The Influence of Release Strategy and Migration History on Capture Rate of Oncorhynchus Mykiss in a Rotary Screw Trap Ian A

This article was downloaded by: [Department Of Fisheries] On: 28 February 2013, At: 22:44 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 Influence of Release Strategy and Migration History on Capture Rate of Oncorhynchus mykiss in a Rotary Screw Trap Ian A. Tattam a d , James R. Ruzycki b , Peter B. Bayley c , Hiram W. Li a & Guillermo R. Giannico c a Oregon Cooperative Fishery Research Unit, Department of Fisheries and Wildlife, Oregon State University, 104 Nash Hall, Corvallis, Oregon, 97331, USA b Oregon Department of Fish and Wildlife, Eastern Oregon University, 203 Badgley Hall, One University Boulevard, LaGrande, Oregon, 97850, USA c Department of Fisheries and Wildlife, Oregon State University, 104 Nash Hall, Corvallis, Oregon, 97331, USA d Oregon Department of Fish and Wildlife, Post Office Box 9, John Day, Oregon, Oregon, 97845, USA Version of record first published: 19 Feb 2013.

To cite this article: Ian A. Tattam , James R. Ruzycki , Peter B. Bayley , Hiram W. Li & Guillermo R. Giannico (2013): The Influence of Release Strategy and Migration History on Capture Rate of Oncorhynchus mykiss in a Rotary Screw Trap, North American Journal of Fisheries Management, 33:2, 237-244 To link to this article: http://dx.doi.org/10.1080/02755947.2012.758202

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 33:237–244, 2013 C American Fisheries Society 2013 ISSN: 0275-5947 print / 1548-8675 online DOI: 10.1080/02755947.2012.758202

ARTICLE

The Inﬂuence of Release Strategy and Migration History on Capture Rate of Oncorhynchus mykiss in a Rotary Screw Trap

Ian A. Tattam*1 Oregon Cooperative Fishery Research Unit, Department of Fisheries and Wildlife, Oregon State University, 104 Nash Hall, Corvallis, Oregon 97331, USA James R. Ruzycki Oregon Department of Fish and Wildlife, Eastern Oregon University, 203 Badgley Hall, One University Boulevard, LaGrande, Oregon 97850, USA Peter B. Bayley Department of Fisheries and Wildlife, Oregon State University, 104 Nash Hall, Corvallis, Oregon 97331, USA Hiram W. Li Oregon Cooperative Fishery Research Unit, Department of Fisheries and Wildlife, Oregon State University, 104 Nash Hall, Corvallis, Oregon 97331, USA Guillermo R. Giannico Department of Fisheries and Wildlife, Oregon State University, 104 Nash Hall, Corvallis, Oregon 97331, USA

Abstract Rotary screw traps are used in rivers throughout the west coast of North America to capture emigrating juvenile salmonids. Calibrating the capture efficiency of each trap is essential for valid estimates of fish passage. We released PIT-tagged Oncorhynchus mykiss upstream of a rotary screw trap in the South Fork John Day River, Oregon, to estimate capture efficiency. We used three strategies for release of fish recently captured in the trap. We recaptured 28% of medium-sized fish (86–145 mm FL) and 14% of large-sized fish (146–230 mm FL) released during daylight 1.6 km upstream from the trap. We recaptured 33% of medium-sized fish and 17% of large-sized fish released

Downloaded by [Department Of Fisheries] at 22:44 28 February 2013 during daylight 4.8 km upstream from the trap. We recaptured 42% of medium-sized fish and 23% of large-sized fish released at twilight 1.8 km upstream from the trap. A PIT tag antenna detected summer-tagged parr (which were PIT-tagged upstream 1–5 months before migration) as they approached the trap to evaluate potential bias from reduced recapture of recently trapped fish. We captured 53% of the medium-sized first-time migrants and 40% of the large-sized first-time migrants. Although average capture efficiencies of first-time migrants were greater than those from any of the recently trapped fish from the three release strategies, twilight releases of recently trapped fish were the least negatively biased, especially for medium-sized fish.

Extensive population monitoring of salmonid ﬁshes across et al. 1991). Returns of adult anadromous salmonids are in- the west coast of North America (Volkhardtet al. 2007) has been ﬂuenced by numerous factors in freshwater and marine life initiated in response to declining salmonid abundance (Nehlsen stages. Survival rates in migratory corridors and in the ocean are

*Corresponding author: [email protected] 1Present address: Oregon Department of Fish and Wildlife, Post Ofﬁce Box 9, John Day, Oregon 97845, USA. Received May 7, 2012; accepted December 6, 2012 237 238 TATTAM ET AL.

variable (Bilton et al. 1982; Achord et al. 2007) and may mask movement patterns. There has been no evaluation of whether the influence of freshwater rearing areas on production. Hence, these protocols result in equal capture efficiency between recent there is a need to determine abundance by life stage (Solazzi releases and migrants that have not been previously captured et al. 2000; Johnson et al. 2005) to measure effects of proximate in an RST. We evaluated these protocols by estimating capture factors, such as marine survival (Pyper et al. 2002). Most of- efficiency for Oncorhynchus mykiss that were PIT-tagged 1–5 ten such life history monitoring involves estimation of numbers months before their downstream migration (hereafter “summer- of (1) out-migrant juveniles emigrating from freshwater, and tagged parr”) and monitored by a PIT tag antenna immediately (2) adults returning to freshwater to spawn. upstream from an RST. Abundance estimates of juvenile salmonids emigrating from The three objectives of this study were: (1) compare esti- rearing habitats require out-migrant traps, except in the few mates of the capture efficiency of an RST among recent releases situations where census counts may be conducted at weirs. A made during daylight hours in two different locations and recent common out-migrant trap throughout the west coast of North releases made after the end of civil twilight, (2) develop a size- America is the rotary screw trap (RST; E.G. Solutions, Corvallis, structured model predicting capture efficiency for juvenile O. Oregon). Rotary screw caps can be nested inside weirs (Scace mykiss, and (3) validate the accuracy of the model by compar- et al. 2007) to increase capture efficiency. This can potentially ing the model-predicted capture efficiency for recent releases result in overcrowding of the holding box, causing mortality against estimates of capture efficiency for summer-tagged parr of biologically and economically valuable fishes (Music et al. detected by an antenna as they approached an RST. 2010). Thus, RSTs are commonly used as a “stand-alone” gear that samples a portion of the channel profile and captures a portion of the emigrant population. Valid estimates of capture METHODS efficiency are required for each RST in each location (Thedinga Site description.—This study was conducted in the South et al. 1994; Roper and Scarnecchia 2000) in order to estimate Fork John Day River (SFJD), a fifth-order basin located in out-migrant abundance. northeastern Oregon (Figure 1). The SFJD supports a natu- Capture efficiency varies depending on stream size, water rally reproducing population of O. mykiss, including resident velocity, water depth, cone rotation speed, and fish size (Roper and anadromous life history types. No hatchery stocking occurs and Scarnecchia 2000). The most commonly used method of in this basin, so all O. mykiss observed in this study were natu- estimating capture efficiency is to capture unmarked fish in the rally produced. Oncorhynchus mykiss are widely distributed in trap, apply a unique mark or tag, such as a PIT tag (Schultz et al. the SFJD and its four main tributaries downstream from Izee 2006; Copeland and Venditti 2009), and release the marked fish Falls, a barrier for anadromous fish movement (Figure 1). Em- upstream of the RST. The proportion of these marked fish (here- igration of juvenile O. mykiss from the SFJD is bimodal, with after referred to as “recent releases”) subsequently recaptured peaks occurring in October–November and April–May. During in the trap estimates capture efficiency for that sample period fall 2005, we estimated that 3,966 O. mykiss migrated past this (Thedinga et al. 1994; Miller et al. 2000). This mark–recapture RST site (I. Tattam, unpublished data). The O. mykiss captured population estimation technique is subject to the assumptions in the trap during this period ranged in fork length (FL) from 82 of the Petersen estimate (Seber 1982). Violation of any of these to 227 mm, with a mean of 140 mm. assumptions can result in erroneous population estimates (Frith We operated a 1.52-m-diameter RST at river kilometer 10 of et al. 1995). The most pertinent, yet seldom evaluated, is the as- the SFJD (Figure 1). An RST comprises a partially submerged sumption of equal capture efficiency of marked and unmarked cone with an interior helical structure that is passively rotated fish. For an estimate of catch efficiency to be unbiased the cap- by water pressure and funnels emigrant fish into a submerged Downloaded by [Department Of Fisheries] at 22:44 28 February 2013 ture efficiency of recent releases, which are fish captured in an holding box on the downstream end of the trap. The same type RST, tagged, released upstream of the RST, and then make a and size trap was used at the same location during fall (October– second migration past the RST (usually occurring in <24 h), December) 2004 and fall (October–December) 2005. We con- must equal the capture efficiency of fish approaching that RST ducted this calibration study during fall because flows are lower for the first time (termed “unhandled na¨ıve” by Scace et al. than during spring, allowing for placement of an in-stream PIT 2007). tag antenna. Stream discharge at this site during the 2005 release Two protocols for recent releases may help meet the assump- experiment ranged from 0.74 to 2.01 m3/s (OWRD 2012). In- tion of equal capture efficiency. The first protocol is to liberate stream PIT tag antenna data were not available for 2005; hence, recent releases close enough to an RST site so that mortality we used the combination of antenna and RST data from 2004 to or delayed migration prior to returning to the RST site is min- validate the predictive model we developed for recent releases imized (Roper and Scarnecchia 2000; Volkhardt et al. 2007). in 2005. The second protocol is to liberate recent releases at or after the The trap was situated at the head of a pool and was adjusted end of civil twilight (when the sun is 6◦ below the horizon line) both longitudinally and laterally to remain in the thalweg as each day. This protocol is based on the assumption that libera- discharge changed. Wetted width at this location was approx- tion after civil twilight will reduce predation and mimic natural imately 6 to 8 m, depending on discharge. We monitored two CAPTURE RATE IN A ROTARY SCREW TRAP 239

FIGURE 1. The South Fork John Day River (SFJD) basin. Locations of release strategies (A, B, and C) used during fall 2005 are indicated. Dashed circles indicate the summer rearing locations where summer-tagged parr were released 1–5 months prior to migration. The dashed arrow denotes streamﬂow direction. Inset shows the location of the SFJD basin in Oregon.

operational variables: stream depth (a surrogate for discharge) assigned to one of three release strategies. For example, every and trap rotation speed (a surrogate for water velocity). Depth first, fourth, seventh individual, and so on, retrieved from the was measured with a staff gauge in the pool downstream of the day’s catch was transported 1.6 km upstream and immediately trap. Speed was the number of seconds required for the cone of liberated during daylight hours (typically around 1100 hours). Downloaded by [Department Of Fisheries] at 22:44 28 February 2013 the trap (the mechanism by which fish are captured) to complete This short-distance release strategy during daylight was labeled three full rotations. Depth and speed were recorded daily. strategy A. Fish assigned to release strategy B were transported Comparison of capture efficiency estimates among different 4.8 km upstream (long distance) and immediately liberated dur- release strategies.—During fall 2005 we used three different ing daylight hours a few minutes after the release of the other recent release strategies. All unmarked O. mykiss captured in fish under strategy A. Finally, fish assigned to release strategy the RST were tagged with 12-mm-long full-duplex PIT tags in- C were transported 1.8 km upstream (short distance) and placed jected intraperitoneally (Prentice et al. 1990) and measured for into a holding device equipped with a timer (see description in FL. Retention rates for smaller Chinook Salmon O. tshawytscha Miller et al. 2000) that was set to release them at the conclusion PIT-tagged with these methods in a hatchery have been esti- of civil twilight. A total of 848 O. mykiss were PIT-tagged and mated at 99.9% over a 4-week period, with mortality rates < 1% released upstream from the SFJD RST on 37 separate days (daily (Dare 2003). We assumed no shedding of tags or tagging-related release by strategy ranged from 1 to 51 individuals) during fall mortality in our study. From October 14 through December 15, 2005 (Table 1). The RST was operated every night except one 2005, on each day that three or more unmarked O. mykiss were during this release experiment because a high volume of floating captured in the RST each fish was tagged and systematically leaf debris prevented RST operation on that night. We excluded 240 TATTAM ET AL.

TABLE 1. Sample sizes of O. mykiss captured, marked with a PIT tag, and released upstream (M) of the South Fork John Day River rotary screw trap over 37 different days during fall 2005. The number of recaptures (R) and recapture rate of marked ﬁsh (E) are presented by size-group and release strategy. Range is the minimum and maximum number of ﬁsh released by strategy on a single day. Strategy A was release during daylight 1.6 km upstream from the trap, strategy B was release during daylight 4.8 km upstream from the trap, and strategy C was release at civil twilight 1.8 km upstream from the trap.

Small (86–115 mm) Medium (116–145 mm) Large (146–230 mm) Strategy MR E M R E MR E Range A 53 16 0.30 130 37 0.28 113 15 0.13 1–51 B 41 16 0.39 140 39 0.28 99 22 0.22 1–49 C 42 16 0.38 133 63 0.47 97 18 0.19 1–49 Total 136 48 0.35 403 139 0.34 309 55 0.18 3–149

data from that day, as there was no potential for recapture on rotations per second. Product signs denote first order interac- the night after release and most recaptures occurred on the first tions. The logit function represents a log odds ratio expression night after release. of E, i.e., log[E/(1 − E)], allowing additive terms on the right We anticipated that length would influence capture efficiency side of equation (1) to be tested by analysis of deviance. We used (e.g., Roper and Scarnecchia 2000). Thus, we partitioned fish drop-in-deviance F-tests (Ramsey and Schafer 2002) to sequen- into three size categories within each recent release strategy: tially compare reduced models with the full model (equation 1). small (86–115 mm FL), medium (116–145 mm FL), and large Significant changes in deviance in reduced models represent (146–230 mm FL). We used a Pearson correlation to test for significant effects on logit(E) and, by association, on E. collinearity among explanatory variables. Depth and speed were Size-structured predictive model of E.—We developed a size- correlated (r = 0.85, n = 38, P < 0.0001). We eliminated depth structured predictive model of E for release strategy C. These and analyzed trap rotation speed, since we had some control over were of the form, speed as trap position was routinely adjusted to maximize it. We used logistic regression (SAS Procedure GenMod with logit link logit(E) = log[E/(1 − E)] = B0 + B1·Isize, (2) function) to model daily capture efficiency as a proportion and estimate significance of our strategies and other variables. Our where I is an indicator for different size-groups based on FL. model assumed a binomial distribution with an overdispersion size The B variables are fitted coefficients. The size ranges of indi- parameter to account for extrabinomial variation. Overdisper- viduals released in strategy C and the summer-tagged parr dif- sion is typical for capture efficiency estimates, probably because fered slightly (Figure 2). There were no summer-tagged parr > fish do not behave as independent and identical units, as a pure 200 mm FL (Figure 2). To account for possible size-based influ- binomial model assumes. Failure to account for overdispersion ences on the comparison of E between strategy C and summer- could have resulted in erroneous error estimates. This model ap- tagged parr, we censored the 11 individuals in strategy C that plies more weight to samples with a larger number of releases. were > 200 mm FL. Thus, the range of sizes was comparable The full model was between groups. Validation of the predictive model for E.—During fall 2004 = / − = + · + · + · logit(E) log[E (1 E)] B0 B1 Ib B2 Ic B3 Ismall we operated a PIT tag detection antenna (inner dimensions, Downloaded by [Department Of Fisheries] at 22:44 28 February 2013 + B4·Imedium + B5 · speed + B6·Ib·Ismall 30.5 cm high × 80.0 cm wide) in the thalweg 78 m upstream

+ B7·Ib·Imedium + B8·Ic·Ismall + B9·Ic·Imedium from the RST. The antenna was coupled to a Destron-Fearing 2001F transceiver that recorded date and time of detection. The + B ·I ·speed + B ·I ·speed + B ·I ·speed 10 b 11 c 12 small stream segment between the antenna and RST included two + · · , B13 Imedium speed (1) meanders and a turbulent riffle. The antenna detected 66% of summer-tagged parr known to have migrated past the array where E is capture efficiency (number recaptured / number re- (based on capture in the RST). Summer-tagged parr were O. leased). The B variables are fitted coefficients; Ib is the indicator mykiss that were PIT-tagged and released upstream of the RST (dummy variable, value 0 or 1) for release strategy B, Ic is during summer 2004. These individuals were primarily tagged the indicator for release strategy C (strategy A is represented in Black Canyon and Murderers creeks and, to a lesser extent, when Ia = Ib = 0, against which strategies B and C are com- in the SFJD upstream from Black Canyon Creek to Izee Falls pared in turn), Ismall is the indicator for the small FL group, (Figure 1). Summer-tagged parr were last handled 1–5 months Imedium is the indicator for the medium FL group (the small and before approaching the RST location. They were captured via medium size-groups are individually compared with the large seining or electrofishing and were unlikely to have had prior size-group in this model), and speed is the number of cone experience with an RST. Thus, we assumed that the migratory CAPTURE RATE IN A ROTARY SCREW TRAP 241

80 was made to determine whether strategy C (twilight-release, making a second migration past the RST) produced unbiased Summer Tagged Parr Release Strategy C estimates of E for O. mykiss. Release Strategy C Exclusions 60 RESULTS Comparison of E among Different Release Strategies 40 Drop-in-deviance tests found none of the ﬁrst-order interactions (release strategy × size-group, release strategy × speed,

Number Observed size-group × speed) significantly contributed to the model 20 (F8, 214 = 1.51, P = 0.16). Trap rotation speed also did not significantly contribute to the model (F1, 215 = 1.31, P = 0.25). Release strategy (F2, 217 = 4.7, P < 0.01) and FL (F2, 217 = 0 12.2, P < 0.001) were significant. Thus, we interpreted a re- 0 0 5 0 5 0 0 4 5 7 8 0 15 3 -11 1 -1 -1 -2 2 -2 < 95 6 6- 1-1 6 1 6 1- 6 duced version of equation (1) with release strategy and FL as 9 2 4 7 0 111-125 1 1 15 1 18 2 21 main effects. Fork Length Predicted E varied significantly among release strategies and FL (Figure 3). When analyzing FL, recent releases in the FIGURE 2. Length-frequency histogram for O. mykiss migrating past a rotary screw trap on the South Fork John Day River. Summer-tagged parr were tagged small FL group had significantly higher E than those in the large upstream from the trap during June–September 2004, 1–5 months before mi- FL group (P < 0.001). Likewise, the E of recent releases in the gration. Fork lengths for summer-tagged parr were those observed or estimated medium FL group was significantly higher than those in the when they migrated past the trap. Release strategy C fish were captured in the large FL group (P < 0.001). There was no significant difference trap during October–December 2005, tagged, and released upstream from the in E between recent releases in the small and medium FL groups trap at civil twilight. Exclusions were those individuals removed from the final = logistic regression model. (P 0.81). The E of recent releases under strategy A was

behavior, diel migration timing, and probability of capture in the RST of these summer-tagged parr were equal to that of O. mykiss that had never been captured before. Summer-tagged parr were categorized into the same FL groups used in equation (1). For summer-tagged parr captured in the RST, we used FL on the day of RST capture to group individuals. Additionally, we used FL data from summer-tagged parr captured in the RST to estimate mean growth rates (mm/d) experienced by each spe- ciﬁc tagging group (i.e., Black Canyon Creek, Murderers Creek, or upper SFJD) from tagging date to their recapture at the RST. For summer-tagged parr that were not captured in the RST, we estimated FL of each individual on the day it migrated past the RST. We estimated FL of summer-tagged parr that were detected

Downloaded by [Department Of Fisheries] at 22:44 28 February 2013 but not captured from the formula: [FL when tagged in upper basin + (mean daily growth rate × number of days at large)]. This was a minor correction and increased FL by a mean of 15% FIGURE 3. Capture efficiencies for PIT-tagged O. mykiss released upstream (range, 4–32%). from the South Fork John Day River rotary screw trap during 2004–2005. For summer-tagged parr, E was estimated by the quotient Estimates are from two binomial logistic regression models of the effect of of the number of O. mykiss captured at the RST divided by release strategy and size-group on capture efficiency. Release strategy A (fish the total number detected migrating past the PIT array. We were released 1.6 km upstream during daylight), release strategy B (fish were released 4.8 km upstream during daylight), and release strategy C (fish were restricted this analysis to nights when both the RST and the PIT released 1.8 km upstream at civil twilight) occurred during fall 2005. Size-groups array were operational, as O. mykiss nearly always migrated for these releases were: small = 86–115 mm FL, medium = 116–145 mm FL, past the PIT array and RST in the same night. Oncorhynchus and large = 146–230 mm FL. Bars with diagonal stripes compare capture mykiss detected at the PIT array on multiple days were censored efficiencies from the final binomial logistic regression model of strategy C as it was unknown whether they migrated past the RST during (Final C) with observed capture efficiencies for summer-tagged parr (Summer Parr). Summer-tagged parr were detected migrating past a PIT tag antenna the study period. Estimates of E for summer-tagged parr were 78 m upstream from the rotary screw trap during fall 2004. Size-groups for this compared with model-predicted 95% confidence intervals of E comparison were: small–medium = 86–145 mm FL, and large = 146–200 mm derived for two FL groups within strategy C. This comparison FL. Error bars are 95% CIs. 242 TATTAM ET AL.

significantly lower than for those under strategy C (P = 0.005). and Scarnecchia 1996). Migration timing, rather than loss of The E of recent releases with strategy B was not significantly na¨ıvete,´ appeared to drive E in our study. Fish < 146 mm different from those with strategy A (P = 0.32). We proceeded captured in the RST and released upstream at civil twilight with our final size structured predictive model (equation 2) (migrating past the trap during darkness) were recaptured at only for strategy C, because it was closest to our summer parr a rate comparable with fish approaching the trap for the first validation data (see below). time. By altering diel migration timing, release strategies A and B produced biased estimates of E compared with the E for Size-structured Predictive Model of E naturally migrating summer-tagged parr. We found no difference in E between small (86–115 mm) and Fish length had a significant effect on rate of recapture in the medium (116–145 mm) FL groups; thus, we combined these two RST. However, the relationship between E, as logit(E), and FL FL groups into a small–medium group. The predicted E for the is not linear. The recapture rate of recent releases in the small small–medium group was 0.45 (95% confidence interval [CI], and medium FL groups was not significantly different when it 0.36–0.55; Figure 3). The predicted E for the large group was is compared within any single release strategy (Figure 3). There 0.19 (95% CI, 0.10–0.31; Figure 3). The size-structured bino- is a threshold length, represented by the large size-group in our mial logistic model (equation 2) was overdispersed, as indicated study, above which O. mykiss have an increased ability to avoid by an estimated overdispersion parameter of 1.26. capture in an RST (Figure 3). This decline in E for individuals > 146 mm was also present for summer-tagged parr, although to Validation of the Predictive Model for E a lesser degree than for recent releases (Figure 3). Dambacher Strategy C predictions were compared with averages for (1991) also found E to decrease with FL. However, he noted summer-tagged parr. The observed E of summer-tagged parr declining E beginning at an FL of only 106 mm when fish- in the RST differed between FL groups. Estimated average E ing a Humphreys trap. Trap placement, operation, stream flow, was 0.53 (24 captured of 45 available for capture) for summer- fish size, and species encountered will influence E uniquely in tagged parr in the small–medium group and 0.40 (14 captured of each trapping situation. For example, Thedinga et al. (1994) 35 available for capture) in the large size-group. For the small– did not find any size-based differences in E when using a 2.4-m medium group, the 95% CI of E from the regression model RST. However, the observed recapture rate was very low (3–6%, (equation 2) for strategy C included the average estimate of E Thedinga et al. 1994), perhaps limiting the power to detect size observed for summer-tagged parr (Figure 3). The 95% CI of E differences in E. Future RST calibration efforts should antici- for the large group (equation 2) did not encompass the observed pate size-based differences in E. If the number of recent releases E for summer-tagged parr (Figure 3). is even among size-groups, fewer large individuals will be recaptured. If fewer, large individuals are recaptured, the estimate DISCUSSION of E for the large size-group will be less precise. Increasing Time of release influenced E for O. mykiss in the South the number of fish in the large size-group released upstream Fork John Day River. Daylight releases (strategy A or B) from an RST is necessary to increase recaptures and, hence, resulted in lower estimates of E than twilight releases (strategy increase the precision of the estimate of E. If few large wild C). Between daylight releases, transporting O. mykiss farther out-migrants can be captured, releasing large hatchery-origin upstream (strategy B) did not significantly change E compared out-migrants upstream from an RST may be a strategy to in- with releases in close proximity to the trap (strategy A). Such crease recaptures. The capture efficiency of hatchery and wild daylight releases (strategies A and B) probably resulted in a out-migrants may differ (Roper and Scarnecchia 1996). Hence, daytime second migration past the RST. During fall 2004, recent statistically comparing capture efficiency of the two groups is Downloaded by [Department Of Fisheries] at 22:44 28 February 2013 releases with strategy A often migrated past the PIT tag antenna necessary before applying capture efficiencies of hatchery fish during daylight (I. Tattam, unpublished data). Conversely, natu- to wild fish. Releasing hatchery fish may not be an option in ral downstream migration of salmonids occurs during darkness basins managed for natural production, such as the South Fork (Roper and Scarnecchia 1996). Of the summer-tagged parr de- John Day River. Nonetheless, it may be a strategy to increase tected at the PIT tag antenna during fall 2004, 94% of detections precision of efficiency estimates in basins that are managed for occurred after evening civil twilight and before the beginning both natural and hatchery production. of civil twilight the following morning. Individuals migrating Strategy C produced estimates of E comparable with the E during daylight were seldom captured in the RST. Similarly, observed for summer-tagged parr. We found evidence that, at Cramer et al. (1992) found capture efficiency of juvenile least for O. mykiss in the small and medium size-groups, esti- Chinook Salmon in an RST to be 15 times higher at night mates of E for recent releases and summer-tagged parr were not than during daylight. We suspect that individuals migrating statistically different when using strategy C (Figure 3). These during daylight might have been lower in the water column and results are similar to those of Scace et al. (2007), who also em- less vulnerable to the RST, which samples the upper portions ployed a PIT tag antenna upstream from an RST. They found of the column. To avoid the RST daytime out-migrants may that when using a weir and RST in combination, E was high and also use visual clues, which are unavailable at night (Roper comparable between summer-tagged parr and twilight-released CAPTURE RATE IN A ROTARY SCREW TRAP 243

