half-duplex passive data could help inform and improve the accuracy of fish pas- integrated transponder (PIT) system. An array of four radio fre- sage assessments. Incorrect classification of barrier status and quency identification (RFID) antennas was constructed within evaluation of the benefits to culvert replacement can either place the test culvert to detect fine-scale movements of each tagged Downloaded by [Department Of Fisheries] at 22:33 28 February 2013 valued fisheries resources at risk or, conversely, waste signifi- individual during the trials. Two types of antennas were used: cant public and private funds with a more costly crossing option. two pass-through antennas at the downstream and upstream ends Because there are hundreds of millions of dollars at stake (US- of the pipe, and two pass-under antennas near the midpoint in GAO 2001) and it is highly improbable that all the suspected the pipe. The pass-through antennas detected fish across the barriers can be corrected in a reasonable period of time, it is entire width of the pipe, and each of the two pass-under an- critical that barrier assessment protocols accurately reflect the tennas was tuned to only detect fish on the left or right side of risk to passability and the benefits for fish populations. the pipe. Pass-through antennas at the culvert entrance and exit Recent studies have identified the need for empirically based were tuned so the fields detected fish as they entered and exited fish passage data, which have the potential to further inform the culvert but not if they were milling about or holding in front assessment protocols currently in use while also facilitating the of the culvert. Detection of tagged fish at pass-under antennas development of a new probabilistic approach (Haro et al. 2004; occurred approximately from the culvert midline outward and Cahoon et al. 2007; Burford et al. 2009). For culverts located within 0.5 m downstream and upstream of the antennas. A zone in headwater streams of the Pacific Northwest, information re- approximately 15 cm wide along the culvert midline between garding the passage performance of Cutthroat Trout could help the pass-under antennas purposefully had no detection for de- to inform passability assessment protocols. Additionally, these termining passage through the midline. Midline passage was 194 PETERSON ET AL.

determined by subtracting passage on the left and right sides handheld reader to assure tag retention. Air and water temper- from all passage detections. A pass-under antenna for detecting atures, as well as headwater and tailwater surface elevations, midline passage was not possible due to the coupling of electri- were recorded both before and immediately after each passage cal fields. Types of movements recorded for each fish included trial. Hydraulic conditions for each trial were set 3 to4hpriorto entries into the pipe, unsuccessful passage attempts, successful placement of fish into the containment cage to allow the system passes through the culvert, and the side (left or right) of the to equilibrate. culvert used for each of these events. We synched five Oregon Passage performance was summarized by reviewing the in- RFID single antenna high performance LF HDX, ISO 11784 formation recorded from each RFID antenna. Each individual compatible readers together to record detections. Half-duplex fish was first identified as either a participant or a nonparticipant. Texas Instrument PIT tags (12 mm × 2.5 mm) were used for Participants included all fish that were detected at a minimum all trials. A previous comparison of the performance of tagged on the most downstream antenna, indicating that they at least and untagged fish in the CTB indicated no tag effect on rates of made an attempt to enter the culvert. A rate of participation for participation and passage. each trial was calculated by dividing the participants by the total Field capture and handling.—Coastal Cutthroat Trout used number of individuals tested. Nonparticipants included fish that in passage trials were collected from headwater tributaries of were never recorded on any of the antennas, and these fish were the Skookumchuck River (latitude 46.756822 N, longitude – dropped from further analysis. Participants were classified as 122.580845 W). All fish capture locations were separated from successful or unsuccessful in passing through the entire length the CTB facility by the Skookumchuck Reservoir and an impass- of the culvert based on detections at the uppermost antenna. able dam. Test fish (20–29 individuals per trial) were collected Elapsed time of ascent and the location of the fish midculvert during separate field samplings so that only na¨ıve fish were used (left or right) were recorded for each successful passage at- in each trial. The trial population size was constrained by the tempt. A rate of successful passage for each trial was calculated level of effort and time required to capture fish as well as by by dividing the number of successfully passing fish by the total keeping densities low in the transport tank to minimize stress. number of participating fish in the trial. Minimizing density-related stress was also important during the Water velocities for testing ranged from 0.6 to 2.4 m/s, while pretrial holding periods and the trials themselves. Fish were col- depths ranged from 0.12 to 0.28 m. A standard-step model of lected with a Smith-Root LR-24 backpack electrofisher using gradually varied flow (J. Cahoon, Montana State University, power output settings that minimized risk of injury to the fish personal communication) was used to determine the combina- (voltage range 600–650, frequency 30 Hz, duty cycle 12%). Any tion of culvert slopes and flows required to produce velocities in individuals that did not immediately recover or that had visible 0.15 m/s increments. The model was set to calculate distance- marks or physical deformities or injuries were not used for the averaged bulk velocity (hereafter “average velocity”) and output trials. Fish were transported to the CTB in an aerated 170-L the discharge required to achieve a target velocity, representing transport tank. the average cross-sectional velocity across the entire longitudi- At the CTB all fish were anaesthetized in a solution of MS- nal distance of the pipe. Due to facility constraints on slope (0– 222 (tricaine methanesulfonate; 40 mg/L), weighed to the near- 10%) and discharge (≤0.57 m3/s), three slopes (0.52%, 3.14%, est 0.1 g, and measured to the nearest millimeter (fork length; and 8.6%) were required to achieve the range of average veloci- FL). All fish were then implanted with half-duplex PIT tags ac- ties for testing (Table 1). Tabular model outputs were used to set cording to protocols from the Columbia Basin Fish and Wildlife up the CTB for each trial (i.e., a discharge value was selected Authority (CBFWA 1999). After tagging, fish were held in one from the output table to achieve the test velocity for any given of four 210-L holding tanks with flow-through rates of approxi- slope the CTB was set on). Test velocities tended to increase Downloaded by [Department Of Fisheries] at 22:33 28 February 2013 mately 38 L/min. Water temperature in each tank was monitored with testing date because the CTB was set at an increasingly and recorded throughout the pretrial holding period. Elapsed steeper slope, but within a slope setting we varied test velocity times between field capture and the start of passage trials varied randomly to avoid potential temporal bias. from 28 to 35 h for individual fish. Discharge was recorded both from the in-line flowmeter and Passage trials.—Each trial was performed over a 2-d period through the use of two sharp-crested weirs developed to pro- during June and July 2010. A trial began with the placement of vide an independent measure of flow. The in-line electromag- test fish into the containment cage in the late afternoon of the netic flowmeter was inspected and certified during the course first test day. The detection system was activated at that time of the study to provide assurance of accurate readings. Several and no coercive actions were taken to induce movement of fish. methods were used to validate modeled velocities during each Care was taken not to introduce light into the test facility during trial and provide assurance that actual test velocities comported evening hours, and all overhead or other external stimuli were with model predictions. Velocities were hand measured using avoided. Fish remained in the test facility for two evenings (a a Global Water velocity meter at three locations in the cross minimum of 36 h). On the morning of the second day, the wa- section of the pipe. These included the assumed locations near ter source was shut off and the headwater and tailwater tanks the left and right (looking downstream) sides of the waters’ were drained to collect the fish. All fish were scanned using a edges that fish were likely to occupy, plus a measurement in the ASSESSING PASSAGE OF CUTTHROAT TROUT 195

TABLE 1. Hydraulic characteristics and passage performance rates for the 11 passage trials (velocities and water depths derived from model output; Q = discharge; FL = fork length in mm; SE = standard error). Rates of success may be higher than participation due to the use of participants as the quotient rather than total N.

Velocity (m/s) Q (m3/s) Slope (%) Water depth (m) N FL (SE) Participation Success Trial sequence 0.6 0.06 0.5 0.15 21 123 (20.2) 1.0 0.86 1 0.8 0.12 0.5 0.21 20 130 (20.1) 0.95 0.89 3 0.9 0.21 0.5 0.28 26 122 (25.8) 0.96 1.0 2 1.4 0.14 3.1 0.15 23 123 (25.8) 0.78 0.61 7 1.6 0.22 3.1 0.18 23 117 (19.9) 0.96 0.77 4 1.7 0.30 3.1 0.21 29 111 (18.0) 1.0 0.83 6 1.8 0.41 3.1 0.25 27 121 (24.0) 0.67 0.39 5 1.9 0.14 8.6 0.12 22 115 (28.1) 0.82 0.33 9 2.2 0.23 8.6 0.15 26 122 (25.2) 0.77 0.75 8 2.3 0.28 8.6 0.16 28 117 (16.6) 0.79 0.27 11 2.4 0.34 8.6 0.18 26 120 (21.5) 0.62 0.31 10

inferred maximum velocity zone (culvert midline). These mea- interaction between velocity and depth that yielded a significant surements were taken at three longitudinal stations (1, 8, and effect for the velocity depth interaction and insignificant main 11.5 m from the downstream end of the pipe). As a secondary effects. While there does appear to be some dependence between check on average velocity, a neutrally buoyant object was timed velocity and depth, the interaction term and depth were dropped as it floated along the midline of the culvert prior to each trial. from the final model. Depth was not a targeted variable in this Finally, the accuracy of the modeled depth profile through the set of experiments and a full range of depth levels were not re- entire length of the culvert was assessed at 15 locations for the flected across the targeted velocities. As a result, interpretation 1.9-m/s trial conditions. These measured depths were averaged of the two-way interaction is difficult and any observed results to calculate an expected velocity for comparison with the mod- may not reflect the true dynamics of the two variables across eled velocity. The 1.9-m/s trial conditions were also selected for a full range of velocity and depth combinations. In turn, target fine-scale velocity mapping using 36 equally spaced point mea- velocity was the only independent variable used in the sub- surements to better understand the cross-sectional variability in sequent model reported here. All model fitting was conducted velocity. Use of an acoustic Doppler velocimeter proved impos- using R version 2–12.0 on Mac OS X. The 11 data records sible due to the excessive amount of noise in the data produced were used to find βˆ0 and βˆ1, estimates of the parameters β0 and by entrainment of air within the highly turbulent water column β1, respectively. Once the parameters were estimated, a fitted (addressed by Mori et al. 2007). Instead, a Swoffer meter (Model logistic curve was generated by estimating passage probabilities 2100) was used for these measurements, which were taken at calculated over a set of target velocities ranging from 0.0 to a cross section located at the longitudinal center of the culvert 3.0 m/s. A wider range of target velocities was used when fitting following the protocol established by Richmond et al. (2007). the logistic curve in order to better extrapolate the shape of the Statistical analysis.—The effect of fish size was evaluated curve for 0% and 100% passage. It is possible that the shape of with several tests. Since the distribution of fish lengths was the curve would be somewhat different if target velocities below strongly right skewed, a log transformation (loge) was applied to 0.6 and above 2.4 m/s had been included in the trials. Pointwise Downloaded by [Department Of Fisheries] at 22:33 28 February 2013 these length data. Differences in mean fish lengths between trials 95% confidence bands were added to the logistic curve using were compared using a one-way analysis of variance (ANOVA). procedures outlined in Hosmer and Lemeshow (1989). Testing for differences in mean length between successful and unsuccessful fish passage across trials was evaluated through a two-way ANOVA. Where trial factor was slightly statistically RESULTS significant in the two-way ANOVA test results, 10 individual A total of 271 Coastal Cutthroat Trout were captured from two-way t-tests were used for within trial comparisons of mean seven tributaries of the upper Skookumchuck River for the 11 length for successful and unsuccessful fish passage using a Bon- passage trials. The median size of all captured fish was 114 mm ferroni adjusted significance level of 0.005. (range = 85–207 mm). The age groups and the size structure Passage performance data were used to fit a logistic of the test population was representative of resident Coastal regression model predicting the probability of passage success. Cutthroat Trout populations in headwater streams of western In addition to velocity, average fish length (by trial group) and Washington. Test groups for each passage trial varied in num- water depth were added in initial model runs. However, these co- ber (range 20–29) based on the number of fish sampled from variates were not a significant determinant in the passage model. the watershed, but mean fish size did not significantly differ An additional model was developed that included a two-way across trials (one-way ANOVA: F10, 216 = 1.33, P = 0.21). 196 PETERSON ET AL.

Water temperature during the holding periods and passage trials participants ranged from 27% to 100% and followed a pattern of ranged from 9.0◦C to 13.5◦C, similar to the water temperature generally lower success rates for trials testing higher velocities within the streams where the fish were captured (range = 8.0– (Table 1). The two-way ANOVA testing for differences in mean 13.5◦C). Temperatures were not regulated during passage trials length between successful and unsuccessful fish passage across and reflected a range commonly found in many Pacific North- trials indicated that trial was slightly statistically significant west stream environments during the early summer season. One (F10 = 1.73, P = 0.08), while passage versus no passage tag shed occurred over the course of the study (during the 1.6- was highly statistically significant (F1 = 11.61, P = 0.0008). m/s trial). Trial-related mortality included one fish during the Although successful individuals were larger on average in 9 1.8-m/s trial and four fish during the 1.7-m/s trial. Causes of of the 10 trials where a comparison could be made, it was death were unknown, although impingement on the downstream only statistically significant for the 1.9-m/s trial (t = 3.55, screen was suspected. df = 12, P = 0.004). It is noteworthy that fish of all sizes, Actual slope and discharge values deviated from those tar- even the smaller individuals in the test population, successfully geted in the initial hydraulic modeling exercise due to facility- navigated the culvert over the entire range of trial velocities related constraints on the fine-scale adjustment of both parame- (Figure 2). ters. Actual water velocities were calculated using actual slope Multiple ascents by at least one fish were recorded in all but and discharge values and are reported in Table 1. Validation a single trial (2.3 m/s). Overall, 37% of the fish that successfully of modeled velocities using point measurements of the inferred passed through the culvert did so more than once, and for some maximum velocity and float tests indicated that a consistent pat- trials and individuals multiple-pass behavior was common. For tern between modeled and measured velocities existed across example, the 12 fish transiting the culvert more than once during the entire range of conditions tested. Both validation techniques the first trial (0.6 m/s) averaged 14 ascents, with two fish passing recorded consistently higher velocities than the modeled aver- 22 times each (FL 114 mm and 132 mm). Fewer multiple ascents age velocity by a similar magnitude, which was to be expected occurred as trial velocities increased, with a noticeable drop because both validation methods targeted above-average veloc- between 1.4 and 1.6 m/s velocities (average of 1.4 ascents per ity locations within the culvert. In addition, the accuracy of the fish for trials ≥1.4 m/s versus 5.2 for trials ≤1.4 m/s). The fish modeled depth profile along the longitudinal axis of the culvert making more than one ascent were larger than those making was found to be highly accurate based on point measurements a single ascent for all but two trials (0.6 and 1.6 m/s), though taken during the 1.9-m/s trial. The calculated bulk average ve- sample sizes were too small for statistical comparison. locity using these point measurements was within 0.31% of the The duration of ascent for successful passage attempts ranged modeled result. Fine-scale velocity mapping indicated cross- from 12 s to 47 min. Larger fish tended to ascend more sectional asymmetry in velocity values (range 1.1–3.5 m/s), quickly than smaller fish in all trials, although this pattern was with a reduced-velocity region (approximately 60% of maxi- highly variable. Passage times generally decreased as veloc- mum velocity) located along the left flow margin in the culvert ity increased, especially for the five test velocities at or above (looking downstream; Figure 1). 1.8 m/s (Figure 2). For the three lowest velocities (0.6–0.9 m/s), The dominant period of passage activity occurred at dusk long passage times were recorded for fish of all sizes (105– with scattered activity throughout the night. Daytime activ- 207 mm). However, long passage times for the next three trials ity generally increased with higher test velocities, especially (1.4–1.8 m/s) were recorded only for fish below 120 mm long ≥1.6 m/s. Rates of participation ranged from 62% to 100% (Figure 2). Duration of passage became relatively uniform and were generally lower for trials testing higher velocities, across all fish sizes during the four trials testing the highest especially ≥1.8 m/s. Similarly, rates of successful passage by velocities (Figure 2). Downloaded by [Department Of Fisheries] at 22:33 28 February 2013

FIGURE 1. Velocity contour proﬁle at a modeled average velocity of 2.4 m/s (Q = 0.34 m3/s; slope = 8.6%) at the center of the culvert looking downstream; the vertical axis is exaggerated to provide greater resolution of the variation in velocities. ASSESSING PASSAGE OF CUTTHROAT TROUT 197

FIGURE 4. Fitted logistic regression curve (solid line) giving average culvert FIGURE 2. Duration (circle size) and culvert side (black = right; grey = left) passage probabilities for various target velocity values and pointwise 95% lower of successfully passing fish by fork length and trial velocity. For fish exhibiting and upper confidence bands (dashed lines). multiple passes, duration was averaged across all successful passages. If fish with multiple passes used the middle or right side of the pipe during at least one ascent, the resulting data point was chosen to reflect this with a black circle. ascended the culvert midline (i.e., not detected on either the left or right antenna: 6 of 473 successful passage records). Four transit paths were identified for fish ascending the cul- The estimated regression line for the logit of passage prob- vert: left, right, both sides, and midline pathways. Low-velocity ability indicated a linear decreasing logit with increasing target trials had roughly even passage distributions between left and velocity (Figure 4). The curve represents the average culvert right sides, but a notable shift occurred to the left side of the cul- passage probability over the domain of target velocities, and the vert at velocities >1.4 m/s (Figure 3). At test velocities of 2.2 m/s confidence bands are for the expected response. Although the and above, no fish were detected successfully passing on the cul- logistic curve was fit over target velocity values ranging from 0 vert’s right side. Few fish occupied both left and right sides of to 3 m/s, the observed velocities ranged from 0.6 to 2.4 m/s. the culvert during a single successful ascent and this occurred only in velocities at or below 1.7 m/s (Figure 3). Even fewer fish DISCUSSION Observed passage performance during volitional trials provided sufficient data to describe a passability likelihood function based on average velocity for CMPs with dimensions similar to the test culvert. This probability function describes the capabil-

Downloaded by [Department Of Fisheries] at 22:33 28 February 2013 ity of Cutthroat Trout to pass through similarly sized corrugated metal culverts with no outfall drop over a range of flow conditions and is broadly applicable to resident Coastal Cutthroat Trout populations. Caution should be taken in attempting to extrapolate passage probabilities beyond the observed velocities in the study (0.6–2.4 m/s). Despite an apparent pattern between water depth and fish passage across trials grouped by slope (Table 1), depth was not a significant determinant in the passage model. While there did appear to be some dependence between velocity and depth based on the additional model run including a two-way interaction between velocity and depth, interpretation of the two-way interaction was difficult because of the narrow range of depths tested across the targeted velocities. Depth was FIGURE 3. Proportion of fish choosing the middle, the right, the right and not a target variable we intended to control but rather the product left, and only the left sides of the culvert during passage trials. of the discharge and slope combinations selected for targeted 198 PETERSON ET AL.

trial velocities. Thus a full range of depth levels was not reflected ate the complex hydraulic environments of engineered structures across the targeted velocities, and observed results may not re- and the correspondingly complex swimming responses of fish flect the true dynamics of the two variables across a full range to these settings. of velocity and depth combinations. While this study did not Observed passage performance of Coastal Cutthroat Trout handle other aspects of culvert passability (e.g., outfall drops) in our study could provide valuable information for both sim- or different culvert dimensions, these are fruitful areas for future ulation and expert system approaches to passage assessment. research that would deepen the themes addressed in this study. An assessment of our experimental conditions using the model Passage participation and success of Coastal Cutthroat Trout FishXing (Version 3.0.17; Furniss et al. 2008) indicated all test decreased as average velocity increased, especially >1.7 m/s. trials contained passable conditions. Two user-defined parame- We expected fish size to play a more significant role in the perfor- ters in particular were found to be very sensitive (leveraging pass mance of test individuals across all passage trials because fish or fail determinations) to user input in the current beta version size is positively correlated with swimming capability (Bain- of FishXing, including minimum depth and a velocity reduction bridge 1958; Webb et al. 1984). However, overall, fish length factor. We populated these fields with data derived from our was not significant in the logistic model, and successful indi- experiments and, using the guidance in FishXing, input factors viduals were statistically larger than their failed counterparts in of 3.7 cm and 0.6 m/s for the depth and velocity factor fields, only one trial when analyzed separately. Successful and failed respectively. Modest deviations from these values failed some of attempts were recorded for fish across the entire size range dur- our trials for both depth and velocity, suggesting that FishXing ing all trials, including the most challenging ones. This general is calibrated realistically if these two parameters are used. We lack of size-specific passage performance and success may be also assessed how passable the experimental conditions were an artifact of the relatively limited size range of the test fish using the expert-based Washington State protocol, which as- (85–207 mm). However, this size range is common for head- sumes culvert slopes in excess of 4% slope are unacceptable water populations of Coastal Cutthroat Trout (excluding young for passing trout >150 mm (WDFW 2009). Our slope of 8.6% of the year), making our results directly applicable to natural documented successful rates of passage ranging from 27% to populations. 71% for fish, 76% of which were smaller than 150 mm. While Fish in trials >1.8 m/s displayed a more focused swimming the Washington State protocol indicates that maximum culvert behavior, as evidenced by shorter and less variable transit time. velocities >1.8 m/s represent nearly insurmountable barriers for These findings are consistent with the weak relationship of mea- trout >150 mm (WDFW 2009), a third of the participants were sured transit times versus average velocity for adult Westslope successful in ascending the pipe where maximum velocity was Cutthroat Trout Oncorhynchus clarkii lewisi described by Solcz doubled (3.7 m/s), with five of the six fish in this trial being (2007). While we can only report the average ground speed of <150 mm long. Our results were expected to exceed Washing- the test fish as measured by detections at exit and entrance anton State fish passage design criteria because they are intended tennas, the most rapid ascents suggest swim speeds in excess to pass the weakest and smallest individuals of each species of the critical speed of 0.6 m/s for juvenile Coastal Cutthroat requiring passage (WDFW 2003), though criteria for trout are Trout reported by Hawkins and Quinn (1996). Ascending fish based on individuals >150 mm (WDFW 2009). displayed a noticeable preference in culvert side only when the The consideration of empirically based data such as those velocity exceeded 1.5 m/s. The reduced-velocity zone on the presented here could further refine the protocols most commonly culvert’s left side that has been noted for helically wound CMPs used for passage assessments. For example, consideration of a (Silberman 1970) was the preferred passage route for fish at reduced-velocity area when applying criteria could lead to a these higher velocities, as only 14% of successful ascents above better understanding of passability, leading to improved prior- Downloaded by [Department Of Fisheries] at 22:33 28 February 2013 1.5 m/s occurred on culvert right. itizations and cost savings. Such data also lay the foundation Despite the wealth of information available on fish swimming for developing new quantitative tools for describing passability capability, its effective application to determine passage perfor- through empirically derived functions. These functions could mance through engineered structures remains elusive. Swim- address difficult management questions such as partial pass- ming performance has usually been studied in a laboratory ap- ability (i.e., how passable a crossing structure is) and better paratus such as a Blazka-type swim chamber (e.g., Smith and inform managers faced with making decisions. Recent stud- Newcomb 1970), with performance characterized in metrics ies have focused on constructing passability models that are normalized for fish size and classified by a swimming behavior based on observed fish passage values. Cahoon et al. (2007) keyed to sustainability (e.g., Hawkins and Quinn 1996). Models utilized RFID to assess fine-scale movements of Yellowstone designed to assess fish passage through high-velocity zones in Cutthroat Trout Oncorhynchus clarkii bouvieri through road engineered structures are parameterized with this type of swim- culverts, while others have conducted volitional movement tri- ming performance data, though their output is often unreliable als through test culverts and flumes that accurately emulate due in part to assumptions of uniform velocities within structures velocity profiles of stream crossing structures (Powers et al. and the swimming behavior of fish when confronted with chal- 1997; Haro et al. 2004; Pearson et al. 2005). This study con- lenging water velocities (Castro-Santos 2006). A more realistic tributes to this body of empirically based research, developing approach, based on actual fish observations, is needed to evalu- a probabilistic model describing the likelihood of successful ASSESSING PASSAGE OF CUTTHROAT TROUT 199