smolts of Atlantic Salmon Salmo salar. Our results differed for O. mykiss in the small and medium size-groups, liberating re- from Scace et al. (2007) for O. mykiss in the large size-group. cent releases at civil twilight created an estimate of E that is not For large O. mykiss, we found a significant difference between statistically different from that of naturally migrating fish. Our estimates of E for summer-tagged parr and recent releases (Fig- results indicate that this release strategy will accurately estimate ure 3). Therefore strategy C did not effectively duplicate the E out-migrant abundance for the small and medium size-groups. of summer-tagged parr O. mykiss in the large size-group. Prior However, it remains unclear whether nighttime upstream re- experience with the RST did not reduce the E of fish released leases will produce a valid estimate of E for O. mykiss in the under strategy C in the small and medium size-groups compared large size-group. Therefore, estimates of E should be qualified with summer-tagged parr. Unless prior experience with the RST by time (Roper and Scarnecchia 2000) and fish length. Alter- differentially influences large fish, we believe this probably did native methods of estimating E for large fish making their first not cause the discrepancy in E. approach to the trap should be further investigated. Placing PIT The summer rearing location of large-sized O. mykiss may tag antennas or dual-frequency identification sonar immediately explain the difference in E between the recent release twilight upstream from an RST may be two approaches that could be group and summer-tagged parr. Summer-tagged parr were all used. PIT-tagged >10 km upstream from the RST (Figure 1). We do The efficiency of any RST needs to be estimated in order to not know from where the fish used for the recent release groups estimate out-migrant abundance. We found evidence that releas- originated. However, it is plausible that some of these individ- ing marked O. mykiss upstream from an RST during daylight uals originated from near the RST and were simply making will result in biased estimates of out-migrant abundance. Releas- home-range movements when captured. Oncorhynchus mykiss ing marked O. mykiss at twilight will create unbiased estimates were present in this location year-round (I. Tattam, personal ob- of out-migrant abundance for small and medium size-groups. servation). Small- and medium-sized O. mykiss dominated the Alternative trap calibration methods, preferably using an inde- population size structure in upstream reaches of the SFJD and pendent measure of migrating fish abundance, should be con- its tributaries. Larger O. mykiss dominated the population near sidered. One option is to PIT-tag juvenile salmonids upstream the RST (Madrinãn´ 2008). Thus, when unmarked O. mykiss in from the RST and use PIT tag antennas near the RST to de- the small and medium size-groups were captured in the RST, tect migrants. Our results suggest this is critical for an accurate it is more likely that they were migrating several kilometers estimate of E for large-sized O. mykiss. or more to reach the RST (similar to the summer-tagged parr) rather than just moving within their home range. Some of the large O. mykiss captured in the RST might have been released ACKNOWLEDGMENTS upstream within their original home range. In this scenario, they We thank W. Wilson, J. Schricker, T. Goby, T. Schultz, R. may not attempt a second migration past the RST. Alternatively, Lamb, D. Bondurant, T. Hartill, and L. Hewlett for their diligent large individuals might have been more effective than small- work operating the SFJD screw trap. S. White, F. Madrinãn,´ and medium-sized individuals at avoiding capture on a second J. Feldhaus, S. Heppell, J. Davis, B. Kingsley, V. Mueller, J. migration past the RST. However, it seems as likely that loca- Togstad, and N. Weber assisted with fish capture and PIT- tion of origin, rather than enhanced trap avoidance on a second tagging in the upper SFJD and tributaries. C. Jordan of Na- pass, influenced the difference in E between large-sized recent tional Oceanic and Atmospheric Administration, Fisheries Ser- releases and summer-tagged parr. The apparent lack of directed vice provided PIT tags. M. Huso of Oregon State University migration by large O. mykiss released upstream from the RST provided advice on statistical analyses. The work of I. Tattam, indicates the importance of RST location within the stream net- P.Bayley, H. Li, and G. Giannico was funded by the U.S. Bureau Downloaded by [Department Of Fisheries] at 22:44 28 February 2013 work. If possible, an RST should be located in a stream section of Reclamation, Pacific Northwest Region through M. Newsom. that is not continuously occupied by juvenile salmonids, so that The involvement of J. Ruzycki and other Oregon Department of only active migrants are captured. However, this may be impos- Fish and Wildlife employees was supported by the Bonneville sible in small subbasins such as the SFJD. Power Administration (Project Number 1998-016-00) through J. Baugher, J. Karnezis, and J. Swan. Reference to trade names Management Implications does not imply endorsement by the U.S. Geological Survey, Strategies A and B resulted in estimates of E that were lower Oregon Cooperative Fishery Research Unit, Oregon State Uni- than estimates of E from summer-tagged parr. The estimated E versity, or the Oregon Department of Fish and Wildlife. for small and medium size-groups in strategies A and B ranged from 28% to 34%. The E from strategy C in these same size- groups was 42–44%. Strategy C best mimicked the E of summer- tagged parr, which was 53% for the small–medium size-group. If REFERENCES Achord, S., R. W. Zabel, and B. P. Sandford. 2007. Migration timing, growth, E is underestimated, population abundance will, in turn, be over- and estimated parr-to-smolt survival rates of wild Snake River spring–summer estimated. An appropriate recent release strategy is crucial for Chinook Salmon from the Salmon River basin, Idaho, to the lower Snake population estimates and management. Our results suggest that River. Transactions of the American Fisheries Society 136:142–154. 244 TATTAM ET AL.

Bilton, H. T., D. F. Alderdice, and J. T. Schnute. 1982. Influence of time and sw/hydro near real time/display hydro graph.aspx?station nbr=14039500. size at release of juvenile Coho Salmon (Oncorhynchus kisutch) on returns at (May 2012). maturity. Canadian Journal of Fisheries and Aquatic Sciences 39:426–447. Prentice, E. F., T. A. Flagg, C. S. McCutcheon, D. F. Brastow, and D. C. Cross. Copeland, T., and D. A. Venditti. 2009. Contribution of three life history types 1990. Equipment, methods, and an automated data-entry station for PIT to smolt production in a Chinook Salmon (Oncorhynchus tshawytscha) pop- tagging. Pages 335–340 in N. C. Parker, A. E. Giorgi, R. C. Heidinger, D. B. ulation. Canadian Journal of Fisheries and Aquatic Sciences 66:1658–1665. Jester Jr., E. D. Prince, and G. A. Winans, editors. Fish-marking techniques. Cramer, S. P., D. Demko, C. Fleming, T. Loera, and D. Neeley. 1992. Effects American Fisheries Society, Symposium 7, Bethesda, Maryland. of pumping by Glenn-Colusa irrigation district on juvenile Chinook migrat- Pyper, B. J., F. J. Mueter, R. M. Peterman, D. J. Blackbourn, and C. C. Wood. ing down the Sacramento River. Annual Report to Glenn-Colusa Irrigation 2002. Spatial covariation in survival rates of northeast Pacific Chum Salmon. District, S. P. Cramer and Associates, Gresham, Oregon. Transactions of the American Fisheries Society 131:343–363. Dambacher, J. M. 1991. Distribution, abundance, and emigration of juvenile Ramsey, F. L., and D. W. Schafer. 2002. The statistical sleuth: a course in Steelhead (Oncorhynchus mykiss), and analysis of stream habitat in the methods of data analysis, 2nd edition. Duxbury, Pacific Grove, California. Steamboat Creek basin, Oregon. Master’s thesis. Oregon State University, Roper, B., and D. L. Scarnecchia. 1996. A comparison of trap efficiencies Corvallis. for wild and hatchery age-0 Chinook Salmon. North American Journal of Dare, M. R. 2003. Mortality and long-term retention of passive integrated Fisheries Management 16:214–217. transponder tags by spring Chinook Salmon. North American Journal of Roper, B., and D. L. Scarnecchia. 2000. Key strategies for estimating pop- Fisheries Management 23:1015–1019. ulation sizes of emigrating salmon smolts with a single trap. Rivers 7: Frith, H. R., T. C. Nelson, and C. J. Schwarz. 1995. Comparison of rotary 77–88. trap mark–recapture outmigration estimates with fence counts for Coho and Scace, J. G., B. H. Letcher, and J. Noreika. 2007. An efficient smolt trap for sandy Steelhead smolts in the Keogh River, 1995. Report to the British Columbia and debris-laden streams. North American Journal of Fisheries Management Ministry of Environment, Land and Parks, LGL Limited Environmental Re- 27:1276–1286. search Associates, Sidney, British Columbia. Schultz, T. L., W. H. Wilson, J. R. Ruzycki, R. Carmichael, J. Schricker, and D. P. Johnson, S. L., J. D. Rodgers, M. F. Solazzi, and T. E. Nickelson. 2005. Effects Bondurant. 2006. Escapement and productivity of spring Chinook Salmon and of an increase in large wood on abundance and survival of juvenile salmonids summer Steelhead in the John Day River basin. 2003–2004 Annual Technical (Oncorhynchus spp.) in an Oregon coastal stream. Canadian Journal of Fish- Report to the Bonneville Power Administration, Project 199801600, Portland, eries and Aquatic Sciences 62:412–424. Oregon. Madrinãn,´ L. F. 2008. Biophysical factors driving the distribution and abundance Seber, G. A. F. 1982. The estimation of animal abundance and related parame- of Redband/Steelhead Trout (Oncorhynchus mykiss gairdneri) in the South ters, 2nd edition. Griffin, London. Fork John Day River basin, Oregon, USA. Doctoral dissertation. Oregon State Solazzi, M. F., T. E. Nickelson, S. L. Johnson, and J. D. Rodgers. 2000. Effects University, Corvallis. Available: oasis.oregonstate.edu/record=b2478888. of increasing winter rearing habitat on abundance of salmonids in two coastal (May 2012). Oregon streams. Canadian Journal of Fisheries and Aquatic Sciences 57:906– Miller, B. A., J. D. Rodgers, and M. F. Solazzi. 2000. An automated device to 914. release marked juvenile fish for measuring trap efficiency. North American Thedinga, J. F., M. L. Murphy, S. W. Johnson, J. M. Lorenz, and K. V. Koski. Journal of Fisheries Management 20:284–287. 1994. Determination of salmonid smolt yield with rotary-screw traps in the Music, P. A., J. P. Hawkes, and M. S. Cooperman. 2010. Magnitude and causes Situk River, Alaska, to predict effects of glacial flooding. North American of smolt mortality in rotary screw traps: an Atlantic Salmon case study. North Journal of Fisheries Management 14:837–851. American Journal of Fisheries Management 30:713–722. Volkhardt, G. C., S. L. Johnson, B. A. Miller, T. E. Nickelson, and D. E. Seiler. Nehlsen, W., J. E. Williams, and J. A. Lichatowich. 1991. Pacific salmon at the 2007. Rotary screw traps and inclined plane screen traps. Pages 235–266 in crossroads: stocks at risk from California, Oregon, Idaho, and Washington. D. H. Johnson, B. M. Shrier, J. S. O’Neal, J. A. Knutzen, X. Augerot, T. A. Fisheries 16(2):4–21. O’Neil, and T. N. Pearsons, editors. Salmonid field protocols handbook: OWRD (Oregon Water Resources Department). 2012. Near real time techniques for assessing status and trends in salmon and trout populations. hydrographic data. OWRD, Salem. Available: apps.wrd.state.or.us/apps/ American Fisheries Society, Bethesda, Maryland. Downloaded by [Department Of Fisheries] at 22:44 28 February 2013 This article was downloaded by: [Department Of Fisheries] On: 28 February 2013, At: 22:45 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 Delayed Hooking Mortality of Blue Catfish Caught on Juglines Joseph D. Schmitt a & Daniel E. Shoup a a Department of Natural Resource Ecology and Management, Oklahoma State University, 008c Ag Hall, Stillwater, Oklahoma, 74077, USA Version of record first published: 19 Feb 2013.

To cite this article: Joseph D. Schmitt & Daniel E. Shoup (2013): Delayed Hooking Mortality of Blue Catfish Caught on Juglines, North American Journal of Fisheries Management, 33:2, 245-252 To link to this article: http://dx.doi.org/10.1080/02755947.2012.754805

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 33:245–252, 2013 C American Fisheries Society 2013 ISSN: 0275-5947 print / 1548-8675 online DOI: 10.1080/02755947.2012.754805

ARTICLE

Delayed Hooking Mortality of Blue Catﬁsh Caught on Juglines

Joseph D. Schmitt* and Daniel E. Shoup Department of Natural Resource Ecology and Management, Oklahoma State University, 008c Ag Hall, Stillwater, Oklahoma 74077, USA

Abstract Growing interest in catfish angling, particularly for trophy-sized fish, has resulted in new regulations that limit the harvestoflargeBlueCatfishIctalurus furcatus in several states. For these regulations to be effective, released fish must survive to further contribute to the fishery, either through reproduction or subsequent recapture. We investigated the effect of capture depth, hook type, water temperature, and fish size on the delayed hooking mortality of Blue Catfish caught on juglines. Blue Catfish (N = 559) were caught from three Oklahoma reservoirs with either 5/0 circle hooks or J-hooks fished for 24-h sets. One experimental fish (captured via jugline) and one control fish (captured via low-frequency electrofishing) were then placed in field enclosures (N = 25) and monitored for mortality after 72 h. Mean mortality was low at 8.50% (range, 0.00–37.50%; SE, 1.81%). Mortality decreased significantly with decreasing water temperatures (P < 0.01; odds ratio 1.1). Mortality was highest (mean = 25.31%) at water temperatures >20◦C and decreased to 3.89% in water temperatures <20◦C. We observed 0% mortality in water temperatures <14◦C. Hook type did not significantly affect mortality, nor did the depth in the water column where the fish was hooked. For every 100-mm increase in total length, fish were six times less likely to die (odds ratio 0.17). Mean mortality for preferred-size fish was low at 2.50%, and no mortalities were observed for memorable or trophy-size fish. These results suggest that length regulations limiting the harvest of preferred-size or larger fish should be effective as a large proportion of released fish should survive to further contribute to the fishery.

Angler harvest has the potential to alter population size struc- only irritate anglers rather than benefiting the fishery. However, ture and reduce the abundance of fishes (Gigliotti and Taylor with a better understanding of the conditions responsible for 1990; Beard and Essington 2000; Faust 2011). To prevent this, mortality, seasonal closures, gear restrictions, or other types managers often implement length restrictions and bag limits, of regulations could be effective management tools when high

Downloaded by [Department Of Fisheries] at 22:45 28 February 2013 which require anglers to release fish. These restrictions assume postrelease mortality occurs. that a large proportion of released fish, otherwise known as Angling for Blue Catfish has become increasingly popular “regulatory discards,” survive to further contribute to the fish- over the past two decades (Michaletz and Dillard 1999; Mauck ery, either through reproduction or subsequent recapture by an and Boxrucker 2006; Kuklinski and Boxrucker 2010; Makinster angler. Delayed hooking mortality of teleosts is highly variable and Paukert 2008), with many anglers being attracted to this and ranges from less than 1% (Dotson 1982; Childress 1989a; species due to its large size potential (Arterburn et al. 2002). Clapp and Clark 1989; Parks and Kraai 1991) to as high as Although historic accounts from the 1800s indicate that Blue 70–90% postrelease mortality (Martin et al. 1987; Childress Catfish had impressive growth potential (e.g., numerous ac- 1989b; May 1990; Siewert and Cave 1990). There have been no counts of 56–140 kg individuals), individuals over 50 kg are less peer-reviewed studies that adequately assess the delayed hook- common in modern times (Graham 1999; Mauck and Boxrucker ing mortality of Blue Catfish Ictalurus furcatus, and, if postre- 2006; Homer and Jennings 2011). This is most likely caused in lease mortality is high, bag limits or length regulations may part by the overharvest of larger individuals (Graham 1999).

*Corresponding author: [email protected] Received July 23, 2012; accepted November 28, 2012 245 246 SCHMITT AND SHOUP

While there has been a recent increase in catfish angling across of mortality (Schisler and Bergersen 1996). Due to prolonged the USA, most agencies have been slow to adopt trophy catfish hooking, posthooking mortality rates for catfish captured with management strategies. In fact, only 2% of the state agencies passive gears likely exceed those of fish captured with rod and surveyed by Arterburn et al. (2002) emphasized trophy man- reel (Muoneke and Childress 1994). If high posthooking mor- agement for any of the three largest North American catfishes tality rates exist for Blue Catfish caught on juglines, this could (Channel Catfish Ictalurus punctatus, Blue Catfish, or Flathead negate the intended benefit of bag restrictions, maximum length Catfish Pylodictus olivaris). This is in direct contrast to the 75% limits, or “one over” limits designed to increase survival of large of the anglers surveyed who favored the development of trophy fish, necessitating the implementation of other regulations to as- catfish fisheries (Arterburn et al. 2002). Therefore, many state sist in the development of trophy fisheries (i.e., gear or seasonal agencies have recently become interested in managing catfish restrictions). populations, but the lack of information on both the life history Whereas delayed hooking mortality has been well studied and general biology of Blue Catfish (Graham 1999; Mauck in species such as Largemouth Bass Micropterus salmoides and Boxrucker 2006) and Flathead Catfish (Makinster and (Rutledge 1978; Schramm et al. 1987; Kwak and Henry Paukert 2008) has been a hindrance. In particular, there is a 1995; Wilde 1998) and Atlantic Salmon Salmo salar (Warner need for information about the utility of length and harvest reg- and Johnson 1978; Warner 1979; Thorstad et al. 2003), very ulations (e.g., maximum length or “one over” length restrictions) little research has addressed delayed hooking mortality of to modify the size structure of Blue Catfish populations through catfishes in general and Blue Catfish are particularly poorly reduced mortality (harvest and discard mortality) of rare, large studied in this regard. This is especially problematic given individuals. the prolonged hooking times associated with the passive gears The growth of Blue Catfish in southern reservoirs can be commonly used to capture catfishes (i.e., trotlines, limb lines, poor (Graham 1999; Mauck and Boxrucker 2006; Boxrucker and juglines). These prolonged hooking times suggest these and Kuklinski 2008), especially when compared with growth in fishes may have higher posthooking mortality rates than other lotic systems (Jolley and Irwin 2011; Rypel 2011). Poor growth species that have been studied. Studies evaluating delayed in reservoirs is often attributed to increases in fish densities, hooking mortality with Channel Catfish or Flathead Catfish which leads to greater intraspecific competition and ultimately have found mortality was variable but could be as high as 50% results in resource limitation (Conder and Hoffarth 1965; Freeze under some conditions (Muoneke 1993; Ott and Storey 1993). 1977). This pattern of poor growth in lentic systems has be- Only one study has measured delayed hooking mortality for come evident in several of Oklahoma’s large impoundments. Blue Catfish (Muoneke 1993). The researcher found a mean In a recent survey of nine major Oklahoma reservoirs, only postrelease mortality of 5.1% for Blue Catfish after a 72-h 0.7% of Blue Catfish sampled were preferred size or larger assessment period; however, that study was conducted on only (762 mm total length [TL]; Gabelhouse 1984) and it took fish, one reservoir, did not standardize hook type or fishing duration, on average, 13–16 years to reach preferred size (Boxrucker and had limited replication (only four sampling dates and 35 or 47 Kuklinski 2008). Additionally, growth of Blue Catfish in reser- total fish for winter and summer trials, respectively), and lacked voirs is highly variable (Boxrucker and Kuklinski 2008), which control fish (Muoneke 1993). Therefore, additional research is may suggest that certain fish are genetically predisposed to grow needed to determine the delayed hooking mortality of catfishes quickly and these fast growing fish are more likely to survive caught with passive fishing gears, particularly those gears with to reach trophy size. By removing these large fish from the prolonged hooking durations. gene pool, fishermen may be artificially selecting for slower- While one study exists evaluating the postrelease mortality growing, smaller fish. Due to limited numbers of large fish, of Blue Catfish captured with trotlines, there is no research Downloaded by [Department Of Fisheries] at 22:45 28 February 2013 several state agencies have recently enacted regulations limiting addressing the postrelease mortality of fish captured with jug- the harvest of large Blue Catfish (Dorsey et al. 2011; Eder 2011; lines. Although the previous trotline study had a prolonged Kuklinski and Patterson 2011; CMTC 2012) in an effort to hooking component, it may not accurately reflect mortality for increase survival of these rare individuals. However, little is fish caught on juglines. Trotlines are generally set horizontally, known concerning the delayed hooking mortality of Blue Cat- whereas juglines are set vertically. Trotlines can be set on the fish, and these types of regulations will only be successful if surface or on the bottom, but anchored juglines will always be a substantial proportion of the regulatory discards survive to in contact with the bottom. This often results in anglers setting further contribute to the fishery. some hooks below the thermocline in hypoxic conditions, pre- Trotlines and juglines are commonly used by catfish anglers sumably increasing mortality of fish that become hooked there and commercial fishermen in the southern United States (White (juglines are often anchored in position with several kilograms 1956; Kuklinski and Boxrucker 2010). Many states only require of weight). Additionally, juglines are quite effective in deep trotlines and juglines to be checked once daily, so fish captured water, particularly during the winter months, and fish captured by these methods may be hooked for up to 24 h. Prolonged from greater depths generally exhibit greater postrelease mortal- hooking can increase stress (Tomasso et al. 1996), lead to poor ity because of rapid depressurization (Muoneke and Childress condition (Thorstad et al. 2003), and increase the probability 1994). To quantify the delayed hooking mortality of Blue Catfish HOOKING MORTALITY OF BLUE CATFISH 247

caught on juglines, field trials were conducted seasonally at three 24 h, which is the maximum time allowed by Oklahoma Depart- different Oklahoma reservoirs over a 2-year period. The objec- ment of Wildlife Conservation. Hooks were removed from all tives were to address the effects of hook type, water temperature, captured catfish after retrieving the jug, even though removing capture depth, and anatomical hooking location on the delayed the hook from a deeply hooked fish can increase the probabil- hooking mortality of Blue Catfish. This information is needed to ity of mortality (Mason and Hunt 1967; Warner and Johnson determine the effectiveness of mandatory release policies used 1978; Warner 1979; Weidlein 1989; Vincent-Lang et al. 1993, for the conservation of large Blue Catfish. Aalbers et al. 2004). This was done based on the anecdotal observation that most jug anglers retrieve their hooks. Hooks were METHODS removed by hand or with pliers, and, after the removing hook, From May 2010 to March 2012, Blue Catfish were captured we recorded total length, anatomical hooking location, capture using weighted juglines at Kaw (6,896 ha), Keystone (9,550 ha), depth, water temperature, and hook type for each fish. Follow- and Sooner (2,182 ha) lakes in northeastern Oklahoma, which ing hook removal, fish were marked by clipping the soft portion are all man-made impoundments. Trials were conducted sea- of the left pectoral fin and placed into individual field enclo- sonally at all reservoirs so that mortality estimates could be sures (square, 1.25 m on each side) constructed of galvanized, made across the full range of water temperatures. More effort 12.5-gauge fencing with 51-mm × 102-mm openings. During was directed towards warmer water temperatures because the sampling dates when thermal stratification created the potential probability of mortality typically increases with water temper- for hypoxic hypolimnetic conditions, DO was monitored us- ature (Muoneke and Childress 1994), but jugs were deployed ing a YSI 556 multi-probe system (Yellow Springs Instruments, on at least five dates in each reservoir below the median water Yellow Springs, Ohio) and cages were never suspended in areas temperature observed in our study (≤16◦C). All juglines were with less than 5.8 mg/L DO, which is well above the minimum baited with freshly killed Gizzard Shad Dorosoma cepedianum, DO suggested for the culture of Channel Catfish (4.0 mg/L; Common Carp Cyprinus carpio,orbuffaloIctiobus spp., de- Tucker 1991). pending upon bait availability. All hooks were baited with the After retrieving all of the jugs, boat-mounted electrofishing same bait type during each replicate, and all bait types were equipment (either Smith-Root Model 7.5 GPP or 5.0 GPP; 15 used proportionally across all reservoirs and seasons. pulses per second DC; ∼4 amps) was used to collect control fish, Juglines were constructed using a 3.79-L jug for floatation which served to identify any mortality related to cage confine- and a 2.3-kg anchor made from QuickCrete with 295-lb-test ment or other unforeseen causes. One control fish (marked by seine twine (#30) for the main line (between jug and anchor) clipping the soft portion of the right pectoral fin) was placed with and 126-lb seine twine (#15) for the dropper lines (lines to which the jugline-captured fish in each field enclosure and mortality hooks are attached). Three dropper lines were attached to the was quantified after 72 h. Because Blue Catfish can be canni- main line at intervals of ≥1 m using trotline clips. To quantify balistic (Schloesser et al. 2011), control and treatment fish in mortality of fish that were hooked in the hypolimnion, where each enclosure were of similar size (−±25% in length) to prevent hypoxic to anoxic conditions can exist, vertical dissolved oxy- predation. To determine if a significant amount of mortality oc- gen (DO) profiles (1-m intervals) were measured using a YSI curs after the initial observation period, 56 fish were observed 556 multi-probe system (Yellow Springs Instruments, Yellow at both 72 h and 7 d. These trials were conducted during the Springs, Ohio) during the summer of 2010. This information winter and summer at both Kaw and Sooner lakes. Whereas was used to ensure two hooks were set above the thermocline most posthooking mortality is expected to occur in the first 24 h and one hook was set below it during times of the year when (Muoneke 1992; Muoneke and Childress 1994; Schill 1996), hypoxic hypolimnetic conditions existed. No trials were con- we used a 72-h mortality evaluation period, because a substan- Downloaded by [Department Of Fisheries] at 22:45 28 February 2013 ducted below the thermocline on Sooner Lake because the main tial amount of mortality can sometimes occur after several days lake did not stratify during our study. No hooks were set in the (Grover et al. 2002). A 72-h observation period is supported by hypolimnion in 2011 because after 11,680 hook-hours of fishing the findings of several mortality studies (Warner and Johnson effort in 2010, no Blue Catfish had been captured. 1978; Ott and Storey 1993; Stunz and McKee 2006; James et al. Dropper lines were equipped with either 5/0 Mustad Biggun’ 2007). J-style hooks or 5/0 Diachii Circle-Chunk Light hooks, both We modeled the binary probability (1 = dead, 0 = alive) of of which had nonoffset points. To quantify the effect of hook survival using Proc Glimmix in SAS 9.3 with a logit link func- type on mortality, equal numbers of jugs with 5/0 J-hooks and tion and a binary probability distribution (SAS 2011). Separate 5/0 circle hooks were set in the same locations during each repeated-measures analyses, where reservoirs were treated as collection period (all three hooks on a given jugline were the subjects, were used to test for significant differences in the mor- same type). In order to intersperse the two hook styles and tality rates for each explanatory variable (water temperature, reduce bias, the juglines were deployed in series that alternated hook type, fish size, capture depth, and anatomical hooking by hook type. location). Hook type, fish size (100-mm groupings), anatomi- We tried to imitate the behaviors of typical jug fishermen as cal hooking location, and depth (above or below the thermo- much as possible throughout this study, so jugs were fished for cline) were treated as categorical variables; water temperature 248 SCHMITT AND SHOUP

FIGURE 1. Length frequencies of 559 Blue Catﬁsh caught on juglines in Kaw, Keystone, and Sooner lakes, Oklahoma, 2010–2012.

was treated as a continuous variable. For significant tests with FIGURE 2. Relationship between temperature and 72-h posthooking mortal- categorical variables that were unordered (i.e., hook type and ity of Blue Catfish caught on juglines at Kaw, Keystone, and Sooner lakes, anatomical hooking location), a Tukey’s post hoc test was used Oklahoma, 2010–2012. to test all pairwise combinations. Odds ratios were used to describe differences between levels of continuous and ordered 100-mm increase in TL, Blue Catfish were about six times categorical variables. Significance was assessed with α = 0.05 less likely to die (odds ratio 0.16; Figure 3). Mortality of fish for all analyses. >762 mm (preferred size) was low at 2.4% (N = 83). No mortalities were observed in memorable or trophy-sized fish (N = 16). RESULTS Hooking location also influenced mortality (Figure 4). The Combining trials from Kaw, Keystone, and Sooner lakes, majority of fish (72.2%) were hooked in the corner of the mouth, 97,200 hook-hours of total effort were expended from May though many of the smaller fish also had damage to the eye due 2010 to March 2012. Water temperatures ranged from 2.30◦C ◦ to its close proximity to the mouth. These “eye-hooked” fish to 31.60 C. A total of 559 Blue Catfish were captured, ranging were categorized separately and accounted for 20.6% of the fish in size from 310 mm to 1,238 mm TL (Figure 1). Jugs were de- caught. Fish that were hooked externally, usually in the cheek, ployed on 54 separate occasions. Hooking mortality rates were the bottom of the jaw, or the opercula, accounted for 4.79% of the = similar across reservoirs so reservoir data were pooled (Kaw fish captured. Only 15 fish (2.4%) were hooked in the stomach = = = = 11.0%, Keystone 9.9%, Sooner 6.4%; F2,511 0.65, P or esophagus. Mortality among all hooking locations was sig- 0.53). Mean overall mortality for hooked fish was low at 8.54% nificantly different, with the exception of externally hooked fish (range = 0.00–37.50%, SE = 1.81%) and subtracting mean control mortality (1.64%) from mean hooking mortality results in an overall adjusted mortality of 6.9%. Of the 56 fish that were observed for mortality for 7 d, no mortalities were observed after Downloaded by [Department Of Fisheries] at 22:45 28 February 2013 72 h, suggesting that 72 h was sufficient for observing delayed hooking mortality in Blue Catfish. A total of 11,680 hook-hours of effort were expended beneath the thermocline. Depth (above or below the thermocline) was not a significant driver of Blue Catfish mortality, as we found Blue Catfish are not frequently captured in the hypolimnion. Whereas Channel Catfish were caught occasionally (N = 6; three of which were dead upon retrieval of the jug), no Blue Catfish were ever captured below the thermocline. Mortality was significantly related to several of the tested factors. Mortality rates increased with increasing water temper- ◦ ature (F1,501 = 12.56, P < 0.01), and for every 1 C increase in temperature, Blue Catfish were 1.1 times more likely to die FIGURE 3. Seventy-two-hour mortality rate as a function of total length (100- (odds ratio 1.1; Figure 2). Mortality also declined significantly mm length classes) for Blue Catfish caught on juglines at Kaw, Keystone, and with increasing fish size (F7,10 = 4.58, P < 0.02). For every Sooner lakes, Oklahoma, 2010–2012. HOOKING MORTALITY OF BLUE CATFISH 249

were individuals <356 mm TL. Additionally, this inverse relationship between increasing fish size and mortality was found in Lake Trout Salvelinus namaycush (Loftus et al. 1988), Rainbow Trout Oncorhynchus mykiss (Schisler and Bergersen 1996), and Dusky Shark Carcharhinus obscurus (Romine et al. 2009), yet the mechanism causing this pattern is unclear. Damage caused by 5/0 hooks may be insignificant to large fish but proportionally more severe to small fish, as there is less space between the mouth and important organs, such as the eye or the esophagus, with decreased body size. Younger, smaller fish have less developed immune systems (Tatner 1986; Tatner 1997), presumably making them more susceptible to posthooking bacterial or viral infections. The low mortality rates observed in memorable and trophy-size fish are noteworthy for government agencies trying to manage for trophy fisheries, as our findings imply that managers can effectively use maximum length limits or restricted bag limits to conserve large Blue Catfish. Mortality rates decreased significantly with decreasing temperature and, at temperatures cooler than 15◦C, posthooking mortality rates were similar to control mortality rates. This pattern of reduced mortality at cooler water temperatures is consistent with previous studies of Brook Trout Salvelinus fontinalis (Dotson 1982), Cutthroat Trout Oncorhynchus clarkii (Marnell and Hunsaker 1970), Largemouth Bass (Rutledge 1975), Striped Bass Morone saxatilis (Childress 1989a), Tiger Muskellunge (Muskellunge Esox masquinongy × Northern Pike E. lucius; Newman and Storck 1986), and Spotted Seatrout Cynoscion FIGURE 4. Total catch (top panel) and 72-h mortality rate (bottom panel) of nebulosus (Matlock and Dailey 1981). Lower occurrences of Blue Catfish hooked in different anatomical locations by juglines at Kaw, Key- mortality at cooler water temperatures likely result from de- stone, and Sooner lakes, Oklahoma, 2010–2012. Variables with nonsignificant creases in metabolic rate, activity level, and bacterial concentra- differences share a letter; P > 0.05. tions (Muoneke and Childress 1994); however, the exact mechanism causing this pattern in Blue Catfish is unknown. Trophy and eye-hooked fish (P = 0.28). Mortality was relatively high Blue Catfish are most susceptible to capture during the win- for esophagus-hooked fish (91%) and externally hooked fish ter months (Kuklinski and Boxrucker 2010), which is when we (20%) and was low for all other anatomical locations (<10%). observed the lowest mortality rates. This further suggests that Circle hooks accounted for 75.5% of the total catch, despite bag limits and maximum length limits should be particularly equal effort with both hook types. Mortality rates for circle effective for conserving large fish (>762 mm TL). hooks and J hooks were not statistically different (F = 0.24, 1,510 Esophagus hooking was rare in this study but usually resulted P = 0.62). Circle hooks did not always set in the corner of the in mortality (90.9% mortality). High mortality of esophagus-

Downloaded by [Department Of Fisheries] at 22:45 28 February 2013 mouth, as they are designed to, which resulted in several fish hooked fish is common with other fish species, especially (3.8%) being hooked externally or deep in the esophagus. when the hook is removed (Diggles and Ernst 1997; Butcher et al. 2006). Circle hooks can significantly reduce instances DISCUSSION of deep hooking in billfish (Sailfish Istiophorus platypterus The mortality of Blue Catfish caught on juglines was very and Blue Marlin Makaira nigricans; Prince et al. 2002) low, particularly at water temperatures less than 20◦C. Man- and can reduce mortality in several other species (Channel agers should be able to effectively use length limits or bag Catfish, Ott and Storey 1993; Striped Bass, Caruso 2000; Coho limits because regulatory discards have a high probability of Salmon Oncorhynchus kisutch, McNair 1997; Chinook Salmon survival. However, mortality was significantly related to several O. tshawytscha, Grover et al. 2002; and Bluefin Tuna Thunnus of the tested factors. Our findings suggest that mortality could thynnus, Skomal et al. 2002). This does not appear to be the case be problematic for some groups of fish, particularly smaller fish with Blue Catfish, as 40% of the deeply hooked Blue Catfish in (i.e., <450 mm) captured in water temperatures ≥25◦C. this study were captured with circle hooks and there was no dif- Mortality decreased with increasing fish size. Higher mortal- ference in mortality rates between the two hook types. The deep ity in smaller fish is consistent with Muoneke (1993), who found hooking of Blue Catfish with circle hooks may be related to feed- that 75% of the mortalities of Blue Catfish caught on trotlines ing behavior. While pelagic piscivores like tuna are constantly 250 SCHMITT AND SHOUP

chasing their prey, Blue Catfish are more opportunistic and often captured on offset circle and J-type hooks. North American Journal of Fish- scavenge on wounded and dead prey (Graham 1999). The seden- eries Management 24:793–800. tary feeding behavior of Blue Catfish may give them time to Arterburn, J. E., D. J. Kirby, and C. R. Berry Jr. 2002. A survey of angler attitudes and biologist opinions regarding trophy catfish and their management. swallow the bait before they swim off and could explain the deep Fisheries 27(5):10–21. hooking with circle hooks. Consequently, restrictions concern- Beard, T. D., Jr., and T. E. Essington. 2000. Effects of angling and life history ing hook type do not appear to be an effective management tool processes on Bluegill size structure: insights from an individual-based model. for this species. However, deeply hooked fish were extremely Transactions of the American Fisheries Society 129:561–568. rare, so gear restrictions pertaining to hook type are unnecessary. Boxrucker, J., and K. Kuklinski. 2008. Abundance, growth, and mortality of selected Oklahoma Blue Catfish populations: implications for management Although Blue Catfish may feed in the hypolimnion un- of trophy fisheries. Proceedings of the Annual Conference Southeastern As- der certain conditions, no fish were captured beneath the ther- sociation of Fish and Wildlife Agencies 60(2006):152–156. mocline in this study. Blue Catfish may be more sensitive to Butcher, P. A., M. K. Broadhurst, and C. P. Brand. 2006. Mortality of Sand DO than Channel Catfish because they surface before Channel Whiting (Sillago ciliata) released by recreational anglers in an Australian Catfish in fish kills caused by depleted oxygen levels (Graham estuary. ICES Journal of Marine Science 63:567–571. Caruso, P. G. 2000. A comparison of catch and release mortality and wound- 1999). Additionally, Grist (2002) found that Blue Catfish in ing for Striped Bass (Morone saxatilis), captured with two baited hook Norman Lake, North Carolina, show a distinct preference for types. Massachusetts Division of Marine Fisheries, Federal Aid in Sport areas with much higher DO concentrations than the lake mean, Fish Restoration, Project F-57-R-12, Boston. particularly during the summer months, and were rarely found Childress, W. M. 1989a. Hooking mortality of White Bass, Striped Bass, White × in areas with DO concentrations less than 7.0 mg/L. Bass Striped hybrid Bass and Red Drum. Texas Parks and Wildlife De- partment, Federal Aid in Sport Fish Restoration, Project F-31-R-15, Final Our results demonstrate that the delayed hooking mortal- Report, Austin. ity of Blue Catfish captured on juglines is low. This suggests Childress, W. M. 1989b. Catch-and-release mortality of White and Black that management agencies can effectively use harvest restric- Crappie. Pages 175–186 in R. A. Barnhart and T. D. Roelofs, editors. Catch- tions to reduce the mortality of Blue Catfish. In locations where and-release fishing—a decade of experience: a national sport fishing sym- density-dependent growth exists, management agencies may posium. California Cooperative Fisheries Research Unit, Humboldt State University, Arcata, California. benefit from restricting the harvest of large fish while encourag- Clapp, D. F., and R. D. Clark Jr. 1989. Hooking mortality of Smallmouth Bass ing the liberal harvest of smaller fish. The removal of small fish caught on live minnows and artificial spinners. North American Journal of from a system will reduce intraspecific competition and, with Fisheries Management 9:81–85. enough removal, improve growth rates and relative weights of CMTC (Catfish Management Technical Committee). 2012. State catfish reg- the remaining fish. Growth of Blue Catfish in Oklahoma reser- ulations table. American Fisheries Society, Southern Division, Bethesda, Maryland. Available: www.sdafs.org/catfish/. (November 2012). voirs is highly variable (Boxrucker and Kuklinski 2008), which Conder, J. R., and R. Hoffarth. 1965. Growth of Channel Catfish, Ictalurus may also indicate that only certain fish are genetically capable punctatus, and Blue Catfish, Ictalurus furcatus, in the Kentucky Lake portion of growing to trophy size. Therefore, the overharvest of large of the Tennessee River in Tennessee. Proceedings of the Annual Confer- individuals in a given system could ultimately eliminate that sys- ence Southeastern Association of Game and Fish Commissioners 16(1962): tem’s potential to produce trophy fish. This further emphasizes 348–354. Diggles, B. K., and I. Ernst. 1997. Hooking mortality of two species of shallow- the need to conserve large fish, and maximum length limits may water reef fish caught by recreational angling methods. Marine and Freshwater be an effective means of doing so. However, these approaches Research 48:479–483. will only be effective if regulatory discards survive to contribute Dorsey, L. G., B. J. McRae, and T. M. Thompson. 2011. Evaluation of an 813- their genetics to subsequent cohorts or to provide recreational mm maximum size limit for Blue Catfish in two North Carolina reservoirs. opportunity to anglers in the future. Pages 177–185 in P.H. Michaletz and V.H. Travnichek, editors. Conservation, ecology, and management of catfish: the second international symposium. American Fisheries Society, Symposium 77, Bethesda, Maryland. Downloaded by [Department Of Fisheries] at 22:45 28 February 2013 Dotson, T. 1982. Mortalities in trout caused by gear type and angler-induced ACKNOWLEDGMENTS stress. North American Journal of Fisheries Management 2:60–65. We thank Coleman Lucado, Suzie Schmitt, John Smith, Eder, S. 2011. Missouri’s catfish: a history of utilization, management, and Candice Miller, and Garrett Steinbrook for their help in col- culture. Pages 55–70 in P. H. Michaletz and V. H. Travnichek, editors. Con- lecting field data. We also thank Kurt Schmitt for his technical servation, ecology, and management of catfish: the second international symposium. American Fisheries Society, Symposium 77, Bethesda, Maryland. assistance with the electrofishing equipment. We thank James Faust, M. D. 2011. Effect of harvest mortality on muskellunge size structure in Long and Michael Tobler for reviewing earlier versions of this Wisconsin’s ceded territory. Master’s thesis. University of Wisconsin, Stevens manuscript. Funding for this project was provided through Fed- Point. eral Aid in Sport Fish Restoration Project F-85-R, administered Freeze, T. M. 1977. Age and growth, condition, and length-weight relationships by the Oklahoma Cooperative Fish and Wildlife Research Unit. of Ictalurus furcatus and Ictalurus punctatus from Barkley and Kentucky lakes, Kentucky. Master’s thesis. Murray State University, Murray, Kentucky. Gabelhouse, D. W., Jr. 1984. A length-categorization system to assess fish stocks. North American Journal of Fisheries Management 4:273–285. REFERENCES Gigliotti, L. M., and W. W. Taylor. 1990. The effect of illegal harvest on Aalbers, S. A., G. M. Stutzer, and M. A. Drawbridge. 2004. The effects of catch- recreational fisheries. North American Journal of Fisheries Management 10: and-release angling on the growth and survival of juvenile White Seabass 106–110. HOOKING MORTALITY OF BLUE CATFISH 251

Graham, K. 1999. A review of the biology and management of Blue Catfish. ceedings of the Annual Conference Southeastern Association of Fish and Pages 37–49 in E. R. Irwin, W. A. Hubert, C. F. Rabeni, H. L. Schramm Wildlife Agencies 58(2004):57–65. Jr., and T. Coon, editors. Catfish 2000: proceedings of the international ic- May, B. E. 1990. Evaluation of angler induced mortality of Striped Bass in talurid symposium. American Fisheries Society, Symposium 24, Bethesda, Maryland. National Marine Fisheries Service, Completion Report PL 89- Maryland. 304, AFC-18-1, Gloucester, Massachusetts. Grist, J. D. 2002. Analysis of a Blue Catfish population in a southeastern reser- McNair, D. 1997. Effect of gear and methods on marine sports salmon voir: Lake Norman, North Carolina. Master’s thesis. Virginia Polytechnic hooking mortality rates. Pacific Fishery Management Council, Northwest Institute and State University, Blacksburg. Emergency Assistance Program, Final Research Report 68:2, Portland, Grover, A. M., M. S. Mohr, and M. L. Palmer-Zwahlen. 2002. Hook-and- Oregon. release mortality of Chinook Salmon from drift mooching with circle hooks: Michaletz, P. H., and J. G. Dillard. 1999. A survey of catfish management in the management implications for California’s ocean sport fishery. Pages 39–56 United States and Canada. Fisheries 24(8):6–11. in J. A. Lucy and A. L. Studholme, editors. Catch and release in marine Muoneke, M. I. 1992. Seasonal hooking mortality of Bluegills caught recreational fisheries. American Fisheries Society, Symposium 30, Bethesda, on natural baits. North American Journal of Fisheries Management 12: Maryland. 645–649. Homer, M. D., Jr., and C. A. Jennings. 2011. Historical catch, age structure, Muoneke, M. I. 1993. Seasonal hooking mortality of Flathead Catfish and Blue sizes, and relative growth for an introduced population of Blue Catfish in Lake Catfish. Proceedings of the Annual Conference Southeastern Association of Oconee, Georgia. Pages 383–394 in P. H. Michaletz and V. H. Travnichek, Fish and Wildlife Agencies 45(1991):392–398. editors. Conservation, ecology, and management of catfish: the second inter- Muoneke, M. I., and W. M. Childress. 1994. Hooking mortality: a review for national symposium. American Fisheries Society, Symposium 77, Bethesda, recreational fisheries. Reviews in Fisheries Science 2:123–156. Maryland. Newman, D. L., and T. W. Storck. 1986. Angler catch, growth, and hooking James, J. T., G. W. Stunz, D. A. McKee, and R. R. Vega. 2007. Catch-and-release mortality of Tiger Muskellunge in small centrarchid-dominated impound- mortality of Spotted Seatrout in Texas: effects of tournaments, seasonality, ments. Pages 346–351 in G. E. Hall, editor. Managing muskies: a treatise and anatomical hooking location. North American Journal of Fisheries Man- on the biology and propagation of muskellunge in North America. American agement 27:900–907. Fisheries Society, Special Publication 15, Bethesda, Maryland. Jolley, J. C., and E. R. Irwin. 2011. Catfish population characteristics in tail- Ott, R. A., Jr., and K. W. Storey. 1993. Hooking mortality of Channel Cat- water and reservoir habitats of the Coosa River, Alabama. Pages 155–166 in fish caught by trotline. Proceedings of the Annual Conference Southeastern P. H. Michaletz and V. H. Travnichek, editors. Conservation, ecology, and Association of Fish and Wildlife Agencies 45(1991):399–406. management of catfish: the second international symposium. American Fish- Parks, J. O., and J. E. Kraai. 1991. Walleye hooking mortality at Lake Meredith. eries Society, Symposium 77, Bethesda, Maryland. Texas Parks and Wildlife Department, Fisheries Division, Management Data Kuklinski, K. E., and J. Boxrucker. 2010. Catfish angling and harvest statistics Series 52, Austin. with an emphasis on trophy Blue Catfish management in Oklahoma. Proceed- Prince, E. D., M. Ortiz, and A. Venizelos. 2002. A comparison of circle hook and ings of the Annual Conference Southeastern Association of Fish and Wildlife “J” hook performance in recreational catch-and-release fisheries for Billfish. Agencies 62(2008):149–153. Pages 66–79 in J. A. Lucy and A. L. Studholme, editors. Catch and release Kuklinski, K. E., and C. P. Patterson. 2011. Development of a Blue Catfish in marine recreational fisheries. American Fisheries Society, Symposium 30, management program in Oklahoma: a case history. Pages 71–79 in P. H. Bethesda, Maryland. Michaletz and V. H. Travnichek, editors. Conservation, ecology, and man- Romine, J. G., J. A. Musick, and G. H. Burgess. 2009. Demographic analyses agement of catfish: the second international symposium. American Fisheries of the Dusky Shark, Carcharhinus obscurus, in the northwest Atlantic incor- Society, Symposium 77, Bethesda, Maryland. porating hooking mortality estimates and revised reproductive parameters. Kwak, T. J., and M. G. Henry. 1995. Largemouth Bass mortality and related Environmental Biology of Fishes 84:277–289. causal factors during live-release fishing tournaments on a large Minnesota Rutledge, W. P. 1975. Hooking mortality study. Texas Parks and Wildlife De- lake. North American Journal of Fisheries Management 15:621–630. partment, Federal Aid in Fish Restoration, Project F-31-R1, Final Report, Loftus, A. J., W. W. Taylor, and M. Keller. 1988. An evaluation of Austin. Lake Trout (Salvelinus namaycush) hooking mortality in the upper Rutledge, W. P. 1978. Hooking mortality of Largemouth Bass captured Great Lakes. Canadian Journal of Fisheries and Aquatic Sciences 45: by artificial lures and natural bait. Texas Parks and Wildlife Depart- 1473–1479. ment, Federal Aid in Fish Restoration, Project F-33-R1, Final Report, Makinster, A. S., and C. P. Paukert. 2008. Effects and utility of minimum Austin. length limits and mortality caps for Flathead Catfish in discrete reaches of Rypel, A. L. 2011. Meta-analysis of growth for five North American catfishes: Downloaded by [Department Of Fisheries] at 22:45 28 February 2013 a large prairie river. North American Journal of Fisheries Management 28: effects of climate, hydrologic habitat, and latitudinal countergradients. Pages 97–108. 661–677 in P. H. Michaletz and V. H. Travnichek, editors. Conservation, Marnell, L. F., and D. Hunsaker II. 1970. Hooking mortality of lure-caught ecology, and management of catfish: the second international symposium. Cutthroat Trout (Salmo clarki) in relation to water temperature, fatigue, and American Fisheries Society, Symposium 77, Bethesda, Maryland. reproductive maturity of released fish. Transactions of the American Fisheries SAS (Statistical Analysis Systems). 2011. SAS/STAT 9.3 user’s guide. SAS Society 99:684–688. Institute, Cary, North Carolina. Martin, J. H., K. W. Rice, and L. W. McEachon. 1987. Survival of three fishes Schill, D. J. 1996. Hooking mortality of bait-caught Rainbow Trout in an Idaho caught on trotlines. Texas Parks and Wildlife Department, Coastal Fisheries trout stream and a hatchery: implications for special-regulation management. Branch, Management Data Series 111, Austin. North American Journal of Fisheries Management 16:348–356. Mason, J. W., and R. L. Hunt. 1967. Mortality rates of deeply hooked Rainbow Schisler, G. J., and E. P. Bergersen. 1996. Postrelease hooking mortality of Trout. Progressive Fish-Culturist 29:87–91. Rainbow Trout caught on scented artificial baits. North American Journal of Matlock, G. C., and J. A. Dailey. 1981. Survival of hook-caught Spotted Seatrout Fisheries Management 16:570–578. held in cages. Texas Park and Wildlife Department, Coastal Fisheries Branch, Schloesser, R. W., M. C. Fabrizio, R. J. Latour, G. C. Garman, B. Greenlee, Management Data Series 15, Austin. M. Groves, and J. Gartland. 2011. Ecological role of Blue Catfish in Chesa- Mauck, P., and J. Boxrucker. 2006. Abundance, growth, and mortality of the peake Bay communities and implications for management. Pages 369–382 Lake Texoma Blue Catfish population: implications for management. Pro- in P. H. Michaletz and V. H. Travnichek, editors. Conservation, ecology, and 252 SCHMITT AND SHOUP

management of catfish: the second international symposium. American Fish- Tomasso, A. O., J. J. Isely, and J. R. Tomasso Jr. 1996. Physiological re- eries Society, Symposium 77, Bethesda, Maryland. sponses and mortality of Striped Bass angled in freshwater. Transactions of the Schramm, H. L., Jr., P. J. Haydt, and K. M. Portier. 1987. Evaluation of pre- American Fisheries Society 125:321–325. release, postrelease, and total mortality of Largemouth Bass caught during Tucker, C. S. 1991. Water quantity and quality requirements for Channel tournaments in two Florida lakes. North American Journal of Fisheries Man- Catfish hatcheries. Southern Regional Aquaculture Center, Publication 461, agement 7:394–402. Stoneville, Mississippi. Siewert, H. F., and J. B. Cave. 1990. Survival of released Bluegill, Lepomis Vincent-Lang, D., M. Alexandersdottir, and D. McBride. 1993. Mortality of macrochirus, caught on artificial flies, worms, and spinner lures. Journal of Coho Salmon caught and released using sport tackle in the Little Susitna Freshwater Ecology 5:407–411. River, Alaska. Fisheries Research 15:339–356. Skomal, G. B., B. C. Chase, and E. D. Prince. 2002. A comparison of circle Warner, K. 1979. Mortality of landlocked Atlantic Salmon hooked on four types hook and straight hook performance in recreational fisheries for juvenile of fishing gear at the hatchery. Progressive Fish-Culturist 41:99–102. Atlantic Bluefin Tuna. Pages 57–65 in J. A. Lucy and A. L. Studholme, Warner, K., and P. R. Johnson. 1978. Mortality of landlocked Atlantic Salmon editors. Catch and release in marine recreational fisheries. American Fisheries (Salmo salar) hooked on flies and worms in a river nursery area. Transactions Society, Symposium 30, Bethesda, Maryland. of the American Fisheries Society 107:772–775. Stunz, G. W., and D. A. McKee. 2006. Catch-and-release mortality of Spotted Weidlein, W. D. 1989. Mortality of released sublegal-sized Smallmouth Bass, Seatrout in Texas. North American Journal of Fisheries Management 26:843– catch and release implications. Pages 217–228 in R. A. Barnhart and 848. T. D. Roelofs, editors. Catch-and-release fishing—a decade of experience: a Tatner, M. F. 1986. The ontogeny of humoral immunity in Rainbow Trout, salmo national sport fishing symposium. California Cooperative Fisheries Research gairdneri. Veterinary Immunology and Immunopathology 12:93–105. Unit, Humboldt State University, Arcata, California. Tatner, M. F. 1997. Natural changes in the immune system of fish. Pages 255–287 White, C. E., Jr. 1956. Fish catches with various types of commercial fishing in G. Iwama and T. Nakanishi, editors. The fish immune system: organism, gear used in T.V.A. lakes from June, 1954 through January, 1955. Proceed- pathogen, and environment. Academic Press, San Diego, California. ings of the Annual Conference Southeastern Association of Game and Fish Thorstad, E. B., T. F. Næsje, P. Fiske, and B. Finstad. 2003. Effects of hook Commissioners 9(1955):80–86. and release on Atlantic Salmon in the River Alta, northern Norway. Fisheries Wilde, G. R. 1998. Tournament-associated mortality in Black Bass. Fisheries Research 60:293–307. 23(10):12–22. Downloaded by [Department Of Fisheries] at 22:45 28 February 2013 This article was downloaded by: [Department Of Fisheries] On: 28 February 2013, At: 22:45 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 Unbiased Methods for Calculating Mortality in Mark- Selective Fisheries Models for Ocean Salmon Robert Conrad a , Angelika Hagen-Breaux b & Henry Yuen c a Northwest Indian Fisheries Commission, 6730 Martin Way East, Olympia, Washington, 98516, USA b Washington Department of Fish and Wildlife, 600 Capitol Way North, Olympia, Washington, 98501, USA c U.S. Fish and Wildlife Service, 1211 Southeast Cardinal Court, Suite 100, Vancouver, Washington, 98683, USA Version of record first published: 19 Feb 2013.