passage of wild-caught fish through bare CMPs similarly sized Hawkins, D. K., and T. P. Quinn. 1996. Critical swimming velocity and associ- to the test culvert (no baffles or natural substrate) using average ated morphology of juvenile Coastal Cutthroat Trout (Oncorhynchus clarki velocity as the explanatory variable. This model quantifies the clarki), Steelhead Trout (Oncorhynchus mykiss), and their hybrids. Canadian Journal of Fisheries and Aquatic Sciences 53:1487–1496. occurrence of partial passage, characterizing passage in a more Hosmer, D. W., and S. Lemeshow. 1989. Applied logistic regression. Wiley, realistic way. The consideration of these data in culvert assess- New York. ments will generate conclusions that more accurately reflect Karle, K. F. 2005. Analysis of an efficient fish barrier assessment protocol the risk to passability and provide greater certainty to potential for highway culverts. Final report to Alaska Department of Transportation, population-level benefits from remedial actions. FHWA-AK-RD-05-02, Juneau. Kemp, P.S., and J. R. O’Hanley. 2010. Procedures for evaluating and prioritising ACKNOWLEDGMENTS the removal of fish passage barriers: a synthesis. Fisheries Management and Ecology 17:297–322. Use of the culvert test facility at the Washington Department Latterell, J. J., R. J. Naiman, B. R. Fransen, and P. A. Bisson. 2003. Physical of Fish and Wildlife Skookumchuck Fish Hatchery was coordi- constraints on trout (Oncorhynchus spp.) distribution in the Cascade Moun- nated by Jim Dill and Dan Adkins in cooperation with Rhonda tains: a comparison of logged and unlogged streams. Canadian Journal of Brooks and Jon Peterson of the Washington Department of Fisheries and Aquatic Sciences 60:1007–1017. Transportation. Financial and other support was provided by the Mori, N., T. Suzuki, and S. Kakuno. 2007. Noise of acoustic doppler velocimeter data in bubbly flows. Journal of Engineering Mechanics 133:122–125. National Council for Air and Stream Improvement, Oregon For- Morita, K., and S. Yamamoto. 2002. Effects of habitat fragmentation by est Industries Council, Association of Oregon Counties, Plum damming on the persistence of stream-dwelling charr populations. Conser- Creek Timber Company, Potlatch Forest Holdings, Washington vation Biology 16:1318–1323. Forest Protection Association, Rayonier, Green Crow, Stimson Park, D., M. Sullivan, E. Bayne, and G. Scrimgeour. 2008. Landscape-level LC, Forest Capital Partners, and Merrill and Ring. Doug Mar- stream fragmentation caused by hanging culverts along roads in Alberta’s boreal forest. Canadian Journal of Forest Research 38:566–575. tin, Joel Cahoon, Pat Powers, Warren Leach, Earl Prentice, Chris Pearson, W., M. Richmond, G. Johnson, S. Sargeant, R. Mueller, V. Cullinan, Jarmer, and Pete Heide provided critical support for this project. Z. Deng, B. Dibrani, G. Guensch, C. May, L. O’Rourke, K. Sobocinski, and Weyerhaeuser Company provided access to streams on their land H. Tritico. 2005. Protocols for evaluation of upstream passage of juvenile for capture of fish. Kyle Meier, Neil Slifka, Spencer Johnson, salmonids in an experimental culvert test bed. Battelle, Pacific Northwest and Adelaide Sibeaux assisted in fish collection, tagging, and Division, Report WA-RD 614.1, Richland, Washington. Powers, P. D., K. Bates, T. Burns, B. Gowen, and R. Whitney. 1997. Culvert facility operation. Four anonymous reviewers provided helpful hydraulics related to upstream juvenile salmonid passage. Washington State comments on an earlier version of this manuscript. Department of Fish and Wildlife, Lands and Restoration Services Program, Project 982740, Olympia. REFERENCES Richmond, M. C., Z. Deng, G. R. Guensch, H. Tritico, and W. H. Pearson. 2007. Bainbridge, R. 1958. The speed of swimming of fish as related to size and to Mean flow and turbulence characteristics of a full-scale spiral corrugated the frequency and amplitude of the tail beat. Journal of Experimental Biology culvert with implications for fish passage. Ecological Engineering 30:333– 35:109–133. 340. Burford, D. D., T. E. McMahon, J. E. Cahoon, and M. Blank. 2009. Assessment Rolls, R. J. 2011. The role of life-history and location of barriers to migration of trout passage through culverts in a large Montana drainage during summer in the spatial distribution and conservation of fish assemblages in a coastal low flow. North American Journal of Fisheries Management 29:739–752. river system. Biological Conservation 144:339–349. Cahoon, J. E., T. McMahon, A. Solcz, M. Blank, and O. Stein. 2007. Fish passage Silberman, E. 1970. Effect of helix angle on flow in corrugated pipes. Journal in Montana culverts: phase II – passage goals. Final Report to Montana of the Hydraulics Division 96:2253–2263. Department of Transportation, FHWA/MT-07-010/8181, Helena. Smith, L. S., and T. W. Newcomb. 1970. A modified version of the Blazka Cahoon, J. E., T. McMahon, O. Stein, D. Burford, and M. Blank. 2005. Fish respirometer and exercise chamber for large fish. Journal of the Fisheries passage at road crossings in a Montana watershed. Final Report to Montana Research Board of Canada 27:1321–1324. Department of Transportation, FHWA/MT-05-002/8160, Helena. Solcz, A. A. 2007. Assessment of culvert passage of Yellowstone Cutthroat Castro-Santos, T. 2006. Modeling the effect of varying swim speeds on fish Trout in a Yellowstone River spawning tributary using a passive integrated Downloaded by [Department Of Fisheries] at 22:33 28 February 2013 passage through velocity barriers. Transactions of the American Fisheries transponder system. Master’s thesis. Montana State University, Bozeman. Society 135:1230–1237. USGAO (U.S. General Accounting Office). 2001. Restoring fish passage CBFWA (Columbia Basin Fish and Wildlife Authority). 1999. PIT tag marking through culverts on Forest Service and BLM lands in Oregon and Washington procedures manual. CBFWA, PIT Tag Steering Committee, Portland, Oregon. could take decades. USGAO, Report to the Ranking Minority Member, Sub- Furniss, M., M. Love, S. Firor, K. Moynan, A. Llanos, J. Guntle, and R. Guber- committee on Interior and Related Agencies, Committee on Appropriations, nick. 2008. FishXing, v. 3.0. U.S. Department of Agriculture Forest Service, House of Representatives, Washington, D.C. San Dimas Technology and Development Center, San Dimas, California. WDFW (Washington Department of Fish and Wildlife). 2003. Design of road Gresswell, R. E., D. S. Bateman, G. Lienkaemper, and T. J. Guy. 2004. Geospa- culverts for fish passage. WDFW, Olympia. tial techniques for developing a sampling frame of watersheds across a re- WDFW (Washington Department of Fish and Wildlife). 2009. Fish passage gion. Pages 515–528 in T. Nishida, P. J. Kailola, and C. E. Hollingworth, barrier and surface water diversion screening assessment and prioritization editors. GIS/spatial analyses in fishery and aquatic sciences: proceedings of manual. WDFW, Olympia. the second international symposium on GIS in fishery and aquatic sciences. Webb, P. W., P. T. Kostecki, and E. D. Stevens. 1984. The effect of size and Fishery-Aquatic GIS Research Group, Saitama, Japan. swimming speed on locomotor kinematics of Rainbow Trout. Journal of Haro, A., T. Castro-Santos, J. Noreika, and M. Odeh. 2004. Swimming perfor- Experimental Biology 109:77–95. mance of upstream migrant fishes in open-channel flow: a new approach to Wofford, J. E. B., R. E. Gresswell, and M. A. Banks. 2005. Influence of barriers to predicting passage through velocity barriers. Canadian Journal of Fisheries movement on within-watershed genetic variation of Coastal Cutthroat Trout. and Aquatic Sciences 61:1590–1601. Ecological Applications 15:628–637. This article was downloaded by: [Department Of Fisheries] On: 28 February 2013, At: 22:33 Publisher: Taylor & Francis Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

North American Journal of Fisheries Management Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/ujfm20 The Effects of Total Dissolved Gas on Chum Salmon Fry Survival, Growth, Gas Bubble Disease, and Seawater Tolerance David R. Geist a , Timothy J. Linley a , Valerie Cullinan b & Zhiqun Deng c a Pacific Northwest National Laboratory, Earth Systems Science Division, Ecology Group, Post Office Box 999, Mailstop K6-85, Richland, Washington, 99352, USA b Pacific Northwest National Laboratory, Coastal Sciences Division, Marine Biotechnology Group, Marine Sciences Laboratory, 1529 West Sequim Bay Road, Sequim, Washington, 98382, USA c Pacific Northwest National Laboratory, Earth Systems Science Division, Hydrology Group, Post Office Box 999, Mailstop K9-33, Richland, Washington, 99352, USA Version of record first published: 01 Feb 2013.

To cite this article: David R. Geist , Timothy J. Linley , Valerie Cullinan & Zhiqun Deng (2013): The Effects of Total Dissolved Gas on Chum Salmon Fry Survival, Growth, Gas Bubble Disease, and Seawater Tolerance, North American Journal of Fisheries Management, 33:1, 200-215 To link to this article: http://dx.doi.org/10.1080/02755947.2012.750634

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material. North American Journal of Fisheries Management 33:200–215, 2013 American Fisheries Society 2013 ISSN: 0275-5947 print / 1548-8675 online DOI: 10.1080/02755947.2012.750634

ARTICLE

The Effects of Total Dissolved Gas on Chum Salmon Fry Survival, Growth, Gas Bubble Disease, and Seawater Tolerance

David R. Geist* and Timothy J. Linley Pacific Northwest National Laboratory, Earth Systems Science Division, Ecology Group, Post Office Box 999, Mailstop K6-85, Richland, Washington 99352, USA Valerie Cullinan Pacific Northwest National Laboratory, Coastal Sciences Division, Marine Biotechnology Group, Marine Sciences Laboratory, 1529 West Sequim Bay Road, Sequim, Washington 98382, USA Zhiqun Deng Pacific Northwest National Laboratory, Earth Systems Science Division, Hydrology Group, Post Office Box 999, Mailstop K9-33, Richland, Washington 99352, USA

Abstract Chum Salmon Oncorhynchus keta alevins developing in gravel downstream of Bonneville Dam on the Columbia River are exposed to elevated total dissolved gas (TDG) when water is spilled to move migrating salmon smolts to the ocean. We studied whether alevins that were exposed to six levels of dissolved gas ranging from 100% to 130% TDG at three development periods between hatch and emergence (hereafter, early, middle, and late stages) experienced differential mortality, growth, gas bubble disease, or seawater tolerance. Each life stage was exposed for 49 d (early stage), 28 d (middle stage), or 15 d (late stage) beginning at 13, 34, and 47 d posthatch, respectively, through emergence. Mortality for all stages was estimated to be 8% (95% confidence interval [CI] = 4–12%) when dissolved gas levels were less than 117% TDG. Mortality increased as dissolved gas levels rose above 117% TDG; the lethal concentration producing 50% mortality was 128.7% TDG (95% CI = 127.2–130.3% TDG) in the early and middle stages. There was no evidence that gas exposure affected growth. The proportion of fish with gas bubble disease increased with increasing gas concentrations, and bubbles occurred most commonly in the nares and gastrointestinal tract. Early stage fish exhibited higher ratios of filamental to lamellar gill chloride cells than late-stage fish, and these ratios increased and decreased for early and late-stage fish, respectively, as gas levels increased; however, there were no significant differences in mortality between life stages after 96 h in seawater. The study results suggest that water Downloaded by [Department Of Fisheries] at 22:33 28 February 2013 quality guidelines for dissolved gas (≤105% TDG) offer a conservative level of protection to Chum Salmon alevins incubating in gravel habitat downstream of Bonneville Dam.

Gas supersaturation generated by spill from hydroelectric salmonids Oncorhynchus spp. were documented, including ex- dams on the Columbia River was first acknowledged as an en- ophthalmia (i.e., popeye), emboli in the gill blood vessels, and vironmental concern in the 1960s, when total dissolved gas bubbles or blisters under the skin on the head, in the lining of (TDG) levels in surface waters were found to be as high as the mouth, along the lateral line, and between the fin rays (Ebel 143% of air saturation (Beiningen and Ebel 1970). Symptoms 1969, 1971; Beiningen and Ebel 1970; Raymond 1970). These of gas bubble disease in dead or moribund juvenile and adult findings prompted considerable research on the effects of gas

*Corresponding author: [email protected] Received June 1, 2012; accepted November 13, 2012

200 EFFECTS OF TOTAL DISSOLVED GAS ON CHUM SALMON FRY 201

supersaturation on salmonids, with a focus on migrating of the lower Columbia River evolutionarily significant unit of salmonid smolts (Coutant 1970; Ebel 1971; Cramer and Chum Salmon listed under the Endangered Species Act of 1973 McIntyre 1975; Dawley and Ebel 1975; Blahm et al. 1976; Ebel (NOAA 2004). Spill operations at Bonneville Dam are designed and Raymond 1976; Nebeker and Brett 1976; Ebel et al. 1979); to move migrating salmon smolts downstream to the Pacific today, we have a relatively good understanding that salmonid Ocean during a time when pre-emergent Chum Salmon alevins smolts are minimally affected by dissolved gas levels between are still present in the gravel in spawning areas downstream of 110% and 120% TDG (reviewed by NOAA 1995, 2000). Con- the dam. To provide protection for pre-emergent Chum Salmon sequently, the U.S. Environmental Protection Agency adopted alevins, managers use a guideline of limiting dissolved gas levels a nationwide standard of 110% TDG for the protection of downstream of Bonneville Dam to 105% TDG after allowing aquatic life (USEPA 1973). However, during the spring smolt for depth compensation; water depth is part of the manage- out-migration season in the Columbia River, the U.S. Environ- ment criterion because with each approximately 1-m increase mental Protection Agency grants waivers that allow up to 115% in depth, gas solubility increases by approximately 10% (Colt or 120% TDG in hydroelectric dam tailraces, depending on 1984). In other words, the dissolved gas level of the river within mixing of spillway and powerhouse discharge (NOAA 1995). the spawning areas could be as high as 115% TDG at the surface Salmonids spawning in the tailraces of hydroelectric dams and still provide protection to Chum Salmon alevins, assuming represent a special case of gas exposure wherein developing em- 1 m of water depth over the redds. During periods of low river bryos (i.e., alevins) in the gravel environment are potentially ex- flow, however, depth compensation can occur in only the deeper posed to elevated dissolved gas at a sensitive life stage (McGrath areas of the riverbed, and dissolved gas concentrations at the et al. 2006; Arntzen et al. 2009a). Unlike migrating smolts, locations of pre-emergent Chum Salmon alevins in shallower which can move within the water column to avoid or mitigate spawning habitat may be higher than 105% TDG (Arntzen et al. the effects of elevated dissolved gas (Meekin and Turner 1974; 2007, 2008, 2009b). Dawley et al. 1976; Stevens et al. 1980), alevins that are develop- Whether current management guidelines for dissolved gas ing within the gravel are not able to easily move to avoid elevated are protective of Columbia River Chum Salmon alevins is un- levels of dissolved gas. In addition, gas bubble disease mani- known because very little research on the effects of elevated dis- fests itself differently in developing alevins than in smolt-size solved gas has been conducted with this species (but see Birtwell fish, and sensitivity to supersaturation varies among salmonid et al. 2001). In one of the few studies on Chum Salmon alevins, species (reviewed by Weitkamp and Katz 1980). For example, symptoms of gas bubble disease were noted at around 119– Sockeye Salmon O. nerka alevins experienced gas bubble dis- 121% TDG, and mortality occurred at around 125% TDG (Hand ease and mortality at concentrations of 108–110% TDG (Harvey et al. 2008, 2009). However, sample sizes in that study were and Cooper 1962), whereas Sockeye Salmon smolts were able small and the results were equivocal. Chum Salmon alevins that to tolerate gas levels of 110% TDG with no mortality (Nebeker were exposed to 113% TDG emerged earlier and were smaller and Brett 1976). Rucker (1975) observed that fingerling-size than untreated fish, but differences were small and were not Coho Salmon O. kisutch were more susceptible to gas bub- consistently observed in other groups exposed to the same gas ble disease than smaller fish of approximately the same age. levels (Hand et al. 2008, 2009). Subtle differences in the size of Wood (1979) recommended that Pacific salmon (species not in- Chum Salmon alevins at emergence could be important to their dicated) in hatcheries should not be exposed to dissolved gas survival because small fish may be more prone to predation in levels greater than 104% TDG at the alevin stage but that gas ex- estuarine habitat or may be less able to transition to exogenous posure of fingerlings could go as high as 112% TDG. Krise and feeding, which could reduce growth. Dissolved gas concentra- Herman (1989) found intracranial hemorrhaging and subcuta- tions of around 108% TDG also appeared to cause epithelial Downloaded by [Department Of Fisheries] at 22:33 28 February 2013 neous bubbles in Lake Trout Salvelinus namaycush alevins after injury (i.e., cell hypertrophy and separation) to the gills of de- a 15-d exposure to 101% TDG; visible bubbles (intra-orbital, veloping Chum Salmon alevins (Hand et al. 2008); this could head, and abdomen) were observed after a 40-d exposure to be particularly problematic if it leads to impaired respiratory 105% TDG. Montgomery and Becker (1980) noted gas bubbles gas transfer because Chum Salmon often spawn in groundwater and mortality of Rainbow Trout O. mykiss alevins at 113% TDG areas where dissolved oxygen concentrations are less than fully in hatchery troughs. Collectively, these studies would suggest saturated (Arntzen et al. 2009a). It is not known whether epithe- that salmonid alevins are more susceptible to dissolved gas than lial injury to gill tissue as observed by Hand et al. (2008) would are smolt-size fish. affect osmoregulation; the TDG concentration causing epithe- The Chum Salmon O. keta is one Pacific salmon species for lial injury could directly or indirectly affect osmoregulation by which little information exists on the tolerance of alevins to gas altering the abundance and distribution of both lamellar and exposure. Columbia River adult Chum Salmon spawn in dam filamental chloride cells, which function as sites of ion uptake tailrace habitat that is affected by elevated dissolved gas caused for Chum Salmon in freshwater (Uchida et al. 1996; Shikano by spilling water at Bonneville Dam, a large hydroelectric dam and Fujio 1998). Moreover, epithelial cell injury from elevated at river kilometer 235 on the lower Columbia River. These dissolved gas exposure occurring at or near emergence may fish collectively represent one of two remaining populations cause impaired respiration, leading to compensatory responses 202 GEISTETAL.

that may indirectly affect preparatory changes for seawater en- round tanks) and were held at an initial temperature of 10◦C and try and increase the likelihood of delay in the onset of seawater a flow rate of 20 L/min until placement in the supersaturated gas tolerance. If this leads to delay in the onset of seawater tolerance. system. The objective of this study was to document the occurrence of Supersaturated gas system and treatment design.—Chum gas bubble disease in Chum Salmon alevins of varying develop- Salmon alevins were exposed to the six dissolved gas levels mental stages at dissolved gas levels ranging up to 130% TDG. by using a treatment system that consisted of two water sup- We studied whether Chum Salmon alevins exposed to increasing plies (i.e., saturated water and supersaturated water), six 50-L dissolved gas levels at three development periods between hatch vertical-flow incubators, and an integrated dissolved gas sensing and emergence suffered differential mortality, growth, gas bub- and regulating platform. A dissolved gas level of approximately ble disease, or seawater tolerance. Three developmental stages 140% TDG was generated with a gas supersaturation column were studied to determine whether susceptibility to increased (Specific Mechanical Systems Ltd., Victoria, British Columbia) dissolved gas levels was influenced by early ontogeny. Organ that consisted of a 0.25-m3 stainless-steel column with inflow structure and function differ notably between posthatch alevins lines for both compressed air and water. Water and compressed and pre-emergent fry, particularly for gas exchange (Wells and air were injected at the top of the column, while a water-level Pinder 1996a, 1996b), and such differences could potentially float switch and a gate valve at the bottom of the tank were used influence the form and degree of response to elevated dissolved to control the air-to-water volume inside the tank, the flow rate, gas. The goal of this study was to determine whether existing and the internal tank pressure. water quality criteria are protective of Chum Salmon alevins Saturated river water (∼100% TDG) was mixed with the developing in gravel habitat downstream of Bonneville Dam gassed water at each incubator to achieve the desired dissolved during the period when water is spilled to promote downstream gas treatment level. The incubators consisted of 50-L, circu- migration of salmonid smolts. lar, cone-bottom tanks with a perforated aluminum plate near the bottom to hold the egg cups and to allow for uniform flow through the outlet. Mixed water was introduced just below the METHODS surface by using solenoid valves (Irritrol Ultra Flow 700-1; Incubation of eggs and embryos.—Eyed eggs (n = 9,000) Irritrol Systems, Riverside, California) to control the flow of su- of Chum Salmon were obtained from the Minter Creek Hatch- persaturated water and saturated water. The dissolved gas levels ery (Washington Department of Fish and Wildlife, Gig Harbor) in each incubator were monitored by a sensor (Model T507; and transferred to the Pacific Northwest National Laboratory’s In-Situ, Inc., Fort Collins, Colorado) and were maintained by a (PNNL) Aquatic Research Laboratory on January 7, 2011. The barometric pressure sensor (Model CS100; Campbell Scientific, eggs had been fertilized on November 29, 2010, and incubated Inc., Logan, Utah), data logger (Model CR1000; Campbell Sci- at a mean water temperature of 9.7◦C. The accumulated ther- entific), and controller (Model SDM-CD16AC; Campbell Sci- mal units (ATU) at the time of transfer equaled 378. Prior to entific) to within a range of ±4.8 mm Hg, which represents an transport, eggs were wrapped in damp burlap and were placed SD less than or equal to 0.25% of the mean depth-compensated in wooden mesh trays inside a cooler with ice packs to main- target TDG. The dissolved gas sensors were calibrated (Fluke tain a temperature of 7–10◦C. Upon arrival at the laboratory, 719-30G; Fluke Corporation, Everett, Washington) to an accu- the eggs were disinfected in 100-µL/L iodophor for 30 min racy of ±2 mm Hg over the range of 400–1,400 mm Hg. A and were then divided into four vertical-flow incubator trays computer program (written in CRBasic and implemented via (MariSource, Fife, Washington) that were supplied with a mix LoggerNet; Campbell Scientific) operated two solenoid valves of river water and well water at a temperature of 10◦C. for each incubator. Temperature for the incubators was con- Downloaded by [Department Of Fisheries] at 22:33 28 February 2013 Four days after transfer, the eggs were distributed into 72 egg trolled by mixing ambient and chilled river water with well cups. The egg cups were constructed from polyvinyl chloride water. Over the study period, the mean water temperature ± (PVC) pipe (diameter = 76 mm; height = 50 mm), with nylon SD was 7.7 ± 0.6◦C (temperature range = 6.3–9.2◦C). screen attached (by use of silicone caulk) to the bottom of the Each of the three development stages of Chum Salmon cup and to the top of the removable lid. Each cup was assigned alevins was exposed to the six treatment levels of dissolved gas to a dissolved gas treatment level (100, 105, 115, 120, 125, or over a period of 49 d (early stage), 28 d (middle stage), or 15 130% TDG), a life stage (early, middle, or late), and a replicate d (late stage). Exposure of the early stage alevins began at 13 that either was used to determine survival from hatch to emer- d posthatch (50% hatch of all alevins occurred on January 19, gence (cups 1, 2, or 3) or was used for sampling to measure 2011), whereas exposure for the middle- and late-stage groups growth from 50% hatch to 50% emergence (cup 4). Fifty eggs was initiated at 34 and 47 d posthatch, respectively. Each life were counted and placed in each cup used for survival analysis, stage was held under its assigned treatment conditions until the whereas 90 eggs were counted and placed in the cups used for predicted date of emergence. The predicted date of emergence growth sampling. After eggs were distributed among the cups, was determined from previous research (Hand et al. 2008, the egg cups were placed into two vertical-flow incubators (50-L 2009) showing that volitional emergence of Minter Creek Chum EFFECTS OF TOTAL DISSOLVED GAS ON CHUM SALMON FRY 203