To cite this article: Robert Conrad , Angelika Hagen-Breaux & Henry Yuen (2013): Unbiased Methods for Calculating Mortality in Mark-Selective Fisheries Models for Ocean Salmon, North American Journal of Fisheries Management, 33:2, 253-264 To link to this article: http://dx.doi.org/10.1080/02755947.2012.754806

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 33:253–264, 2013 C American Fisheries Society 2013 ISSN: 0275-5947 print / 1548-8675 online DOI: 10.1080/02755947.2012.754806

ARTICLE

Unbiased Methods for Calculating Mortality in Mark-Selective Fisheries Models for Ocean Salmon

Robert Conrad* Northwest Indian Fisheries Commission, 6730 Martin Way East, Olympia, Washington 98516, USA Angelika Hagen-Breaux Washington Department of Fish and Wildlife, 600 Capitol Way North, Olympia, Washington 98501, USA Henry Yuen U.S. Fish and Wildlife Service, 1211 Southeast Cardinal Court, Suite 100, Vancouver, Washington 98683, USA

Abstract Mark-selective fisheries (MSF) are increasingly being used as a strategy for managing fisheries for Coho Salmon Oncorhynchus kisutch and Chinook Salmon O. tshawytscha on the west coast of North America. Mark-selective fisheries allow anglers to keep legal-size Coho Salmon or Chinook Salmon with a missing adipose fin (typically hatchery fish) and require the release of those with an adipose fin (unmarked fish, which are usually wild fish). The objective of MSF is to provide meaningful fisheries on abundant stocks of hatchery salmon while reducing the impact on wild (unmarked) salmon stocks. As has been previously shown, the model currently used in the Pacific Fishery Management Council’s preseason planning process to project mortalities for proposed Coho Salmon and Chinook Salmon fisheries underestimates the number of unmarked salmon mortalities occurring in MSF and concurrent nonselective fisheries. We propose equations that provide unbiased estimates of salmon mortalities that occur in these fisheries due to the release of fish. The performance of the proposed methods is evaluated and compared to the current methods using a simulation model. The methods are shown to provide unbiased calculations of total mortalities for unmarked salmon in both mark-selective and concurrent nonselective fisheries. The unbiased methods are able to incorporate different release-mortality and mark-recognition rates for the fisheries modeled.

Downloaded by [Department Of Fisheries] at 22:45 28 February 2013 In 1998, the State of Washington introduced mark-selective eries for Chinook Salmon in Washington have been conducted fisheries (MSF) as a salmon Oncorhynchus spp. fisheries man- in the inner marine waters of the Strait of Juan de Fuca and agement tool (PFMC 1999). Regulations for MSF permit anglers Puget Sound since 2003 (WDFW 2008a, 2008b) while MSF for to retain legal-size Coho Salmon O. kisutch or Chinook Salmon Chinook Salmon were conducted in Washington coastal marine O. tshawytscha that have had their adipose fin removed (marked) waters for the first time in 2010 (WDFW 2011). The harvest and require the release of all Coho or Chinook Salmon with an of marked Chinook Salmon in Washington’s marine waters has adipose fin (unmarked). The objective of these fisheries is to increased greatly since the inception of MSF and reached a total provide meaningful fisheries on abundant (generally marked) of about 25 thousand fish annually in recent years (Figure 1B). hatchery salmon while reducing the impact on wild (unmarked) The Fishery Regulation Assessment Model (FRAM) is salmon. The annual harvest of marked Coho Salmon by MSF used annually by the Pacific Fishery Management Council off the coast of Washington State has ranged from 28 to 249 ([PFMC] 2008a) to evaluate the potential impact to major thousand fish (Figure 1A; PFMC 2011). Mark-selective fish- stocks of Coho and Chinook salmons from a proposed set

*Corresponding author: [email protected] Received July 24, 2012; accepted November 26, 2012 253 254 CONRAD ET AL.

the base-period data used for the Coho and Chinook Salmon FRAMs are provided in Packer and Cook-Tabor (2007) and PFMC (2008c), respectively. Models similar to the FRAM are used for salmon management in other forums such as the Chi- nook Model by the Pacific Salmon Commission (JCTC 2012) and the Klamath Harvest Rate Model (Prager and Mohr 2001). Prior to the implementation of MSF, a key assumption in these models was that the exploitation rates for specific coded- wire-tagged salmon stocks (often called indicator stocks) were representative of the exploitation rates for wild and hatchery stocks (typically from the same watershed) with similar life histories and ocean distributions. However, with the advent of MSF this basic assumption was violated because in MSF unmarked fish are released while the corresponding marked hatchery stocks are retained and removed from the population. These models were built with the assumption that there was no release of legal-sized fish caught during the base period used to estimate average fishery-specific exploitation rates for each stock. A time-period-specific exploitation rate is, therefore, assumed to represent the encounter rate of a stock in a fishery. The models were restructured with the advent of MSF. The time- period-specific exploitation rates were still used to estimate the exploitation rates on the marked stocks, but these same rates were used as surrogates for the encounter rates for the unmarked stocks (PFMC 2008b). These encounter rates are used to produce stock-specific estimates of the number of encounters with unmarked fish in a mark-selective fishery, which, combined with an estimate of the release-mortality rate, provide estimates of the mortalities due to the catch and release of unmarked fish. In these models, the exploitation rate on the unmarked stock component in a mark-selective fishery is a linear function of the time-period- specific average exploitation rate from the base period (µ¯ Base), a scalar that relates current year expected effort to base period δ FIGURE 1. Annual catches of marked salmon by mark-selective fisheries effort, and a release mortality rate ( ). The “simple” exploitation conducted in the marine waters of Washington State for (A) Coho Salmon rates for the marked and unmarked cohorts in a mark-selective caught by recreational and commercial troll fisheries and (B) Chinook Salmon fishery (µM and µU , respectively) are calculated as caught by recreational fisheries. µM = µ¯ Base · scalar of fishery regulations for a management season. Based upon Downloaded by [Department Of Fisheries] at 22:45 28 February 2013 FRAM projections, proposed fishery regulations are adjusted and to achieve escapement goals or stay below exploitation rate targets for stocks of concern (often stocks listed under the µU = µ¯ Base · scalar · δ, U.S. Endangered Species Act). The FRAM is a single-pool, deterministic model with discrete time periods varying in where µ¯ Base is the same for the marked and unmarked length from 1 to 7 months (PFMC 2008a). All fisheries during components of a specific stock. a modeled time period are assumed to operate simultaneously Lawson and Sampson (1996) demonstrated that in a mark- on a single pool of fish. The pool consists of all major stocks selective fishery, the actual mortality rate for unmarked fish is that have been caught historically in the fishery as estimated an increasing function of the time-period-specific exploitation from coded wire tag (CWT) recoveries (Nandor et al. 2010). rate. This causes the total number of unmarked mortalities in Exploitation rates estimated from CWTs recovered during a MSF to be underestimated by models relying on the linear re- base period when salmon abundances were relatively high and lationship between exploitation rate and release-mortality rate. fisheries were widely distributed in both time and area are Yuen and Conrad (2011) demonstrated that this bias (underes- the basis for the predictions of fishery mortalities by stock timation of unmarked mortalities) also occurs in any modeled (PSC 2005). Details for the methods and algorithms used in nonselective fisheries (NSF) that take place during the same the FRAM are presented in PFMC (2008b). Descriptions of model time period as the MSF. These biases are the result of MORTALITY IN MARK-SELECTIVE FISHERIES MODELS 255

approximating the nonlinear Baranov catch equation with a lin- related to drop-off mortality to zero, the probability of a marked ear model. When MSF operate during a modeled time period, fish dying ( pM) in a mark-selective fishery during time span t is unmarked mortalities are underestimated because released fish that survive may encounter the fishing gear more than once dur- pM (t) = 1 − exp{−λ · t · [γ + (1 − γ) · δ]} (1) ing the time period and the unmarked-to-marked fish ratio for all fish in the pool increases as a result of the selective removal of and the probability of an unmarked fish dying (pU) during time marked fish in MSF. Neither of these processes is currently cap- span t is tured by the algorithms used to calculate unmarked mortalities when MSF are operating during a modeled time period. pU (t) = 1 − exp{−λ · t · [(δ · ζ) + (1 − ζ)]}, (2) In most current management models, time period and fishery- specific exploitation rates that are derived from base period data where λ is the instantaneous encounter rate of a fish with the and adjusted for current year fishery projections are (1) equiv- gear. alent to mortality rates for a marked cohort in MSF and NSF During a single, discrete time period, if the recognition rate and (2) equivalent to encounter rates for an unmarked cohort in for marked fish (γ) is 100%, equation (1) simplifies to MSF and mortality rates in NSF. Mortalities in fisheries are typ- M ically projected using either quotas or exploitation rate scalars. pM = 1 − exp(−λ) = µ (3a) For a quota fishery, a total catch for the fishery (summed across all stocks) is specified. Fishery effort or season length needed or to achieve that quota is then projected using µ¯ Base. For an exploitation rate scalar fishery, the exploitation rate in the fishery exp(−λ) = 1 − µM . (3b) is scaled relative to µ¯ Base using a user-defined scalar. The most

common scaling mechanism is fishing effort relative to the aver- In this equation, pM is equivalent to an exploitation rate for age level during the base period (PFMC 2008b). In this paper we the marked cohort (µM ) in a fishery during the time period. When M describe unbiased methods for calculating total mortalities for pM is multiplied by the number of marked fish (N ) present at marked and unmarked fish when MSF are modeled using either the beginning of the time period (but after natural mortality) it scalars or quotas. We also describe unbiased methods for both provides the number of marked fish encountered by the gear and allocating total fishery-related mortalities in a modeled time pe- landed in the fishery. riod to each fishery and calculating the number of mortalities Similarly, when the recognition rate for unmarked fish (ζ)is in each fishery attributed to landed and nonlanded catch (e.g., 100% equation (2) simplifies to mortalities due to intentional release or mark-recognition error).

pU = 1 − exp(−λ · δ). (4)

METHODS By substituting the right side of equation (3b) into equation We first developed equations that provide unbiased calcula- (4) for exp(–λ), the exploitation rate for an unmarked cohort tions for the number of unmarked mortalities when there are (µˆ U ) becomes multiple MSF and a nonselective fishery operating concurrently during a time period. Because most of the management models µˆ U = 1 − (1 − µM )δ. (5) using equations of this type are accounting models that provide calculations of catch, total mortalities, and exploitations rates Equation (5) provides an unbiased method for calculating Downloaded by [Department Of Fisheries] at 22:45 28 February 2013 for specific stocks of interest by fishery and time period, we the total exploitation rate for an unmarked cohort when the only choose not to refer to them as estimates but as calculations. fishery in the time period is a mark-selective fishery with a Many of the model inputs are assumed (such as natural mortal- release mortality rate of δ (superscript ˆ indicates an unbiased ity rates and release mortality rates) and there are currently no exploitation rate calculation). methods incorporated into the models to provide estimates of As demonstrated by Yuen and Conrad (2011), additional the uncertainty associated with any model output. complexities are introduced when multiple mark-selective and The simulation model and methods used to evaluate the per- nonselective fisheries impact an unmarked cohort during the formance of the proposed unbiased methods are also briefly de- same time period in a single-pool model (i.e., all fisheries impact scribed. Finally, we describe how the proposed unbiased meth- a cohort simultaneously). As they demonstrated, even when ods would be implemented in the FRAM. assuming a constant δ in each mark-selective fishery Basic unbiased equations.—Table 1 defines the notation used in the development of the equations and parameters of the − − µM δ < − − µM + µM +···µM δ , simulation model. Following the notation of Lawson and 1 1 i 1 1 1 2 i Sampson (1996), but with separate mark-recognition rates for i marked (γ) and unmarked (ζ) fish and setting the parameters (6) 256 CONRAD ET AL.

TABLE 1. Deﬁnitions of parameters used in equations.

Parameter Definition N j Initial number of marked ( j = M) or unmarked ( j = U) fish after natural mortality has occurred in a time step. j = = NS Initial number of marked ( j M) or unmarked ( j U) fish from stock s after natural mortality has occurred in a time step. λ Instantaneous encounter rate of a fish with the gear. δi Release-mortality rate: probability that a fish that is caught and released dies due to the encounter in fishery i. δji Overall release mortality rate for marked ( j = M) or unmarked ( j = U) fish that accounts for mark-recognition error in fishery i. δW = = jI Weighted release mortality rate for marked ( j M) or unmarked ( j U) fish calculated across all fisheries in a modeled time step. γi Recognition rate for marked fish: the probability that a caught marked fish is properly identified as a marked fish and retained in fishery i. ζi Recognition rate for unmarked fish: the probability that a caught unmarked fish is properly identified as unmarked and released in fishery i. µ¯ Base Exploitation rate for a cohort estimated as an average of base period exploitation rates defined as the total number of fishery mortalities divided by the cohort size at the beginning of a time period. µM i Exploitation rate for the marked cohort: the total number of marked fish mortalities occurring in fishery i divided by the marked fish cohort size at the beginning of a time period (NM). µM µM ˜ i and ˆ i Biased and unbiased calculations, respectively, of the exploitation rate on a marked cohort in fishery i. µU i Exploitation rate for the unmarked cohort: the total number of unmarked fish mortalities occurring in fishery i divided by the unmarked fish cohort size at the beginning of a time period (NU). µU µU ˜ i and ˆ i Biased and unbiased calculations, respectively, of the exploitation rate on an unmarked cohort in fishery i. πi Proportion of total unmarked fish mortalities in all fisheries during a time period that occurred in fishery i. j = = Di Total number of marked ( j M) or unmarked ( j U) fish mortalities in fishery i. j = = DLi Number of marked ( j M) or unmarked ( j U) fish mortalities in fishery i that were landed catch. j = = DNi Number of marked ( j M) or unmarked ( j U) fish mortalities in fishery i that were nonlanded mortalities. j = = DLSi Number of marked ( j M) or unmarked ( j U) fish mortalities from stock s that were landed catch in fishery i.

µM δ where i is the exploitation rate for the marked component of a However, the proper for equation (7) when multiple MSF stock in ﬁshery i during the time period and is a surrogate for the (with different release mortality rates) and NSF operate concur- encounter rate for the unmarked component of the stock. The rently during a modeled time period must be determined. This sum of the individual ﬁshery unbiased calculations (the left side is described in the next section. of the equation) will always be less than the right side of equa- Weighted release mortality rate equations.—As described tion (6) because of the nonlinearity resulting from the quantities previously, the release of salmon is the source of the bias in inside the parentheses being raised to the δ power. This is an im- the exploitation rate calculations unless there is 100% release

Downloaded by [Department Of Fisheries] at 22:45 28 February 2013 portant distinction, because initial bias correction attempts were mortality. The release of a fish can be a consequence of fisheries performed separately on each fishery and then summed over all regulations or mark-recognition error. Many stocks are subject to fisheries in a time step (left side of equation 6). These bias- a mixture of nonselective and mark-selective fisheries, as well as adjusted calculations worked well within the FRAM’s compu- mark-recognition error, during the same time step. To accurately tational structure and did not necessitate finding a solution for δ compute exploitation rates on these stocks, a weighted release when multiple fisheries with different release mortality occurred mortality rate is needed. Nonselective fisheries can be viewed as in a time step. This initial method, however, underestimated the MSF with δ = 1.0, i.e., a fishery with a 100% release mortality true exploitation rate on the unmarked cohort. rate. Recognition errors for marked fish in MSF (i.e., when γ is The right side of equation (6) correctly calculates the total <1.0) reduce the exploitation rate on a marked cohort (because exploitation rate on an unmarked cohort for a single-pool model some marked fish are released and survive) and introduce bias when δ is constant across all fisheries: into the exploitation rate calculation for the marked component of a stock. Similarly, recognition errors for unmarked fish in δ MSF (i.e., when ζ is <1.0) increase the exploitation rate on µ = − − µM . ˆ I 1 1 i (7) an unmarked cohort (because some unmarked fish are kept). i An overall release mortality rate in fishery i that accounts for MORTALITY IN MARK-SELECTIVE FISHERIES MODELS 257

mark-recognition error for a marked cohort (δMi) is defined by When there is imperfect mark recognition for marked fish in MSF (γ < 1.0), additional complexities are introduced be- µM δMi = γi + [(1 − γi ) · δi ] . (8) cause i no longer provides an unbiased estimate of either the exploitation rate or encounter rate for the marked cohort Similarly, for the unmarked cohort due to the release of marked fish (some of which will die upon release). Note that γ < 1.0 can also be used to account for legal- sized marked fish that are released intentionally by anglers. The δUi = (1 − ζi ) + (ζi · δi ). (9) marked cohort exploitation rate is now subject to the same bias as the unmarked cohort but it is usually much smaller because Weighted release mortality rates across all fisheries in the recognition rates for marked fish are typically high (≥0.90). modeled time period for marked and unmarked cohorts (δW MI When there is error in identifying marked fish, an unbiased cal- and δW , respectively) can be calculated using fishery-specific UI culation of the total exploitation rate for the marked cohort is release mortality rates and µM to weight each rate. Specifically, i provided by for the marked cohort δW I MI I µM W i µM = − − µM . δ = · δ ˆ I 1 1 i (14) MI I M Mi µ i i i i I M M µ · {γi + ((1 − γi ) · δi )} µ˜ = i i = I (10) When equation (14) is solved for µM the result is I µM µM I i i I 1/δW µM = 1 − 1 − µˆ M MI . (15) and for the unmarked cohort I I This definition of µM can be substituted back into equation I µM I δW = i · δ (13) to define the unbiased exploitation rate for the unmarked UI I Ui µM cohort relative to the unbiased exploitation rate on the marked i i i I M U cohort: µ · {(1 − ζi ) + (ζi · δi )} µ˜ = i i = I (11) I µM µM δW /δW i i I µU = − − µM UI MI . ˆ I 1 1 ˆ I (16)

when there are I total fisheries in the modeled time period Equations (13) and (16) are the key equations needed to pro- (subscript I indicates summed across all fisheries and super- vide unbiased calculations of unmarked exploitation rates and ∼ script indicates a biased exploitation rate calculation). The mortalities when there are MSF and NSF operating concurrently weighted release mortality rates are simply the ratio of the sum during a modeled time step and there is mark-recognition error of the biased exploitation rate calculations, which account for in the MSF. Equations (12) and (14) are the analogous equations µM µU µM release mortalities (either ˜ I or ˜ I ) and the sum of the i . needed to provide unbiased calculations of marked exploitation These weighted release mortality rates are needed to provide rates and mortalities. unbiased calculations of marked and unmarked exploitation Allocating total mortalities to fisheries and sources of mortal- rates when fish are released due to mark-recognition error or ity.—For management purposes, the total number of unmarked MSF. U · µU Downloaded by [Department Of Fisheries] at 22:45 28 February 2013 mortalities that are projected to occur in all fisheries (N ˆ I ) Calculating total marked and unmarked fishery mortali- must be apportioned to each fishery in the time period. The pro- ties.—When there is perfect mark recognition of marked fish portional contribution of a cohort’s “simple” (biased) exploita- γ = in MSF ( 1.0), the total exploitation rate for a marked cohort tion rate for each fishery to the sum of the individual “simple” in a modeled time period is calculated as exploitation rates for all fisheries in the time period can be used to apportion total mortalities to each fishery: I 1 I µM = − − µM = µM ˆ I 1 1 i i (12) µM · δ µ˜ U Dˆ U = N U · µˆ U · π with π = i Ui = i , i i i I i i I µM · δ I µU i i Ui i ˜ i and the total exploitation rate for an unmarked cohort when ζ is (17) 1.0 as ˆ U where Di is the total number of unmarked mortalities occurring δW I UI in fishery i and the sum of the proportions of unmarked mor- µU = − − µM . π = ˆ I 1 1 i (13) talities occurring in each fishery will equal 1 ( i 1). This i is equivalent to the unbiased exploitation rate for the unmarked 258 CONRAD ET AL.

cohort in fishery i being 90% for both the unmarked and marked cohorts, µU = µU · π . ˆ i ˆ I i (18) 90% mark recognition for the unmarked cohort and 95% for the marked cohort, and A similar procedure can be used to apportion total marked 95% mark recognition for the unmarked mortalities to fisheries. cohort and 90% for the marked cohort; and Once total mortalities for a fishery have been calculated • proportions of the total exploitation rate of the marked (Dˆ U ), they can be apportioned to landed (catch) and nonlanded cohort occurring in MSF: i = µM mortalities using methods similar to those used above to ap- scenario 1 about 25% of I for the marked portion total mortalities to fisheries. For an unmarked cohort, cohort in MSF, ˆ U scenario 2 = about 43% of µM for the marked landed catch for fishery i (DLi) is calculated as I cohort in MSF, and − ζ scenario 3 = about 75% of µM for the marked ˆ U = ˆ U · 1 i I DLi Di (19) cohort in MSF. (1 − ζi ) + (ζi · δi ) A total of 72 different simulations were run for these analyses ˆ U and nonlanded mortality (DNi) is calculated as (6 combinations of mark-recognition rates for the unmarked and marked cohorts × 3 different proportions of the total exploita- (ζ · δ ) × ˆ U = ˆ U · i i . tion rate on the marked cohort occurring in MSF 4 different DNi Di (20) µM (1 − ζi ) + (ζi · δi ) levels of I at each combination of the previous factors). The simulations conducted for this study did not include any increase Similarly for a marked cohort in the release-mortality rate with successive encounters. For the unmarked cohort, we calculated mean relative bias γ ˆ M = ˆ M · i as the difference between the exploitation rates calculated DLi Di (21) γi + [(1 − γi ) · δi ] using the equations described above and simulation (model) exploitation rates expressed relative to the model result: ˆ M and nonlanded mortality (DNi) is calculated as µU −µU EQ model µU ( − γ ) · δ model ˆ M = ˆ M · 1 i i . relative bias = , (23) DNi Di (22) n γi + [(1 − γi ) · δi ] where n = number of replications of a simulation (100 for Simulation model description.—The simulation model we all evaluations reported here) and the equations are either the used was very similar to the individual-based simulation model simple biased equation (for each fishery or summed over all three described by Yuenand Conrad (2011). There were three fisheries fisheries) or the unbiased versions of the equations (13 and 16). simulated, two MSF with different release-mortality rates and Relative bias was calculated similarly for the marked cohort. one nonselective fishery. One difference between the models Application to the FRAM.—In the FRAM, as long as fish- was that the catch target for the nonselective fishery was based eries are modeled as rates, the solutions for computing unbiased on a number of marked fish landed in our simulations and not exploitation rates for the unmarked cohort presented here work the combined landed catch of marked and unmarked fish (Yuen well and produce results that match expectations. The FRAM Downloaded by [Department Of Fisheries] at 22:45 28 February 2013 and Conrad 2011). A target landed catch of marked fish for each output deviates from expectations for scenarios where MSF are fishery modeled was established using µM . i modeled as quotas and recognition of the mark status of captured Starting sizes for the marked and unmarked cohorts were fish is imperfect, e.g., a proportion of the marked fish encoun- always 200,000 fish each. Performance of the unbiased methods tered are released or a portion of the unmarked fish encountered was evaluated using simulations in which we varied three main are retained. In the FRAM, mark-selective quotas are modeled categories of parameters as follows: as the landed catch of marked and unmarked fish. The unmarked • µM landed catch is the result of anglers mistakenly retaining a por- I for the marked cohort (ranging from 0.08 to 0.80); • combinations of mark-recognition rates (γi and ζi): tion of the unmarked fish encountered during MSF. All quota 100% mark recognition for the unmarked and fisheries are initially modeled at base period exploitation rates. marked cohorts, For a marked cohort, the total landed catch of stock s in fishery 90% mark recognition for the unmarked i is calculated using the base period exploitation rate and the cohort and 100% for the marked cohort, projected current year abundance of stock s: 100% mark recognition for the unmarked ˜ M = µBase · M · γ cohort and 90% for the marked cohort, DLsi ¯ si Ns i (24) MORTALITY IN MARK-SELECTIVE FISHERIES MODELS 259

and for an unmarked cohort unbiased calculations presented earlier in this paper: ⎡ ⎤ δW D˜ U = µ¯ Base · N U · (1 − ζ ). (25) I Lsi si s i µM = ⎣ − − µM ⎦ · π , ˆ si 1 1 si si (29) i The FRAM then sums landed catch over all stocks encountered in a fishery: µM where si for quota fishery i is the base period exploitation rate for stock s (µ¯ Base). Then the landed catch of marked fish is j si Base Period Catchi = D˜ , (26) Lsi γ s ˜ M = µM · M · i DLsi ˆ si Ns (30) γi + [(1 − γi ) · δi ] where j = M or U depending upon the mark status of the cohort. In a final step, the FRAM compares the current year catch quota and the landed catch of unmarked fish for the fishery with this base period catch and scales the base 1 − ζ period exploitation rate by the ratio of quota catch to base period D˜ U = µˆ U · N U · i , Lsi si s − ζ + ζ · δ (31) catch: (1 i ) ( i i )

µU Catch Quota Fishery where ˆ si is calculated using equations (13) or (16). Scalar = i (27) Total catch and a scalar are then computed as in equations i Base Period Catch i (26) and (27) using these new estimates of landed catch. The scalar is applied to the base period exploitation rate and the iter- and ative loop (equations 29, 30, and 31 and the computation of total catch and a new scalar) is repeated. Iterations are continued until µM = · µBase. the newly computed value is within a predetermined number of si Scalari ¯ si (28) ﬁsh of the target quota.

The resulting stock-specific exploitation rate for fishery i µM ( si ) can then be used to compute the unbiased exploitation rate RESULTS for the unmarked component of stock s. Even though the unbi- Because the calculation of fishery-specific exploitation rates ased equations are used to calculate the unmarked exploitation for the unmarked cohort is a two-step process using the proposed µM rate, this rate can be biased because the computation of si may unbiased methods (first estimate the total exploitation rate and be biased if there is imperfect mark recognition. then partition this rate to the contributing fisheries), the results This problem cannot be addressed solely on the stock level, are presented in two sections. First, we compare the calcula- since a fishery quota is made up of the catch of many stocks. tions of the total exploitation rate for the unmarked cohort by However, an iterative solution can be found, where the quota is the biased and unbiased methods to simulation results. We then first modeled using base period exploitation rates. Finding the examine fishery-specific calculations for subsets of the simula- µM correct ˆ si for a quota fishery can be accomplished by using the tion results to compare results of the two calculation methods Downloaded by [Department Of Fisheries] at 22:45 28 February 2013

µU µM FIGURE 2. Total exploitation rates on the unmarked cohort ( I ) as a function of I for the biased and unbiased calculations compared with the simulation µM results. Three different scenarios for the proportion of I occurring in mark-selective ﬁsheries (1) 25%, (2) 43%, and (3) 75% are examined (numbers to the right indicate the scenario). Mark-recognition rates of (A) 100% for both unmarked and marked salmon, (B) 90% for unmarked and 100% for marked salmon, and (C) 100% for unmarked and 90% for marked salmon. 260 CONRAD ET AL.

to the simulation results, illustrate common trends, and demonstrate that the unbiased methods perform well across the range of ﬁshery parameters explored.

Estimation of Total Exploitation Rate for the Unmarked Cohort There was strong agreement between total exploitation rates on the unmarked cohort calculated using the unbiased methods and the simulation results. Relative biases for the unbiased calculations were between −0.5% and 0.5% for all simulations. For comparison, relative biases for the biased calculations were between −1.1% and −35.8%. The differences between the biased calculations and simu- µM lation results increased as I increased and also increased as µM the proportion of I in MSF increased (Figure 2). For a given µM I , increasing the proportion occurring in MSF is mathemat- ically equivalent to decreasing the weighted release mortality rate which has been shown to increase bias. The unbiased calculations of the total exploitation rate on the unmarked cohort were essentially the same as the simulation results (Figure 2). By comparing Figures 2A, 2B, and 2C it can be seen that a low level of mark-recognition error (10%) for either the unmarked (Figure 2B) or marked (Figure 2C) cohorts has little effect on the results when total exploitation rate is examined. As has been previously demonstrated by Lawson and Sampson (1996) and Yuen and Conrad (2011), for a fixed set of release mortality rates, relative bias in the calculation of total exploitation rate on the unmarked cohort increases as the exploitation rate increases on the marked cohort (Figure 3). The effect of increasing the proportion of the total exploitation rate on the marked cohort in MSF is seen by comparing Figures 3A, 3B, and 3C. Noting that the three panels have different scales for the y-axis, maximum relative bias increased from about 12% µM = (Figure 3A) to nearly 35% (Figure 3C) for I 0.80. Rela- tive bias decreases slightly as the recognition rate for unmarked fish decreases because a lower mark-recognition rate for the unmarked cohort in MSF reduces the number of unmarked fish released and available for recapture. Relative bias for the unbiased FIGURE 3. Comparison of bias (relative to simulation results) for biased and unbiased calculations of µU over a range of µM , two different scenarios for calculations is zero or nearly zero in every instance (Figure 3). I I marked and unmarked recognition rates (% unmarked | % marked), and three Downloaded by [Department Of Fisheries] at 22:45 28 February 2013 µM different scenarios for the proportion of I occurring in mark-selective fisheries (A) 25%, (B) 43%, and (C) 75%. Estimation of Fishery-specific Exploitations Rates for the Unmarked Cohort Agreement between fishery-specific exploitation rates on the unmarked cohort calculated using the unbiased methods and (<50%). At low levels of exploitation in MSF, the simulation the simulation results was very good. For the two MSF simu- results are more variable than at higher exploitation rates. The lated, relative biases for the unbiased calculations were between higher relative biases for the fishery-specific unbiased calcula- −1.2% and 1.6% for all simulations. This is an improvement tions of exploitation rates on the unmarked cohort are a result of as the relative biases for the biased calculations were between random variation in the simulation results. At higher levels of −0.1% and −35.9% (Table 2). The relative biases that were exploitation in MSF, the relative biases for the unbiased calcula- furthest from zero for the unbiased calculations, and the relative tions of fishery-specific exploitation rates are in the same range biases that were closest to zero for the biased calculations, all as those for the unbiased calculations for the total unmarked µM occurred for the simulations with the lowest I exploitation cohort exploitation rate discussed in the previous section, i.e., < µM < ± rates ( 0.20) and smallest proportions of I occurring in MSF 1%. MORTALITY IN MARK-SELECTIVE FISHERIES MODELS 261

TABLE 2. Summary of the relative bias in ﬁshery-speciﬁc exploitation rate calculations for the unmarked cohort by the biased and unbiased methods compared µM with simulation results. Summarized for each level of the percentage of the total marked cohort exploitation rate ( I ) occurring in MSF and across the six µM combinations of mark-recognition rates for the unmarked and marked cohorts and four different values of I used in the simulations. There are a total of 24 observations in each level summarized.

Biased calculations Unbiased calculation µM %of I in MSF Summary statistic MSF #1 MSF #2 NSF MSF #1 MSF #2 NSF 25% Mean −5.2% −4.9% −5.1% −0.1% 0.2% −0.1% Median −3.5% −3.6% −3.5% 0.0% 0.1% 0.0% Minimum −13.0% −12.9% −13.0% −1.2% −0.4% −0.2% Maximum −1.0% −0.1% −1.2% 1.4% 0.9% 0.1% 43% Mean −6.9% −7.1% −7.0% 0.2% −0.1% 0.0% Median −5.0% −5.3% −5.1% 0.0% −0.1% 0.0% Minimum −17.4% −17.6% −17.5% −0.7% −0.4% −0.1% Maximum −0.5% −1.2% −1.4% 1.6% 0.3% 0.2% 75% Mean −14.3% −14.5% −14.2% 0.0% −0.2% 0.1% Median −10.1% −10.3% −10.0% 0.1% −0.2% 0.1% Minimum −35.8% −35.9% −35.8% −0.7% −0.3% 0.0% Maximum −2.9% −3.0% −2.6% 0.5% 0.1% 0.5% Downloaded by [Department Of Fisheries] at 22:45 28 February 2013

µM FIGURE 4. Exploitation rates on the unmarked cohort as a function of I in MSF (combined) and NSF for the biased and unbiased calculations compared to the µM simulation results. Three different scenarios for the proportion of I occurring in MSF (1) 25%, (2) 43%, and (3) 75% are examined (numbers to the right indicate the scenario). Panels (A)and(B) show MSF with 100% mark recognition, and panels (C)and(D) show MSF with 90% mark recognition (mark-recognition rates speciﬁed are for both marked and unmarked ﬁsh). 262 CONRAD ET AL.

As was shown previously for all fisheries combined (Fig- ure 2), the bias in the calculated exploitation rate on the un- µM marked cohort in each fishery increases with I (Figure 4). For µM µM a specific I , the bias also increases as the proportion of I in MSF increases for both the mark-selective and nonselective fisheries. The unbiased calculations of exploitation rates on the unmarked cohort were essentially the same as the simulation results for both the MSF and NSF (Figure 4). Mark-recognition error in MSF increased the exploitation rate on the unmarked cohort in the MSF but had little effect on relative bias. This increase in the exploitation rate on the unmarked cohort in MSF due to mark-recognition error is due to the following: • unmarked fish being mistakenly landed instead of released, and • the fishery sorting through more marked fish (compared with the same fishery with γ = 1.0) to reach the target for landed marked fish because marked fish are being mistakenly released. Figure 5 compares fishery-specific relative bias for different scenarios when the mark-recognition rate for unmarked fish is 90% and 100% for marked fish. Relative bias is the same in each of the mark-selective fisheries and the nonselective fishery for µM µM agiven I and proportion of I in MSF (Figure 5). As shown µM previously, bias increases with I . The slightly higher level of bias for the unbiased calculations ( ±1.5%) at low levels of total exploitation on the marked cohort (≤0.10) due to variability of the simulation results is seen in Figure 5. For all other cases, relative bias for the unbiased calculations is not discernable from zero.

DISCUSSION The relative bias for the simulation model used by Yuen and Conrad (2011) was dependent on the unmarked-to-marked fish ratio for the starting cohorts. This was because their simulation model used the total landed catch of both marked and unmarked fish to establish a stopping rule in NSF (essentially a combined FIGURE 5. Comparison of bias (relative to simulation results) for biased quota for marked and unmarked fish). For the simulation model µU µM and unbiased calculations of i in MSF and NSF over a range of I , 90% Downloaded by [Department Of Fisheries] at 22:45 28 February 2013 used for this paper, the stopping rule for both MSF and NSF mark recognition for the unmarked cohort, and three different scenarios for the µM was based on the landed catch of marked fish (a marked fish proportion of I occurring in MSF (A) 25%, (B) 43%, and (C) 75%. only quota). This simplifies the development of unbiased calculations and results in relative bias no longer being dependent on original FRAM calculations of exploitation rates for each un- the unmarked-to-marked fish ratio for the starting cohorts. The marked stock in the model. Relative bias for total exploitation actual number of unmarked mortalities associated with a given rate calculations (defined as total fishery mortalities divided by landed catch of marked fish is identical for the two models. The total fishery mortalities plus escapement) across all model time differences between the two simulation models arise because periods was between 0.07% and −3.5% for all stocks and av- the NSF are ended during the simulations at different numbers eraged −0.89% and −0.68% in 2009 and 2010, respectively. of landed marked fish. Applying unbiased methods to the model output resulted in Conrad and Hagen-Breaux (2011) examined the relative bias an exploitation rate that exceeded the guidelines for two listed in estimates of exploitation rates on unmarked stocks for the Coho Salmon stocks. For one listed stock, fisheries were struc- 2009 and 2010 preseason Coho FRAM runs used by the PFMC tured to fish up to the exploitation rate guideline; thus, after (2009, 2010). They applied the unbiased methods proposed here accounting for bias the exploitation rate for this stock exceeded to the FRAM output and compared unbiased calculations with the guideline in both years (Conrad and Hagen-Breaux 2011). MORTALITY IN MARK-SELECTIVE FISHERIES MODELS 263

Several current salmon fishery management models, used change, such as economics or population biology, is familiar for both preseason planning and postseason assessment of fish- with the basic concepts presented here. This paper adapts the eries, calculate mortalities for unmarked cohorts in MSF using basic equation to specifically deal with problems commonly en- methods that rely on a linear relationship between a base period countered in fisheries management, such as multiple sources exploitation rate (µ¯ Base) and a release-mortality rate. Previous of gear-related mortalities and complex regulatory regimes. research has demonstrated that the single-pool paradigm that Mark-selective fisheries are becoming increasingly popular in most of these models rely on results in an underestimate of the response to ESA listings and conservation concerns. When ap- unmarked mortalities occurring in MSF (Lawson and Sampson plied in the FRAM, the equations in this paper use histori- 1996) and any concurrent NSF (Yuen and Conrad 2011). We cal exploitation rate data from retention fisheries to estimate have proposed methods that correct for this bias and demon- mortalities on populations exposed to a sophisticated regime of strated through simulation that they provide unbiased calcula- retention and MSF. Mark-selective fisheries require the release tions of the unmarked mortalities in both MSF and NSF. of unmarked salmon, subjecting populations to a range of values A key assumption of these single-pool models is that all for surviving gear encounters that are a function of release rates salmon of the species of interest are randomly mixed throughout (intentional and unintentional) as well as release mortality rates. all fishery areas during the time period being modeled. In our The algorithms presented here capture these processes by using simulations, this process is modeled by allowing an unmarked a weighted release mortality rate in conjunction with the basic fish encountered and released by a mark-selective fishery to be exponential algorithm. immediately available for recapture by the same fishery or any The FRAM is the main model used to estimate salmon mor- other fisheries in the same time period (as long as the marked talities in marine MSF under the jurisdiction of the PFMC. fish quota for a fishery has not been met). This assumption is The Coho FRAM is currently being modified to incorporate the not met in reality as, for example, a fish released in Puget Sound methods described in this paper to account for the bias intro- cannot be immediately available for recapture by a fishery in duced by MSF for Coho Salmon. Other salmon fishery man- northern California. agement models are also subject to the bias described in this But this assumption is necessary as alternatives to the single- paper. Spreadsheet-based models, for instance, are commonly pool model require detailed information on the migration routes used to evaluate the impacts of recreational and commercial and migration timing for a large number of hatchery and wild MSF on salmon in terminal freshwater systems (e.g., in Puget stocks that is currently not available. The large number of stocks Sound and the Columbia River). Bias in these fisheries can involved in some of the management models (e.g., 123 Coho arise due to inconsistencies between the idealized continuous Salmon stocks and 38 Chinook Salmon stocks for the manage- mortality process and a combination of temporal and spatial ment models used by the PFMC) presents a daunting challenge model resolution effects. More broadly, any model used to eval- for the collection of the data needed to parameterize these mod- uate total fishery impacts when there is a significant nonreten- els. Zhou (2004) describes a pipeline model that provides es- tion component (e.g., trout catch-and-release fisheries, fisheries timates of unmarked mortalities in MSF using change-in-ratio with mandatory release of nontarget species, or release due to methods similar to those used to estimate animal abundance minimum size limits) is subject to the biases described in this (Seber 1982). Two important assumptions of his model are that paper. For salmon, this includes the Pacific Salmon Commis- the marked and unmarked stocks migrate together through a sion’s exploitation rate analysis (JCTC 2012), which accounts series of mark-selective and nonselective fisheries in a known for landed and nonretention fishery mortalities in a variety of sequence (the pipeline) and that the fisheries are of sufficiently U.S. and Canadian fisheries, and the Klamath Ocean Harvest short duration so that they maintain a constant ratio of un- Model, which accounts for landed and nonretention mortali- Downloaded by [Department Of Fisheries] at 22:45 28 February 2013 marked to marked fish in the catch. Zhou’s equations do not ties in Chinook Salmon fisheries off Oregon and California allow for parallel or concurrent fisheries or for multiple encoun- (M. Mohr, National Marine Fisheries Service, personal commu- ters in MSF. Given the large number of stocks and fisheries nication). Although we are not intimately familiar with planning involved in the FRAM, it is not possible to specify migration models used beyond the salmon arena, we suspect the issues ad- pipelines with currently available migration information. Given dressed in this paper are germane to a variety of other exploited its assumptions and requirement for additional migration infor- fish and shellfish species. mation, it is not clear that this model offers any advantages over the current single-pool model once unbiased methods have been implemented. ACKNOWLEDGMENTS Models using continuous equations to project population The authors thank their colleagues on the following PFMC growth have a long history of use in population dynamics and work groups for their feedback and exchange of ideas on the harvest management (e.g., Malthusian growth model, Baranov’s subject of mark-selective fisheries and calculating fishery mor- catch equation). While these calculations are very useful in the talities in the FRAM: Model Evaluation Workgroup, Salmon realm of fisheries management, their application is not limited Technical Team, and the Scientific and Statistical Committee. to this field. Any discipline dealing with instantaneous rates of Marianna Alexandersdottir, Peter McHugh, and Laurie Peterson 264 CONRAD ET AL.

provided valuable comments on the initial draft of this paper. PFMC (Pacific Fishery Management Council). 2008c. Chinook fishery regu- James Packer programmed the unbiased exploitation rate equa- lation assessment model (FRAM): base data development. PFMC, Model tion into the Coho FRAM. We also thank Peter Lawson and one Evaluation Workgroup, Portland, Oregon. Available: www.pcouncil.org/ salmon/background/document-library/fishery-regulation-assessment-model- anonymous reviewer for providing valuable suggestions, which fram-documentation/. (November 2012). improved this manuscript. PFMC (Pacific Fishery Management Council). 2009. Preseason report III: analysis of council adopted management measures for 2009 ocean salmon fisheries. PFMC, Salmon Technical Team, Portland, Oregon. Available: www. REFERENCES pcouncil.org/salmon/stock-assessment-and-fishery-evaluation-safe-documents/ Conrad, R. H., and A. Hagen-Breaux. 2011. Application of bias-corrected meth- preseason-reports/. (November 2012). ods for estimating mortality in mark-selective fisheries to Coho FRAM. PFMC (Pacific Fishery Management Council). 2010. Preseason report III: Pacific Fishery Management Council, Report to the November Meeting, analysis of council adopted management measures for 2010 ocean salmon Portland, Oregon. fisheries. PFMC, Salmon Technical Team, Portland, Oregon. Available: www. JCTC (Joint Chinook Technical Committee). 2012. 2011 exploitation rate anal- pcouncil.org/salmon/stock-assessment-and-fishery-evaluation-safe-documents/ ysis and model calibration. Pacific Salmon Commission, Report TCCHI- preseason-reports/. (November 2012). NOOK (12)-2, Vancouver. Available: www.psc.org/publications tech PFMC (Pacific Fishery Management Council). 2011. Review of 2010 ocean techcommitteereport.htm#TCCHINOOK/. (November 2012). salmon fisheries. PFMC, Portland, Oregon. Available: www.pcouncil.org/ Lawson, P. W., and D. B. Sampson. 1996. Gear-related mortality in selective salmon/stock-assessment-and-fishery-evaluation-safe-documents/. (Novem- fisheries for ocean salmon. North American Journal of Fisheries Management ber 2012). 16:512–520. Prager, M. H., and M. S. Mohr. 2001. The harvest rate model for Klamath Nandor, G. F., J. R. Longwill, and D. L. Webb. 2010. Overview of the coded wire River fall Chinook Salmon, with management applications and comments on tag program in the greater Pacific region of North America. Pages 5–46 in model development and documentation. North American Journal of Fisheries K. S. Wolf and J. S. O’Neal, editors. Tagging, telemetry and marking measures Management 21:533–547. for monitoring fish populations—a compendium of new and recent science PSC (Pacific Salmon Commission). 2005. Report of the expert panel on the fu- for use in informing technique and decision modalities. Pacific Northwest ture of the coded wire tag recovery program for Pacific salmon. PSC, Vancou- Aquatic Monitoring Partnership, Special Publication 2010-002, Portland, ver. Available: www.psc.org/info codedwiretagreview finalreportintro.htm. Oregon. (November 2012). Packer, J. F., and C. Cook-Tabor. 2007. Coho FRAM base period de- Seber, G. A. F. 1982. The estimation of animal abundance and related parame- velopment. Pacific Fishery Management Council, Portland, Oregon. ters, second edition. Macmillan, New York. Available: www.pcouncil.org/salmon/background/document-library/fishery- WDFW (Washington Department of Fish and Wildlife). 2008a. A multi-year as- regulation-assessment-model-fram-documentation/. (November 2012). sessment of the marine areas 5 and 6 selective Chinook fishery: 2003–2007— PFMC (Pacific Fishery Management Council). 1999. Review of 1998 ocean final working draft 3-14-08. WDFW, Olympia. Available: wdfw.wa.gov/ salmon fisheries. PFMC, Portland, Oregon. Available: www.pcouncil.org/ publications/00495/. (November 2012). salmon/stock-assessment-and-fishery-evaluation-safe-documents/. (Novem- WDFW (Washington Department of Fish and Wildlife). 2008b. A multi-year ber 2012). assessment of the marine areas 8-1 and 8-2 selective Chinook fishery: 2005– PFMC (Pacific Fishery Management Council). 2008a. Fishery regulation assess- 2007—final working draft. WDFW, Olympia. Available: wdfw.wa.gov/ ment model (FRAM)—an overview for Coho and Chinook. PFMC, Model publications/00496/. (November 2012). Evaluation Workgroup, Portland, Oregon. Available: www.pcouncil.org/ WDFW (Washington Department of Fish and Wildlife). 2011. 2010 ocean se- salmon/background/document-library/fishery-regulation-assessment-model- lective fishery sampling report. WDFW, draft report, Olympia. fram-documentation/. (November 2012). Yuen, H., and R. Conrad. 2011. Bias in the estimation of impacts of simultaneous PFMC (Pacific Fishery Management Council). 2008b. Fishery regula- mark-selective and nonselective fisheries on ocean salmon. North American tion assessment model (FRAM)—technical documentation for Coho Journal of Fisheries Management 31:1043–1051. and Chinook. PFMC, Model Evaluation Workgroup, Portland, Oregon. Zhou, S. 2004. A pipeline model for estimating fishing mortality in salmon Available: www.pcouncil.org/salmon/background/document-library/fishery- mark-selective fisheries. North American Journal of Fisheries Management regulation-assessment-model-fram-documentation/. (November 2012). 24:979–989. Downloaded by [Department Of Fisheries] at 22:45 28 February 2013 This article was downloaded by: [Department Of Fisheries] On: 28 February 2013, At: 22:46 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 Lightweight Battery for Backpack Electrofishing Eric E. Hockersmith a , Gabriel Brooks b , Nathan D. Dumdei b & Stephen Achord b a U.S. Army Corps of Engineers, Northwestern Division, Walla Walla District, 201 North 3rd, Walla Walla, Washington, 99362-1875, USA b National Oceanic and Atmospheric Administration, National Marine Fisheries Service, Northwest Fisheries Science Center, Fish Ecology Division, 2725 Montlake Boulevard East, Seattle, Washington, 98112-2097, USA Version of record first published: 26 Feb 2013.

To cite this article: Eric E. Hockersmith , Gabriel Brooks , Nathan D. Dumdei & Stephen Achord (2013): A Lightweight Battery for Backpack Electrofishing, North American Journal of Fisheries Management, 33:2, 265-268 To link to this article: http://dx.doi.org/10.1080/02755947.2013.765526

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 33:265–268, 2013 American Fisheries Society 2013 ISSN: 0275-5947 print / 1548-8675 online DOI: 10.1080/02755947.2013.765526

MANAGEMENT BRIEF

A Lightweight Battery for Backpack Electroﬁshing

Eric E. Hockersmith* U.S. Army Corps of Engineers, Northwestern Division, Walla Walla District, 201 North 3rd, Walla Walla, Washington 99362-1875, USA Gabriel Brooks, Nathan D. Dumdei, and Stephen Achord National Oceanic and Atmospheric Administration, National Marine Fisheries Service, Northwest Fisheries Science Center, Fish Ecology Division, 2725 Montlake Boulevard East, Seattle, Washington 98112-2097, USA

generator-powered models often have higher power output with Abstract more duration, compared with battery-powered models. A lithium ion battery was modified to replace the conventional Battery-powered backpack electrofishing models require fa- sealed lead acid battery used to operate a backpack electrofishing cilities and equipment to recharge batteries, the sampling time unit. Specifications and performance of the lithium ion battery were compared with those of a lead acid battery of similar capacity. is limited by the capacity of or number of available batteries, The lithium ion battery was 76% lighter in weight than the lead and traditional lead acid electrofishing batteries are heavy. De- acid battery, reducing the overall weight of a Smith-Root model pending on the model, a backpack electrofishing unit, powered 12 backpack electrofishing unit and battery by 55%. Including the by a 24-V, 12-ampere-hour (Ah) conventional lead acid bat- cost of a charger and parts to make the battery compatible with tery, weighs 14–18 kg, and the battery contributes 47–60% to an electroshocking unit, the lithium ion battery and charger cost was 26% less than that of the lead acid battery and charger. Bench the overall weight. Lithium ion batteries of comparable capac- tests indicated the lithium ion battery provided 91% and 98% of the ity and voltage are lighter in weight, require less maintenance, operating time of the lead acid battery per charge when the settings and have a higher recharge cycle capacity (number of charge– were 300 V at 90 Hz and 500 V at 90 Hz, respectively. The fuel gauge discharge cycles) compared with traditional lead acid batteries (battery discharge indicator) on the lithium ion battery, which was (Albright et al. 2012). absent in the lead acid battery, provided the ability for a user to assess the remaining charge level while in the field. The lithium ion In this study, we described the modification and use of a battery provided similar performance with a significant reduction lithium ion battery, which was used to power a backpack elec- in weight and cost compared with a conventional lead acid battery trofishing unit, and compared these with a conventional lead for backpack electrofishing. The lighter weight of the backpack acid battery. The goal of this modification was to reduce the electrofishing unit using the lithium ion battery can reduce fatigue weight of the electrofishing unit, and therefore reduce the risk

Downloaded by [Department Of Fisheries] at 22:46 28 February 2013 and the risk of fatigue-related injuries to field crews. of fatigue-related injuries to field crews, with minimal change in performance of the electrofishing unit. Backpack electrofishing is a common technique for actively sampling small streams (Onorato et al. 1998). Modern backpack electrofishing units typically use pulsed DC and are powered by METHODS either a battery or gasoline generator (Reynolds 1996). Battery- A 12.6-Ah, 25.9-V lithium ion battery and charger were pur- powered models are preferred to generator-powered models be- chased from BatterySpace.com. The battery weighed 2.00 kg cause they are quieter, can be used in restricted use areas such as and cost US$376 (Table 1). To charge the lithium ion battery we National Wilderness Areas, and do not expose sampling crews to used a model CH-L2596N charger with a maximum output of carbon monoxide, which is produced by the generator motor. In 180 W and cost of $86. The lithium ion battery had a fuel gauge addition, the gasoline needed to power a generator is flammable (battery discharge indicator), which allowed the user to assess and hazardous to the environment if spilled. However, the remaining charge level.

*Corresponding author: [email protected] Received March 23, 2012; accepted January 7, 2013 265 266 HOCKERSMITH ET AL.

TABLE 1. Physical characteristics and cost (US$) of a conventional lead acid a backpack electrofishing unit. The bench tests were conducted battery and a lithium ion battery used to power a backpack electrofishing unit. indoors, where variables such as depth, substrate, conductivity, The cost of the battery charger alone and combined battery and charger are also and temperature could be controlled across tests that could be shown. replicated under identical conditions. Characteristic Lead acid battery Lithium ion battery The first bench test used a Smith-Root BAT-01 battery analysis tool. This tool measured the amount of time (in seconds) Manufacturer Smith-Root Powerizer required to discharge a battery to below 19.6 V at a continuous Model 6682 PLB-8570170-7S-FG load of 5.76 Ω. During each test, the batteries were charged to Voltage (V) 24 25.9 capacity, and time to discharge (below 19.6 V) was measured. Ampere-hour (Ah) 12 12.6 The battery analysis test was replicated six times for each battery × × × × Dimensions (mm) 200 150 90 200 120 75 type, and the results were compared using a t-test. Weight (g) 8,437 1,996 The second bench test used a Smith-Root model 12 backpack Fuel gauge None Included electrofishing unit operated at two different settings (300 V at $ $ Battery cost 384.00 375.95 90 Hz and 500 V at 90 Hz). For each test, the batteries were $ Parts for None 92.00 charged to capacity. Tests were conducted using a traditional compatibility rat-tail cathode and a wand type anode with a 28-cm-diameter Charger model UBC24 CH-L2596N aluminum ring. Tests were done in a nonmetallic container con- $ $ Charger cost 360.00 85.95 taining 208 L of freshwater at a water temperature of 18◦C. $ $ Total cost with 744.00 553.90 Space between the anode and cathode was fixed at 55 cm and the parts and charger anode and cathode were submerged to a depth of 60 cm. Setting levels of the electrofishing unit were then set for the particular test, and the unit was set to run continuously by depressing the The lithium ion battery was modified using a Smith-Root anode activation switch. Time until the battery stopped oper- battery adapter (part 07459, cost $50) in order to connect it to ating the electrofishing unit was measured using the seconds a Smith-Root model 12 electrofishing unit. The modifications counter on the unit. This test was replicated three times for each required removing the terminal ends of the battery adapter and battery type and power level, and results were compared using connecting the wiring to the terminals of the lithium ion battery. a t-test. Both the battery analysis and model 12 electrofishing A similar modification was made to the lithium ion battery unit testing were conducted under the same conditions (e.g., charger using a Smith-Root equipment adapter (part 07458; cost temperature, depth, distance between the anode and cathode, $42). The Smith-Root model 12 electrofishing battery holder conductivity) with the same equipment, and only the specific did not require modification because the lithium ion battery was battery was changed between tests. similar in size and shape to the lead acid battery. The characteristics and performance of the lithium ion battery, used to power a backpack electrofishing unit, were com- RESULTS AND DISCUSSION pared with those of a sealed lead acid battery of similar capacity. The lithium ion battery cost 2% less and its weight was For comparison we used a Smith-Root 12-Ah, 24-V model lead 76% lighter than a sealed lead acid battery of similar capac- acid battery that weighed 8.44 kg and cost $384 (Table 1). ity (Table 1). When we included the cost of the charger and The lead acid battery had no method to assess charge level ex- modifications, the cost of the lithium ion system ($553.90) was cept during recharging. To charge the lead acid battery we used approximately 26% less than that of the conventional lead acid Downloaded by [Department Of Fisheries] at 22:46 28 February 2013 Smith-Root model UBC24 charger that had a maximum output battery system ($744.00). of 60 W and cost $360. During the battery analysis tool test the sealed lead acid We used two bench tests to compare the performance of battery discharge time averaged 10,327 s (2.9 h) and the lithium similar capacity lithium ion and lead acid batteries used to power ion battery averaged 8,887 s (2.5 h) (Table 2). During the bench

TABLE 2. Mean operating time (s) for two battery types during performance testing used to power a backpack electroﬁshing unit (SD in parentheses).