Salmon fry from egg tubes occurred at approximately 932 observers who looked for external signs of gas bubble disease ATU. Based on a mean incubation temperature ± SD of 7.7 ± (bubbles in the nares, mouth, gills, fins, yolk sac, and eyes); the 0.84◦C during the study, the predicted date for 50% emergence alevins were then necropsied to identify internal signs of gas was March 22, 2011 (62 d posthatch). At that time, the dissolved bubble formation (gastrointestinal tract and kidney). Bubbles in gas level in each test incubator was returned to 100% TDG, each of the examined areas were recorded as either mild (<50% and the remaining alevins were either euthanized or held at of surface area), severe (>50% of surface area), or not present. saturation for an additional 3 d prior to the seawater challenge. For each treatment, we calculated the binomial probability (P) Survival and growth.—During the period of exposure, the of observing at least as many of the tested fish (x out of n ex- egg cups were checked daily for mortality. Dead eggs and alevins posed fish) that showed bubbles in any of the combined 11 body were removed, recorded, and externally examined for signs of locations examined: gas bubble formation. The mortality from initial exposure to the estimated date of emergence (March 22, 2011) was totaled n n − P (X ≥ x) = pk (1 − p)n k . and expressed as a proportion (cups 1–3 only). A mild outbreak k of the fungus Saprolegnia occurred in the exposure system, k=x which affected mortality primarily in the fish from the middle and late stages. Therefore, individual fish that were noted to be The estimated “natural” probability of observing bubbles in the affected by fungus were removed and recorded as mortalities body ( p) was either the control proportion showing bubbles for separate from the mortalities due to gas bubble disease and are each test group or the minimum of all the observed control pro- not reported here. A segmented regression of two simple linear portions greater than zero. Significance was determined with a models connected by a join point (spline model) was fitted Bonferroni correction for the number of comparisons conducted to the proportional mortality of each development stage as a in each test group (i.e., P < α/5, where α = 0.05 and 5 is the function of dissolved gas. The slope of the first linear model number of treatments [n = 6] minus the control). The binomial was constrained to be zero. The spline model was probability tests the null hypothesis that the number of exposed fish with observed bubbles is not greater than would be expected by chance alone. αˆ for x < x0 y = , Five fish sampled from each treatment group on the date of αˆ + βˆ (x − x ) for x ≥ x 0 0 50% emergence were x-rayed in an effort to quantify internal gas bubble formation. Images were acquired by using an x-ray where αˆ is the intercept, x0 is the join point, and βˆ is the slope tube with a molybdenum anode (Comet MCD100H-1x) and a of the second linear model. The spline models were used to detector (RadIcon Shad-o-box 4k) placed approximately 2 m estimate the lethal concentration of dissolved gas that produced from the source. Sampled fish were mounted vertically on a 50% mortality (i.e., LC50) for each life stage. 50-cm2 Plexiglas plate by simple adhesion and were placed 2– Alevins were sampled for growth at three points during the 4 cm from the detector. The x-ray images were analyzed by course of the study: on the date of 50% hatch, immediately three individuals separately using an image processing program prior to exposure, and on the predicted date of 50% emergence. (ImageJ; National Institutes of Health) to determine the ratio Alevins were removed (n = 15 from each life stage and treat- of total bubble area to fish body area. The surface area of each ment), euthanized in tricaine methanesulfonate (MS-222), blot- fish was calculated by using ImageJ to trace the perimeter of ted dry, weighed, measured for FL, and then preserved in a 10% the fish. The surface area for each bubble was measured by solution of neutral buffered formalin (NBF). After at least 80 d tracing each observable bubble separately and then summing all Downloaded by [Department Of Fisheries] at 22:33 28 February 2013 in the NBF (Heming and Preston 1981), the preserved samples of the areas to obtain a total bubble area for that fish. Bubbles were dissected to measure the dry weight of body tissue sep- in the fish were identified as light-gray or white circular shapes arated from the yolk. The dissected body tissues were pooled in the x-ray image. The swim bladder and other internal organs into separate aluminum weigh boats (4.5-cm diameter), were were also identified in all images and were not included in driedinanovenat60◦C for 48 h, and were weighed (±1mg). the bubble measurements. Any fish with major discrepancies To determine differences in growth rates among treatments, the (measurements with an SD > 0.1) either in bubble area or total differences between wet weights and FLs within each life stage fish body area among the three observers were then remeasured at initial exposure and at emergence were regressed against dis- by all three observers. solved gas level. A similar analysis of the change in dry tissue The consistency in measuring the total bubble area for each weight between hatch and emergence was performed. fish was tested with a general linear model that included the Gas bubble disease.—Alevins from each life stage and treat- main effects of life stage, dissolved gas level, and observer and ment (n = 10) were sampled at 48 h postexposure and on the the interaction of life stage and dissolved gas level. The average predicted date of 50% emergence for gross histopathology to de- total bubble area measured by an observer was then divided by termine the incidence of gas bubble disease. Alevins euthanized the given average fish surface area to produce the bubble area : with MS-222 were examined under stereomicroscopes by four body area ratio. For each life stage, the bubble area : body area 204 GEISTETAL.

ratio was fitted with a sigmoidal dose–response model using temperature of approximately 8.5◦C in the exposure system. log10(TDG). We used F-tests and residual plots to compare Dissolved oxygen at the outlet (pump) end of the seawater trough models with separate and common parameter values. was checked daily. Seawater tolerance.—Chloride cell distribution and abun- The seawater challenge was initiated on March 25, 2011, dance on gill tissue were determined for a subsample of alevins which was 3 d after the estimated date of 50% emergence. Fifty that were sampled prior to the seawater challenge. Ten alevins fish (25 fish/cup; two replicates from each life stage and treat- from each life stage and treatment were euthanized in MS-222, ment) were transferred into horizontally perforated PVC pipe fixed for 2 h in a modified Bouin’s solution (24 mL of ethanol, (diameter = 76 mm; height = 150 mm) with a screened bot- 12 mL of formalin, 3 mL of glacial acetic acid, and 6 mL of tom and were placed randomly throughout the seawater trough. picric acid), and then stored in 90% ethanol. These samples Mortalities were recorded and removed daily. The seawater chal- were dissected to remove the second gill arch, which was de- lenge was terminated on March 29, 2011 (96 h after the fish hydrated and embedded in paraffin. Two gills representing two were transferred to seawater). Percentage mortality data from fish per treatment were embedded in each block and were ori- the two replicates were arcsine transformed to reduce the het- ented to allow sectioning parallel to the gill surface. Sections erogeneity of the within-class variances. Regression analysis of 5–8-mm thickness were cut through the whole gill, result- was performed on the transformed data to determine whether ing in four to six slides with 8–16 sections/slide. Sections were mortality in seawater increased with increasing dissolved gas arranged on the slides so that the entire thickness of the gill level. Finally, in the analysis of all data, transformations were was represented on each slide. One slide for each life stage and made (when appropriate) based on the pattern of residuals, non- treatment was stained with hematoxylin and eosin. Immuno- parametric analyses (Kruskal–Wallis test followed by Dunn’s cytochemical localization of Na +,K +-ATPase in gill chloride multiple comparison test) were conducted when an assumption cells was performed as described by Nearing et al. (2002). The could not be met, and Minitab version 16 (Minitab, Inc., State anti-Na +,K +-ATPase antibody α5, which was used to local- College, Pennsylvania) and Prism version 4 (GraphPad Soft- ize immune-positive gill chloride cells, was obtained from the ware, La Jolla, California) were used to conduct the analyses. Developmental Studies Hybridoma Bank (Department of Bi- ology, University of Iowa, Iowa City). The relative abundance RESULTS of gill chloride cells on sagittal gill sections that were stained for Na +,K +-ATPase was determined for a minimum of 8 fish Survival and Growth (range = 8–10 fish) from two life stages (early and late) and The three life stages (early, middle, and late) showed a mor- three dissolved gas treatment levels (100, 115, and 120% TDG). tality response to increasing dissolved gas levels, with mor- Counts were made for up to three gill sections from each fish. tality increasing as dissolved gas increased above 115% TDG Each section was examined to determine the number of primary 1 (Table 1). The baseline mortality between 100% and 117% (filamental) and secondary (lamellar) chloride cells that were TDG was estimated to be 8% (95% confidence interval [CI] = 4– present on and between 10 lamellae on each side of a filament. 12%). The individual spline mortality models for the early (mean The cell types were distinguished by size and location following LC50 ± 95% CI = 129.7 ± 1.9% TDG) and middle (127.9 ± Uchida et al. (1996) and Shikano and Fujio (1998). Thus, two 1.9% TDG) life stages were not significantly different (P = counts were made for each gill section. The number of primary 0.37) from one another and were thus combined and used to cells and secondary cells in each count was then expressed as estimate the LC50 for these two groups, even though the maxi- a ratio, and the mean ± SD of these ratios was determined for mum proportion mortality in the early life stage was only 0.48. each fish. A generalized linear model of the main effects and in- For the early life stage, the mean proportion mortality values for Downloaded by [Department Of Fisheries] at 22:33 28 February 2013 teraction of life stage and dissolved gas treatment level was used all but the 100% TDG treatment group increased monotonically to test the significance of the interaction effect on the chloride (Figure 1). Using the common model for the early and middle cell ratio. For each life stage, we conducted (1) a Dunnett’s test life stages, the estimated LC50 was 128.7% TDG (95% CI = comparing the mean cell ratio from dissolved gas treatment lev- 127.2–130.3% TDG). The maximum proportion mortality at the els with the mean control response and (2) a simple linear regres- highest dissolved gas treatment in the late life stage was only sion of the loge(chloride cell ratio) on the dissolved gas level. 0.35, and the mean proportion mortality did not monotonically After the gill histology sampling, a 96-h seawater challenge increase with dissolved gas level (Figure 1); therefore, an LC50 was conducted using a recirculating seawater bath, which was was not estimated. created by mixing evaporated sea salt (Instant Ocean; Spectrum Approximately 20% to nearly 50% of the fish from the early Brands, Madison, Wisconsin) with river water in a 300-L fiber- stage died in dissolved gas levels of 120–130% TDG by the end glass trough to achieve a salinity of 30‰. The final salinity of the study. However, these fish were able to tolerate 120–130% was determined by using a refractometer (Model ATC-S/Mill- TDG for several days before suffering any mortality and were E; ATAGO U.S.A., Inc., Bellevue, Washington). A pressurized able to tolerate these levels for 14–44 d before experiencing heat exchange system consisting of polyethylene tubing laid in at least 10% mortality (Figure 2). In contrast, middle- and late- a trough of 5◦C river water was used to maintain a constant stage fish that were exposed to the three highest gas levels began EFFECTS OF TOTAL DISSOLVED GAS ON CHUM SALMON FRY 205

TABLE 1. Absolute number of fish that died (n) and the mean (±SE) percentage mortality (%) through 50% emergence (March 22, 2011) for Chum Salmon fry in all subreplicates (i.e., cups 1–3). The absolute number of deaths and the percentage mortality do not include fungus-related mortality and culls. Exposure to total dissolved gas (TDG) began on February 1, 2011, for the early stage fish; February 22, 2011, for the middle-stage fish; and March 7, 2011, for the late-stage fish.

100% TDG 105% TDG 115% TDG 120% TDG 125% TDG 130% TDG Life stage n % n % n % n % n % n % Early 8 6.1 2 1.3 11 7.6 29 19.2 68 45.6 69 46.0 (5.1) (0.7) (0.4) (2.8) (1.5) (1.5) Middle 7 5.1 12 10.8 7 4.9 21 14.9 50 33.8 77 60.8 (0.9) (8.5) (2.8) (5.7) (13.0) (9.3) Late 20 17.6 9 7.9 10 7.2 16 11.2 43 30.9 37 26.0 (6.1) (4.4) (3.2) (1.6) (9.0) (6.4)

to die within the first day of exposure, and a cumulative mortality a zero intercept was fitted to the cumulative mortality for each of 10% was reached between days 7 and 22 in the middle stage life stage in the 130% TDG treatment divided by the maximum and between days 4 and 14 in the late stage (Figure 2). These mortality (i.e., 61.1% for the middle stage at 130% TDG). The results show that 90% of the pre-emergent Chum Salmon fry resulting slopes were significantly different among the early from the early stage were able to tolerate 120% TDG for about (slope ± SE = 0.017 ± 0.0002), middle (slope = 0.036 ± twice as long as fish from the middle stage and almost three times 0.0004), and late (slope = 0.036 ± 0.0006) life stages (F-test as long as fish from the late stage. A simple linear model with of common slope: F2, 92 = 967; P < 0.001) but not between Downloaded by [Department Of Fisheries] at 22:33 28 February 2013

FIGURE 1. Proportional mortality of Chum Salmon fry as a function of the level of total dissolved gas (TDG). Each point represents the mean of the three subreplicates, with the SE indicated by the vertical line. A combined ﬁtted mortality model (solid line) with 95% conﬁdence interval (dotted lines) isshownfor the early and middle life stages. [Figure available in color online.] 206 GEISTETAL. Downloaded by [Department Of Fisheries] at 22:33 28 February 2013

FIGURE 2. Cumulative mortality by exposure day for three Chum Salmon life stages exposed to six total dissolved gas (TDG) treatments. Exposure was initiated on day 1 of the treatment, which varied by life stage (day 1 was February 1, 2011, for the early stage; February 22, 2011, for the middle stage; and March 7, 2011, for the late stage), and continued through emergence (March 22, 2011, for all three stages). For reference, the horizontal dashed line in each panel represents 10% cumulative mortality. EFFECTS OF TOTAL DISSOLVED GAS ON CHUM SALMON FRY 207

function of dissolved gas concentration, with values decreasing from approximately 69–83 mg at hatch to 24–54 mg at emergence (Table 2); the slope of the regression line (early stage: slope ± SE = 0.35 ± 0.32; middle stage: slope = 0.15 ± 0.16; late stage: slope =−0.20 ± 0.13) was not statistically different from zero (P > 0.21 for all three life stages).

Gas Bubble Disease Significantly more bubbles were found in the bodies of fish exposed to increased dissolved gas than in control fish (binomial test with a Bonferroni correction: P < 0.01), and the proportion of fish affected with bubbles increased with dissolved gas concentration (Figure 4). The bubble area : body area ratio appeared to increase noticeably when concentrations were at or above 120% TDG (Figure 5). Fish were found to have half the maximum ratio of bubble area to body area at 118% TDG (95% CI = 116–120% TDG) in the middle and late stages and at 120% TDG (95% CI = 119–121%) in the early stage. Gas bubbles were observed in the nares and gastrointestinal tracts of fish immediately after exposure was initiated and also after long- term exposure. Within 48 h of exposure to dissolved gas levels of 115% TDG or higher, 20–40% of the fish examined were found to contain bubbles in the nares; similar results were observed at emergence. Bubbles in the gastrointestinal tract were not obvious in the early stage fish at 48 h postexposure, but the other life stages showed a significant dose–response between gas levels and bubbles in the gastrointestinal tract both at 48 h FIGURE 3. (A) Change in fork length (delta length) and (B) change in wet postexposure and at emergence. More than half of the fish that weight (delta weight) of Chum Salmon alevins sampled immediately prior to were exposed to dissolved gas levels of 120% TDG or greater dissolved gas exposure and at 50% emergence. Three life stages (early stage = were observed to have gas bubbles in their gastrointestinal tracts. diamonds; middle stage = squares; late stage = triangles) were exposed to six In addition to gas bubbles in the nares and gastrointestinal tract, levels of total dissolved gas (100, 105, 115, 120, 125, and 130%). Each point 50–70% of the early stage fish exposed to gas levels of 115% represents a mean of the samples (in most cases, n = 15); the solid line represents the linear regression, and the dotted lines represent the 95% confidence interval TDG or higher were found to have gas bubbles in their yolk around the regression. [Figure available in color online.] sacs at 48 h postexposure; none of the fish that were exposed to 100% or 105% TDG were found with gas bubbles in their yolks. the middle and late stages (F-test of common slope: F1, 43 = By the time these early stage fish reached emergence, however, 0.0778; P = 0.78). These results suggest that the middle and the bubbles in the yolk sacs had disappeared, and bubbles were late life stages succumb to gas bubble disease at a faster rate not present in the yolk sacs of fish from the other life stages. than the early life stage. Only a few fish had bubbles in the gill filaments and eyes, and Downloaded by [Department Of Fisheries] at 22:33 28 February 2013 The growth rate based on the change in FL between initia- no fish were found to have bubbles in the kidneys. tion of exposure and emergence (delta length, Figure 3A) did not change significantly at any dissolved gas level among the Seawater Tolerance life stages (early stage: mean slope ± SE =−0.05 ± 0.02, P = The interaction term between life stage and dissolved gas 0.102; middle stage: slope = 0.01 ± 0.02, P = 0.748; late stage: treatment level in the generalized linear model was significant slope = 0.06 ± 0.03, P = 0.111). Mean FL ranged from ap- (P = 0.005), so the analysis was conducted by life stage to as- proximately 28 to 37 mm at the initiation of exposure and from sess treatment effects on chloride cell ratios. In the early stage 35 to 38 mm at emergence (Table 2 shows values at emergence). fish, the ratio of primary to secondary chloride cells increased Mean wet weight ranged from around 240 to 380 mg at initia- significantly with increasing dissolved gas level from 100% to tion of exposure and from around 260 to 350 mg at emergence 120% TDG (Figure 6; regression of log10 transformed ratios: er- (Table 2), and there was no significant change in wet weight ror df = 24, P = 0.006). Further, the median ratio of primary to (delta weight, Figure 3B) during exposure at any gas level for secondary chloride cells in fish that were exposed to 120% TDG the early stage (slope ± SE =−0.0002 ± 0.0003; P = 0.581), was significantly greater than the median ratio in fish that were middle stage (slope =−0.0003 ± 0.0008; P = 0.747), or exposed to 100% TDG (Kruskal–Wallis test: P = 0.031; Dunn’s late stage (slope = 0.0010 ± 0.0013; P = 0.457). The change multiple comparison test: P < 0.05). Within the late-stage fish, in dry tissue weight from hatch to emergence was also not a there was a significant decrease in the ratio of primary to 208 GEISTETAL.

TABLE 2. Mean (±SD) wet weight (mg) and fork length (FL, mm) of Chum Salmon fry sampled at 50% emergence (March 22, 2011) after exposure to six levels of total dissolved gas (TDG). Also shown are mean dry weights for Chum Salmon fry at emergence (total weight divided by the sample size of 15 ﬁsh); variances could not be calculated because the dry weights of individual ﬁsh were not determined.

Wet weight (mg) FL (mm) Life stage TDG (%) na Mean SD Mean SD Mean dry weight (mg) Early 100 15 308.3 52.9 36.7 1.7 43.1 105 15 281.1 38.2 35.7 1.8 40.4 115 15 281.7 45.1 35.8 1.7 41.6 120 15 291.7 37.9 35.9 1.7 38.5 125 15 281.1 27.5 36.1 0.9 39.1 130 10 259.2 38.2 34.7 1.6 24.4 Middle 100 15 312.3 40.1 37.1 1.3 46.5 105 15 310.1 53.4 37.0 1.3 44.4 115 15 258.9 36.6 34.5 1.7 36.5 120 15 306.1 37.5 37.0 1.5 44.9 125 15 300.4 40.7 36.8 1.2 44.8 130 15 273.6 34.4 36.1 1.3 40.4 Late 100 14 272.5 39.3 35.4 1.2 34.8 105 15 317.3 45.9 36.9 1.2 44.4 115 15 354.4 44.4 38.2 1.5 53.6 120 15 299.3 40.8 36.7 1.2 42.9 125 15 313.7 35.0 37.4 1.5 47.3 130 15 292.8 35.9 37.1 1.3 43.2

aSample size (n) is listed for wet weight and FL; sample size for dry weights was 15 for all treatments and life stages.

secondary cells in the gills with increasing exposure to dis- about 92% survival. However, survival rates for all develop- solved gas levels (regression of log10 transformed ratios: error mental stages decreased significantly when the gas concentra- df = 26, P = 0.043; Figure 6). The Kruskal–Wallis compari- tion was about 117% TDG or above. In previous studies with son of the median responses between treatment levels, however, this same stock of Chum Salmon, we showed that dissolved gas was only nearly significant (P = 0.067). Mortality after the 96- levels up to 113% TDG resulted in very little mortality, whereas h exposure to 30‰ seawater for each life stage and gas level mortality increased once gas levels exceeded 120% TDG (Hand ranged from 0% to 18% (Table 3), and there was no relation- et al. 2008, 2009). Our findings are consistent with observations ship between dissolved gas level and mortality at any life stage from studies of smolt-size juveniles of other salmonid species (P-values ranged from 0.213 to 0.828). (e.g., Chinook Salmon O. tshawytscha, Steelhead [anadromous Rainbow Trout], Coho Salmon, and Sockeye Salmon), which DISCUSSION have shown that mortality increases rapidly once dissolved gas Chum Salmon fry that were exposed to gas concentrations levels rise above 115% TDG (Dawley and Ebel 1975; Nebeker

Downloaded by [Department Of Fisheries] at 22:33 28 February 2013 between 100% and 115% TDG at various posthatch develop- and Brett 1976; Nebeker et al. 1979). Although Chum Salmon mental stages survived equally well to emergence, averaging alevins were as tolerant of elevated dissolved gas levels as

TABLE 3. Absolute number of ﬁsh that died (n) and the mean (±SE) percentage mortality (%) after a 96-h exposure to 30‰ seawater for two subreplicates (i.e., cups 1–2) of three life stages of Chum Salmon fry that had been exposed to six total dissolved gas (TDG) treatment levels through emergence. Initial sample size was 25 fry/cup, for a total of 50 fry per dissolved gas level.

100% TDG 105% TDG 115% TDG 120% TDG 125% TDG 130% TDG Life stage n % n % n % n % n % n % Early4836489183636 (0) (2) (4) (2) (6) (2) Middle003612362436 (0) (6) (2) (2) (0) (6) Late362400120048 (2) (0) (0) (2) (0) (4) EFFECTS OF TOTAL DISSOLVED GAS ON CHUM SALMON FRY 209

FIGURE 4. Four x-ray images collected from early stage Chum Salmon at the time of emergence. The images are arranged from left to right in order of increasing <

Downloaded by [Department Of Fisheries] at 22:33 28 February 2013 total dissolved gas (TDG) exposure level: (A) control fish (exposed to 100% TDG), (B) fish exposed to 115% TDG, (C) fish exposed to 120% TDG, and (D) fish exposed to 130% TDG. The large, elongated, light-colored area in the anterior half of the fish is the swim bladder. Gas bubbles in the gastrointestinal tract, gills, and nares appear as smaller, light-colored circles.

smolt-size fish of similar species, we did find that the earliest relate to the physiological development of the alevin respiratory developmental stage of fish in our study was more tolerant to system. As the gills of posthatch alevins develop, the young fish elevated dissolved gas levels than the middle or late stage; 90% rely on passive diffusion of oxygen through the skin, yolk sac, of pre-emergent Chum Salmon fry from the early stage were and other membranous tissue for respiration (Wells and Pinder able to tolerate 120% TDG twice as long as the middle stage 1996a, 1996b). This diffusive respiratory process continues until and three times as long as the late stage. This finding was con- the gills are more developed (i.e., occupy a larger surface area sistent with results of other studies on salmonid alevins, which relative to the skin and yolk sac) and able to carry blood and have shown that mortality increased as the fish developed toward respiratory gases to and from body tissue (Rombough 1999). emergence (Shirahata 1966; Meekin and Turner 1974; Rucker Once the gills become the primary organ for respiration, the fish 1975; Nebeker et al. 1978; Jensen et al. 1986). may be more susceptible to bubbles from supersaturated gas, One explanation for why the tolerance to elevated gas is which block small capillaries of the gill tissue, thus leading to greater for posthatch fry than for fry nearing emergence may death by asphyxiation. Counihan et al. (1998) found that White 210 GEISTETAL.

0.12 Japanese Medaka Oryzias latipes fry immediately after hatching were more resistant to elevated dissolved gas levels than older 0.10 fry. Collectively, these results suggest that the time it takes for ﬁsh to convert from a passive diffusion process to gill respiration 0.08 will depend on the species and environmental conditions. Whether the impermeability of skin to water and ions protects 0.06 young ﬁsh from dissolved nitrogen is unknown, but the skin of young Sea Trout (anadromous Brown Trout Salmo trutta) and 0.04 Atlantic Salmon Salmo salar fry immediately after hatch was suspected to be highly impermeable to water and ions, allowing Bubbles-to-body area ratio area Bubbles-to-body 0.02 the fry to survive in seawater concentrations that killed older trout that were closer to emergence (Talbot et al. 1982). In addi- 0.00 tion, the yolk sacs of young alevins could act as a nitrogen sink

100 105 115 120 125 130 and protect the fish from bubble formation (Hand et al. 2009). TDG (%) Although we do not know whether this pertains to fish, body fat influences gas bubble formation in humans who have undergone FIGURE 5. Ratio of bubble area to fish body area for five Chum Salmon rapid decompression; apparently, the lipids in the fat solubilize from each life stage (data pooled for early, middle, and late stages) exposed to nitrogen faster than other tissue (Philp and Gowdey 1964). This six levels of total dissolved gas (TDG) and then sampled at emergence (i.e., concept is highly speculative and should be confirmed in fish. termination of exposure). The horizontal line through the middle of each box represents the median; the lower and upper boundaries of the box represent the Dissolved gas levels and exposure time did not result in sig- 25th and 75th percentiles, respectively; the whiskers represent the 10th and 90th nificantly lower growth rates between the initiation of exposure percentiles; and asterisks represent the outliers. and emergence for the treated and untreated fish. In fact, the range of FLs at emergence for our treated and untreated fry (35– 38 mm) was similar to that reported for nine British Columbia Sturgeon Acipenser transmontanus larvae became susceptible to stocks (33–36 mm) held at similar temperatures, although the gas bubble disease soon after they began gill respiration; those British Columbia stocks were heavier than our fish (363–430 mg authors proposed that the opening of the buccal cavity might versus 260–350 mg; Beacham and Murray 1987). The devel- provide sites where gas bubbles could form, followed by rapid opment rate of Chum Salmon fry from British Columbia was diffusion into the branchial system. Egusa (1959) suggested that dependent mostly on incubation temperature, which varied ge- the elasticity and tenacity of the body wall tissues explained why ographically from north to south, but the development rate was also under genetic control through variations in female egg size (Beacham and Murray 1987). The finding that exposure to dis- 2.4 solved gas did not affect growth rates in our study is similar to 2.2 Early the results of some other studies. For example, surviving Steel- Late 2.0 head showed no difference in growth rates after exposure to dissolved gas levels up to 126% total gas pressure (TGP) for 1.8 90 d (Nebeker et al. 1978). There were also no differences in 1.6 the growth rates of juvenile Lake Trout (TL = 2.9–3.9 cm) that 1.4 were exposed for 30 d to five gas levels ranging from 101.3% Downloaded by [Department Of Fisheries] at 22:33 28 February 2013 to 122.0% TDG (Krise and Smith 1991). However, Dawley and 1.2 Ebel (1975) showed that long-term (35-d) exposure of juvenile 1.0 Chinook Salmon and Steelhead to sublethal levels of dissolved 0.8 nitrogen and argon (105% to 115% of saturation) resulted in Mean cell ratio primary/secondary primary/secondary ratio cell Mean mean weights and lengths that were significantly lower than 0.6 those of control fish; those authors noted that after 30 d of test- 0.4 100 115 120 ing at 115% saturation, the Chinook Salmon fingerlings became lethargic in their feeding response. TDG (%) Gas bubble disease was more prevalent in treated versus un- FIGURE 6. Distribution of the mean ratio of primary to secondary chloride treated fish, and there was a general increase in gas bubble cells in gill sections from Chum Salmon fry (early and late stages) exposed to disease as the dissolved gas level increased. Fish were found three levels of total dissolved gas (TDG) and then sampled at emergence. The to have half the maximum ratio of bubble area to body area at horizontal line through the middle of each box represents the median; the lower about 118% TDG (95% CI = 116–120% TDG) in the middle and upper boundaries of the box are the 25th and 75th percentiles, respectively; = the whiskers represent the 10th and 90th percentiles (for sample sizes > 8); and and late stages and at about 120% TDG (95% CI 119–121% asterisks represent the outliers. TDG) in the early stage. The gastrointestinal tract and the nares EFFECTS OF TOTAL DISSOLVED GAS ON CHUM SALMON FRY 211

were the most significant locations on the body where gas bubbles formed. Moreover, gas bubble disease also appeared to be developmentally dependent. In early stage fish, bubbles were also found in the yolk sac after the fish had been in the exposure system for 48 h, whereas 48-h postexposure examinations of yolk sacs from middle- and late-stage fish did not reveal significant numbers of bubbles. In extreme cases, gas bubbles have been observed in the gut, mouth, exterior surface, or yolk sac (Rucker and Kangas 1974; Wood 1979), with variable effects on behavior, physiology, and survival. Birtwell et al. (2001) and Harvey and Cooper (1962) observed that the effects of gas bubble trauma included impaired swimming performance and sensory capabilities; affected individuals floated and swam head up or abdomen up. Immediate causes of death from bubble formation in the tissues of these fish might include (1) bubbles in the buccal cavity, leading to suffocation (Fidler 1988); or (2) rupture of the perivitelline membrane (Stroud et al. 1975). Gas bubble disease, as indicated by exophthalmia and by bub- FIGURE 7. Primary chloride cells (positioned on the gill filament) and secondary chloride cells (positioned on the gill lamellae) on a sectioned gill arch bles in the mouth, nares, and the rim of the eye, was present in from a Chum Salmon fry that was exposed to 115% total dissolved gas dur- young Lake Trout at 117% and 122% TGP (Krise and Smith ing the early life stage and was sampled immediately after emergence (40× 1991). Lake Trout sac fry were tolerant of dissolved gas levels magnification). [Figure available in color online.] up to 120.2% TGP; the death rate was not affected through a 40-d study (Krise and Herman 1989). However, signs of gas late-stage fish and that the early stage fish were potentially more bubble disease, including a distended abdomen or a hyperin- tolerant to seawater as gas levels increased; an example of a sec- flated swim bladder, were evident during the entire Krise and tioned gill arch from an early stage fish exposed to 115% TDG, Herman (1989) study; at the end of that study, many fish were which shows a higher number of chloride cells on the primary moribund due to the high gas levels but were still alive. Histo- filament versus the secondary lamellae, is depicted in Figure 7. logical lesions “were inconsistent within and between groups, This trend, however, was not reflected in survival after 96 h showing no particular patterns other than a general increase in in seawater, as we found no significant differences among life lesions with increasing time of exposure and gas level. Intracra- stages or treatment groups. The reason for this may relate to the nial hemorrhages were the only lesions that were potentially life fact that as Chum Salmon adapt to seawater, much of the in- threatening” (Krise and Herman 1989: p. 271). We have previ- crease in Na +,K +-ATPase activity is associated primarily with ously noted the inconsistent histological lesions relative to gas theincreaseinsizeofNa+,K +-ATPase immunopositive chloride levels in Chum Salmon fry (Hand et al. 2008, 2009). cells on the gill filament (Uchida et al. 1996) or the specific chlo- We examined seawater tolerance by using the abundance and ride cell isoform (Bystriansky et al. 2006); however, neither the distribution of chloride cells on gill tissue because a change in size of Na +,K +-ATPase immunopositive chloride cells nor the chloride cell ratios from gill filaments and lamellae represents chloride cell isoform was measured in this study. Early and late- the combined change that is known to occur as anadromous fish stage exposed fry may well have had similar Na +,K +-ATPase transition from freshwater to seawater (Foskett and Scheffey´ activity, which could have obviated any observed differences Downloaded by [Department Of Fisheries] at 22:33 28 February 2013 1982; Richman et al. 1987; McCormick 1995). Within chlo- in chloride cell ratios and presumptive importance for seawater ride cells, the enzyme Na +,K +-ATPase is the major molecule adaptation. Moreover, the salinity challenge of 30‰ may have that is responsible for this function; in general, chloride cells been too low to detect a difference in survival between the two in freshwater-adapted fish are found on both gill filaments and groups. Shikano and Fujio (1998) reported that an increase in the lamellae, whereas in seawater-adapted fish the lamellar chlo- number of immunopositive chloride cells on the gill filaments ride cells are either absent or greatly reduced in number (Pisam and a reduction in the number of those cells on the lamellae et al. 1987; Uchida et al. 1996; Shikano and Fujio 1998; Evans were both significantly correlated with seawater survival but at et al. 2005). More recently, Bystriansky et al. (2006) showed salinities ranging from 34‰ to 46‰. that gill Na +,K +-ATPase in salmonids exists as two specific These contrasting results are not entirely unexpected because isoforms (α1a and α1b) that are reciprocally expressed in fresh- earlier studies of the effects of increased dissolved gas levels on water and seawater; these isoforms are respectively localized to the survival of salmonids transferred to seawater produced sim- the filamental and lamellar chloride cells of fish in freshwater ilarly equivocal results. Dawley et al. (1976) reported that fall and are primarily found in the filamental chloride cells of fish Chinook Salmon that were transferred to 25‰ seawater after in seawater (McCormick et al. 2009). Our findings suggest that 127 d of exposure to dissolved gas levels up to 120% TDG suf- early stage fish were potentially more tolerant to seawater than fered a total mortality of 98%, whereas Steelhead experienced 212 GEISTETAL.

mortality of up to 16%. In both cases, the surviving fish were salmon develop from yolk sac fry to alevins, they tend to become significantly larger than those that died. Dawley et al. (1976) more susceptible to mortality from asphyxiation due to gas concluded that much, if not most, of the observed mortality bubbles that form in developing gill filaments; yolk sac fry was more likely related to the fish not achieving the thresh- at earlier stages of development may be able to tolerate higher old size for seawater adaptation rather than to any effect from dissolved gas levels because their respiration is diffusive, pri- dissolved gas level. Despite the apparent size effect on seawa- marily through the skin. We do not know exactly when this ter survival, these results are surprising given that the brackish susceptibility to asphyxiation begins to increase, but some gen- water salinity (25‰) and time of transfer for Chinook Salmon eral thresholds can be seen within the cumulative mortality of and Steelhead (July and June, respectively) are near optimal for the early stage fish (Figure 2, top panel). During the period be- the direct transfer of these two species into seawater, particu- tween hatch and 25 d posthatch (exposure day 12), early stage larly for the Steelhead, which were about 180 mm in length. fish were generally tolerant of all gas levels (i.e., cumulative However, water temperature during the tests was substantially mortality was less than 10%). Between 25 and 53 d posthatch lower than normal conditions in the Columbia River during mi- (exposure days 12–40), the relative mortality varied, as gas con- gration of these fish groups and may have inhibited the typical centrations of 125% and 130% TDG were increasingly lethal physiological smoltification processes prior to seawater transi- and mortality at dissolved gas levels of 120% TDG and below tion (E. Dawley, National Oceanic and Atmospheric Adminis- was relatively low and stable. Between 53 and 61 d posthatch tration [retired], personal communication). By contrast, Bouck (exposure days 40–50), when emergence was estimated to oc- et al. (1976) observed no mortality in spring Chinook Salmon, cur, mortality rates increased substantially in the 120% TDG Sockeye Salmon, and Steelhead smolts that were held in 30‰ group but stayed low in the groups that were exposed to levels seawater for 5 d after a 48-h exposure to 120% TDG. The results less than 120% TDG. These observations suggest that in the of Bouck et al. (1976) are similar to ours (mortality = 0–18%) early stage fish, there were at least two distinct mortality thresh- and reflect the fact that exposure to high dissolved gas levels olds that were a function of exposure time and dissolved gas does not seem to impair osmoregulation in salmonids when concentration: one threshold occurred at about 25 d posthatch they are at the appropriate size and time for seawater entry. in all exposure groups (exposure day 12; February 12, 2011), In general, there were few problems during our study. Dis- and another threshold occurred at about 53 d posthatch in the solved gas levels remained constant throughout the study. We 120% TDG group (exposure day 40; March 12, 2011). believe that the system we used for dissolved gas exposure of Whether the thresholds in this laboratory study are trans- fry in this study was more representative than the system we ferable to the field is uncertain. Chum Salmon emergence in used in previous years (e.g., Hand et al. 2008, 2009). As such, spawning areas of the Columbia River downstream of Bon- we believe that the mortality rates in the present study are more neville Dam is spatially and temporally variable and is a func- representative than our previous results. The exception was that tion of hyporheic water temperature (Murray et al. 2011). In a in mid-March, we documented an outbreak of bacterial gill dis- case study conducted from 2009 to 2010, Murray et al. (2011) ease in Chum Salmon fry held as reserves in incubators that showed that emergence could begin as early as February 10 were kept separate from the exposure system. Prior to the diag- and could extend all the way to April 9 (50% emergence was nosis of bacterial gill disease, some of these reserve fish were estimated to be March 25). Thus, there is more than likely a used in the middle and late stages in our study. Although we did range of developmental stages of Chum Salmon fry present in not test for bacterial gill disease in the fish within the exposure the spawning areas at any given time, and there are likely few system, there was an increase in mortalities among the early windows of time during which gas levels above 117% TDG will and late-stage control fish in the exposure system during mid- not result in mortality for at least a portion of the alevins present. Downloaded by [Department Of Fisheries] at 22:33 28 February 2013 March, and we cannot entirely attribute this mortality increase Critically, however, for alevins and fry that survive exposure to to gas bubble disease. It is possible that some of these fish died dissolved gas above this level, the latent effects on seawater from undetected Saprolegnia infection, undetected bacterial gill tolerance and estuarine survival are likely to be negligible, irre- disease, or both; these unexplained mortalities were included in spective of the stage of development at which exposure occurs. the mortalities reported here. Early and late-stage fish that were exposed to dissolved gas up to 120% TDG survived equally well upon direct entry into 30‰ Management Implications seawater, despite the variability in abundance and distribution of The results from this study support the existing management gill chloride cells, which suggested potential differences in sea- guideline that limits depth-compensated dissolved gas levels to water tolerance between the groups. After their emergence from 105% TDG at Chum Salmon spawning locations. This guide- the gravel, Chum Salmon fry are highly dependent on estuaries line appears to provide a conservative level of protection to to forage and grow, and environmental effects that delay entry Chum Salmon sac fry incubating in the gravel downstream of or extend estuarine residence time have the potential to reduce Bonneville Dam. Our results would suggest that exposure to subsequent survival. Our results indicate that the current depth- elevated dissolved gas levels early in the development period compensated guideline of 105% TDG for Chum Salmon spawn- would be better than exposure occurring later. In general, as ing areas provides sufficient protection to limit such losses. EFFECTS OF TOTAL DISSOLVED GAS ON CHUM SALMON FRY 213

Monitoring of dissolved gas in the Chum Salmon spawn- Arntzen, E. V., K. J. Murray, J. L. Panther, D. R. Geist, and E. M. Dawley. 2008. ing areas downstream of Bonneville Dam revealed that depth- Assessment of total dissolved gas within Chum Salmon spawning areas in compensated dissolved gas levels exceeded the 105% TDG man- the Columbia River downstream of Bonneville Dam. Pages 1.1–1.38 in E. V. Arntzen, K. D. Hand, D. R. Geist, K. J. Murray, J. L. Panther, V. I. Cullinan, agement guideline when water levels were low (up to 7% of the E. M. Dawley, and R. A. Elston, editors. Effects of total dissolved gas on incubation period in 2007 and up to 4% of the incubation pe- Chum Salmon fry incubating in the lower Columbia River. Final Report to riod in 2008; Arntzen et al. 2007, 2008, 2009b). The relative U.S. Army Corps of Engineers, PNNL-17132, Pacific Northwest National risk to the Chum Salmon population in these spawning areas Laboratory, Richland, Washington. can be determined through knowledge of Chum Salmon redd Arntzen, E. V., J. L. Panther, D. R. Geist, and E. M. Dawley. 2007. Total dissolved gas monitoring in Chum Salmon spawning gravels below Bonneville elevations, the hyporheic dissolved gas levels, and the river sur- Dam. Final Report to U.S. Army Corps of Engineers, PNNL-16200, Pacific face elevation. The elevation of Chum Salmon redds within the Northwest National Laboratory, Richland, Washington. tailwater spawning area can be variable, and shallow redds are Beacham, T. D., and C. B. Murray. 1987. Adaptive variation in body size, not always protected from elevated dissolved gas because water age, morphology, egg size, and developmental biology of Chum Salmon depths cannot provide adequate compensation. Further, due to (Oncorhynchus keta) in British Columbia. Canadian Journal of Fisheries and Aquatic Sciences 44:244–261. groundwater interactions within the hyporheic zone, the surface Beiningen, K. T., and W. J. Ebel. 1970. Effect of John Day Dam on dissolved ni- water concentrations of dissolved gas are not representative of trogen concentrations and salmon in the Columbia River, 1968. Transactions the dissolved gas level experienced by the alevins incubating of the American Fisheries Society 99:664–671. in the gravel. These findings suggest that Chum Salmon redd Birtwell, I. K., J. S. Korstrom, M. Komatsu, B. J. Fink, L. I. Richmond, and location, surface water depth within the spawning areas, and R. P. Fink. 2001. The susceptibility of juvenile Chum Salmon (Oncorhynchus keta) to predation following sublethal exposure to elevated temperature and dissolved gas at the depth of an egg pocket should be mon- dissolved gas supersaturation in seawater. Canadian Technical Report of itored during the period before Chum Salmon emergence in Fisheries and Aquatic Sciences 2343. order to more fully understand the impacts of dissolved gas on Blahm, T. H., B. McConnell, and G. R. Snyder. 1976. Gas supersaturation re- Chum Salmon alevins incubating in the gravel downstream of search National Marine Fisheries Service Prescott Facility—1971 to 1974. Bonneville Dam. Pages 11–19 in D. H. Fickeisen and M. J. Schneider, editors. Gas bubble disease: proceedings of a workshop held at Richland, Washington (CONF- 741033). Technical Information Center, Office of Public Affairs, Energy Re- ACKNOWLEDGMENTS search and Development Administration, Oak Ridge, Tennessee. This study was funded by the U.S. Army Corps of Engi- Bouck, G. R., A. V. Nebecker, and D. G. Stevens. 1976. Mortality, saltwater neers, Portland District; Dennis Schwartz oversaw the contract adaptation and reproduction of fish during gas supersaturation. U.S. Environ- mental Protection Agency, Environmental Research Laboratory, EPA-600/3- and provided financial support to all aspects of the study. Earl 76-050, Duluth, Minnesota. Dawley (National Oceanic and Atmospheric Administration, Bystriansky, J. S., J. G. Richards, P. M. Schulte, and J. S. Ballantyne. 2006. retired) provided assistance in setting up the study and review- Reciprocal expression of gill Na + /K +-ATPase α-subunit isoforms α1a and ing written products. Marlies Betka (self employed, Portland, α1b during seawater acclimation of three salmonid fishes that vary in their Maine) completed the gill preparations and staining that were salinity tolerance. Journal of Experimental Biology 209:1848–1858. Colt, J. 1984. Computation of dissolved gas concentrations in water as functions used in the gill chloride histology assessment. Jill Janak, Andy of temperature, salinity, and pressure. American Fisheries Society, Special LeBarge, Kathleen Carter, Kris Hand, Erin Miller, and Tylor Publication 14, Bethesda, Maryland. Abel (PNNL) all helped in collecting the data used in this study. Counihan, T. D., A. I. Miller, M. G. Mesa, and M. J. Parsley. 1998. The effects Andrea Currie (PNNL) greatly improved the document through of dissolved gas supersaturation on White Sturgeon larvae. Transactions of her editorial oversight. Finally, two anonymous reviewers and the American Fisheries Society 127:316–322. Coutant, C. C. 1970. Exploratory studies of the interactions of gas supersatu- the associate editor provided comments that substantially im- ration and temperature on mortality of juvenile salmonids. Battelle Pacific proved the final product. Pacific Northwest National Laboratory Northwest Laboratories, Report BNWL-SA-3585, Richland, Washington. Downloaded by [Department Of Fisheries] at 22:33 28 February 2013 is operated by Battelle for the U.S. Department of Energy under Cramer, S. P., and J. D. McIntyre. 1975. Heritable resistance to gas bubble Contract DE-AC05-76RL01830. disease in fall Chinook Salmon, Oncorhynchus tshawytscha. U.S. National Marine Fisheries Service Fishery Bulletin 73:934–938. Dawley, E. M., and W. J. Ebel. 1975. Effects of various concentrations REFERENCES of dissolved atmospheric gas on juvenile Chinook Salmon and Steel- Arntzen, E. V., D. R. Geist, K. J. Murray, J. Vavrinec, E. M. Dawley, and head Trout. U.S. National Marine Fisheries Service Fishery Bulletin 73: D. E. Schwartz. 2009a. Influence of the hyporheic zone on supersaturated gas 787–796. exposure to incubating Chum Salmon. North American Journal of Fisheries Dawley, E. M., M. Schiewe, and B. Monk. 1976. Effects of long-term expo- Management 29:1714–1727. sure to supersaturation of dissolved atmospheric gases on juvenile Chinook Arntzen, E. V., K. J. Murray, D. R. Geist, E. M. Dawley, and J. Vavrinec III. Salmon and Steelhead Trout in deep and shallow test tanks. Pages 1–10 in 2009b. Assessment of total dissolved gas within Chum Salmon spawning D. H. Fickeisen and M. J. Schneider, editors. Gas bubble disease: proceedings areas in the Columbia River downstream of Bonneville Dam. Pages 1.1–1.27 of a workshop held at Richland, Washington (CONF-741033). Technical In- in E. V. Arntzen, K. D. Hand, K. M. Carter, D. R. Geist, K. J. Murray, formation Center, Office of Public Affairs, Energy Research and Development E. M. Dawley, V. I. Cullinan, R. A. Elston, and J. Vavrinec III, editors. Administration, Oak Ridge, Tennessee. Total dissolved gas effects on incubating Chum Salmon below Bonneville Ebel, W. J. 1969. Supersaturation of nitrogen in the Columbia River and its Dam. Final Report to U.S. Army Corps of Engineers, PNNL-18081, Pacific effect on salmon and Steelhead Trout. U.S. National Marine Fisheries Service Northwest National Laboratory, Richland, Washington. Fishery Bulletin 68:1–11. 214 GEISTETAL.