Mean operating time (s) Difference between Test Lead acid battery Lithium ion battery the means t-value P-value Battery analysis tool 10,327 (324) 8,887 (610) 1,440 5.28 0.0003 Electroﬁshing unit, 300 V at 90 Hz 19,141 (83) 17,408 (511) 1,733 5.77 0.0045 Electroﬁshing unit, 500 V at 90 Hz 8,233 (298) 8,038 (235) 195 0.078 0.4802 MANAGEMENT BRIEF 267

test of the electrofishing unit at 300 V and 90 Hz the sealed lead Electrofishing operations can be dangerous and require ad- acid battery discharge time averaged 19,141 s (5.3 h) and the equate planning and preparation to be conducted safely (PSC lithium ion battery averaged 17,408 s (4.8 h). In both the battery 2008). Stream sampling using a backpack electrofishing unit analysis tool test and the electrofishing unit bench test at 300 V can expose the user to a variety of hazards. The most seri- and 90 Hz, the sealed lead acid battery had a significantly longer ous of these hazards are due to the performance of poten- discharge time than the lithium ion battery (Table 2). In these tially strenuous activity while moving upstream, against the tests, the lithium ion battery had 24–29 min less operating time water current, and on substrate that is often slippery and un- (9–14% less) than the lead acid battery. When the electrofishing even (Berry 1996). The additional weight of a backpack elec- unit was tested at 500 V and 90 Hz, average operating time trofishing unit may increase the risk of injury to field staff, was similar for the lead acid and lithium ion batteries (<4min particularly as fatigue associated with long periods of sam- difference). Differences in operating times between the lithium pling effort develops. In our study, replacing a conventional ion and lead acid batteries, at differing voltage settings, were due 12-Ah lead acid battery with a 12.6-Ah lithium ion battery re- to differences in the discharge profiles between the battery types. duced the weight of a Smith-Root model 12 backpack elec- The discharge profile of a lithium ion battery is relatively flat trofishing unit by 55% from 14.2 to 7.4 kg. The backpack compared with the sloping profile of a lead acid battery, which electrofishing unit with a lithium ion battery was lighter and results in similar discharge profiles at higher loads (Albright easier to carry; thus, it provided safer sampling conditions for et al. 2012). field crews by reducing fatigue and the risk of fatigue-related We field tested the lithium ion battery while collecting Chi- injuries. nook Salmon Oncorhynchus tshawytscha parr from headwater Since the completion of this study, Smith-Root began tributaries of the South Fork Salmon River in central Idaho offering a lithium ion battery for their backpack electrofishing during the summer of 2011. The primary goal of field testing units. The Smith-Root lithium ion battery specifications are was to determine whether the lithium ion battery during ac- 24 V and 9.6 Ah, and as of August 2012, it sells for $995. The tual use in the field would provide sufficient power for a day Smith-Root’s lithium ion battery charger cost $360. Smith-Root of sampling. Water temperature ranged from 7.3◦C to 13.4◦C estimates their lithium ion battery has a recharge cycle capacity and conductivity ranged from 21.3 to 25.6 µS/cm during field of at least 2,000 times, which is more than for their lead acid testing. Procedures for fish collection during field testing used a batteries that have recharge cycle capacity of 250–300 times single-pass technique (Meador et al. 2003) and were described (www.Smith-Root.com/electrofishers/batteries). by Achord et al. (1996, 2007). In our field test, a Smith-Root model 12 backpack electrofishing unit powered by the lithium ACKNOWLEDGMENTS ion battery provided between 21,506 and 25,681 s (6–7 h) This work was funded in part by the U.S. Department of sampling time when operated at 400 V and 90 Hz. This of Energy, Bonneville Power Administration (BPA), Portland, amount of time is sufficient for most field sampling. The time Oregon (contract 004981). We thank the following for their required to discharge a battery depends on the resistance of assistance in conducting this study: Deborah Docherty from the circuit and the voltage applied; lower resistance and higher BPA, and Gordon Axel, Shane Collier, John Ferguson, Bruce voltage will discharge a battery quicker. Discharge times for Jonasson, Jesse Lamb, Kenneth McIntyre, Jeff Moser, Richard other users may be different from the results reported here Nelson, Matthew Nesbit, Neil Paasch, Sam Rambo, and Louis due to differences in circuit resistance and the voltage applied. Tullos from the National Marine Fisheries Service. Refer- Due to the high variability in sampling conditions (i.e., con- ence to trade names does not imply endorsement by the U.S. ductivity, temperature, depth, target species, stream discharge,

Downloaded by [Department Of Fisheries] at 22:46 28 February 2013 Government. experience of sampling crew, and electrofishing unit settings), which can influence resistance and the voltage applied, it was not possible in this evaluation to address the range of dis- REFERENCES charge times that others may experience. However, the fuel Achord, S., G. M. Matthews, O. W. Johnson, and D. M. Marsh. 1996. Use gauge on the lithium ion battery is a useful tool for sampling of passive integrated transponder (PIT) tags to monitor migration timing of Snake River Chinook Salmon smolts. North American Journal of Fisheries crews to monitor the remaining capacity of a battery while sam- Management 16:302–313. pling in the field. We did not compare the performance of the Achord, S., R. W. Zabel, and B. P. Sandford. 2007. Migration timing, growth, lead acid and lithium ion batteries in the field because of the and estimated parr-to-smolt survival rates of wild Snake River spring– number of variables we could not control such as sampling summer Chinook Salmon from the Salmon River basin, Idaho, to the depth, substrate, conductivity, temperature, distance between lower Snake River. Transactions of the American Fisheries Society 136: 142–154. the anode and cathode, and crew experience, any of which Albright, G., J. Edie, and S. Al-Hallaj. 2012. A comparison of lead could change the load on the battery and thus the operating acid to lithium-ion in stationary storage applications. Alternative En- time. ergy eMagazine Industry [online serial] (April 12):article 1. Available: 268 HOCKERSMITH ET AL.

www.altenergymag.com/emagazine/2012/04/a-comparison-of-lead-acid-to- Onorato, D. P., R. A. Angus, and K. R. Marion. 1998. Comparison of a small- lithium-ion-in-stationary-storage-applications/1884. (August 2012). mesh seine and a backpack electroshocker for evaluating fish populations Berry, C. R., Jr. 1996. Safety in fisheries work. Pages 63–81 in B. R. Murphy and in a north-central Alabama stream. North American Journal of Fisheries D. W. Willis, editors. Fisheries techniques, 2nd edition. American Fisheries Management 18:361–373. Society, Bethesda, Maryland. PSC (Professional Safety Committee). 2008. Fisheries safety handbook. Meador, M. R., J. P. McIntyre, and K. H. Pollock. 2003. Assessing the American Fisheries Society, Bethesda, Maryland. efficacy of single-pass backpack electrofishing to characterize fish com- Reynolds, J. B. 1996. Electrofishing. Pages 221–253 in B. R. Murphy and D. W. munity structure. Transactions of the American Fisheries Society 132: Willis, editors. Fisheries techniques, 2nd edition. American Fisheries Society, 39–46. Bethesda, Maryland. Downloaded by [Department Of Fisheries] at 22:46 28 February 2013 This article was downloaded by: [Department Of Fisheries] On: 20 March 2013, At: 00:00 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 Anesthesia of Juvenile Pacific Lampreys with MS-222, BENZOAK, AQUI-S 20E, and Aquacalm Helena E. Christiansen a , Lisa P. Gee a & Matthew G. Mesa a a U.S. Geological Survey, Western Fisheries Research Center, Columbia River Research Laboratory, 5501 Cook-Underwood Road, Cook, Washington, 98605, USA Version of record first published: 05 Mar 2013.

To cite this article: Helena E. Christiansen , Lisa P. Gee & Matthew G. Mesa (2013): Anesthesia of Juvenile Pacific Lampreys with MS-222, BENZOAK, AQUI-S 20E, and Aquacalm, North American Journal of Fisheries Management, 33:2, 269-276 To link to this article: http://dx.doi.org/10.1080/02755947.2012.754807

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 33:269–276, 2013 American Fisheries Society 2013 ISSN: 0275-5947 print / 1548-8675 online DOI: 10.1080/02755947.2012.754807

ARTICLE

Anesthesia of Juvenile Paciﬁc Lampreys with MS-222, BENZOAK, AQUI-S 20E, and Aquacalm

Helena E. Christiansen, Lisa P. Gee, and Matthew G. Mesa* U.S. Geological Survey, Western Fisheries Research Center, Columbia River Research Laboratory, 5501 Cook-Underwood Road, Cook, Washington 98605, USA

Abstract Effective anesthetics are a critical component of safe and humane fish handling procedures. We tested three concentrations each of four anesthetics—Finquel (tricaine methanesulfonate, herein referred to as MS-222), BENZOAK (20% benzocaine), AQUI-S 20E (10% eugenol), and Aquacalm (metomidate hydrochloride)—for efficacy and safety in metamorphosed, outmigrating juvenile Pacific Lampreys Entosphenus tridentatus. The anesthetics MS-222 (100 mg/L) and BENZOAK (60 mg/L) were the most effective for anesthetizing juvenile Pacific Lampreys to a handleable state with minimal irritation to the fish. Fish anesthetized with BENZOAK also had lower rates of fungal infection than those exposed to MS-222, AQUI-S 20E, or no anesthetic. Exposure to AQUI-S 20E irritated juvenile Pacific Lampreys, causing them to leap or climb out of the anesthetic solution, and Aquacalm anesthetized fish to a handleable state too slowly and incompletely for effective use with routine handling procedures. Our results indicate that MS-222 and BENZOAK are effective anesthetics for juvenile Pacific Lampreys, but field studies are needed to determine whether exposure to MS-222 increases risk of fungal infection in juvenile Pacific Lampreys released to the wild.

Pacific Lamprey Entosphenus tridentatus populations are de- lost equilibrium and sensory perception (stage 4 anesthesia as clining (Close et al. 2002), and many studies have focused on described by Summerfelt and Smith 1990). At this level of anes- identifying limiting factors for this species and increasing our thesia, they can also be subjected to minor surgical procedures knowledge of their life history and biology. Preservation of this such as PIT tagging (Mesa et al. 2012). Juvenile lampreys re- species will require research that involves handling procedures sponded slowly to anesthetic concentrations typically used with such as tagging. We recently developed a surgical passive inte- salmonids (50 mg/L MS-222), so a much higher dose (250 mg/L) grated transponder (PIT) tagging protocol for metamorphosed, has been used (Mueller et al. 2006; Meeuwig et al. 2007; Mesa Downloaded by [Department Of Fisheries] at 00:00 20 March 2013 outmigrating juvenile Pacific Lampreys (9-mm tag; Mesa et al. et al. 2012). We observed that when juvenile Pacific Lampreys 2012) that could provide critical information about this under- were exposed to this higher concentration (even when buffered studied life stage, but to use this tagging method in the field and to a neutral pH) they were highly agitated. The behavior of juve- minimize handling-related fish mortality, a safe and effective nile lampreys exposed to MS-222 suggested that it was irritating anesthetic is needed. and stressful to them, and many fish presented with fungal infec- Anesthesia is a key aspect of fish handling because it reduces tions after handling and anesthetic exposure both in our study stress caused by invasive procedures (Ross and Ross 2008). (Mesa et al. 2012) and those of others (Schreck et al. 1999; Tricaine methanesulfonate (MS-222) is the most common anes- Mueller et al. 2006). thetic used in aquaculture and fisheries research, and it has There are no empirical data in the published literature testing previously been used to anesthetize juvenile Pacific Lampreys the response of juvenile Pacific Lampreys to different concen- for handling and tagging procedures. Juvenile Pacific Lampreys trations of MS-222 or to other anesthetics. A lower dose of MS- are not easily handled (measured for length) until they have 222 or a different anesthetic may provide safer, less irritating

*Corresponding author: [email protected] Received August 28, 2012; accepted November 26, 2012 269 270 CHRISTIANSEN ET AL.

anesthesia for juvenile lampreys. In this study, we determined Efficacy trials were conducted in the following manner. We the efficacy of four anesthetics (MS-222, BENZOAK, AQUI-S made a 10-L bulk batch of the desired anesthetic concentration 20E, and Aquacalm) for juvenile Pacific Lampreys. We mea- containing 0.025% Stress Coat (Aquatic Eco-systems, Apopka, sured the time to a handleable state (i.e., when fish could be Florida) and measured its temperature, dissolved oxygen (DO), measured for length) and time to recovery for three concen- and pH. For each trial, 1 L of anesthetic solution was poured trations of each anesthetic and identified concentrations that into a 2-L exposure container, and a single juvenile Pacific Lam- provided the quickest induction and recovery while minimizing prey was placed in the container. We recorded the time to reach irritation. We also tested the relationship between fungal infec- a handleable state and the general behavior of the fish during tions and anesthetic exposure by anesthetizing juvenile lampreys sedation, including the level of agitation, jumping or climbing until handleable and then monitoring development of fungal in- behavior, and head shaking. The fish was considered handleable fections and survival for 30 d. when it could be easily caught by hand and measured for length (similar to stage 4 anesthesia as described by Summerfelt and Smith 1990). The fish were limp upon removal from the anes- METHODS thetic solution. Juvenile lampreys were then placed in a 19-L Animal collection and husbandry.—Juvenile Pacific Lam- recovery bucket with aerated freshwater, and time to recovery preys were collected from the John Day Dam juvenile bypass and general behavior after recovery were recorded. We made facility on the Columbia River in May 2011 and transported to qualitative assessments of fish activity during recovery. The fish the Columbia River Research Laboratory in Cook, Washington, was considered recovered when it resumed normal swimming in an insulated tank with supplemental oxygen. Fish were behavior and swam away from a hand inserted into the recovery bucket. We kept the bulk anesthetic solutions, exposure contain- maintained in four 1.2-m-diameter circular tanks (about 240 ◦ lampreys per tank) with aerated water from the Little White ers, and recovery containers in a water bath at 12 C to maintain Salmon River under a simulated ambient photoperiod with temperature and monitored the temperature and DO of solutions overhead incandescent lights. Water was heated to 12◦C (range: in the exposure and recovery containers during all trials. Wa- ◦ ± ◦ ± 11.9–12.5 C) with single-pass electric heaters and then passed ter temperature was 12.0 0.2 C (mean SD) for all trials, ± through packed columns to dissipate excess dissolved gas. and DO content was 10.9 0.4 mg/L. We tested 15 juvenile Juvenile lampreys were acclimated in the laboratory for a Pacific Lampreys per concentration, and there was no signifi- minimum of 1 week before starting experiments. cant difference in length of fish used for trials between different Anesthetic efficacy trials.—We tested the efficacy of the fol- concentrations of the same anesthetic or between different anes- ± lowing four anesthetics: Finquel (herein referred to as MS- thetics as determined by one-way ANOVA. Mean SD length ± 222; Argent Chemical Laboratories, Redmond, Washington), was 145 10 mm. A fresh aliquot of anesthetic solution from BENZOAK (20% benzocaine; Frontier Scientific, Logan, the bulk solution was used for each fish, and we made a fresh Utah), AQUI-S 20E (10% eugenol; Western Chemical, bulk solution half way through the trials for each concentration. Ferndale, Washington), and Aquacalm (metomidate hydrochlo- Anesthetic health and survival trials.—The most effective ride; Western Chemical). BENZOAK and AQUI-S 20E were concentration for each anesthetic, based on time to a handleable used under the authorization of the U.S. Fish and Wildlife Ser- state and time to recovery from the three concentrations tested vice’s (USFWS) Investigational New Animal Drug (INAD) pro- in the efficacy trials, was selected for further analysis of short- gram (http://www.fws.gov/fisheries/aadap/national.htm). Both term juvenile Pacific Lamprey health, specifically susceptibility anesthetics are undergoing testing for U.S. Food and Drug Ad- to development of fungal infections, and survival after anesthetic exposure. We exposed fish to 100 mg/L of MS-222, 60 mg/L of Downloaded by [Department Of Fisheries] at 00:00 20 March 2013 ministration (FDA) approval for aquaculture applications. Preliminary trials were used to determine the appropriate BENZOAK, or 100 mg/L of AQUI-S 20E as described above range of doses for testing, and all anesthetic concentrations were with minor modifications (Aquacalm was not included in this calculated as milligrams per liter of active ingredient. Anesthet- experiment due to its poor performance in the efficacy trials). ics were tested at the following doses: Three juvenile lampreys were exposed at a time in 2 L of anesthetic until they were handleable. We recorded time in anesthetic MS-222: 50, 100, and 200 mg/L for each fish and measured fish length to the nearest millimeter BENZOAK: 40, 60, and 80 mg/L and mass to the nearest 0.1 g. Juvenile lampreys were placed AQUI-S 20E: 50, 100, and 200 mg/L in randomly selected aquaria containing about 19 L of aerated Aquacalm: 50, 75, and 100 mg/L water flowing at 0.7 L/min to recover. Control fish were placed in an exposure container with freshwater containing 0.025% The MS-222 was buffered 1:1 (w/w) with sodium bicarbonate Stress Coat for 4.6 min (the average time to a handleable state to maintain a neutral pH. High concentrations (100 mg/L) of for MS-222, BENZOAK, and AQUI-S 20E from the efficacy tri- Aquacalm were also acidic, so 1.2 mL/L of a 100-mg/L sodium als) and then transferred directly to an aquarium. We could not bicarbonate stock solution was added to the anesthetic bulk weigh or measure them since they were not anesthetized. Fif- solution to reach a near neutral pH. teen fish were placed in each aquarium with three groups of 15 ANESTHESIA IN PACIFIC LAMPREYS 271

fish per each anesthetic treatment or control. For comparisons honestly significant difference test was used to identify differ- of fish size between anesthetics, we first compared fish from ences between pairwise comparisons. All data sets used for different groups exposed to the same anesthetic and confirmed ANOVA were tested for normality using the D’Agostino and that there was no significant difference in size. We then pooled Pearson omnibus normality test, and the majority (75%) had fish treated with the same anesthetic for comparisons between a normal distribution. We used Kaplan–Meier curves to de- treatments. There was no significant difference in initial length scribe fish survival and compared curves between treatments or mass between fish exposed to different anesthetics as deter- using the log-rank (Mantel–Cox) test. GraphPad Prism software mined by one-way ANOVA. Lamprey length was 146 ± 9mm (GraphPad Software, La Jolla, California) was used for all (mean ± SD) and mass was 3.9 ± 0.8 g; the mean density of analyses, and the level of significance for all tests was juvenile lampreys in each aquarium was 3.1 g/L. We held the 0.05. lampreys for 30 d at an average temperature of 12.1◦C, ranging ◦ ◦ from 11.5 C to 14.6 C, and visually monitored the development RESULTS of fungal infections and survival. At the end of 30 d, surviving fish were weighed, measured, inspected for signs of infection, Identification of Effective Anesthetics for Juvenile Pacific and then euthanized. Lampreys Data analysis.—One-way ANOVA was used to compare MS-222.—Juvenile Pacific Lampreys were anesthetized sig- mean time to a handleable state and time to recovery between nificantly faster using 100 and 200 mg/L MS-222 than with different concentrations of the same anesthetic in efficacy trials 50 mg/L (ANOVA: F = 81.78; df = 2, 42; P < 0.0001; and time in anesthetic for health and survival trials. Tukey’s Figure 1). There was no significant difference in time to a Downloaded by [Department Of Fisheries] at 00:00 20 March 2013

FIGURE 1. Box and whisker plots of time to a handleable state and time to recovery for juvenile Paciﬁc Lampreys exposed to four different anesthetics. The boxes represent the 25th and 75th quartiles, the whiskers represent the 5th and 95th percentiles, and the line is the median. Different doses of the same anesthetic were compared using one-way ANOVA. Doses of the same anesthetic accompanied by different letters are signiﬁcantly different. 272 CHRISTIANSEN ET AL.

handleable state in lampreys between doses of 100 and dleable state and time to recovery, 100 mg/L MS-222 was the 200 mg/L, but time to recovery was significantly faster at 50 most effective anesthetic dose of those tested. and 100 mg/L than at 200 mg/L (ANOVA: F = 48.60; df = 2, 42; P < 0.0001; Figure 1). The most effective concentration Short-term Health and Survival of Juvenile Pacific of MS-222 tested was 100 mg/L with a time to a handleable Lampreys Exposed to Different Anesthetics state of 3.9 ± 0.5 min (mean ± SD) and a recovery time Mean ( ± SD) time in anesthetic for Pacific Lampreys ex- of 2.4 ± 0.5 min. At this concentration, fish were moderately posed to MS-222 was 4.8 ( ± 0.8) min and time ranged from agitated upon initial exposure. After recovery from anesthesia, 3.2 to 6.6 min. For fish exposed to BENZOAK and AQUI-S the fish resumed normal active swimming behavior and avoided 20E, mean time in anesthetic was 4.1 ( ± 0.8) min with a range capture. of 2.5–5.7 min and 7.5 ( ± 1.2) min with a range of 5.0–10.5 min. BENZOAK.—Time to a handleable state for lampreys treated There was no significant difference in exposure time between with BENZOAK was significantly faster at 60 and 80 mg/L than fish in different groups exposed to the same anesthetic except at 40 mg/L (ANOVA: F = 10.08; df = 2, 42; P = 0.0003; Fig- between two of the three groups exposed to AQUI-S 20E. Lam- ure 1). There was no significant difference in time to a handleable preys in one group were exposed 1.3 min longer on average than state between 60 and 80 mg/L or in time to recovery between all fish in the other group. three concentrations. The most effective concentration of BEN- Fungal infections appeared in one group of fish exposed ZOAK tested was 60 mg/L. Mean time to a handleable state to each anesthetic by day 6 or 7 postexposure (Figure 2). No was 3.8 ± 0.6 min and mean recovery time was 5.9 ± 1.0 min. control fish presented with a fungal infection until day 13. The Juvenile Pacific Lampreys were slightly to moderately agitated number of fish with fungal infections in groups exposed to upon exposure to BENZOAK, and after anesthetic exposure, the same anesthetic and in control groups was variable, but fish tended to remain attached to the recovery container wall by day 30 postexposure, more than 88% of juvenile Pacific unless prompted to swim. Lampreys exposed to MS-222 or AQUI-S 20E, or not exposed AQUI-S 20E.—Time to a handleable state for lampreys was to any anesthetic, either had a fungal infection or had died significantly faster with 100 and 200 mg/L AQUI-S 20E than with a fungal infection (Figure 2). Fish exposed to BENZOAK with 50 mg/L (ANOVA: F = 64.13; df = 2, 42; P < 0.0001; had the lowest incidence of fungal infection (57% at day 30), Figure 1), but there was no difference in induction time between primarily due to one group in which only one fish presented 100 and 200 mg/L. Time to recovery was significantly faster at with a fungal infection. All fish that died in all treatments had 50 and 100 mg/L than at 200 mg/L (ANOVA: F = 34.81; df = fungal infections. 2, 42; P < 0.0001; Figure 1). The most effective concentration Survival was highly variable between groups of Pacific Lam- of AQUI-S 20E tested was 100 mg/L with a mean time to a preys exposed to the same anesthetic and between control groups handleable state of 6.2 ± 0.8 min and a mean recovery time (Figure 3). For groups of control fish, survival ranged from 0% of 6.1 ± 1.8 min. Juvenile Pacific Lampreys were extremely to 93% at 30 d postexposure, and two groups had significantly agitated upon exposure to AQUI-S 20E, but fish showed normal higher survival than the third group (log-rank tests: P < 0.05). behavior after recovery. Fish exposed to BENZOAK had slightly higher survival, rang- Aquacalm.—Time to a handleable state using Aquacalm did ing from 27% to 93% for the three groups. Two groups of fish not differ significantly between the concentrations tested, but the exposed to BENZOAK had significantly higher survival than time to recovery was slower at 75 and 100 mg/L than at 50 mg/L the third group (log-rank tests: P < 0.05). Survival for groups (ANOVA: F = 20.23; df = 2, 42; P < 0.0001; Figure 1). No of lampreys exposed to MS-222 ranged from 0% to 60%, and effective concentration was identified since the fastest time to survival was significantly different between all groups (log-rank Downloaded by [Department Of Fisheries] at 00:00 20 March 2013 a handleable state was nearly 10 min with a recovery time of tests: P < 0.05). For groups of fish exposed to AQUI-S 20E, nearly 9 min. Juvenile Pacific Lampreys showed slight agitation survival ranged from 7% to 53%. Survival was not significantly upon initial exposure to Aquacalm. After recovery, fish were different between the two groups exposed to AQUI-S 20E for lethargic and slow to avoid capture. different lengths of time (log-rank test: P > 0.05), and these Comparison of the most effective anesthetic doses.—To de- two groups had significantly higher survival than the third group termine the most effective anesthetic dose overall, we compared (log-rank tests: P < 0.05). the most effective concentrations of MS-222, BENZOAK, and When survival data from all groups for each anesthetic were AQUI-S 20E and found that time to a handleable state for lam- combined (Figure 3), there was no significant difference be- preys did not differ between 100 mg/L MS-222 and 60 mg/L tween survival of control fish and fish exposed to BENZOAK BENZOAK. Both MS-222 and BENZOAK induced anesthe- (log-rank test: P > 0.05). Survival of juvenile Pacific Lam- sia more quickly than 100 mg/L AQUI-S 20E (ANOVA: F = preys exposed to AQUI-S 20E was significantly lower than that 64.86; df = 2, 42; P < 0.0001). Time to recovery for juvenile of control fish or fish exposed to BENZOAK (log-rank tests: Pacific Lampreys was fastest with 100 mg/L MS-222 (ANOVA: P < 0.05). Survival of lampreys exposed to MS-222 was signif- F = 42.77; df = 2, 42; P < 0.0001) and did not differ between icantly lower than that of fish exposed to any other anesthetic BENZOAK and AQUI-S 20E. In terms of both time to a han- tested or control fish (log-rank tests: P < 0.05). ANESTHESIA IN PACIFIC LAMPREYS 273

FIGURE 2. Number of juvenile Pacific Lampreys with fungal infections during a 30-d monitoring period after exposure to 100 mg/L MS-222, 60 mg/L BENZOAK, 100 mg/L AQUI-S 20E, or no anesthetic (control). Different shading patterns in each bar represent replicate groups of 15 fish each exposed to the same anesthetic. Daily values include live fish with a fungal infection and a cumulative total of fish that died with a fungal infection up to that day.