Ebel, W. J. 1971. Dissolved nitrogen concentrations in the Columbia and Snake Engineers, PNNL-15525, Pacific Northwest National Laboratory, Richland, rivers in 1970 and their effect on Chinook Salmon and Steelhead Trout. Washington. NOAA (National Oceanic and Atmospheric Administration) Technical Re- Meekin, T. K., and B. K. Turner. 1974. Tolerance of salmonid eggs, juveniles and port NMFS (National Marine Fisheries Service) SSRF (Special Scientific squawfish to supersaturated nitrogen. Washington Department of Fisheries Report Fisheries) 646. Technical Report 12:78–126. Ebel, W. J., K. T. Beiningen, G. R. Bouck, W. R. Penrose, and D. E. Weitkamp. Montgomery, J. C., and C. D. Becker. 1980. Gas bubble disease in Small- 1979. Gases, total dissolved. Pages 113–118 in R. V. Thurston, R. C. Russo, mouth Bass and Northern Squawfish from the Snake and Columbia rivers. C. M. Fetterolf Jr., T. A. Edsall, and Y. M. Barber Jr., editors. A review of the Transactions of the American Fisheries Society 109:734–736. EPA red book: quality criteria for water. American Fisheries Society, Water Murray, C. J., D. R. Geist, E. V. Arntzen, Y. J. Bott, and M. A. Nabelek. Quality Section, Bethesda, Maryland. 2011. Development of a conceptual Chum Salmon emergence model for Ives Ebel, W. J., and H. L. Raymond. 1976. Effect of atmospheric gas supersaturation Island. Final Report to U.S. Army Corps of Engineers, PNNL-20035, Pacific on salmon and Steelhead Trout of the Snake and Columbia rivers. Marine Northwest National Laboratory, Richland, Washington. Fisheries Review 38:1–14. Nearing, J., M. Betka, S. Quinn, H. Hentschel, M. Elger, M. Baum, M. Egusa, S. 1959. The gas disease of fish due to excess of nitrogen. Journal of the Bai, N. Chattopadyhay, E. M. Brown, S. C. Hebert, and H. W. Harris. Faculty of Fisheries and Animal Husbandry, Hiroshima University 2:157– 2002. Polyvalent cation receptor proteins (CaRs) are salinity sensors in fish. 182. (In Japanese with English summary.) Proceedings of the National Academy of Sciences of the USA 99:9231– Endangered Species Act of 1973. U.S. Code, volume 16, sections 1531–1544. 9236. Evans, D. H., P. M. Piermarini, and K. P. Choe. 2005. The multifunctional fish Nebeker, A. V., J. D. Andros, J. K. McCrady, and D. G. Stevens. 1978. Sur- gill: dominant site of gas exchange, osmoregulation, acid–base regulation, vival of Steelhead Trout (Salmo gairdneri) eggs, embryos, and fry in air- and excretion of nitrogenous waste. Physiological Reviews 85:97–177. supersaturated water. Journal of the Fisheries Research Board of Canada Fidler, L. E. 1988. Gas bubble trauma in fish. Doctoral dissertation. University 35:261–264. of British Columbia, Vancouver. Nebeker, A. V., and J. R. Brett. 1976. Effects of air-supersaturated water on Foskett, J. K., and C. Scheffey. 1982. The chloride cell: definitive identification survival of Pacific salmon and Steelhead smolts. Transactions of the American as the salt-secretory cell in teleosts. Science 215:164–166. Fisheries Society 105:338–342. Hand, K. D., K. M. Carter, D. R. Geist, V. I. Cullinan, and R. A. Elston. 2009. Nebeker, A. V., A. K. Hauck, and F. D. Baker. 1979. Temperature and oxygen- Bioassays on the formation of gas bubble disease in Chum Salmon fry at total nitrogen gas ratios affect fish survival in air-supersaturated water. Water dissolved gas levels ranging up to 129% saturation. Pages 2.1–2.29 in E. V. Research 13:299–303. Arntzen, K. D. Hand, K. M. Carter, D. R. Geist, K. J. Murray, E. M. Dawley, NOAA (National Oceanic and Atmospheric Administration). 1995. Item 2. V. I. Cullinan, R. A. Elston, and J. Vavrinec III, editors. Total dissolved gas Pages 104–110 in Endangered Species Act - section 7, consultation, biological effects on incubating Chum Salmon below Bonneville Dam. Final Report to opinion, federal Columbia River power system (FCRPS). NOAA, National U.S. Army Corps of Engineers, PNNL-18081, Pacific Northwest National Marine Fisheries Service, Northwest Regional Office, Seattle. Laboratory, Richland, Washington. NOAA (National Oceanic and Atmospheric Administration). 2000. Risk assess- Hand, K. D., D. R. Geist, V. I. Cullinan, and R. A. Elston. 2008. Bioassays on ment for spill program described in 2000 draft biological opinion. Appendix E the formation of gas bubble disease in Chum Salmon fry at total dissolved gas in Endangered Species Act – section 7, biological opinion on the re-initiation levels ranging up to 113% saturation. Pages 2.1–2.20 in E. V. Arntzen, K. D. of consultation on operation of the federal Columbia River power system, Hand, D. R. Geist, K. J. Murray, J. L. Panther, V. I. Cullinan, E. M. Dawley, including the juvenile fish transportation program, and 19 Bureau of Recla- and R. A. Elston, editors. Effects of total dissolved gas on Chum Salmon fry mation projects in the Columbia Basin. NOAA, National Marine Fisheries incubating in the lower Columbia River. Final Report to U.S. Army Corps of Service, Northwest Regional Office, Seattle. Engineers, PNNL-17132, Pacific Northwest National Laboratory, Richland, NOAA (National Oceanic and Atmospheric Administration). 2004. Endangered Washington. Species Act – section 7, consultation, biological opinion, consultation on Harvey, H. H., and A. C. Cooper. 1962. Origin and treatment of a supersaturated remand for operation of the Columbia River power system and 19 Bureau river water. International Pacific Salmon Fisheries Commission, Progress of Reclamation projects in the Columbia basin (National Wildlife Federation Report 9, Vancouver. versus National Marine Fisheries Service, Civil Case 01-640-RE [May 23, Heming, T. A., and R. P. Preston. 1981. Differential effect of formalin preserva- 2007; revised and reissued pursuant to court order]). NOAA, National Marine tion on yolk and tissue of young Chinook Salmon (Oncorhynchus tshawytscha Fisheries Service, Northwest Regional Office, Seattle. Walbaum). Canadian Journal of Zoology 59:1608–1611. Philp, R. B., and C. W. Gowdey. 1964. Experimental analysis of the relation Jensen, J. O. T., J. Schnute, and D. F. Alderdice. 1986. Assessing juve- between body fat and susceptibility to decompression sickness. Aerospace Downloaded by [Department Of Fisheries] at 22:33 28 February 2013 nile salmonid response to gas supersaturation using a general multivariate Medicine 35:351–356. dose–response model. Canadian Journal of Fisheries and Aquatic Sciences Pisam, M., A. Caroff, and A. Rambourg. 1987. Two types of chloride cells in 43:1694–1709. the gill epithelium of a freshwater-adapted euryhaline fish: Lebistes reticula- Krise, W. F., and R. L. Herman. 1989. Tolerance of Lake Trout, Salvelinus tus; their modifications during adaptation to saltwater. American Journal of namaycush (Walbaum), sac fry to dissolved gas supersaturation. Journal of Anatomy 179:40–50. Fish Diseases 12:269–273. Raymond, H. L. 1970. A summary of the 1969 and 1970 outmigration of juvenile Krise, W. F., and R. A. Smith. 1991. Tolerance of juvenile Lake Trout exposed Chinook Salmon and Steelhead Trout from the Snake River: progress report. to gas supersaturation. Progressive Fish-Culturist 53:17–20. U.S. Bureau of Commercial Fisheries Biological Laboratory, Seattle. McCormick, S. D. 1995. Hormonal control of gill Na +,K+-ATPase and chloride Richman, N. H., III, S. Tai de Diaz, R. S. Nishioka, P. Prunet, and H. A. Bern. cell function. Pages 285–315 in C. M. Wood and T. J. Shuttleworth, editors. 1987. Osmoregulatory and endocrine relationships with chloride cell mor- Cellular and molecular approaches to fish ionic regulation: fish physiology, phology and density during smoltification in Coho Salmon (Oncorhynchus volume 14. Academic Press, San Diego, California. kisutch). Aquaculture 60:265–285. McCormick, S. D., A. M. Regish, and A. K. Christensen. 2009. Distinct fresh- Rombough, P. J. 1999. The gill of fish larvae: is it primarily a respiratory water and seawater isoforms of Na +/K +-ATPase in gill chloride cells of or an ionoregulatory structure? Journal of Fish Biology 55(Supplement A): Atlantic Salmon. Journal of Experimental Biology 212:3994–4001. 186–204. McGrath, K. E., E. M. Dawley, and D. R. Geist. 2006. Total dissolved gas effects Rucker, R. R. 1975. Gas-bubble disease: mortalities of Coho Salmon, On- on fishes of the lower Columbia River. Final Report to U.S. Army Corps of corhynchus kisutch, in water with constant total gas pressure and different EFFECTS OF TOTAL DISSOLVED GAS ON CHUM SALMON FRY 215

oxygen-nitrogen ratios. U. S. National Marine Fisheries Service Fishery Bul- Talbot, C., F. B. Eddy, and J. Johnston. 1982. Osmoregulation in salmon and letin 73:915–918. Sea Trout alevins. Journal of Experimental Biology 101:61–70. Rucker, R. R., and P. M. Kangas. 1974. Effect of nitrogen supersaturated Uchida, K., T. Kaneko, K. Yamauchi, and T. Hirano. 1996. Morphometrical water on Coho and Chinook Salmon. Progressive Fish-Culturist 36:152– analysis of chloride cell activity in the gill filaments and lamellae and changes 156. in Na +,K+-ATPase activity during seawater adaptation in Chum Salmon fry. Shikano, T., and Y. Fujio. 1998. Immunolocalization of Na + ,K+-ATPase and Journal of Experimental Zoology 276:193–200. morphological changes in two types of chloride cells in the gill epithelium USEPA (U.S. Environmental Protection Agency). 1973. Water quality criteria during seawater and freshwater adaptation in a euryhaline teleost, Poecilia 1972. USEPA, EPA-R3-73033, Government Printing Office, Washington, reticulata. Journal of Experimental Zoology 281:80–89. D.C. Shirahata, S. 1966. Experiments on nitrogen gas disease with Rainbow Trout Weitkamp, D. E., and M. Katz. 1980. A review of dissolved gas supersaturation fry. Bulletin of the Freshwater Fisheries Research Laboratory (Tokyo) 15: literature. Transactions of the American Fisheries Society 109:659–702. 197–211. Wells, P. R., and A. W. Pinder. 1996a. The respiratory development of Atlantic Stevens, D. G., A. V. Nebeker, and R. J. Baker. 1980. Avoidance responses of Salmon: I. morphometry of gills, yolk sac and body surface. Journal of salmon and trout to air-supersaturated water. Transactions of the American Experimental Biology 199:2725–2736. Fisheries Society 109:751–754. Wells, P. R., and A. W. Pinder. 1996b. The respiratory development of Atlantic Stroud, R. K., G. R. Bouck, and A. V. Nebeker. 1975. Pathology of acute and Salmon: II. partitioning of oxygen uptake among gills, yolk sac and body chronic exposure of salmonid fishes to supersaturated water. Pages 435–449 surfaces. Journal of Experimental Biology 199:2737–2744. in W. A. Adams, editor. Chemistry and physics of aqueous gas solutions. Wood, J. W.1979. Diseases of Pacific salmon: their prevention and treatment, 3rd Electrochemical Society, Princeton, New Jersey. edition. Washington Department of Fisheries, Hatchery Division, Olympia. Downloaded by [Department Of Fisheries] at 22:33 28 February 2013 This article was downloaded by: [Department Of Fisheries] On: 28 February 2013, At: 22:34 Publisher: Taylor & Francis Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

North American Journal of Fisheries Management Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/ujfm20 Route Use and Survival of Juvenile Chinook Salmon through the San Joaquin River Delta Rebecca A. Buchanan a , John R. Skalski a , Patricia L. Brandes b & Andrea Fuller c a Columbia Basin Research, School of Aquatic and Fishery Sciences, University of Washington, 1325 4th Avenue, Suite 1820, Seattle, Washington, 98101-2509, USA b U.S. Fish and Wildlife Service, 850 Guild Avenue, Suite 105, Lodi, California, 96240, USA c FISHBIO Environmental, LLC, 1617 South Yosemite Avenue, Oakdale, California, 95361, USA Version of record first published: 01 Feb 2013.

To cite this article: Rebecca A. Buchanan , John R. Skalski , Patricia L. Brandes & Andrea Fuller (2013): Route Use and Survival of Juvenile Chinook Salmon through the San Joaquin River Delta, North American Journal of Fisheries Management, 33:1, 216-229 To link to this article: http://dx.doi.org/10.1080/02755947.2012.728178

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material. North American Journal of Fisheries Management 33:216–229, 2013 C American Fisheries Society 2013 ISSN: 0275-5947 print / 1548-8675 online DOI: 10.1080/02755947.2012.728178

ARTICLE

Route Use and Survival of Juvenile Chinook Salmon through the San Joaquin River Delta

Rebecca A. Buchanan* and John R. Skalski Columbia Basin Research, School of Aquatic and Fishery Sciences, University of Washington, 1325 4th Avenue, Suite 1820, Seattle, Washington 98101-2509, USA Patricia L. Brandes U.S. Fish and Wildlife Service, 850 Guild Avenue, Suite 105, Lodi, California 96240, USA Andrea Fuller FISHBIO Environmental, LLC, 1617 South Yosemite Avenue, Oakdale, California 95361, USA

Abstract The survival of juvenile Chinook Salmon through the lower San Joaquin River and Sacramento–San Joaquin River Delta in California was estimated using acoustic tags in the spring of 2009 and 2010. The focus was on route use and survival within two major routes through the Delta: the San Joaquin River, which skirts most of the interior Delta to the east, and the Old River, a distributary of the San Joaquin River leading to federal and state water export facilities that pump water out of the Delta. The estimated probability of using the Old River route was 0.47 in both 2009 and 2010. Survival through the southern (i.e., upstream) portion of the Delta was very low in 2009, estimated at 0.06, and there was no significant difference between the Old River and San Joaquin River routes. Estimated survival through the Southern Delta was considerably higher in 2010 (0.56), being higher in the Old River route than in the San Joaquin route. Total estimated survival through the entire Delta (estimated only in 2010) was low (0.05); again, survival was higher through the Old River. Most fish in the Old River that survived to the end of the Delta had been salvaged from the federal water export facility on the Old River and trucked around the remainder of the Delta. The very low survival estimates reported here are considerably lower than observed salmon survival through comparable reaches of other large West Coast river systems and are unlikely to be sustainable for this salmon population. More research into mortality factors in the Delta and new management actions will be necessary to recover this population. Downloaded by [Department Of Fisheries] at 22:34 28 February 2013 The Central Valley of California marks the southern limit of extract water from the southern portion of the Delta (Figure 1) Chinook Salmon Oncorhynchus tshawytscha in North America for agricultural, industrial, and municipal use throughout Cal- (Healey 1991). Chinook Salmon population abundances in ifornia. The Delta provides drinking water for approximately this region have been much reduced from the 19th century 27 million Californians and irrigation water for more than in response to a number of factors, including habitat loss, 1,800 agricultural users, and 4.6–6.3 million acre-feet of water hatcheries, and water development (e.g., pumping water are exported from the Delta annually (DSC 2011). This intense out of the basin; Healey 1991; Fisher 1994). Today, the exporting combined with tidal fluctuations can sometimes Sacramento–San Joaquin River Delta is a highly modified cause net flows in the Delta to be directed upstream rather environment with levees and drained fields replacing tidal than downstream (Brandes and McLain 2001). Pollution from wetlands, and riprap replacing natural shoreline. Demand for industry, agricultural and urban runoff, and erosion are also Delta waters is high. State and federal water export facilities concerns (DSC 2011). Both native and nonnative species of

*Corresponding author: [email protected] Received March 28, 2012; accepted September 5, 2012

216 JUVENILE CHINOOK SALMON IN THE SAN JOAQUIN RIVER 217

FIGURE 1. Acoustic telemetry receiver sites throughout the San Joaquin River Delta for the juvenile Chinook Salmon tagging studies in 2009 and 2010. The region included in each major route through the study area is shaded for the Southern Delta for the (a) San Joaquin River and (b) Old River routes and through the entire Delta for the (c) San Joaquin River and (d) Old River routes. Sites in the San Joaquin, Old, and Middle rivers are labeled A, B, and C, respectively. The label for site B2 includes the study years 2009 (09) and 2010 (10). Sites A7, C1, and G1 were used only in 2010. Mossdale is denoted by A2, Chipps Island at river kilometer 0 by G1, the federal water export facilities by E1 and E2, and state water export facilities by D1 and D2. The city of Stockton is near sitesA5and A6. Sites B3 and C2 are located near California Highway 4. Release sites are designated as follows: DF = Durham Ferry (2009, 2010), OR = Old River (2010), STK = Stockton (2010), and R = release after salvage and trucking. Route-specific survival and route entrainment probability were estimated for the Southern Delta in 2009 and 2010 and for the entire Delta in 2010. [Figure available in color online.] Downloaded by [Department Of Fisheries] at 22:34 28 February 2013 predatory fish (e.g., Striped Bass Morone saxatilis, Largemouth long-term goal is to relate Delta survival to changes in river flow Bass Micropterus salmoides, White Catfish Ameiurus catus) (discharge) and water export levels in the presence of a tempo- inhabit these areas and feed on migrating smolts, as do avian rary barrier at the head of the Old River, which was designed to predators including double-crested cormorants Phalacrocorax prevent salmon from entering the Old River (Figure 1). Prior to auritus and white pelicans Pelecanus erythrorhynchos.Allof 2006, VAMP tagging studies relied on coded wire tags (CWTs), these factors lower survival of migrating salmon smolts relative which provided information on salmon survival on a large spa- to historical conditions. tial scale using 100,000–300,000 study fish each year (Newman The Vernalis Adaptive Management Plan (VAMP) is a large- 2008). Starting in 2006, the tagging studies began using micro- scale, long-term (12-year) experimental management program acoustic tags, which provide more precise survival information begun in 2000 that was designed to protect juvenile Chinook on a smaller spatial scale with much smaller releases groups Salmon as they migrate from the San Joaquin River through the (e.g., about 1,000 fish). Coded wire tags were discontinued in Sacramento–San Joaquin River Delta (Figure 1; SJRGA 2005, 2007. Study years 2006 and 2007 were pilot studies provid- 2007, 2010, 2011). Part of the VAMP is a multiyear tagging ing feedback on design and implementation of the acoustic tag study to monitor juvenile salmon survival through the Delta; the studies. The 2008 study deployed an extensive array of acoustic 218 BUCHANAN ET AL.

hydrophones throughout the Delta but suffered from a high junction with the San Joaquin River. The Southern Delta region degree of premature tag failure (Holbrook et al. 2013). Thus, was entirely contained within the Delta region (Figure 1). In 2009 and 2010 were the first years that provided sufficient infor- 2009, no acoustic receivers were deployed at Chipps Island, so mation to estimate salmon survival through portions of the Delta the study area was limited to the Southern Delta. In 2010, a more on a relatively detailed spatial scale, yielding the first estimates extensive detection field was installed, including dual receivers of how fish distribute across various migration routes. Further, at Chipps Island (G1) (Figure 1). Thus, in 2010, the study area these 2 years represent different hydrologic conditions—very included the entire migration path through the Delta region. low flows in 2009 and above normal flows in 2010—thus pro- Two migration routes were monitored through both the South- viding preliminary information needed to identify a relationship ern Delta and Delta regions: the San Joaquin Route (route A in between survival and flow. Survival through the southern portion Figure 1a, c) and the Old River route (route B in Figure 1b, d). of the Delta was estimated in both 2009 and 2010, and survival Since the 1990s, a temporary physical or nonphysical bar- through the entire Delta was estimated in 2010 (described be- rier (sound, strobe lights, and a bubble curtain) has often been low; Figure 1). In both years, survival estimates were compared installed at the head of the Old River with the aim of pre- through two major migration routes: the San Joaquin River route venting migrating smolts from entering that river. In 2009 and and the Old River route. We present here the first spatially de- 2010, a nonphysical barrier was installed there, and its smolt- tailed estimates of survival and route use by juvenile Chinook guidance effectiveness was evaluated in studies concurrent with Salmon through the lower San Joaquin River into the Delta. the VAMP studies (Bowen et al. 2009; Bowen and Bark 2012). The nonphysical barrier was operated during passage of approximately half of each VAMP release group in 2009 or 2010. No STUDY AREA physical barrier was installed. Historically, focus has been on the survival of fish through the Delta to Chipps Island, located in Suisan Bay at the confluence of the San Joaquin and Sacramento rivers near Pittsburg, METHODS California, at river kilometer (rkm) 0 (Figure 1). Fish moving Tagging and release methods.—Both study years used the through the Delta toward Chipps Island may use any of several Hydroacoustic Technology, Inc. (HTI) Model 795 microacous- routes. The simplest route follows the San Joaquin River until tic tag (diameter = 6.7 mm, length = 16.3–16.4 mm, average it joins the Sacramento River near Chipps Island (Figure 1a, weight in air = 0.65 g). In 2009 a total of 933 juvenile Chinook c; route A). An alternative route uses the Old River from its Salmon (fall–spring-run hybrids) originating from the Feather head on the San Joaquin River to Chipps Island, either via its River Fish Hatchery were tagged and released between 22 April confluence with the San Joaquin River just west of Mandeville and 13 May (fork length = 85.0–110.0 mm, mean = 94.8 mm; Island, or through Middle River or the state and federal water ex- Table 1). Difficulties in rearing fish to size resulted in an average port facilities (Figure 1b, d; route B). Additional subroutes were tag burden (i.e., the ratio of tag weight to body weight) of 7.1% monitored for fish use but were contained within either route A (range = 4.4–10.2%), which was higher than desired (≤5.5%; or route B. Subroute C consists of the Middle River from the Brown et al. 2006). Six fish died in 2009 between tagging and Old River to the San Joaquin downstream of Medford Island. release. In 2010, a total of 993 juvenile fall-run Chinook Salmon Two other subroutes were the water export facilities off the Old originating from the Merced River Fish Hatchery were tagged River: fish entering either the State Water Project (subroute D) and released between 27 April and 20 May (fork length = or the Central Valley Project (subroute E) had the possibility of 99.0–121.0 mm, mean = 110.5 mm). Tag burden in 2010 was being trucked from those sites and released upstream of Chipps 2.8–5.8% (mean = 4.2%; Table 1). Four fish died in 2010 be- Downloaded by [Department Of Fisheries] at 22:34 28 February 2013 Island. Subroutes C, D, and E were all contained in route B tween tagging and release. (Old River). Finally, fish that remained in the San Joaquin River In both years, tagging was performed at the Tracy Fish Fa- past Stockton may have entered Turner Cut and maneuvered to cility located in the Delta approximately 30–45 km from the Chipps Island through the interior of the Delta (subroute F). Fish release site(s). Tagging procedures followed those outlined in in routes B, C, and F all had multiple unmonitored pathways Adams et al. (1998) and Martinelli et al. (1998). Fish were available for passing through the Delta toward Chipps Island. anesthetized in a 70-mg/L tricaine methanesulfonate solution, Survival through the study area was estimated on two spatial buffered with an equal concentration of sodium bicarbonate, and scales: (1) the southern portion of the Delta, which is bounded surgically implanted with programmed acoustic transmitters. downstream by the federal and state water export facilities, Cal- Typical surgery times were less than 3 min. Nonfunctioning tags ifornia Highway 4, and the Turner Cut junction with the San were removed from the study. After surgery, fish were placed Joaquin River (the “Southern Delta”; Figure 1a, b) and (2) the in 19-L containers with high dissolved oxygen (DO) concen- entire Delta, which is bounded downstream by Chipps Island trations (110–130%) for recovery. Each holding container was (the “Delta”; Figure 1c, d). Both the Southern Delta and Delta perforated to allow partial water transfer and held no more than regions were bounded upstream by the acoustic receiver (site three tagged fish. After initial recovery from surgery, tagged A2) located near Mossdale Bridge, upstream of the Old River fish were transported in buckets to the release site in transport JUVENILE CHINOOK SALMON IN THE SAN JOAQUIN RIVER 219

TABLE 1. Release data for groups of Chinook salmon smolts used in the 2009 and 2010 Vernalis Adaptive Management Plan studies, where DF = Durham Ferry, STK = Stockton, and OR = Old River. In 2009, releases were pooled into strata for analysis; in 2010, releases from separate locations were jointly analyzed for a single release occasion.

Release Release Release Mean (range) Tag burden Release location date number fork length (mm) (%) stratum/occasion Study year 2009 DF Apr 22 133 96.1 (86–108) 6.9 (5.2–9.0) 1 Apr 25 134 93.4 (88–105) 7.3 (5.2–9.6) 1 Apr 29 134 97.1 (87–110) 6.8 (4.5–3.6) 2 May 2 134 96.6 (87–108) 6.6 (4.4–9.3) 2 May 6 132 92.6 (85–102) 7.7 (5.5–10.2) 2 May 9 133 93.9 (88–100) 7.3 (5.4–9.1) 2 May 13 133 93.8 (90–104) 7.2 (5.3–8.8) 3 Study year 2010 DF Apr 27–28 74 108.0 (102–110) 4.4 (3.5–5.7) 1 Apr 30–May 1 74 109.1 (103–115) 4.3 (3.1–5.4) 2 May 4–5 73 109.4 (102–118) 4.3 (3.4–5.6) 3 May 7–8 70 111.1 (101–119) 4.1 (3.1–5.4) 4 May 11–12 70 112.0 (99–121) 4.1 (3.1–5.4) 5 May 14–15 73 112.6 (101–119) 4.0 (3.1–5.3) 6 May 18–19 70 112.1 (103–119) 3.9 (2.8–5.3) 7 STK Apr 28–29 35 107.5 (100–115) 4.5 (3.5–5.6) 1 May 1–2 36 108.5 (100–115) 4.4 (3.4–5.4) 2 May 5–6 35 110.3 (104–118) 4.2 (3.4–5.0) 3 May 8–9 36 109.6 (102–117) 4.3 (3.5–5.6) 4 May 12–13 35 111.2 (105–119) 4.2 (3.3–5.4) 5 May 15–16 34 112.9 (102–119) 4.0 (3.0–5.2) 6 May 19–20 31 113.4 (108–119) 3.9 (3.1–5.0) 7 OR Apr 28–29 36 108.2 (102–117) 4.5 (3.6–5.3) 1 May 1–2 36 108.5 (102–115) 4.5 (3.5–5.6) 2 May 5–6 36 108.6 (100–118) 4.5 (3.4–5.6) 3 May 8–9 36 110.4 (104–118) 4.2 (3.5–5.1) 4 May 12–13 36 111.8 (104–120) 4.2 (2.9–5.8) 5 May 15–16 35 113.3 (105–119) 4.0 (3.0–5.2) 6 May 19–20 32 112.3 (101–119) 3.9 (3.2–5.3) 7

tanks designed to guard against ﬂuctuations in water tempera- and the other located in the San Joaquin River near the city of

Downloaded by [Department Of Fisheries] at 22:34 28 February 2013 ture and DO. Transport to the release site took approximately Stockton (Figure 1). The supplemental releases were designed 45–60 min. At the release site, tagged fish were held in either to provide enough tagged fish in the lower reaches of the study 1-m3 net pens (3-mm mesh; first release in 2009) or in perfo- area to estimate survival all the way to Chipps Island, even if rated 121.1-L plastic garbage cans (2010) for a minimum of survival was low from Durham Ferry. 24 h before release. For each study year, an in-tank tag life study was performed In 2009, all fish were released on the San Joaquin River at to measure the rate of tag failure under the tag operating param- Durham Ferry, located at approximately rkm 110 (measured eters (i.e., encoding, range, and pulse width) used in the study. from the river mouth at Chipps Island) approximately 20 km Stratified random sampling of tags across manufacturing lots upstream of the boundary of the study area (Mossdale Bridge; and tag codes was used to ensure that tags in the tag-life study Figure 1). The release site was located upstream of the study area represented the population of tags released in study fish. to allow fish to recover from handling and distribute naturally In both study years, tag effects on short-term (48-h) survival in the river channel before entering the study area. In 2010, were assessed using dummy (i.e., inactive)-tagged and untagged each of seven release occasions consisted of an initial release fish that were handled using the same procedures as fish with at Durham Ferry and two supplemental releases, one located active transmitters. No significant difference in survival was in the Old River near the junction with the San Joaquin River observed between dummy-tagged and untagged fish over the 220 BUCHANAN ET AL.

48-h period (SRJGA 2010, 2011). Tag effects on longer-term The locations of the hydrophones were dictated by the pos- (≤21 d) survival and predator avoidance were expected to be sible migration routes (San Joaquin [A], and Old River [B]) and small based on existing studies on effects of acoustic tags on subroutes, and by the two spatial scales on which inference was juvenile Chinook Salmon with comparable tag burden (e.g., to be made (Southern Delta and Delta). The acoustic receivers Anglea et al. 2004). located in Turner Cut (F1) and at the channel markers in the San Water temperatures at the release locations were <20◦C dur- Joaquin River near the Turner Cut junction (A6) monitored the ing most releases, ranging from 16.1◦C to 21.1◦C in 2009 and exit of the San Joaquin route through the Southern Delta region from 14.2◦C to 18.8◦C in 2010. Temperature increased as a in both 2009 and 2010 (Figure 1a). Likewise, the exit of the Old function of distance downstream from Durham Ferry in both River route through the Southern Delta region was monitored the San Joaquin River main stem and the Delta and increased by receivers at the state and federal water facilities and near throughout the season. Temperatures in the study area exceeded Highway 4 in both 2009 and 2010 (Figure 1b). In 2010, the exit 20◦C starting in mid-May in 2009 and in early June in 2010. of both the San Joaquin route (Figure 1c) and the Old River Hydrophone placement.—An extensive array of acoustic hy- route (Figure 1d) through the entire Delta region was monitored drophones and receivers was deployed throughout the Delta by dual receivers at Chipps Island. in each study year, with 19 receivers and hydrophones being Signal processing.—The raw tag detection data generated by deployed in 2009 and 32 receivers (35 hydrophones) in 2010 the acoustic telemetry receivers were processed by identifying (Figure 1). Acoustic receivers were named according to mi- the date and time of each tag detection. Unique tags were identi- gration route (A–G). Chipps Island, the final destination of all fied by the period (1/frequency) of the acoustic signal. The 2009 routes in 2010, was assigned its own route name (G). At each data were processed manually using the HTI proprietary soft- location, one to four hydrophones were deployed to achieve full ware MarkTags. The 2010 data were processed using a combi- cross-sectional coverage of the channel. nation of automatic and manual processing, manual processing Acoustic receivers were located at the Delta entrance being limited to key detection sites (SJRGA 2011). (Mossdale, site A2) in both 2009 and 2010, at the Delta exit The San Joaquin River Delta is home to several populations (Chipps Island, G1) in 2010, and at key points in between in of predatory fish that are large enough to feed on juvenile both years (Figure 1). The Mossdale site was moved 1.4 km salmonids, including Striped Bass, Largemouth Bass, and White downstream in 2010 to an acoustically quieter site. All avail- Catfish. A predatory fish that has eaten an acoustic-tagged juve- able migration routes were monitored at the Old River (sites nile salmon and then moves past a hydrophone may introduce A3 and B1) and Turner Cut (A6 and F1) diversions from the misleading tag detections into the data. Thus, it was necessary to San Joaquin River (Figure 1). Receivers were located on the identify and remove those detections that came from predators. San Joaquin River in Stockton near the Stockton Waste Water Likely predator detections were identified in a decision process Treatment Facility (A4) and near the Navy Drive Bridge just that used up to three levels of spatial–temporal analysis, based upstream of the Stockton Deep Water Ship Channel (A5) be- on the methods of Vogel (2010, 2011): near-field, mid-field, and cause of concern about salmon survival past the water treatment far-field. Near-field analysis required manual processing of the plant. Receivers were also located at the entrance to the state raw acoustic telemetry data, and interpreted the pattern of the and federal water export facilities on the Old River (Figure 1). acoustic signal during detection as an indicator of fish move- At the federal facility (Central Valley Project, CVP), receivers ment near the receiver. Mid-field analysis focused on residence were placed just upstream and downstream of the trash racks time within the detection field of each receiver, and transitions (E1) and in the holding tank (E2), where salvaged fish were held between neighboring receivers. Far-field analysis examined before transportation by truck to release sites in the lower Delta transitions on the scale of the study area. All available detection Downloaded by [Department Of Fisheries] at 22:34 28 February 2013 on the San Joaquin and Sacramento rivers (R). At the state facil- data were considered in identifying likely predator detections, ity, receivers were placed both outside (D1) and inside (D2) the as well as environmental data such as river flow and tidal radial entrance gates to the Clifton Court Forebay (CCF), the stage, measured at several gaging stations throughout the Delta reservoir from which the State Water Project draws water. Both (downloaded from the California Data Exchange Center Web the CVP trash racks and the CCF radial gates are known feeding site: http://cdec.water.ca.gov). The predator decision process areas for piscine predators (Vogel 2010, 2011). Receivers were was based on the assumptions that Chinook Salmon smolts were also located downstream in the Old (B3) and Middle (C2) rivers emigrating and so were directed downstream, and that they were near the Highway 4 bridge. Dual receiver arrays were placed unlikely to move between acoustic receivers (≥2 km) against at some sites to provide data to estimate detection probabili- river flow. Movements directed upstream against the flow were ties, typically at the downstream boundary of the study area and considered evidence of predation, although short-term upstream at sites just downstream of river junctions. Both acoustic lines movements under reverse flow or slack tide conditions were within each dual array (average 0.3 km apart) were designed for deemed consistent with a salmon smolt. Unusually fast or slow full coverage of the channel. The nonphysical barrier located at transitions between detection sites or particularly long residence the head of the Old River was evaluated via a separate network time at a detection site were also considered evidence of pre- of hydrophones that were not used in the VAMP study (Bowen dation. In 2009, the near-field analysis comprised the majority et al. 2009; Bowen and Bark 2012). of the predation decision process. In 2010, more emphasis JUVENILE CHINOOK SALMON IN THE SAN JOAQUIN RIVER 221

FIGURE 2. Model schematic for the 2009 Chinook Salmon smolt tagging study. Horizontal lines indicate acoustic receivers; parallel lines indicate dual receiver arrays. Model parameters are salmon reach survival (S), detection probabilities (P), route entrainment probabilities (ψ), transition probabilities (φ = ψS), and “last reach” parameters (λ = φP). Downloaded by [Department Of Fisheries] at 22:34 28 February 2013

was placed on travel time, residence time, and movements in tection probabilities, and route-use (“entrainment”) probabili- relation to river flow (mid-field and far-field analysis). ties (Figures 2, 3). The release–recapture model was similar to After removing the suspected predator detections, the the model developed by Perry et al. (2010), with states rep- processed data were converted to individual detection histories resenting the various routes through the Delta. Detection sites for each tagged fish. The detection history identified the (acoustic receivers) were named according to route. chronological sequence of sites where the tag was detected. The release–recapture models used for both study years used In the event that a tag was detected at a site or river junction parameters that denoted the probability of detection (Phi), route multiple times, the last path past the site or river junction was entrainment probability (ψhl), salmon reach survival (Shi), and used in the detection history as the best depiction of the final transition probabilities (φkj,hi) equivalent to the joint probability fate of the fish in the region. of movement and survival, where h and k represent route, i Statistical survival and migration model.—A multistate sta- and j represent detection sites within a route, and l represents tistical release–recapture model (Buchanan and Skalski 2010) junctions within a route (Figures 2, 3). The transition probability was developed and used to estimate salmon smolt survival, de- φkj,hi from site j in route k to site i in route h included all 222 BUCHANAN ET AL. Downloaded by [Department Of Fisheries] at 22:34 28 February 2013

FIGURE 3. Model schematic for the 2010 Chinook Salmon smolt tagging study. See Figure 2 for additional information. JUVENILE CHINOOK SALMON IN THE SAN JOAQUIN RIVER 223

possible routes between the two sites and was used when it S A5(R) was not possible to separately estimate the route entrainment SA5, for R =SD and survival probabilities. Unique transition parameters were = SA5(ψA2φA6,A7φA7,G1 + [1−ψA2]φF1,G1), for R = D estimated at receiver D1 located outside the radial gates of the Clifton Court Forebay depending on gate status at the time of and fish arrival (open or closed) in the 2010 study. Gate status data were unavailable for the 2009 study. In some cases, it was not possible to separately estimate the SB2(R⎧) ⎪ φ + φ + φ + φ , = transition probability to a site and the detection probability at ⎨⎪ B2,B3 B2,C2 B2,D1 B2,E1 for R SD φ φ + φ φ = . the site. This occurred primarily at the entrances to the water = B2,B3 B3,G1 B2,C2 C2,G1 for R D = = ⎪ + φ , φ , φ , export facilities (E1 CVP trash racks, and D1 first CCF ⎩⎪ B2 D1 D1 D2 D2 G1 receiver) due to sparse data. In these cases, the joint probability + φB2,E1φE1,E2φE2,G1, of survival from the previous receiver to receiver i in route h was estimated as λhi = φkj,hi Phi. We assumed that the detec- For fish that reached the interior receivers at the Clifton Court tion probability was 100% at the radial gate receivers inside Forebay or CVP in 2010, the parameters φD2,G1 and φE2,G1 Clifton Court Forebay and in the holding tank at the Central included survival during and after collection and transport. Al- Valley Project. These assumptions, necessary in the absence of though a subroute of the Old River route to Chipps Island, receivers located downstream of those detection sites and unique through Middle River from the junction with the Old River to those routes, were reasonable as long as the receivers were (subroute C) was monitored in 2010, no salmon were observed operating. leaving the Old River at that junction (site C1). Thus, the proba- A multinomial likelihood model was constructed based on bility of a smolt taking the Middle River route to Chipps Island possible capture histories under the assumptions of common was estimated to be zero. survival, route entrainment, and detection probabilities and in- In 2009, release groups were pooled into three strata based dependent detections among the tagged fish in each release on release timing, common environmental conditions, and mon- group. The likelihood model was fit using maximum likelihood itoring equipment status: stratum 1 = releases 1–2, stratum in the software Program USER (Lady and Skalski 2008), pro- 2 = releases 3–6, and stratum 3 = release 7 (Table 1). Malfunc- viding point estimates and standard errors of model parameters tioning acoustic receivers meant that some parameters could and derived performance measures. not be estimated for some strata. Model selection was used to In addition to the model parameters, performance at the mi- assess the effect of stratum on model parameters common to gration route level was estimated as functions of the model multiple strata. In 2010, data from each of the seven release parameters. The probability of a smolt taking the San Joaquin occasions (initial release at Durham Ferry combined with sup- River route (route A) was ψA1, while the probability of using plemental releases) were analyzed separately. For each release the Old River route (route B) was 1 − ψA1. Regional passage occasion, several alternative survival models were fit, differ- survival (SR for region R) was estimated on two spatial scales: ing in whether the initial (Durham Ferry) and supplemental the southern Delta (R = SD; 2009 and 2010) and the entire San release groups shared common detection, route entrainment, Joaquin River delta (R = D) from Mossdale Bridge to Chipps and survival parameters over common reaches. Model selec- Island (2010) (Figure 1). Regional passage survival for region R tion was used to find the most parsimonious model that fit all (R = SD or D) was defined in terms of both the route entrainment the data, following the general approach described in Burnham probability (ψA1) and the route-specific survival probabilities: et al. (1987) for comparing treatment groups. Detection prob- Downloaded by [Department Of Fisheries] at 22:34 28 February 2013 abilities were parameterized first, with survival, transition, and

SR = ψA1 SA(R) + (1 − ψA1)SB(R). route entrainment probabilities parameterized next. Backwards selection was used to identify the farthest reach upstream for The route-specific survival probabilities through region R which parameters from the initial and supplemental releases could be equated without reducing model fit. The most general (i.e., SA(R) and SB(R) for R = SD or D) were defined as models were considered first, with unique parameters for each release group for all reaches, and tested against simpler models S = S S S S A(R) A2 A3 A4 A5(R) with common parameters across the initial and supplemental release groups for the downstream reaches. All models used and unique survival and transition probabilities in the first reach downstream of the supplemental release sites. Model selection = . SB(R) SA2 SB1 SB2(R) was performed using the Akaike Information Criterion (AIC) as described in Burnham and Anderson (2002). Final param- The survival probabilities through the final reaches of each eter estimates were weighted averages of the release-specific route (i.e., SA5(R) and SB2(R)) were defined as estimates from the selected model, with weights equal to the 224 BUCHANAN ET AL.

number of fish from the release group present at the supplemen- Delta in the two routes were not significantly different (Z-test, tal release site (estimated for the initial release group). Goodness P = 0.4788). The route entrainment probabilities at the junction of fit was assessed using Anscombe residuals (McCullagh and of the Old River with the San Joaquin River were estimated at Nelder 1989: p. 38). ψˆ A1 = 0.47 (SE = 0.03) for the San Joaquin River, and 1−ψˆ A1 = 0.53 (SE = 0.03) for the Old River. Consequently, overall survival through the Southern Delta in 2009 was estimated as RESULTS SˆSD = 0.06 (SE = 0.01; Table 2). 2009 Results The first two release groups in 2009 (stratum 1) showed a None of the 50 tags in the 2009 tag-life study failed before higher probability of entering the Old River (1 − ψˆ A1 = 0.64; day 21. Because all detections of tagged salmon smolts occurred SE = 0.04) than remaining in the San Joaquin (P = 0.0002). well before day 21 after tag activation, no adjustment for tag Release groups 3–6 (stratum 2) showed no preference for either failure was made to the survival estimates from the release– route (P > 0.05), with 1 − ψˆ A1 = 0.48 (SE = 0.04) for the Old recapture model. River route entrainment probability. No estimates of the route Initial survival after release was low in 2009, with estimates entrainment probabilities were available for group 7 (stratum 3) of survival from Durham Ferry to the Mossdale Bridge (site because of equipment malfunction. A2, approximately 20 rkm) averaging 0.47 (SE = 0.02). The Median travel time through the Southern Delta reaches majority of the acoustic-tag detections downstream of Durham ranged from 0.2 d (SE = 0.2) from the Stockton USGS gauge Ferry were at the upstream sites in the San Joaquin (A2, A3) (A4) to the Navy Drive Bridge in Stockton (A5; approximately and in the Old River (B1). Very few tagged salmon smolts 3km),to2.1d(SE = 0.3) from Lathrop (A3) to the Stockton were detected at the exit points of the Southern Delta region in USGS gauge (A4; approximately 15 km). either the San Joaquin River route or the Old River route. No tagged salmon were detected at the Turner Cut receivers (F1), 2010 Results the Middle River receivers at Highway 4 (C2), or the interior Failure times of the 48 tags in the tag-life study ranged receivers at Clifton Court Forebay (D2). from 10 to 36 d. The early failure of several tags in the tag-life Total salmon survival through the Southern Delta region study made it necessary to incorporate tag-life adjustments (SSD) was estimable only for stratum 2 (releases 3–6) because the into survival estimates (Townsend et al. 2006). The estimated failure of certain acoustic receivers resulted in missing data from probability of tag survival to the time of arrival at each the three other release groups. Estimated route-specific survival detection site ranged from 0.987 to Chipps Island (G1) to 0.995 through the Southern Delta was SÂ(SD) = 0.05 (SE = 0.02) in to Mossdale (A2). Tag survival estimates for the supplemental the San Joaquin route and SˆB(SD) = 0.08 (SE = 0.02) in the Old releases at the Old River and Stockton were generally higher River route (Table 2). Survival estimates through the Southern than for the initial releases at Durham Ferry.

TABLE 2. Estimates of route-speciﬁc survival (S; standard errors in parentheses) of Chinook Salmon smolts through the Southern Delta (SD) and the entire Delta to Chipps Island (D) in the San Joaquin River (A) and Old River (B) and route entrainment probability into the San Joaquin River (A) at the head of theOld River for study years 2009 and 2010. Estimates of survival through the entire Delta are not available for 2009.