DISCUSSION initial exposure, and further work will be needed to document Our goal was to identify an anesthetic that would provide the physiological stress responses of juvenile Pacific Lampreys rapid induction and recovery in juvenile Pacific Lampreys and exposed to these chemicals. BENZOAK was effective at a lower would not increase disease susceptibility. We tested four anes- concentration than was MS-222 (60 mg/L versus 100 mg/L) and thetics used for anesthesia in fish—MS-222 (Finquel), benzo- caused less initial agitation, but fish recovered more slowly from caine (BENZOAK), eugenol (AQUI-S 20E), and metomidate exposure to BENZOAK than to MS-222. Rainbow Trout On- (Aquacalm; Ross and Ross 2008)—and showed that MS-222 corhynchus mykiss, hybrid Striped Bass (White Bass Morone chrysops × Striped Bass M. saxatilis), and Largemouth Bass

Downloaded by [Department Of Fisheries] at 00:00 20 March 2013 and BENZOAK were the most effective anesthetics (based on induction and recovery times and sedation behavior) for juve- Micropterus salmoides also recover more slowly from exposure nile Pacific Lampreys. The only anesthetic approved by the FDA to BENZOAK than to MS-222 (Gilderhus and Marking 1987; Center for VeterinaryMedicine for use in aquaculture is MS-222 Trushenski et al. 2012a, 2012b), but the mechanism for this and it is the most common anesthetic used in fisheries research. delayed recovery is unclear. However, it requires a 21-d withdrawal period after exposure Unlike MS-222 or BENZOAK, AQUI-S 20E excessively ag- during which fish must be held in isolation so that they metabol- itated juvenile Pacific Lampreys. They swam frantically often ically clear the anesthetic before they are returned to the wild. splashing and leaping or climbing out of the solution. Agita- This limits the use of MS-222 in field applications. tion has also been reported in Rainbow Trout treated with a Juvenile Pacific Lampreys exposed to MS-222 and related compound, AQUI-S (50% isoeugenol) at doses greater BENZOAK behaved similarly. This is probably because these than 40 mg/L active ingredient (Wagner et al. 2002), and Sink two compounds are chemically related. Benzocaine, the active et al. (2007) reported thrashing and violent head shaking in ingredient in BENZOAK, is the parent compound of MS-222 Rainbow Trout exposed to clove oil. Clove oil is the compound (Neiffer and Stamper 2009), and both induce general anesthe- from which eugenol and isoeugenol, the active ingredients in sia by depressing the central nervous system (Ross and Ross AQUI-S 20E and AQUI-S, are derived and has been recom- 2008). These anesthetics caused some agitation in the fish upon mended for anesthesia in a wide variety of fishes (Ross and 274 CHRISTIANSEN ET AL.

FIGURE 3. Survival for 30 d of juvenile Paciﬁc Lampreys exposed to 100 mg/L MS-222, 60 mg/L BENZOAK, 100 mg/L AQUI-S 20E, or no anesthetic (control). Gray Kaplan–Meier curves show survival of three replicate groups of 15 ﬁsh each per anesthetic. The smooth black lines represent the mean survival of the three groups exposed to the same anesthetic.

Ross 2008). It is not approved by the FDA, and all fish must solution would often begin moving again when handled or ex- be euthanized after exposure to clove oil, precluding its use posed to air, and it was difficult to achieve stage 4 anesthesia in field applications. Clove oil has been used to sedate adult (Summerfelt and Smith 1990) with this anesthetic. Also, the Pacific Lampreys with no reported agitation (Moser and Close recovery time was slow suggesting that even if more rapid in- Downloaded by [Department Of Fisheries] at 00:00 20 March 2013 2003) and was recommended for adult lampreys because they duction could be achieved at higher doses, the recovery time appeared to recover quickly (Moser et al. 2007). Based on our would still be beyond the desired range. Aquacalm is not ap- results, however, juvenile Pacific Lampreys were anesthetized proved by the FDA for use in aquaculture, but it can be legally slowly with AQUI-S 20E and recovered more slowly than when used for ornamental finfish. Its active ingredient, metomidate, exposed to MS-222. Slow recovery times have also been ob- is a hypnotic and does not depress the central nervous system served for other species of fish anesthetized with clove oil, in- like general anesthetics do (see Neiffer and Stamper 2009 for cluding juvenile Rainbow Trout (Anderson et al. 1997; Keene review). It is also notable for its suppression of cortisol synthesis et al. 1998) and Red Pacu Piaractus brachypomus (Sladky et al. through inhibition of the hypothalamo-pituitary-interrenal axis 2001), and for hybrid Striped Bass and Largemouth Bass anes- (Thomas and Robertson 1991; Olsen et al. 1995). These charac- thetized with AQUI-SE (50% eugenol; Trushenski et al. 2012a, teristics may explain why juvenile Pacific Lampreys were very 2012b). calm when exposed to metomidate but could not be adequately Of all the anesthetics tested, Aquacalm agitated the fish least, anesthetized for handling and tagging. but it was eliminated from consideration due to slow induction For all anesthetics tested, juvenile Pacific Lampreys required and recovery and inadequate anesthesia for handling. During higher concentrations for anesthesia to reach a handleable state the efficacy trials, fish that were motionless in the Aquacalm than the doses recommended for salmonids, which may be due ANESTHESIA IN PACIFIC LAMPREYS 275

to their ancient evolutionary status. Less-evolved species often personal communication). High-density holding conditions in require higher anesthetic doses to achieve adequate anesthesia, laboratory tanks provide excellent breeding grounds for disease which may be related to a decrease in number or specificity (Munro 1990), and a small increase in disease susceptibility of cellular receptors for anesthetic compounds (Ross and Ross could result in a large increase in fungal infections as fish easily 2008). An ideal anesthetic, as defined by Marking and Meyer spread disease within the confines of a laboratory tank. This (1985), is one that induces anesthesia within 3 min with minimal may also explain the variability we observed between groups hyperactivity or stress and from which the fish can fully recover within the same treatment. Until field studies provide further to normal swimming behavior within 5 min. These time limits insight, our results support the use of 100 mg/L MS-222 and may not be realistic for less-evolved species such as lampreys, 60 mg/L BENZOAK for handling procedures that require the but we should strive to use the lowest anesthetic concentrations lamprey to be completely still with the caveat that exposure possible with reasonable induction and recovery times to mini- to MS-222 may increase the incidence of fungal infections in mize risk to the fish and user and to reduce costs. juvenile Pacific Lampreys held in the laboratory in freshwater. In this study we tested only three doses per anesthetic and focused on a specific level of anesthesia. The level of anesthesia (and thus anesthetic concentration) will depend upon the ACKNOWLEDGMENTS requirements of the handling procedure. Lower concentrations We thank Lisa Weiland for technical assistance, Greg may be adequate for handling procedures that do not require Kovalchuk for assistance in lamprey collection, and Sean Tack- the fish to be completely still. Temperature is also an important ley of the U.S. Army Corps of Engineers for administering this factor in anesthetic efficacy, and for MS-222 and BENZOAK, project. Any use of trade names is for descriptive purposes only anesthetic effects may be enhanced at lower temperatures (Ross and does not imply endorsement by the U.S. Government. and Ross 2008). We held the temperature constant in these experiments to allow for comparison between anesthetics. More REFERENCES work will be needed to determine whether anesthetic concen- Anderson, W.G., R. S. McKinley, and M. Colavecchia. 1997. The use of clove oil trations for juvenile Pacific Lampreys should be varied with as an anesthetic for Rainbow Trout and its effects on swimming performance. temperature. North American Journal of Fisheries Management 17:301–307. Besides testing the effectiveness of anesthetics (i.e., time to a Close, D. A., M. S. Fitzpatrick, and H. W. Li. 2002. The ecological and cultural handleable state, time to recovery, and sedation and recovery be- importance of a species at risk of extinction, Pacific Lamprey. Fisheries 27(7):19–25. havior), we also tested MS-222, BENZOAK, and AQUI-S 20E Gilderhus, P. A., and L. L. Marking. 1987. Comparative efficacy of 16 anes- for their effects on development of fungal infections and survival thetic chemicals on Rainbow Trout. North American Journal of Fisheries in the weeks after exposure. Fish exposed to BENZOAK had Management 7:288–292. the fewest fungal infections and the highest survival, suggesting Keene, J. L., D. L. G. Noakes, R. D. Moccia, and C. G. Soto. 1998. The that it is a safer anesthetic (in terms of long-term survival) for efficacy of clove oil as an anesthetic for Rainbow Trout, Oncorhynchus mykiss (Walbaum). Aquaculture Research 29:89–101. juvenile lampreys than MS-222 or AQUI-S 20E. Fish exposed Marking, L. L., and F. P. Meyer. 1985. Are better anesthetics needed in fisheries? to AQUI-S 20E also survived better and had lower rates of Fisheries 10(6):2–5. fungal infection than fish exposed to MS-222, but the improve- Meeuwig, M. H., A. L. Puls, and J. M. Bayer. 2007. Survival and tag retention ment in survival did not outweigh a major disadvantage of this of Pacific Lamprey larvae and macrophthalmia marked with coded wire tags. anesthetic—excessive agitation. Unless it can be demonstrated North American Journal of Fisheries Management 27:96–102. Mesa, M. G., E. S. Copeland, H. E. Christiansen, J. L. Gregg, S. R. Roon, and that the extreme agitation observed is not physiologically detri- P. K. Hershberger. 2012. Survival and growth of juvenile Pacific Lampreys mental to the fish, our results do not support the use of AQUI- Downloaded by [Department Of Fisheries] at 00:00 20 March 2013 tagged with passive integrated transponders (PIT) in freshwater and seawater. S 20E for anesthesia of juvenile Pacific Lampreys. Lampreys Transactions of the American Fisheries Society 141:1260–1268. exposed to MS-222 performed poorly in the health and sur- Moser, M. L., J. M. Butzerin, and D. B. Dey. 2007. Capture and collection of vival trial, though this compound is highly effective in inducing lampreys: the state of the science. Reviews in Fish Biology and Fisheries 17:45–56. anesthesia. Moser, M. L., and D. A. Close. 2003. Assessing Pacific Lamprey status in the The reason for poor survival of juvenile Pacific Lampreys Columbia River basin. Northwest Science 77:116–125. exposed to MS-222 in the laboratory is unknown. We think Mueller, R. P., R. A. Moursund, and M. D. Bleich. 2006. Tagging juvenile that fungal infections may be specific to laboratory holding Pacific Lamprey with passive integrated transponders: methodology, short- conditions, but they could be worsened by anesthetic expo- term mortality, and influence on swimming performance. North American Journal of Fisheries Management 26:361–366. sure. Field studies will be needed to clarify whether the fun- Munro, A. L. S. 1990. Salmon farming. Fisheries Research 10:151–161. gal infections observed after anesthetic exposure are a risk Neiffer, D. L., and M. A. Stamper. 2009. Fish sedation, anesthesia, analgesia, for juvenile lampreys that are handled and then returned to and euthanasia: considerations, methods, and types of drugs. ILAR (Institute the wild. Wild, outmigrating juvenile Pacific Lampreys col- for Laboratory Animal Research) Journal 50:343–360. lected in the same year and from the same site (John Day Olsen, Y. A., I. E. Einarsdottir, and K. J. Nilssen. 1995. Metomidate anaesthesia in Atlantic Salmon, Salmo salar, prevents plasma cortisol increase during Dam juvenile bypass facility) as the fish in our study had stress. Aquaculture 134:155–168. low incidence (0.2% of 4,200 fish examined) of fungal infec- Ross, L. G., and B. Ross. 2008. Anaesthetic and sedative techniques for aquatic tions (G. Kovalchuk, Pacific States Marine Fish Commission, animals, 3rd edition. Blackwell Publishing, Oxford, UK. 276 CHRISTIANSEN ET AL.

Schreck, C. B., M. S. Fitzpatrick, and D. L. Lerner. 1999. Determination Thomas, P., and L. Robertson. 1991. Plasma cortisol and glucose stress re- of passage of juvenile lamprey: development of a tagging protocol. Ore- sponses of Red Drum (Sciaenops ocellatus) to handling and shallow water gon Cooperative Fish and Wildlife Research Unit, Oregon State University, stressors and anesthesia with MS-222, quinaldine sulfate, and metomidate. Corvallis. Aquaculture 96:69–86. Sink, T. D., R. J. Strange, and R. E. Sawyers. 2007. Clove oil used at lower Trushenski, J. T., J. D. Bowker, B. R. Gause, and B. L. Mulligan. 2012a. concentrations is less effective than MS-222 at reducing cortisol stress re- Chemical and electrical approaches to sedation of hybrid Striped Bass: in- sponses in anesthetized Rainbow Trout. North American Journal of Fisheries duction, recovery, and physiological responses to sedation. Transactions of the Management 27:156–161. American Fisheries Society 141:455–467. Sladky, K. K., C. R. Swanson, M. K. Stoskopf, M. R. Loomis, and G. A. Trushenski, J. T., J. D. Bowker, B. L. Mulligan, and B. R. Gause. 2012b. Lewbart. 2001. Comparative efﬁcacy of tricaine methanesulfonate and clove Induction, recovery, and hematological responses of Largemouth Bass to oil for use as anesthetics in Red Pacu (Piaractus brachypomus). American chemo- and electrosedation. North American Journal of Aquaculture 74: Journal of Veterinary Research 62:337–342. 214–223. Summerfelt, R. C., and L. S. Smith. 1990. Anesthesia, surgery, and re- Wagner, E., R. Arndt, and B. Hilton. 2002. Physiological stress responses, egg lated techniques. Pages 213–272 in C. B. Schreck and P. B. Moyle, ed- survival and sperm motility for Rainbow Trout broodstock anesthetized with itors. Methods for ﬁsh biology. American Fisheries Society, Bethesda, clove oil, tricaine methanesulfonate or carbon dioxide. Aquaculture 211: Maryland. 353–366. Downloaded by [Department Of Fisheries] at 00:00 20 March 2013 This article was downloaded by: [Department Of Fisheries] On: 20 March 2013, At: 00:01 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 Optimal Effort Intensity in Backpack Electrofishing Surveys L. W. Stanfield a , N. P. Lester b & I. C. Petreman c a Ontario Ministry of Natural Resources, Southern Science and Information Section, 41 Hatchery Lane, Picton, Ontario, K0K 2T0, Canada b Ontario Ministry of Natural Resources, Aquatic Research and Development Section, Harkness Laboratory of Fisheries Research, 2140 East Bank Drive, Peterborough, Ontario, K9J 7B8, Canada c Ontario Ministry of Natural Resources, Southern Science and Information Section, 300 Water Street, Peterborough, Ontario, K9J 8M5, Canada Version of record first published: 05 Mar 2013.

To cite this article: L. W. Stanfield , N. P. Lester & I. C. Petreman (2013): Optimal Effort Intensity in Backpack Electrofishing Surveys, North American Journal of Fisheries Management, 33:2, 277-286 To link to this article: http://dx.doi.org/10.1080/02755947.2012.758200

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 33:277–286, 2013 C American Fisheries Society 2013 ISSN: 0275-5947 print / 1548-8675 online DOI: 10.1080/02755947.2012.758200

ARTICLE

Optimal Effort Intensity in Backpack Electroﬁshing Surveys

L. W. Stanﬁeld* Ontario Ministry of Natural Resources, Southern Science and Information Section, 41 Hatchery Lane, Picton, Ontario K0K 2T0, Canada N. P. Lester Ontario Ministry of Natural Resources, Aquatic Research and Development Section, Harkness Laboratory of Fisheries Research, 2140 East Bank Drive, Peterborough, Ontario K9J 7B8, Canada I. C. Petreman Ontario Ministry of Natural Resources, Southern Science and Information Section, 300 Water Street, Peterborough, Ontario K9J 8M5, Canada

Abstract We evaluated the effect of backpack electrofishing effort intensity on the precision of estimates of fish density in a southern Ontario stream. Single-pass electrofishing was conducted on 30 sites at three electrofishing effort intensities (5, 10 and 15 s/m2). We found that an effort of 5 s/m2 yielded an average catch that was 78% of the 15-s/m2 effort. An asymptotic model effectively described the catch–effort relationship for most fish taxa in the stream (i.e., Rainbow Trout Oncorhynchus mykiss,BrownTroutSalmo trutta, sculpin Cottus sp., dace Rhinichthys sp., and darter Etheostoma sp.). Using the catchability parameters of this model, we evaluated the trade-off between sampling many sites at low intensity or fewer sites at higher intensity. The survey design that maximizes precision of fish-density estimates depends on the average time spent traveling between sites and the average area of sites. For southern Ontario streams, where sample sites were approximately 300 m2 and travel time between sites was 75 min, the optimum electrofishing effort intensity was approximately 5 s/m2. The applicability of these results to other systems was demonstrated by showing how this optimum intensity was affected by differences in catchability rates of fish and travel distances. These findings will be used to standardize single pass electrofishing catches in nonshield areas of Ontario, and the approach may prove useful in other areas where this fishing technique is effective. Downloaded by [Department Of Fisheries] at 00:01 20 March 2013

In southern Ontario streams, there is a tradition of conduct- and more recent findings (Mitro and Zale 2000; Arnason et al. ing backpack electrofishing surveys on mesoscale habitat units 2005; Bateman et al. 2005), routine sampling in Ontario is now in the range of 2–10 stream widths or 40–100 m in length. based on single-pass surveys (Stanfield et al. 1999). These sur- Traditionally, researchers applied removal methods intended to veys are intended to provide reliable estimates of fish density ensure catchability of fish in numbers that are appropriate to and species richness at a much reduced cost. generate population estimates (Zippin 1958). These methods Defining the reliability of a measure of species richness re- are intensive and require considerable effort to obtain data for a quires spatial context because rare species are always difficult to single site. Fortunately, several studies have shown that single- count. For long stretches of river (35 bankfull widths), Simonson pass catch data provide reliable estimates of three-pass results and Lyons (1995) showed that single intensive versus multiple- for species with high catchabilities, like salmonids (Jones and pass surveys can produce reliable measures of species rich- Stockwell 1995; Kruse et al. 1998). As a result of those studies ness, even for those species with generally lower catchability.

*Corresponding author: les.stanﬁ[email protected] Received January 5, 2012; accepted December 7, 2012 277 278 STANFIELD ET AL.

Kennard et al. (2006) found that species richness can be quite effectively divides stratum A, which additionally contains Rain- variable when measured at smaller spatial scales and suggested bow Darter Etheostoma caeruleum, Longnose Dace Rhinichthys that multiple-pass electrofishing surveys combined with seine cataractae, and Western Blacknose Dace R. obtosus from stra- hauls be conducted to capture an accurate measure of this met- tum B, which is largely devoid of these species. Strata A and B ric in riffles or pools. Given that the Ontario protocol dictates both contain Mottled Sculpin Cottus bairdii. Stratum C signifies sampling sites that are generally shorter in length than are rec- the beginning of the colder headwaters zone where the substrate ommended by Simonson and Lyons (1995), but are larger and sharply changes from till to morainal materials. Here, Mottled intended to capture at least one riffle–pool sequence, the degree Sculpin is replaced by Slimy Sculpin C. cognatus, and American to which measures of species richness are reliable is unknown. Brook Lamprey Lampetra appendix is more common. To be reliable, single-pass surveys must still meet the as- We sampled 30 randomly selected sites each of which had sumptions of equal probability of capture during a survey event a beginning and ending at a crossover (a location in the stream (Zippin 1958). However, we know that catchability may de- where the thalweg was located in the center of the channel) crease with increased density, because the success of surveyor and was a minimum of 40 m in length (Stanfield et al. 1999). netting declines due to swamping effects (Peterson et al. 2004). The length of sites ranged from 40 to 69 m (mean = 48.9 m), In Ontario, this challenge was addressed by recommending in- width ranged from 3.3 to 8.8 m (mean = 6.4 m), and surface tensive levels of effort (14–21 s/m2) to ensure swamping ef- area ranged from 134 to 407 m2 (mean = 308 m2). Sites were fects were minimized and all habitats were sampled sufficiently electrofished three times, once at each of three effort intensities (Jones and Stockwell 1995). In practice, however, the actual (5, 10, or 15 s/m2). Sampling was conducted between July 3 and effort exerted by crews has rarely approached the lower limit September 2, and the time interval between successive sampling of the Jones and Stockwell (1995) recommendation and has at each site was approximately 1 month. The order of effort been as low as 2 s/m2 (OMNR 2008). We were therefore con- intensity was randomized at each site, but balanced across all cerned that sampling at such low intensities may not provide sites to control for possible order effects (i.e., temporal trend in reliable measures of fish abundance or species richness. Also, fish abundance). if these lower sampling intensities were found to be effective, All electrofishing was performed using a Smith-Root model then managers should be made aware of that so that optimum 12b backpack electrofisher. Sampling crews included a back- survey designs can be implemented. We therefore designed a pack operator, two netters, and an additional person on shore study to measure the effect of electrofishing effort intensity on who sorted and tallied the catch by species and then released the catches of different species. them back into the stream. Sampling of a site began at the down- Our study had two primary objectives. First, we attempted stream end and proceeded upstream and sufficient power was to model the relationship between catch and effort intensity so applied so that fishes of all sizes were sufficiently stunned for that we could develop correction factors to standardize histori- netting at a rate that depended on the prescribed sampling inten- cal catches and produce comparable indices of fish abundance. sity (5, 10, or 15 s/m2). Crews were free to allocate electrofishing Second, we evaluated the implications of this relationship in duration of microhabitats to maximize catches within a site and terms of optimum survey design. We recognized that the overall to ensure total effort intensity in each site met the target intensity. precision of abundance indices depends on the total number of Target effort durations were within 2% of observed for all sites. sites sampled and the precision of estimates obtained at each Block nets were not used to constrain fish within site boundaries. site. Higher effort intensity will produce more precise estimates Mortalities occasionally occurred in age-0 fish (1.1% of total at each site, but reduce the total number of sites that could be catch); however, the length of time between sampling events sampled given a fixed survey duration; more sites could be sam- was sufficient to enable replacement of these few fish. Downloaded by [Department Of Fisheries] at 00:01 20 March 2013 pled if effort intensity was reduced. We modeled this trade-off Effect of effort intensity on catch.—We assumed the catch to estimate the optimum effort intensity. (Ci) obtained from electrofishing a site was related to fish abundance as

METHODS Ci = qYi (1) Study area.—Sampling was conducted in 1999 on Wilmot 2 Creek, a moderately sized (89 km watershed area), wadeable, where Yi is the number of fish inhabiting site i and q is the catcha- coldwater stream flowing into Lake Ontario (Figure 1). Wilmot bility coefficient, which is expected to vary with effort intensity. Creek’s fish community is stratified by three unique habitat We assumed effort intensity (E) would have an asymptotic effect zones (identified by sites codes demarcated as A, B, or C; Fig- on catchability: ure 1). All zones contain resident migratory salmonids consist- ing of (in order of greatest to least abundance): Rainbow Trout q = 1 − e−bE, (2) Oncorhynchus mykiss, Brown Trout Salmo trutta, Coho Salmon O. kisutch, and Chinook Salmon O. tshawytscha. A low-head where b dictates how quickly Ci approaches Yi as E increases. barrier located approximately 4.5 km upstream from the mouth In this respect, b is analogous to relative catchability when fish BACKPACK ELECTROFISHING SURVEYS 279 Downloaded by [Department Of Fisheries] at 00:01 20 March 2013

FIGURE 1. Study area and electroﬁshing site locations on Wilmot Creek, Ontario. 280 STANFIELD ET AL.

species are compared. The catch obtained from site i when Because catch variance in fish surveys is typically related to fishing occurred at intensity E is then mean catch by a power function (Taylor 1961; Green 1979; Lester et al. 1996), we assume −bE Ci,E = (1 − e )Yi , (3) 2 = ¯ d sC aC (10) and the expected mean catch from fishing intensity E is and, given C¯ = qY¯, −bE C¯ E = (1 − e )Y¯. (4) 2 = ¯ d sC a(qY ) (11) Given that each site was sampled at three intensities (E = 5, 10, and 15 s/m2), a regression of catch and species richness versus 2 Substituting for sC in equation (9) gives fishing intensity (E) was performed to estimate the values of b and Y¯. This analysis assumed a lognormal error distribution 2 aY¯ d N (because catch variance increases with the mean) and therefore var(Y ) = . (12) − −bE 2−d used a logarithmic model to estimate parameters: n(1 e )

− Given that total time (T) to conduct the survey is fixed, the log(C¯ ) = log((1 − e bE)Y¯) + ε. (5) E number of sites sampled (n) will depend on effort intensity (E) We used the Levenberg–Marquardt, least-squares, nonlinear re- applied to the (mean) area of each site A and the travel time gression method (StatSoft Statistica version 10) to estimate the between sites Ttravel as follows: parameters for the following fish groups: Rainbow Trout, Brown T Trout, sculpin, dace, darter, and lamprey. Estimations were also n = . (13) T + EA performed for the total catch of all species and for species rich- travel ness. In this analysis, we only included the sites where the Substituting for n in equation (12) yields modeled fish group was caught at each effort intensity. Optimum effort intensity.—With a survey objective of obtain- + ¯ d 2 ing an index of fish abundance or richness, the optimum effort = (Ttravel AE) aY N var(Y ) − − (14) intensity is the value that minimizes variance of this index. We (1 − e bE)2 d T developed a formula describing how variance depended on catchability and the number of sites sampled (i.e., sample To account for the effect of effort intensity on variance, we ig- size). Because catchability and sample size both depend on nored the expression in the right hand parentheses because these effort intensity, we expanded this formula to obtain an explicit variables (N, T, and Y¯) are fixed. This results in the following expression relating effort intensity to variance of the abundance “relative variance” formula: estimate. + Given that the survey area contains N sites of which n are (Ttravel AE) relvar(Y ) = . (15) randomly selected for sampling, abundance (Y) is estimated as (1 − e−bE)2−d