Southern Delta survival Entire Delta survival

Release date Route entrainment ψˆ A1 SˆA(SD) SˆB(SD) SˆSD SˆA(D) SˆB(D) SˆD

Downloaded by [Department Of Fisheries] at 22:34 28 February 2013 Study year 2009 Apr 22–25 0.36 (0.04) Apr 29–May 9 0.52 (0.04) 0.05 (0.02) 0.08 (0.02) 0.06 (0.02) May 13 0.05 (0.03) Average 0.47 (0.03) 0.05 (0.02) 0.08 (0.02) 0.06 (0.01) Study year 2010 Apr 27–29 0.48 (0.06) 0.47 (0.07) 0.78 (0.06) 0.63 (0.05) 0.07 (0.03) 0.00 (0.00) 0.03 (0.02) Apr 30–May 2 0.44 (0.06) 0.40 (0.06) 0.90 (0.04) 0.68 (0.05) 0.01 (0.01) 0.03 (0.02) 0.02 (0.01) May 4–6 0.39 (0.06) 0.16 (0.04) 0.75 (0.06) 0.52 (0.06) 0.01 (0.01) 0.01 (0.01) 0.01 (0.01) May 7–9 0.52 (0.07) 0.24 (0.05) 0.56 (0.09) 0.39 (0.06) 0.04 (0.02) 0.10 (0.03) 0.06 (0.02) May 11–13 0.45 (0.06) 0.49 (0.06) 0.88 (0.08) 0.71 (0.06) 0.06 (0.03) 0.13 (0.04) 0.10 (0.03) May 14–16 0.43 (0.06) 0.11 (0.04) 0.68 (0.29) 0.43 (0.17) 0.01 (0.01) 0.07 (0.02) 0.05 (0.02) May 18–20 0.59 (0.07) 0.35 (0.06) 0.83 (0.21) 0.55 (0.10) 0.07 (0.03) 0.15 (0.05) 0.10 (0.03) Average 0.47 (0.02) 0.32 (0.02) 0.77 (0.06) 0.56 (0.03) 0.04 (0.01) 0.07 (0.01) 0.05 (0.01) JUVENILE CHINOOK SALMON IN THE SAN JOAQUIN RIVER 225

All releases in the 2010 study had high initial survival, with Among the 29 salmon released at Durham Ferry in 2010 that estimates of survival from Durham Ferry to the Mossdale Bridge were subsequently detected at Chipps Island, 31% (9 fish) used receiver (site A2; approximately 21 km) averaging 0.94 (range = the San Joaquin route and 69% used the Old River route. The 0.86–1.00). The Old River supplemental release groups had an median travel time from the head of the Old River to Chipps average estimated survival to the head of Middle River (sites B2, Island was 5.7 d (migration rate = 14.0 km/d) through the San C1) of 0.89 (range = 0.84–0.97). The Stockton supplemental Joaquin route, compared with 7.2 d (7 km/d) for the single fish release groups had an average estimated survival to the Navy in the Old River route that migrated in-river past Highway 4, and Bridge in Stockton (site A5) of 0.82–1.07 (average = 0.95). Only 2.6 d for the 19 fish in the Old River route that passed through a single tag released at either Durham Ferry or the Old River was the Central Valley Project. Travel time for the CVP fish included detected in Middle River, so Middle River was omitted from the time spent in holding tanks and truck transport to release sites survival model. None of the 14 tags detected at Turner Cut were just upstream of Chipps Island, as part of the salvage operation subsequently detected at Chipps Island. at the facility. It appears that the fastest route through the San Estimates of the probability of fish remaining in the San Joaquin River Delta to Chipps Island in 2010 was through the Joaquin River at the head of the Old River in 2010 ranged from Old River and the CVP. 0.39 to 0.59 across the seven release groups (average = 0.47; SE = 0.02; Table 2). Only for release 3 did fish show a statistically significant (α = 0.05) preference for the Old River over DISCUSSION the San Joaquin River (P = 0.0443; one-sided Z-test). The results of 2 years of acoustic-tagging studies reported Route-specific survival through the Southern Delta region in here shed light on the survival of juvenile fall Chinook Salmon ˆ in the San Joaquin River Delta. Although estimated survival 2010 had an average estimate of S¯A(SD) = 0.32 (SE = 0.02) in was considerably higher in 2010 than in 2009, overall survival the San Joaquin route and S¯ˆ = 0.77 (SE = 0.05) in the B(SD) was low in both years, and survival and migration rates tended Old River route. For each release occasion, survival through the to be higher upstream and lower downstream. This pattern was Southern Delta was significantly higher in the Old River route observed throughout the Southern Delta in both 2009 and 2010 (P ≤ 0.003; one-sided Z-test on the lognormal scale), which and throughout the entire Delta in 2010. Some reduction in ended at the water export facilities and Highway 4. Combined migration rate is expected as fish move downstream because salmon survival through the Southern Delta region in 2010 was the cyclic tidal environment may reverse the direction of river estimated at S¯ˆ = 0.56 (SE = 0.03), averaged over all seven SD flow and temporarily push smolts upstream. Slower migration release groups (Table 2). rates, in turn, may lead to lower survival in downstream reaches, Survival through the entire San Joaquin River Delta region with slower-moving smolts being less able to evade predators (from Mossdale to Chipps Island, approximately 89 km) was (Anderson et al. 2005). considerably lower than through only the Southern Delta region When survival estimates were adjusted for reach length (i.e., ¯ˆ = = −1 in 2010, the average overall estimate being S D 0.05 (SE survival rate = Sˆ(km )), two regions displayed consistently low 0.01; Table 2). Estimated survival from Mossdale to Chipps survival rates. The San Joaquin River reach from the receiver ¯ˆ = = Island averaged S A(D) 0.04 (SE 0.01) in the San Joaquin near the Navy Drive Bridge in Stockton to the Turner Cut junc- ˆ route, and S¯B(D) = 0.07 (SE = 0.01) in the Old River route. Only tion had an estimated survival rate of 0.85 in 2009 and 0.94 the first release group showed a significant difference in survival in 2010. The reaches in the southwestern portion of the Old to Chipps Island between the two routes, survival through the River route (i.e., from the head of Middle River to the entrances San Joaquin route (SÂ(D) = 0.07, SE = 0.31) being higher than of the CVP and Clifton Court Forebay and to the Old River

Downloaded by [Department Of Fisheries] at 22:34 28 February 2013 through the Old River route (SˆB(D) = 0.00, SE = 0; P = 0.0100; at Highway 4) had comparable survival rate estimates in both Table 2). Lack of signiﬁcance for other release groups may have years, ranging from 0.83 to 0.90 in 2009 and 0.94–0.95 in 2010. been a result of low statistical power. Pooled over release groups, All other Southern Delta reaches had higher estimated survival however, estimated survival to Chipps Island was signiﬁcantly rates, while the only reach in the full Delta study area with higher through the Old River route than through the San Joaquin lower survival rate was the San Joaquin River reach from the River route (P = 0.0133). Turner Cut junction to Medford Island (0.86 in 2010). The San For tags released at Durham Ferry, the median travel time Joaquin River reaches from Stockton to the Turner Cut junction through the reaches ranged from 0.1 d (SE = 0.01) between the and Medford Island and the western portions of the Old River two Stockton receivers (A4 to A5; approximately 3 km) to 3.2 d route warrant further investigation into mortality factors. (SE = 0.5) from Medford Island (A7) to Chipps Island (G1); of The estimated probability of survival throughout the South- the multiple paths between A7 and G1, the path that used only ern Delta region was generally higher in 2010 than in 2009 in the San Joaquin River was approximately 46 km long. No tags both the San Joaquin River route and the Old River route. In par- were observed to move from Turner Cut to Chipps Island, and ticular, survival in the Old River from the junction with Middle the median transition from Old River South (B2) to the CVP River to the entrance of the water export facilities and Highway 4 trash racks (E1) was 0.9 d (SE = 0.1). appeared considerably higher in 2010 (average estimate = 0.92) 226 BUCHANAN ET AL.

than in 2009 (average = 0.16). Overall, the survival estimates export facilities occurred at a slightly higher and more variable through the Southern Delta region in 2009 (average = 0.06) were rate in 2009, the combined average export level being 56.4 m3/s comparable to the survival estimates through the entire Delta re- (range = 38.2–73.3 m3/s; SJRGA 2010). In 2010, the combined gion in 2010 (average = 0.05). Although no direct estimates of average export level was 43.0 m3/s (range = 37.4–44.2 m3/s) survival through the entire Delta were available in 2009, we can (SJRGA 2011). Both lower flows and higher exports may have conclude that total survival was <0.06. The drop in survival in contributed to the lower survival observed in 2009, although 2010 from the Southern Delta (0.56) to the entire Delta (0.05) the difference in average export level between 2009 and 2010 suggests that total survival through the entire Delta in 2009 may is small compared with possible daily variation in export levels have been as low as 0.005. Even considering the uncertainty (42.5–322.8 m3/s). Differences in the source and condition of inherent in the predator decision process, we can conclude that the study fish may also have contributed to performance differ- survival through the Delta was very low in 2009. If the survival ences between the 2 years. The 2009 study fish were hybrids probability estimated in 2009 was similar to survival in other of spring and fall-run Chinook Salmon from the Feather River low-flow years, current recovery efforts for San Joaquin River Fish Hatchery (FRH), located in the Sacramento River basin. Chinook Salmon may be inadequate during dry years. These hybrid fish tended to be smaller than the 2010 study Despite interannual survival differences, the average esti- fish, which were fall-run Chinook Salmon from the Merced mated probability of fish entering the Old River from the San River Fish Hatchery (MRH; located in the San Joaquin River Joaquin (0.53) did not differ between 2009 and 2010. This basin). Historically, experiments in the San Joaquin Delta have route’s entrainment probability was estimated in the presence used MRH fish. In 2009, however, low numbers of MRH fish of the nonphysical barrier operated at the head of the Old River. prompted the switch to the FRH for that year’s tagging study, The barrier was found to be effective at deterring smolts from despite concern that FRH fish (genetically from the Sacramento entering the Old River in 2010, but not in 2009 (Bowen et al. River) may not adequately represent survival of San Joaquin 2009; Bowen and Bark 2012, “protection efficiency”). Never- fall-run Chinook Salmon (Brandes and McLain 2001). In 2010, theless, the effect of the barrier on the overall VAMP study rebounding numbers at the MRH allowed us to return to MRH results was limited because the barrier was operated only for fish for that year’s tagging study. approximately half of each release group, and estimates of the The smaller size of the 2009 fish resulted in an average tag Old River route entrainment probability probably decreased by burden that was higher than in 2010, and also higher than desired <0.1 because of the barrier study. (≤5.5%; Brown et al. 2006). The higher tag burden in 2009 may The 2009 and 2010 survival estimates reported here depend have contributed to the high mortality in the first reach after re- partly on the decision process used to identify and remove pos- lease (Durham Ferry to Mossdale Bridge), where an estimated sible predator detections. Without removing any suspect detec- 53% of study fish died in 2009. However, differences in river tions, overall survival through the Southern Delta region would conditions and predator distribution may also have contributed be estimated at 0.34 in 2009 and 0.79 in 2010 and at 0.11 to differences in estimated mortality in this reach between the through the entire Delta region in 2010. Thus, estimated sur- 2 years. Dry conditions and low flows in 2009 may have con- vival would be higher in both years, but the comparisons be- centrated predators and prey (smolts) in a smaller volume of tween 2009 and 2010 and between the Southern Delta and the water. Higher water temperatures in 2009 may have kept the entire Delta would remain. However, many of the detections predators more active (e.g., Niimi and Beamish 1974), and also producing these higher survival estimates came from tags with more likely to reside in the San Joaquin River between Durham considerably longer residence times (e.g., up to 810 h) or longer Ferry and Mossdale Bridge, where water temperatures tend to travel times than expected for emigrating juvenile salmonids be cooler than in the Delta. Downloaded by [Department Of Fisheries] at 22:34 28 February 2013 (e.g., average residence time of approximately 0.5 h at most Despite the differences in survival between the 2009 and detection sites). Additionally, the fit of the statistical survival 2010 study years, both studies found that juvenile fall run model declined when the presumed predator detections were Chinook Salmon have very low survival through the San included, suggesting that they were unlikely to have come from Joaquin River Delta, well under 0.10. Our 2010 estimates were emigrating salmonids. The results presented here are based on similar to the lower range of previous survival estimates of our current understanding of behavior differences between ju- San Joaquin smolts based on CWT data (Brandes and McLain venile salmon and predators such as striped bass. Nevertheless, 2001). However, the extremely low survival potentially expe- more work needs to be done to develop methods for distinguish- rienced through the Delta in 2009 would have been lower than ing between detections of salmon and detections of predators, the lowest CWT estimates. Even the higher survival observed especially for acoustic tagging studies in highly complex envi- in 2010 was considerably lower than survival estimates of ronments such as the Delta. juvenile late fall-run Chinook Salmon from the Sacramento There are several possible explanations for the differences in River through the Delta, which ranged from 0.35 to 0.54 in Southern Delta survival observed between 2009 and 2010. River the winter of 2007 (Perry et al. 2010). The Perry study used flows in 2009 were very low, whereas 2010 had considerably comparable methods, with similar study design, tagging, and higher flows (Figure 4). Water exports from the federal and state analysis. However, the late fall run Chinook Salmon used in JUVENILE CHINOOK SALMON IN THE SAN JOAQUIN RIVER 227

FIGURE 4. Mean daily discharge of the San Joaquin River at the U.S. Geological Survey gauge near Vernalis, California (rkm 113 from Chipps Island), during Chinook Salmon tagging studies in 2009 and 2010. [Figure available in color online.]

the Perry study migrate in winter, whereas the fall-run Chinook over a distance comparable to the VAMP study area (approxi- Salmon used in the VAMP study migrate months earlier in mately 89 rkm). Over this distance, a population with a survival spring. Thus, not only were the VAMP fish smaller than the rate of 0.989/km would have an overall survival probability of Perry study fish, they also migrated when higher predator 0.37, as opposed to the 2010 estimate of 0.05. Although di- activity is expected because of warmer temperatures and the rect comparison with other basins is difficult, it appears that the striped bass spring spawning migration (Radtke 1966). Thus, salmon smolts used in the 2009 and 2010 VAMP studies are not there are several possible explanations why the VAMP study surviving as well on their seaward migration as other salmon may be expected to estimate lower survival than the Perry study. population on the western coast of North America. Estimates of juvenile Chinook Salmon survival through com- Part of the VAMPis a management plan based on the assump-

Downloaded by [Department Of Fisheries] at 22:34 28 February 2013 parable environments in other basins tend to be higher than those tion that salmon survival to Chipps Island is higher through the observed in the 2009 and 2010 VAMP studies. McMichael et al. San Joaquin River route than through the Old River route. This (2010) used acoustic tags to estimate survival of Chinook salmon assumption is based on CWT studies between 1985 and 1990 smolts through the lower 192 rkm of the Columbia River to that consistently found higher (but not statistically signiﬁcant) the river mouth; scaled by distance, the survival rate estimates point estimates of survival for smolts released in the San Joaquin − (Sˆ(km 1)) were 0.999 for yearlings and 0.998 for subyearlings. River downstream of the Old River than for those released in Acoustic-tagged spring Chinook Salmon from the Thompson– the Old River (Brandes and McLain, 2001). Modeling of these Fraser river system had estimated survival rates of 0.989–0.997 data and other CWT data indicated that keeping salmon out (average = 0.995) through more than 330 rkm to the Fraser of the Old River improved their survival (Newman 2008). The River mouth in 2004–2006 (Welch et al. 2008). These survival 2008 VAMP acoustic tag study results, although hampered by rates are considerably higher than both the VAMP-estimated a high degree of premature tag failure, suggest that survival to Southern Delta survival rate of 0.92 in 2009 and the estimated Chipps Island was also higher through the San Joaquin River entire Delta survival rate of 0.97 in 2010. Even the lowest sur- than through the Old River route in 2008 (Holbrook et al. 2009). vival rate estimate reported by Welch et al. (2008) for the Fraser Furthermore, there is evidence that salmon from the Sacramento River (0.989 in 2004) corresponds to much higher total survival River have a higher probability of reaching Chipps Island if they 228 BUCHANAN ET AL.

remain in the Sacramento River rather than entering the central for their helpful suggestions. Funding for the VAMP study was Delta (Newman and Brandes 2010, Perry et al. 2010). Since the provided by the signatories to the San Joaquin River Agree- 1990s, management has experimented with efforts to keep fish ment, and funding for analysis and manuscript preparation was in the San Joaquin River and out of the Old River by installing provided by the San Joaquin River Group Authority. a barrier (physical or nonphysical) at the head of the Old River. Our results suggest that prevailing ideas about relative survival REFERENCES in the two routes may be too simple, given that we found no Adams, N. S., D. W. Rondorf, S. D. Evans, and J. E. Kelly. 1998. Effects of conclusive evidence that survival was higher in the San Joaquin surgically and gastrically implanted radio transmitters on growth and feeding River route than in the Old River route. One difference between behavior of juvenile Chinook Salmon. Transactions of the American Fisheries the 2009 and 2010 study years and previous years was the switch Society 127:128–136. from a physical barrier to testing a nonphysical barrier at the Anderson, J. J., E. Gurarie, and R. W. Zabel. 2005. Mean free-path length the- head of the Old River in 2009 and 2010. Historically, the phys- ory of predator–prey interactions: application to juvenile salmon migration. Ecological Modelling 186:196–211. ical barrier at the Old River routed both fish and river flow into Anglea, S. M., D. R. Geist, R. S. Brown, K. A. Deters, and R. D. McDonald. the San Joaquin River (SJRGA 2005). In contrast, the nonphys- 2004. Effects of acoustic transmitters on swimming performance and predator ical barrier used in 2009 and 2010 routed fish but not flow into avoidance of juvenile Chinook Salmon. North American Journal of Fisheries the San Joaquin (Bowen et al. 2009; Bowen and Bark 2012). Management 24:162–170. With salmon smolt survival in the San Joaquin River thought to Bowen, M. D., and R. Bark. 2012. 2010 effectiveness of a non-physical fish barrier at the divergence of the Old and San Joaquin rivers (CA). U.S. De- increase with flow (SJRGA 2007), it is possible that the non- partment of the Interior, Bureau of Reclamation, Technical Memorandum physical barrier deprived smolts routed to the San Joaquin River 86-68290-10-07, Denver. of the increased flows necessary for improved survival (Perry Bowen, M. D., S. Hiebert, C. Hueth, and V. Maisonneuve. 2009. 2009 effec- et al. 2013). There is also a concern that the larger in-water struc- tiveness of a non-physical fish barrier at the divergence of the Old and San ture associated with the nonphysical barrier may create habitat Joaquin rivers (CA). U.S. Department of the Interior, Bureau of Reclamation, Technical Memorandum 86-68290-09-05, Denver. for increased predation at the site. More study is needed. Brandes, P. L., and J. S. McLain. 2001. Juvenile Chinook Salmon abundance, The San Joaquin River Delta represents just a small por- distribution, and survival in the Sacramento–San Joaquin Estuary. Pages 39– tion of the entire juvenile out-migration of San Joaquin Chi- 138 in R. L. Brown, editor. Contributions to the biology of Central Valley nook Salmon and in recent years has typically been traversed salmonids, volume 2. California Department of Fish and Game, Fish Bulletin in <2 weeks (SJRGA 2011; Holbrook et al. 2013). With sur- 179, Sacramento. Brown, R. S., D. R. Geist, K. A. Deters, and A. Grassell. 2006. Effects of survival through only a portion of the juvenile migration estimated gically implanted acoustic transmitters >2% of body mass on the swimming at <0.10, management efforts in the lower San Joaquin River performance, survival and growth of juvenile Sockeye and Chinook Salmon. and Delta must be more protective if salmon populations are Journal of Fish Biology 69:1626–1638. to persist in this region. However, effective management must Buchanan, R. A., and J. R. Skalski. 2010. Using multistate mark–recapture meth- be based on a better understanding of the factors influencing ods to model adult salmonid migration in an industrialized river. Ecological Modelling 221:582–589. mortality than is currently available. More research into salmon Burnham, K. P., and D. R. Anderson. 2002. Model selection and multimodel use of and survival in the Delta is needed, especially in dry years inference: a practical information-theoretic approach, 2nd edition. Springer, that may represent future conditions under climate change. In New York. light of increasing human demands for Central Valley water, it Burnham, K. P., D. R. Anderson, G. C. White, C. Brownie, and K. H. Pol- is unlikely that salmon survival will improve on its own. If the lock. 1987. Design and analysis methods for fish survival experiments based on release-recapture. American Fisheries Society, Monograph 5, Bethesda, survival estimates observed in these two studies are represen- Maryland. tative of the future, only extreme measures have a chance of DSC (Delta Stewardship Council). 2011. Available: http://www. Downloaded by [Department Of Fisheries] at 22:34 28 February 2013 saving San Joaquin River Chinook Salmon. deltacouncil.ca.gov/. (13 December 2011). Fisher, F. W. 1994. Past and present status of Central Valley Chinook Salmon. Conservation Biology 8:870–873. Healey, M. C. 1991. Life history of Chinook Salmon (Oncorhynchus ACKNOWLEDGMENTS tshawytscha). Pages 311–393 in C. Groot and L. Margolis, editors. Pacific Many people and groups were involved in designing and salmon life histories. University of British Columbia Press, Vancouver. implementing the VAMP study. We thank Natural Resource Holbrook, C. M., R. W. Perry, and N. S. Adams. 2009. Distribution and joint fish- Scientists, Inc., California Department of Fish and Game, Cal- tag survival of juvenile Chinook Salmon migrating through the Sacramento– San Joaquin River Delta, California, 2008. U.S. Geological Survey, Open-File ifornia Department of Water Resources, the U.S. Fish and Report 2009-1204, Reston, Virginia. Available: pubs.usgs.gov/. (November Wildlife Service (Stockton office), the U.S. Bureau of Recla- 2011). mation, Normandeau and Associates (Stevenson, Washington), Holbrook, C. M., R. W. Perry, P. L. Brandes, and N. S. Adams. 2013. Adjusting and the U.S. Geological Survey (Columbia River Research Lab survival estimates for premature transmitter failure: a case study from the and Sacramento office) for their contributions to the study. We Sacramento–San Joaquin Delta. Environmental Biology of Fishes 96:165– 173. also thank Mark Bowen from the U.S. Bureau of Reclamation Lady, J. M., and J. R. Skalski. 2008. USER 4: user-specified estimation routine. for access to detection data from the nonphysical barrier study School of Aquatic and Fishery Sciences, University of Washington, Seattle. at the head of the Old River, and three anonymous reviewers Available: www.cbr.washington.edu/paramest. (July 2012). JUVENILE CHINOOK SALMON IN THE SAN JOAQUIN RIVER 229

Martinelli, T. L., H. C. Hansel, and R. S. Shively. 1998. Growth and physiologi- ment and the Vernalis Adaptive Management Plan. SJRGA, Report to the cal responses to surgical and gastric radio transmitter implantation techniques California Water Resources Control Board, Davis, California. Available: in subyearling Chinook Salmon (Oncorhynchus tshawytscha). Hydrobiologia www.sjrg.org/technicalreport. (February 2012). 371/372:79–87. SJRGA (San Joaquin River Group Authority). 2007. 2006 annual technical McCullagh, P., and J. A. Nelder. 1989. Generalized linear models, 2nd edition. report: on implementation and monitoring of the San Joaquin River agree- Chapman and Hall/CRC Press, Boca Raton, Florida. ment and the Vernalis Adaptive Management Plan. SJRGA, Report to the McMichael, G. A., R. A. Harnish, B. J. Bellgraph, J. A. Carter, K. D. Ham, California Water Resources Control Board, Davis, California. Available: P. S. Titzler, and M. S. Hughes. 2010. Migratory behavior and survival of www.sjrg.org/technicalreport. (February 2012). juvenile salmonids in the lower Columbia River and estuary in 2009. Pacific SJRGA (San Joaquin River Group Authority). 2010. 2009 annual technical Northwest National Laboratory, PNNL-19545, Richland, Washington. report: on implementation and monitoring of the San Joaquin River agree- Newman, K. B. 2008. An evaluation of four Sacramento–San Joaquin River ment and the Vernalis Adaptive Management Plan. SJRGA, Report to the Delta juvenile salmon survival studies. CALFED Bay-Delta Program, California Water Resources Control Board, Davis, California. Available: Project SCI-06-G06-299, Sacramento, California. Available: www.science. www.sjrg.org/technicalreport. (February 2012). calwater.ca.gov./ (November 2011). SJRGA (San Joaquin River Group Authority). 2011. 2010 annual technical Newman, K. B., and P. L. Brandes. 2010. Hierarchical modeling of juvenile report: on implementation and monitoring of the San Joaquin River agree- Chinook Salmon survival as a function of Sacramento–San Joaquin Delta ment and the Vernalis Adaptive Management Plan. SJRGA, Report to the water exports. North American Journal of Fisheries Management 30:157– California Water Resources Control Board, Davis, California. Available: 169. www.sjrg.org/technicalreport. (February 2012). Niimi, A. J., and F. W. H. Beamish. 1974. Bioenergetics and growth of Large- Townsend, R. L., J. R. Skalski, P. Dillingham, and T. W. Steig. 2006. Cor- mouth Bass (Micropterus salmoides) in relation to body weight and temper- recting bias in survival estimation resulting from tag failure in acoustic and ature. Canadian Journal of Zoology 52:447–456. radiotelemetry studies. Journal of Agricultural, Biological, and Environmen- Perry, R. W., P. L. Brandes, J. R. Burau, A. P. Klimley, B. MacFarlane, C. tal Statistics 11:183–196. Michel, and J. R. Skalski. 2013. Sensitivity of survival to migration routes Vogel, D. A. 2010. Evaluation of acoustic-tagged juvenile Chinook Salmon used by juvenile Chinook Salmon to negotiate the Sacramento–San Joaquin movements in the Sacramento–San Joaquin Delta during the 2009 Vernalis River Delta. Environmental Biology of Fishes 96:381–392. Adaptive Management Program. San Joaquin River Group Authority, Tech- Perry,R.W.,J.R.Skalski,P.L.Brandes,P.T.Sandstrom,A.P.Klimley, nical Report to the California Water Resources Control Board, Davis, Cali- A. Ammann, and B. MacFarlane. 2010. Estimating survival and migra- fornia. Available: www.sjrg.org/technicalreport/. (July 2012). tion route probabilities of juvenile Chinook Salmon in the Sacramento–San Vogel, D. A. 2011. Evaluation of acoustic-tagged juvenile Chinook Salmon Joaquin River Delta. North American Journal of Fisheries Management 30: and predatory fish movements in the Sacramento–San Joaquin Delta during 142–156. the 2010 Vernalis Adaptive Management Program. San Joaquin River Group Radtke, L. D. 1966. Distribution of adult and subadult Striped Bass, Roccus Authority, Technical Report to the California Water Resources Control Board, saxatilis, in the Sacramento–San Joaquin Delta. Pages 15–27 in J. L. Turner Davis, California. Available: www.sjrg.org/technicalreport./ (July 2012). and D. W. Kelley, editors. Ecological studies of the Sacramento–San Joaquin Welch, D. W., E. L. Rechisky, M. C. Melnychuk, A. D. Porter, C. J. Walters, S. Delta, part II: fishes of the Delta. California Department of Fish and Game, Clements, B. J. Clemens, R. S. McKinley, and C. Schreck. 2008. Survival of Fish Bulletin 136, Sacramento. migrating salmon smolts in large rivers with and without dams. PLoS (Pub- SJRGA (San Joaquin River Group Authority). 2005. 2004 annual technical lic Library of Science) Biologyy [online serial] 6(10):e265. DOI: 10.1371/ report: on implementation and monitoring of the San Joaquin River agree- journal.pbio.0060265. Downloaded by [Department Of Fisheries] at 22:34 28 February 2013 This article was downloaded by: [Department Of Fisheries] On: 28 February 2013, At: 22:35 Publisher: Taylor & Francis Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

To cite this article: (2013): Guide for Authors, North American Journal of Fisheries Management, 33:1, 230-235 To link to this article: http://dx.doi.org/10.1080/02755947.2013.771529

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material. North American Journal of Fisheries Management 33:230–235, 2013 c American Fisheries Society 2013 ISSN: 0275-5947 print / 1548-8675 online DOI: 10.1080/02755947.2013.771529

NORTH AMERICAN JOURNAL OF FISHERIES MANAGEMENT Guide for Authors (3) Comments are critiques of papers published by this journal, responses to which will be invited from the original authors.

Submission Procedures Editorial Policy Manuscripts and associated correspondence should be sub- This journal’s mission is to promote communication among mitted at the journal’s online submission and tracking site, managers at all levels of the fisheries profession. We encourage http://mc.manuscriptcentral.com/najfm (this site may also be the submission of original papers on the management of finfish accessed through the Publications section at the American Fish- and exploitable shellfish in marine and fresh waters. Contribu- eries Society’s Web site, www.fisheries.org). Detailed instructions should relate to the means by which species, habitats, and tions, including acceptable file formats, are available at the site. harvests may be managed to protect and enhance fish and fishery resources for societal benefit. Well-presented case histories Review Process of the successes, failures, and side effects of fisheries programs are invited as one means of conveying practical management ex- Submitted papers will be critically reviewed by at least two perience to others. Examples of other appropriate subjects are experts in the relevant discipline(s) and evaluated by one of the new techniques for monitoring populations or allocating har- journal’s associate editors and one of its editors. A manuscript vest; the economics of exploitation and sociology of resource may be returned to its author without review if it is judged to be users; criteria for species protection and habitat manipulation; of poor quality or inappropriate for this journal. and management-related research, law, policy, or philosophy. In submitting a paper, you are stipulating that, except where Acceptance of a paper for publication is based on its util- explicitly indicated otherwise, all of the statements, data, and ity to managers from a broad range of perspectives. Papers other elements reflect your own work and not that of others. concerning fisheries science per se may be submitted to the References to the work of others should be properly cited; exact American Fisheries Society’s (AFS) sister publication Trans- quotations from other sources should be in quotation marks. actions of the American Fisheries Society; those dealing with Failure to follow these requirements may result in rejection of aquaculture may be submitted to the North American Journal the paper and, in extreme cases, restrictions on publishing in of Aquaculture; those dealing with the health of fish and other this journal. aquatic organisms may be submitted to the Journal of Aquatic Authors have the option of anonymity; if they wish to exercise Animal Health; and those with a focus on marine and estuarine it, they should remove any indication of their identity from the species and habitats may be submitted to Marine and Coastal manuscript. Fisheries: Dynamics, Management, and Ecosystem Science. Review of manuscripts relies on volunteers and can be a fairly Authors are cautioned not to republish original data without lengthy process. However, we strive to get decisions to authors full attribution and explicit permission; see “Dual Publication of in 9–12 weeks. If revisions are requested, authors should make Scientific Information” in Transactions of the American Fish- them promptly, normally within 30 days of receiving the editor’s Downloaded by [Department Of Fisheries] at 22:35 28 February 2013 eries Society 110:573–574, 1981. decision (short extensions will be allowed if there are justifiable delays). If a revision is not received within the allowed time, the paper will be considered withdrawn; late revisions will be Manuscript Submission and Review treated as new submissions and may have to go through the review process again.

Manuscript Categories Publication Charges Manuscripts may be submitted in any of the following cate- Publication charges are US$100 per printed page and will gories: (1) Articles are full reports, critical reviews, or in-depth be billed when the paper is in proof. Full and partial subsidies essays. Manuscripts in this category may be up to 100 pages are available to voting members of the American Fisheries So- long, including tables, ﬁgures, and other supporting material ciety who certify that grant or agency funds are not available. (the equivalent of about 30 printed pages). (2) Management Manuscript reviews are not affected by requests for subsidies; Briefs are short papers describing new techniques or observa- however, at least one author must be (or become) an AFS mem- tions, policy changes, management assessments, and the like. ber by the time that a paper is published. Every paper published

230 GUIDE FOR AUTHORS 231

in the journal is subject to a $30 fee to offset handling costs. Results.—As a rule, it is preferable to present detailed re- Authors will receive a free PDF of the published article and may sults in tables and/or figures and to devote the text to summary purchase reprints of their papers when they receive their proofs. statements and analyses. Display data in tables if numerical precision is important, in figures if trends are paramount. Although the presentation of a large amount of raw data is generally not Manuscript Preparation meaningful, data should not be refined to the point that the reader cannot verify the analyses or use the information for other purposes. In presenting the results of statistical tests, report the type Components of test, the test statistic, the degrees of freedom, and the signif- A typical manuscript will have the following components: icance level (P -value). Although the value 0.05 is commonly Title page.—The title page should give the title of the pa- used as the threshold in hypothesis testing, we have no specific per and the name(s) and complete mailing address(es) of the requirements in this area; in the interest of providing useful author(s). In addition to accurately reflecting the content of the information, authors should report all P -values. It is very im- paper, the title should be short (preferably no more than 12 portant that statistical designs and models be appropriate for the words) and to the point. See a recent issue of the journal for the studies in which they are used; we encourage authors to have a format to use for authors’ names and addresses. A suggested statistician review their work before submitting a paper for pub- running head (shortened version of the title) may also be in- lication. Lastly, statistical results should be presented in biolog- cluded on the title page. Keywords are not used in this journal, ically meaningful terms rather than in purely statistical jargon. however, and so should not be included. Discussion.—The merits of a paper can be greatly enhanced Abstract.—Articles and management briefs require abstracts; by a good discussion. In it authors should indicate the signif- comments do not. The abstract should consist of one paragraph icance of their research, how it relates to current knowledge, (up to 300 words for an article and up to 200 words for a brief) and any avenues that it suggests for further research. Informed that concisely states why and (generally) how the study was speculation is acceptable as long as it is clearly identified as done as well as what the results were and what they mean. It such. Authors should avoid merely restating their results and/or should not simply outline the contents (e.g., avoid statements (re)summarizing the literature. to the effect that such-and-such is presented) or present the Acknowledgments.—In this section authors may acknowl- methods in detail. Citations and footnotes are not allowed in edge the sources of their funding and thank those who con- abstracts, and abbreviations should be used sparingly. Because tributed directly to the project or the preparation of the abstracts tend to be more widely read than complete papers, manuscript. Dedications and acknowledgment of emotional sup- authors should take care to make them comprehensive, clear, port from family and friends are not appropriate. If all authors and interesting. are employees of the U.S. Government, this section should state Introduction.—The introduction should provide a context for that the mention of specific products does not constitute en- the work to be reported, particularly its purpose and importance. dorsement by their agency. In doing so, it should present at least a summary review of References.—References should be selected with a view relevant literature on the subject. to relevance and availability, with preference given to peer- Study site.—Brief descriptions of study sites should be in- reviewed publications that are widely available. Internal reports, cluded in the Methods section (discussed below). If a more papers presented at conferences, articles in preparation, and so detailed description is necessary, it should be given in a separate forth should be treated as unpublished and cited like personal section after the introduction. In either case, only information communications (i.e., parenthetically in the text alone). Authors Downloaded by [Department Of Fisheries] at 22:35 28 February 2013 that is essential for understanding and interpreting the results should obtain written permission to cite such material. Common should be given. Well-designed maps can reduce the need for reference formats are given below; a more complete list is given detailed descriptions of study sites. in Chapter 8 of the AFS style guide, which is available at the Methods.—Descriptions of the methods employed in the AFS Web site as well as the manuscript submission site. study should be detailed enough to enable readers to repeat Footnotes.—Text footnotes should be kept to a minimum. the work. Previously published descriptions may be cited in lieu Typically, they are used to report changes of address for au- of presenting complete new ones provided that the sources are thors, identify additional sources of data, or explain technical readily available (in general, avoid citations to theses, disserta- nomenclature (e.g., ages of anadromous fish and structures of tions, agency reports, and similar sources in this instance). If fatty acids). more than one method was used or a particular method entails a Tables.—In general, tables should be designed to present re- series of major steps, present each method or step in a separate lated information as simply and directly as possible. A good subsection. Appropriate tables and figures can reduce the need rule of thumb is to establish the point(s) that the table is infor detailed verbal descriptions of methods. Papers focusing en- tended to make, then to select the information required to do that tirely on techniques or models do not require a separate section and determine the most logical order in which to present it. De- on methods. tailed guidelines for the preparation of tables may be found 232 GUIDE FOR AUTHORS

in Chapter 12 of the AFS style guide, but a few of the more number of the figure and the name of the corresponding au- important ones may be mentioned here: thor should be given outside the image area of each figure for purposes of identification). Like table captions, figure captions 1. We prefer to print tables in “portrait” orientation but will should generally be detailed enough that the figure can be under- allow ones in “landscape” orientation as long as they take stood apart from the text. To the extent possible, however, panel up no more than one page. descriptions, (full) variable names, units of measure, legends, 2. Tables that are too long or too wide to fit on one page can be and so forth should be included in the figure itself rather than carried over to a facing page, but authors should try to avoid in the caption; in no case should they be given in both places. creating tables that span more than two pages. In general, Different panels may be designated “A,” “B,” and so forth, but it very large tables should appear in the online version of the is preferable to give them substantive labels (e.g., “Treatment” article only. and “Control”). 3. Tables should contain only three horizontal rules (lines)— Figures.—Figures include visual materials such as graphs, one before the column headings, one after those headings, maps, diagrams, and photographs. Figures have proved to be and one at the bottom of the table—and no vertical rules. one of the most troublesome aspects of the publishing process. 4. As a rule, captions should be detailed enough that the ta- As the Journals Department has only limited ability to modify ble can be understood apart from the text (if there is more figures, they frequently have to be sent back to the authors for than one table with the same general structure, only the correction. first needs to have a detailed caption). Captions should At the most fundamental level, figure design should follow be written so as to stress the purpose of the table and not certain commonsense principles: figures should be as simple merely list its contents in a mechanical way. and straightforward as possible; have a high enough resolution 5. There should be only one set of column headings. If the to be easily readable (300 dpi or more); and be consistent in information to be presented seems to require more than the use of lettering, line widths, and other graphic elements. In that, the table should be redesigned (e.g., by switching the addition, they need to conform to AFS style. It is particularly rows and columns) or split into two or more tables. important to remember that most figures will be reduced by up 6. Bold, centered headings may be used within the body of the to 50% when printed and thus need to be designed with this in table to distinguish different types of data as long as they mind. We recommend that authors use a copier to reduce each do not conflict with the column headings. figure to the width of one or two printed columns (3.50 and 7. Only the first letter of a row or column heading should be 7.25 inches, respectively), depending on the dimensions of the capitalized (along with words or symbols that would be particular figure, and verify that all elements are still legible. The capitalized in ordinary text). following are particularly problematical: bold type (which tends 8. The data within the body of the table should not be crowded; to blur), italic type (which tends to become less visible), dashed if need be, blank rows can be inserted to separate data into lines (which tend to appear continuous), dotted lines (which logical groups or provide guides for the eye. tend to disappear entirely), and shading in which the different 9. Significant differences among multiple means should be shades are not distinct enough. Additional guidelines for the indicated by lowercase letters, beginning with the letter “z” preparation of figures can be found in the AFS style guide. and moving backwards (“z” may mark either the highest or There is no additional charge for the reproduction of color the lowest value[s], but subsequent letters have to follow figures in the online version of the paper. However, all figures suit); there should be no omissions in the sequence of the in the print version will be in black and white unless specific letters. The letters should be set on the same lines as the arrangements have been made with the Journals Department Downloaded by [Department Of Fisheries] at 22:35 28 February 2013 values to which they pertain (not as superscripts) and be to cover the additional costs of color printing. Because color separated from those values by single spaces. printing is expensive, authors are advised not to use color to 10. Values less than 1.00 should be preceded by zeroes (e.g., distinguish phenomena when other means (different shading, 0.78). symbols, and so forth) are adequate. 11. Values need not be reported to all significant digits if a lesser Digital files in EPS, TIFF, and PSD formats are preferred; number of digits conveys the information in a meaningful figures should be submitted as separate files rather than being way. embedded in text files. 12. Footnotes should be indicated by superscripted lowercase Mathematical and statistical expressions.—Chapter 4 of the letters, beginning with the letter “a”; the letters may appear AFS style guide covers the treatment of these expressions in in the row and column headings as well as the body of the detail, but a few general points may be mentioned here: table but not in the caption. The footnotes per se should be listed on separate lines at the bottom of the table. 1. Symbols representing variables and parameters should be italicized if they consist of single letters in the Latin al- Figure captions.—Figure captions should appear together phabet (e.g., K and F ). All other symbols except Greek in a list rather than separately with each figure (however, the letters may be italicized or not, provided that the treatment is GUIDE FOR AUTHORS 233

consistent (e.g., CPUE or CPUE); Greek letters should never latest edition of Merriam-Webster’s Collegiate Dictionary.Ap- be italicized. pendix A of the AFS style guide shows the proper way to spell 2. Natural logarithms may be expressed as loge or ln; logarithms many of the terms used in fisheries writing (some of which are with other bases should identify the base (e.g., log10). not in the dictionary), including terms for which our preferred 3. Long equations should be “broken” at logical points, nor- spelling differs from that in the dictionary. mally after an operator such as a plus or minus sign. The standard resource for the common and scientific names 4. Definitions of variables and parameters may be run into the of North American fish species is the current edition of Common text if only a few such terms are involved. If there are a and Scientific Names of Fishes from the United States, Canada, number of them or they are used in more than one equation, and Mexico (American Fisheries Society, Bethesda, Maryland). a list is preferable (see section 4.8 of the style guide). For other aquatic species, authors should follow the companion 5. Avoid the expressions “the mean length was 45.2 ± 3.84 publications World Fishes Important to North Americans and mm” and “the mean (±SD) length was 45.2 ± 3.84 mm” Common and Scientific Names of Aquatic Invertebrates from because they are at best awkward and at worst inaccurate. the United States and Canada (the volumes Mollusks, Deca- Use the expressions “the mean ± SD length was 45.2 ± 3.84 pod Crustaceans, and Cnidaria and Ctenophora are currently mm” or “the mean length was 45.2 mm (SD, 3.84)” instead. available in the latter series). In most cases, scientific names should be included only at first mention in both the abstract and text; full common names (e.g., “Coho Salmon” rather than simply “Coho”) should be Style and Format used elsewhere. The format for the first mention is Published articles represent the culmination of research efforts, often lengthy and highly sophisticated ones. To do those Coho Salmon Oncorhynchus kisutch, efforts justice, however, the articles must be well written; poorly in which all parts of the common name are capitalized and the written articles not only place an unnecessary burden on read- scientific name follows the common name but is not given in ers, they also cast doubt on the quality of the research itself. parentheses. See Chapter 9 of the AFS style guide for additional Although some people naturally write better than others, most information on the treatment of species’ names; the accepted can develop the ability to write well through practice and atten- plurals of fish names are given in Appendix C of the guide. tion to detail. The introduction to the AFS style guide should be In papers about population dynamics, we prefer the notation a particularly valuable resource in this regard; in a few pages, it used by W. E. Ricker in Computation and Interpretation of Bio- identifies the errors in composition mostly commonly encoun- logical Statistics of Fish Populations (Fisheries Research Board tered in the papers submitted to AFS journals and shows how to of Canada Bulletin 191, 1975). However, all symbols should be correct them. We also encourage authors to have other fisheries defined anew in every paper. Our standard sources for chemical professionals critique their initial drafts with respect to presen- and enzyme names are the current editions of the Merck Index tation as well as substance. Authors whose native language is (Merck & Co., Rahway, New Jersey) and Enzyme Nomencla- not English should make a point of having English speakers ture (Academic Press, San Diego, California), respectively. The review their manuscripts before submission. preferred treatment of allozymes is noted in the article “Gene In writing for AFS journals, authors are also expected Nomenclature for Protein-Coding Loci in Fish” by J. B. Shaklee to follow certain style conventions pertaining to capitaliza- et al. (Transactions of the American Fisheries Society 119:2– tion, spelling, punctuation, mathematical expressions, technical 15, 1990). Additional information on the treatment of these and P terms, and so forth. For instance, we require that the letter other technical matters may be found in Chapter 11 of the AFS Downloaded by [Department Of Fisheries] at 22:35 28 February 2013 (indicating the degree of statistical significance) be capitalized style guide. as well as italicized, whereas some journals require that it be Manuscript format.—As an aid to reviewers and editors, au- lowercased. Although some of the more important style con- thors should ventions are noted below, all of them are discussed in detail in the AFS style guide. Authors would be well advised to become 1. use a line spacing of at least space and a half for all com- familiar with the main elements of AFS style and to consult the ponents of the paper, including the title page, footnotes, and guide frequently in preparing their manuscripts. tables; Resources for authors.—As suggested above, the principal 2. number all pages sequentially and provide continuous line resource on matters of style is the AFS style guide. Authors numbering beginning with the title page; may also find it helpful to consult the Chicago Manual of Style 3. use a 12-point font throughout; (University of Chicago Press, Chicago) and Scientific Style and 4. use three levels of headings, as follows: for the major sec- Format (Council of Science Editors, Chicago), though the AFS tions of the paper (Methods, Results, Discussion, Acknowl- style guide always takes precedence. edgments, and References), type them flush left with initial The standard resource for word usage and spelling is Web- letters capitalized (except for prepositions and conjunctions) ster’s Third New International Dictionary, as updated by the in ordinary type, preceded by “” (e.g., Methods); 234 GUIDE FOR AUTHORS

for subsections in Results and Discussion, type them flush Note that with one exception citations should be listed in left with initial letters capitalized in ordinary type, preceded chronological order; the exception is that all citations to the same by “” (e.g., Treatment 1); and for subsections in author(s) should be grouped together (see the fourth example Methods and sub-subsections in Results and Discussion, run above). them into the text with only the initial letter of the first word In reference lists, references should be in strict alphabetical capitalized, all words italicized, and followed by a period order by authors’ last names; if there are two or more references and a long dash (e.g., Sampling design.—); and with the same authors, those references should then be listed 5. turn off automatic hyphenation and justification. chronologically. All authors must be named in references. Detailed information on reference formats may be found in General style conventions.—A detailed presentation of AFS Chapter 8 of the AFS style guide. The more common types are style is beyond the scope of these guidelines. The following as follows: conventions, however, are so general as to apply to virtually every paper: Articles in journals Pace, M. L., and J. D. Orcutt. 1981. The relative importance of protozoans, ro- 1. Only symbols and abbreviations included in Webster’s dic- tifers, and crustaceans in a freshwater zooplankton community. Limnology tionaries or listed at the end of these guidelines (as well as and Oceanography 26:822–830. at the back of each printed issue of the journal) may be used without definition. All others should be defined at first use Note that (1) except for the first author, authors’ initials come (e.g., index of biotic integrity [IBI]). Abbreviations should before their last names; (2) only the first word of the title of the not be introduced unless they are used at least two more article is capitalized (along with any other words that would be times. capitalized in ordinary text); and (3) the name of the journal is 2. As a rule, either metric or English units may be used, but given in full. not both. The only exceptions are a few quantities that are typically expressed only one way (e.g., g [of medication]/lb Books [of feed]). 3. Single-digit numbers should be spelled out unless they are Krebs, C. J. 1989. Ecological methodology. Harper and Row, New York. used with units of measure or in conjunction with larger values (e.g., 8 Walleyes and 16 Saugers). Numbers with four Chapters in books or more digits should contain commas; those less than 1.00 should be preceded by zeroes. Omernik, J. M. 1995. Ecoregions: a spatial framework for environmental management. Pages 49–62 in W. S. Davis and T. P. Simon, editors. Biological 4. Ratios involving two values or units of measure should be assessment and criteria: tools for water resource planning and decision indicated by forward slashes (e.g., 0.30 g/d); ratios involving making. Lewis Publishers, Boca Raton, Florida. three such terms should be indicated by negative exponents (e.g., 0.01 g·g−1·d−1). Government reports 5. Ages of fish should be expressed by Arabic numerals and not contain plus signs (e.g., a fish is age 1 [not age 1+] from the Reports that are issued on a regular basis are treated much January 1 after it hatches to the following December 31). like articles in journals (the principal difference being that page 6. Dates may be expressed as either day–month–year (e.g., 11 numbers should not be given); other reports are treated like January 2011) or month–day–year (e.g., January 11, 2011), books:

Downloaded by [Department Of Fisheries] at 22:35 28 February 2013 provided that the same format is used throughout the paper. Note that the term “Julian day” does not mean day of the Everest, F. H., C. E. McLemore, and J. F. Ward. 1980. An improved tri-tube cryogenic gravel sampler. U.S. Forest Service Research Note PNW-350. year and should not be used in that context. [journal format] 7. Time should be expressed in terms of the 24-hour clock followed by the word “hours” (e.g., 1435 hours rather than USEPA (U.S. Environmental Protection Agency). 1998. Water quality crite- 2:35 p.m.). ria and standards plan: priorities for the future. USEPA, 822-R-98-003, Washington, D.C. [book format] Reference formats.—Text citations should conform to the author–year system. Examples of common types are as follows: Electronic publications

(Johnson 1995) Formats for references to electronic publications are still (Johnson and Smith 1996) evolving. The important thing is to give the reader enough in- (Johnson et al. 1997, 1998) [three or more authors] formation to be able to locate the reference easily. (Johnson et al. 1999, 2001; Smith 2000) If a book or report is only available online or is available (Johnson 2000a, 2000b) (Johnson, in press) in print form but was accessed online, the reference should be (E. M. Johnson, National Marine Fisheries Service, personal communication) formatted as follows: GUIDE FOR AUTHORS 235

Baldwin, N. A., R. W. Saalfield, M. R. Dochoda, H. J. Buettner, and R. L. depends on the way(s) in which articles are designated. Two Eshenroder. 2000. Commercial fish production in the Great Lakes, 1867– possible formats are as follows: 1996. Great Lakes Fishery Commission, Ann Arbor, Michigan. Available: www.glfc.org/databases/. (September 2000). Gallagher, M. B., and S. S. Heppell. 2010. Essential habitat information for age- 0 rockfish along the central Oregon coast. Marine and Coastal Fisheries: The month and year in parentheses indicate when the site Dynamics, Management, and Ecosystem Science [online serial] 2:60–72. was last accessed. DOI: 10.1577/C09-032.1 If a journal is available in print form, authors should use Kimmerer, W. J. 2004. Open-water processes of the San Francisco Estuary: the standard reference format even if they accessed the article from physical forcing to biological responses. San Francisco Estuary and online. If a journal is only available electronically, the format Watershed Science [online serial] 2(1):article 1. Downloaded by [Department Of Fisheries] at 22:35 28 February 2013