N n This formula clearly identifies the parameters that must be es- Y = Y , (6) n i timated to assess how effort intensity affects survey efficiency. Downloaded by [Department Of Fisheries] at 00:01 20 March 2013 i=1 These parameters include the coefficient of the variance–mean relationship (d), as well as physical parameters related to the and variance of this estimate is stream sample design (area of sample sites and travel time be- 2 2 tween sites). For Wilmot Creek, mean area of sample sites was N sY var(Y ) = , (7) 308 m2 and mean travel time between sites was estimated as n 4,500 s, including preparation time at each site. 2 To estimate coefficients of the variance–mean relationship, where sY is the among-site variance in abundance. Given equation (1), this variance can be expressed as we computed the mean and variance of the catch from sites within each stratum (downstream, midstream, and upstream) s2 for each effort intensity. These calculations were done for the s2 = c . (8) Y q2 six frequently encountered fish taxa previously listed. We performed least-squares linear regression of log-transformed values Substituting equation (8) into equation (7) gives to estimate parameters of the variance–mean relationship (a, d). This analysis pooled mean catch and variance estimates for dif- N 2s2 ferent kinds of fish because the number of strata (3) was too var(Y ) = C . (9) nq2 small to describe the relationship for each kind. An ANCOVA BACKPACK ELECTROFISHING SURVEYS 281

was used to evaluate whether the variance–mean relationship Optimum Effort Intensity was affected by electrofishing intensity. As expected, variance of the catch was positively related to the mean (Figure 3) and the variance–mean relationship was RESULTS not affected by electrofishing intensity. Regression of the log- transformed values (for all fish groups and effort intensities) Effect of Effort Intensity on Catch resulted in The surveys captured 24 fish species, but the catch was dom- 2 = . + . ¯, inated by relatively few taxa (Table A.1). Six fish “groups” log10 sC 0 40 1 47 log10 C (16) accounted for 94% of the catch: Rainbow Trout (34%), sculpin 2 (23%), dace (20%), Brown Trout (8%), darter (6%) and lam- for which F1, 41 = 1,214, r = 0.96, SE = 0.217, and P < 0.001. prey (1%). Mean catch per site increased with effort intensity Separate regressions for each effort intensity produced three but this increase was not proportional to the change in effort models with very similar slopes and intercepts, suggesting that (Figure 2). As effort increased from 5 to15 s/m2, the mean catch effort intensity did not affect the variance–mean relationship. of all species increased from 132 to 182. The catch at an effort The homogeneity of slopes (ANCOVA) corroborated that ef- 2 2 of 5 s/m was 73% of the catch at an effort of 15 s/m , implying fort had no detectable effect on this relationship (F2 = 0.078, a threefold increase in effort was accompanied by a 1.4-fold P = 0.93). increase in the catch. The slope of equation (16) estimates parameter d, which is The effect of fishing effort intensity on catch was well de- needed to estimate relative variance. Substituting a slope of 1.47 scribed by an asymptotic model of catchability (q = 1 − e−bE) for d into equation (15) gives for several fish groups (Table 1; Figure 2). Regression of the (T + AE) catch against effort produced significant estimates of b that relvar(Y ) = travel . (17) ranged from 0.144 for sculpin to 0.256 for Rainbow Trout and (1 − e−bE)0.53 0.300 for dace. For these taxa, mean catch increased consistently with effort intensity. Estimates of b were not significant The contour map of this function in respect to b and E where for Brown Trout, darter, and lamprey, indicating that the asymp- Wilmot Creek’s average between-site travel time and average = = 2 totic model was not a good fit. For Brown Trout, the maximum site area were used (Ttravel 4,500 s, A 308 m ) is displayed mean catch was observed at the intermediate effort intensity; for in Figure 4A. Minimum relative variance (optimum electrofish- darters, the minimum mean catch was observed at the interme- ing effort) can be followed for any value of the b parameter. diate intensity. For lampreys, catch increased consistently with Relative variance is seen to be minimal when electrofishing ef- 2 effort intensity but in a seemingly linear manner; there was no fort intensity is 3–7 s/m for our pertinent range of b values evidence of the asymptotic relationship that was seen for other (Table 1). A tripling of the travel time (Figure 4B) shows a dis- taxa. Species richness was also asymptotic. In fact, it held the tinct shift in optimum effort intensity at the lowest catchability 2 highest significant value for b of 0.536 as its mean rose subtly coefficients (i.e., E = 14 s/m for b = 0.1). For most of the target from 6.4 species per site at low effort intensity to 6.8 at medium fish in our study, optimal effort intensity would still be low. For effort intensity. Moving from medium to high effort showed no example, the Rainbow Trout optimum electrofishing intensity is 2 increase in mean species richness. 8s/m in this scenario. Alternatively, when the travel time : site area ratio is reversed (and travel time is a third of its original value; Figure 4C) the optimal effort drops lower than 5 s/m2. TABLE 1. Estimates of the coefficients b and Y¯ in the catchability model and Downloaded by [Department Of Fisheries] at 00:01 20 March 2013 resultant optimal effort intensity for each b. Number of sites (n) is the number where the specified fish group was captured at all effort intensities. P-values DISCUSSION = provided for b display significance for the chi-square test (Ho: b 0). Model Effect of Effort Intensity on Catch results are not reported for lampreys because the model consistently failed to estimate both parameters. We demonstrated that total fish catch from electrofishing a site increased nonlinearly with effort intensity. For most taxa, ¯ 2 Fish group nb YPb Eopt (s/m ) the relationship between numbers of fish caught and effort Rainbow Trout 29 0.256 64.419 0.008 5.18 intensity was modeled by a power function, and generally more than 70% of the ascent to the asymptote resulted from fishing Brown Trout 22 0.443 17.521 0.242 3.84 2 Sculpin 30 0.144 51.638 0.053 6.85 only 5 s/m . As expected, there was considerable variation in Darter 11 3.169 27.030 >0.999 <3.43 the rate of ascent to the asymptote that is a reflection of each Dace 11 0.300 96.043 0.057 4.76 taxa’s catchability. Lamprey 5 Fish that reside in hydraulic refuge areas, such as darters, are Species richness 30 0.536 6.834 0.033 3.43 highly vulnerable to electrofishing and had steep ascent rates, Total catch 30 0.246 183.25 0.002 5.29 such that there was no difference between catches at the three effort intensities tested here. Conversely, the ascent rate was 282 STANFIELD ET AL. Downloaded by [Department Of Fisheries] at 00:01 20 March 2013

FIGURE 2. Modeled electrofished catch and species richness. Function denotes best fit of parameters b and Y¯ from equation (13) as estimated by Levenberg– Marquardt, least-squares, nonlinear regression for each fish group (and species richness). Data points for species richness are frequency-sized for frequencies ranging from 1 to 7 (smallest to largest circle sizes). The model failed to fit to the lamprey catch data, probably due to low degrees of freedom. Refer to Table 1 for precise model results. BACKPACK ELECTROFISHING SURVEYS 283

FIGURE 3. Relationship between the variance and mean for electrofishing catch by site. Variance and mean were calculated for different fish groups and different effort intensities within each of three spatial strata. Values for different effort intensities are designated by different symbols. Least-squares linear regression results for the combined data set are presented with the 95% confidence interval represented by the dotted lines. low for benthic specialists, such as sculpins, and only 51% of this catch were captured at the lowest effort. Sculpins are difficult to sample by electrofishing because they lack air bladders and also often do not exhibit galvanotaxis in response to the electric current, often staying at on or near the river bed when shocked, making them difficult to see or catch. This slow rise to the asymptote is consistent with observations by Mahon (1980), who noted that removal surveys typically failed to measure declining populations for this group of fish. Fish that are associated with cover, such as young salmonids and dace, displayed moderately rapid ascent to the asymptote with typically 72–89% of this catch occurring at 5 s/m2 effort. These fish tend to “hold” in areas with cover until they are forced by electrofishing to move and exhibit strong galvanotaxis. That no asymptotic response was observed for lampreys was not a surprise as special techniques (pulsing power on and off) are employed to coax these fish out of their burrows, prior to attempting to catch them. Our results

Downloaded by [Department Of Fisheries] at 00:01 20 March 2013 suggest that these approaches are not effective means to quantify lamprey abundance and we would therefore recommend more specialized equipment for targeted surveys for this fish group. Finally, that around 93% of the catch contributing to the species richness metric was captured at the lowest effort intensity is re- assuring to surveyors interested in this commonly used metric. FIGURE 4. Surface plot of relative variance calculations (equation 16). Re- sultant relative variance (y-axis) is pictured as a function of “catchability” (b, Our modeling of the relationship between catch and effort in- x-axis) and electrofishing effort intensity (E, z-axis). For any given b, optimum tensity was successful for our most commonly encountered fish electrofishing effort is found at its minimal relative variance value. (A)The groups. The resultant model equations may be used to standard- function as calculated with Wilmot Creek’s values for average travel time and ize historical catches with varying electrofishing effort intensi- site area (4,500 s and 308 m2, respectively). (B) The same function as in (A), ties in Ontario streams, thereby fulfilling the primary objective but the travel time has been tripled. (C) Travel time was reverted to 4,500 s, but the mean site area has been tripled. White lines highlight the three treatment of this study. intensities (E = 5, 10, and 15 s/m2). Optimum Effort Intensity The optimum electrofishing effort intensity was approximately 5 s/m2 for Wilmot Creek. There are four variables within 284 STANFIELD ET AL.

the derived formula of optimum effort intensity for managers project is a continuation of the many years of work that have to consider when attempting to transport this method to other been carried out by various field technicians, data analysts, and systems: (1) the “catchability” of the metric of interest (b), (2) biologists associated with the Wilmot Creek Fisheries Project. the variance–mean relationship of the catch, (3) average travel Scott Gibson and Bruce Kilgour assisted with some preliminary time between sites, and (4) average site area. The catchability analyses. Helpful comments to an earlier draft of this manuscript parameter, or more specifically, how quickly the metric of inter- were received from Tim Haxton, three anonymous reviewers, est reaches an asymptote, is difficult to predict without some a and the associate editor of the journal. priori knowledge of the system. This parameter may always be high for simplified species counts (species richness) and may consistently be low for estimating populations of rare fishes. REFERENCES Given this, a general understanding of the stream and its fish Arnason, F., T. Antonsson, and S. M. Einarsson. 2005. Evaluation of single- pass electric fishing to detect changes in population size of Atlantic Salmon community may be required before one can design an optimum (Salmo salar L.) juveniles. Icelandic Agricultural Sciences 18:67–73. sampling strategy using this method. The second consideration, Bateman, D. S., R. E. Gresswell, and C. E. Torgersen. 2005. Evaluating single- the variance–mean relationship of catch, is always expected to pass catch as a tool for identifying spatial pattern in fish distribution. Journal be log-linear, as is typical of abundance metrics of aggregated of Freshwater Ecology 20:335–345. animals (Taylor 1961; Gaston and McArdle 1994). The slope Gaston, K. J., and B. H. McArdle. 1994. The temporal variability of animal abundances: measures, methods and patterns. Philosophical Transactions of of this relationship (d) is consequential to the calculation of the Royal Society of London B 345:335–358. optimum effort intensity, yet the magnitude of d is expected Green, R. H. 1979. Sampling design and statistical methods for environmental to be largely unchanging for backpack electrofishing catches. biologists. Wiley, New York. O’Brien et al. (2001) found it to be in the range of 1.28–1.46 for Jones, M. L., and J. D. Stockwell. 1995. A rapid assessment procedure for the catches of North Sea Atlantic Cod Gadus morhua. Lester et al. enumeration of salmonine populations in streams. North American Journal of Fisheries Management 15:551–562. (1996) estimated d to be 1.82 for trap-net catches in nearshore Krebs, C. J. 1989. Ecological methodology. Harper Collins, New York. fish communities in freshwater lakes. This value could be refined Kruse, C. G., W. A. Hubert, and F. J. Rahel. 1998. Single-pass electrofishing iteratively during a long-term study. A larger d value will reduce predicts trout abundance in mountain streams with sparse habitat. North relative variance and cause travel time and site size to have less American Journal of Fisheries Management 18:940–946. of an effect on optimum effort intensity (further favoring smaller Lester, N. P., W. I. Dunlop, and C. C. Willox. 1996. Detecting changes in the nearshore fish community. Canadian Journal of Fisheries and Aquatic effort intensities), whereas a smaller d value will do the opposite. Sciences 53(Supplement 1):391–402. Lastly, mean travel time between sites and mean surface area Mahon, R. 1980. Accuracy of catch-effort methods for estimating fish density can be simplified by considering them as a ratio. Wilmot Creek’s and biomass in streams. Environmental Biology of Fishes 5:343–363. mean travel time of 4,500 s and 308 m2 can be thought of as Mitro, M. G., and A. V. Zale. 2000. Predicting fish abundance using single- “15 s travel time/m2 of stream sampling.” The effect of tripling pass removal sampling. Canadian Journal of Fisheries and Aquatic Sciences 57:951–961. this value is displayed in Figure 4B, and the effect of reducing it O’Brien, C. M., C. D. Darby, B. D. Rackham, D. L. Maxwell, H. Degel, S. Flat- to one-third its original size is shown in Figure 4C. Again, some man, M. Mathewson, M. A. Pastoors, E. J. Simmonds, and M. Vinther. 2001. a priori knowledge of the stream sites and their access routes The precision of international market sampling for North Sea cod (Gadus are required before these values can be considered. morhua L.) and its influence on stock assessment. International Council for Despite these limitations, we remain optimistic that the the Exploration of the Sea, C.M. 2001/P:14, Copenhagen. OMNR (Ontario Ministry of Natural Resources). 2008. Ontario stream infor- method we developed for calculating optimum electrofishing mation dataset: a composite of data collected on wadeable streams through effort intensity can be applied dynamically to monitoring pro- a cooperative approach. OMNR, Peterborough. Available: comap.ca\fwis. grams that are already in place, or where a stream survey is being (January 2012). Downloaded by [Department Of Fisheries] at 00:01 20 March 2013 designed to optimize electrofishing effort in a familiar system. Peterson, J. T., R. F. Thurow, and J. W. Guzevich. 2004. An evaluation of It is now calculable that if a Rainbow Trout survey on Wilmot multipass electrofishing for estimating the abundance of stream-dwelling 2 salmonids. Transactions of the American Fisheries Society 133:462–475. Creek was formerly taking place using 15 s/m of electrofish- Simonson, T. D., and J. Lyons. 1995. Comparison of catch per effort and removal ing, then a 43% savings in time (and a similar percentage of procedures for sampling stream fish assemblages. North American Journal of cost) could be recouped by moving to 5 s/m2 without losing any Fisheries Management 15:419–427. precision. Alternatively, this 43% time savings could be used to Stanfield, L. W., M. Jones, M. Stoneman, B. Kilgour, J. Parish, and G. Wichert. increase the number of sites sampled and thus provide a more 1999. Stream assessment protocol for Ontario, version 3. Ontario Ministry of Natural Resources, Peterborough. Available: http://mnr.gov.on.ca. (January precise estimate of abundance. 2012). ACKNOWLEDGMENTS Taylor, L. R. 1961. Aggregation, variance and the mean. Nature (London) 189:732–735. The authors acknowledge the field support provided by Lau- Zippin, C. 1958. The removal method of population estimation. Journal of rie Allin, E. Cleveland, J. Darisi, and M. O’Halloran. This Wildlife Management 22:82–90. BACKPACK ELECTROFISHING SURVEYS 285

APPENDIX: SUPPLEMENTAL DATA ON FISH CATCHES

TABLE A.1. Catch summaries for fish groups from sites on Wilmot Creek from the summer survey in 1999. Catch for each fish group is displayed for each electrofishing effort for each site. Site names contain the stratum where the site is located (A, B, or C); see Figure 1 for site locations. NA = not applicable.

Species group Rainbow Trout Brown Trout Sculpin Dace Darter Effort (s/m2): 5 10 15 5 10 15 5 10 15 5 10 15 5 10 15 Site A01423335010611101006884911059 A02 61 53 73 0 0 0 36 51 43 121 145 167 44 22 91 A03 42 55 36 0 0 2 16 26 26 146 126 85 75 40 28 A0448911000456985062139172040 A05 56 57 68 3 3 5 10 10 16 38 70 84 20 9 34 A0630512331052125591052764812 A07 33 53 60 1 0 0 59 44 64 47 75 64 23 25 42 A08 65 37 72 0 0 0 27 44 76 70 37 75 21 4 38 A09 76 42 48 3 1 4 25 36 26 61 93 107 7 21 9 A10 62 82 110 0 0 1 28 40 89 109 120 161 5 15 8 A11406679232399290423996116 A15 77 100 180 6 7 5 19 50 73 0 2 1 0 1 0 B01 102 115 119 7 3 3 86 102 86 0 0 0 0 0 0 B02 56 93 107 3 3 5 54 84 108 0 0 0 0 0 0 B03 72 103 77 1 1 3 68 91 97 0 0 0 0 0 0 B04 587460118 87948840 0 0 0 0 0 B05 78 133 88 10 22 12 27 48 34 0 0 0 0 0 0 B10 4746602627271119130 0 0 0 0 0 B11 2448392128191725330 0 0 0 0 0 B12 323015142117811130 0 0 0 0 0 B14 372719251423713230 0 0 0 0 0 B15 4242434530332212110 0 0 0 0 0 C02 4797613151334343510 0 0 0 0 0 C03 243723161381840360 0 0 0 0 0 C04 15639918172441480 0 0 0 0 0 C05 3125761628232451840 0 0 0 0 0 C06 3329692534331541460 0 0 0 0 0 C07 2431331811191016290 0 0 0 0 0 C08 9 18 16 37 38 46 16 15 21 0 0 0 0 0 0 Downloaded by [Department Of Fisheries] at 00:01 20 March 2013 C10 00012293092023100000 Total 1,349 1,724 1,828 345 399 383 814 1,154 1,386 844 942 1,090 310 216 367 Mean 45 57.5 60.9 11.5 13.3 12.8 27.1 38.5 46.2 28.1 31.4 36.3 10.3 7.2 12.2 SD 23.5 31.7 37.8 12.3 14.2 13 22 25.9 30.7 43.2 46.7 54.5 22.3 12.8 22.2 286 STANFIELD ET AL.

TABLE A.1. Extended.

Species group Lamprey Other Fishes Total species richness Total ﬁsh captured Effort (s/m2): 5 10 15 5 10 15 5 10 15 5 10 15 Site A01 1 0 6 33 11 17 11 9 9 273 134 211 A02 0 0 0 11 14 24 9 10 10 273 285 398 A030109539810288253180 A04 0 0 1 2 15 5 8 11 11 123 201 298 A05 0 2 0 11 43 12 9 12 11 138 194 219 A06 0 0 1 45 28 25 10 11 12 148 254 113 A07 0 0 2 12 12 24 11 7 9 175 209 256 A08011002557183123264 A09000373898175200197 A100010735108204264373 A11 0 5 1 3 28 6 8 11 10 127 234 280 A15 0 0 0 7 11 16 6 8 8 109 171 275 B01010503545200221211 B02000010343113181220 B03000001334141195178 B04000280563150138152 B05 2 0 6 9 10 5 6 6 7 126 213 145 B10 0 0 0 14 11 16 4 5 5 98 103 116 B11100363644 6610794 B12 0 2 2 8 11 10 4 6 6 62 75 57 B14 0 3 3 20 13 9 5 5 6 89 70 77 B15 1 3 6 37 50 21 7 7 6 147 137 114 C02 4 11 10 7 16 8 7 8 8 132 218 163 C03016253657 609676 C04 6 26 21 1 11 11 7 6 6 41 152 136 C05 7 15 39 13 10 18 6 6 6 91 129 240 C06 1 2 11 4 6 31 6 6 6 78 112 190 C07 0 1 0 10 9 12 4 6 4 62 68 93 C08000000433 627183 Downloaded by [Department Of Fisheries] at 00:01 20 March 2013 C10000141433 235354 Total 23 74 117 272 352 292 NA NA NA 3,957 4,861 5,463 Mean 0.8 2.5 3.9 9.1 11.7 9.7 6.4 6.8 6.8 131.9 162 182.1 SD 1.8 5.6 8.1 11.3 11.7 8.8 2.3 2.6 2.6 68 66.3 89.7 This article was downloaded by: [Department Of Fisheries] On: 20 March 2013, At: 00:02 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 Stranding of Spawning Run Green Sturgeon in the Sacramento River: Post-Rescue Movements and Potential Population-Level Effects Michael J. Thomas a , Matthew L. Peterson a b , Nick Friedenberg c , Joel P. Van Eenennaam d , Joseph R. Johnson e , Jan Jeffrey Hoover f & A. Peter Klimley a a Biotelemetry Laboratory, Department of Wildlife, Fisheries and Conservation Biology, University of California, Davis, One Shields Avenue, Davis, California, 95616, USA b FISHBIO, 180 East 4th Street, Suite 160, Chico, California, 95928, USA c Applied Biomathematics, 100 North Country Road, Setauket, New York, 11733, USA d Finfish Reproductive Physiology Laboratory, Department of Animal Sciences, University of California, Davis, One Shields Avenue, Davis, California, 95616, USA e California Department of Fish and Game, North Central Region, 1701 Nimbus Road, Rancho Cordova, California, 95670, USA f U.S. Army Engineer Research and Development Center, 3909 Halls Ferry Road, Vicksburg, Mississippi, 39180-6199, USA Version of record first published: 06 Mar 2013.

To cite this article: Michael J. Thomas , Matthew L. Peterson , Nick Friedenberg , Joel P. Van Eenennaam , Joseph R. Johnson , Jan Jeffrey Hoover & A. Peter Klimley (2013): Stranding of Spawning Run Green Sturgeon in the Sacramento River: Post-Rescue Movements and Potential Population-Level Effects, North American Journal of Fisheries Management, 33:2, 287-297 To link to this article: http://dx.doi.org/10.1080/02755947.2012.758201

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 33:287–297, 2013 C American Fisheries Society 2013 ISSN: 0275-5947 print / 1548-8675 online DOI: 10.1080/02755947.2012.758201

ARTICLE

Stranding of Spawning Run Green Sturgeon in the Sacramento River: Post-Rescue Movements and Potential Population-Level Effects

Michael J. Thomas* Biotelemetry Laboratory, Department of Wildlife, Fisheries and Conservation Biology, University of California, Davis, One Shields Avenue, Davis, California 95616, USA Matthew L. Peterson Biotelemetry Laboratory, Department of Wildlife, Fisheries and Conservation Biology, University of California, Davis, One Shields Avenue, Davis, California 95616, USA; and FISHBIO, 180 East 4th Street, Suite 160, Chico, California 95928, USA Nick Friedenberg Applied Biomathematics, 100 North Country Road, Setauket, New York 11733, USA Joel P. Van Eenennaam Finﬁsh Reproductive Physiology Laboratory, Department of Animal Sciences, University of California, Davis, One Shields Avenue, Davis, California 95616, USA Joseph R. Johnson California Department of Fish and Game, North Central Region, 1701 Nimbus Road, Rancho Cordova, California 95670, USA Jan Jeffrey Hoover U.S. Army Engineer Research and Development Center, 3909 Halls Ferry Road, Vicksburg, Mississippi 39180-6199, USA A. Peter Klimley Biotelemetry Laboratory, Department of Wildlife, Fisheries and Conservation Biology, University of California, Davis, One Shields Avenue, Davis, California 95616, USA Downloaded by [Department Of Fisheries] at 00:02 20 March 2013

Abstract The lower portion of the Sacramento River, California, has been highly engineered to protect low-lying surrounding communities from annual flood events. While engineered floodplains have provided adequate protection for the surrounding communities, there remain unintended consequences to migratory fish that become stranded during high flow events. In April 2011, we rescued 24 threatened Green Sturgeon Acipenser medirostris that were stranded in two flood diversions along the Sacramento River. We tagged these 24 Green Sturgeon with acoustic tags and analyzed their survival and migration success to their spawning grounds. Additionally, we provided a population viability analysis to show the potential impacts of stranding and the benefits of conducting rescues at the population level. We found that 17 of these 24 individuals continued their upstream migration to the spawning grounds. Modeling suggests that recurrent stranding of a similar magnitude without rescue could affect the long-term viability of Green Sturgeon

*Corresponding author: [email protected] Received May 4, 2012; accepted December 6, 2012 287 288 THOMAS ET AL.

in the Sacramento River. Population viability analyses of rescue predicted a 7% decrease below the population baseline model over 50 years as opposed to 33% without rescue. Despite the mitigated impact to the population with rescue, ﬁsh passage improvements should be considered as a long-term goal for preventing population risks at ﬂood control diversions.

Over the past century, the environments of many fishes have been highly altered to meet overwhelming anthropogenic needs. The construction of dams and flood control structures has direct and indirect consequences to fish populations. With respect to intensely studied and managed species such as salmonids, effects from environmental changes have been identified and appropriate management actions have been implemented for many populations. In contrast, species of sturgeon (Acipenseridae) are generally understudied, resulting in relatively little information with which to manage the species. Green Sturgeon Acipenser medirostris in particular were historically understudied, and despite a recent surge in publications (see Jaric´ and Gessner 2012), there remains a considerable void in our understanding of their basic biology. The hydrology of the Sacramento River and Sacramento–San Joaquin Delta, California, has been highly altered through the construction of dams, levees, and water diversions. Large engineered floodplains have been constructed to divert floodwaters away from major urban areas during sustained periods of rain- fall and snowmelt. Flooding of the two biggest diversions, Sutter Bypass and Yolo Bypass, are controlled by a concrete feature commonly referred to as a weir. During high water events in the main-stem Sacramento River, water is either released via gate operation or enters the bypass by spilling. The Fremont Weir located at river kilometer (RKM, measured from Golden Gate Bridge) 226 runs east and west at the northernmost point of Yolo Bypass (Figure 1). Flooding of Yolo Bypass is controlled by the operation of a flood gate when flows are below bank-full levels or by overtopping. To the north of Yolo Bypass is Tisdale Weir (RKM 286.0). Tisdale Weir runs north and south, parallel to the FIGURE 1. Study area of Green Sturgeon rescue sites in the Sacramento River, Sacramento River, and controls inundation only when the main California, showing river channel, monitor locations, and major cities.

Downloaded by [Department Of Fisheries] at 00:02 20 March 2013 stem is below spill stage (Figure 1). Tisdale Weir is inundated when flows of the Sacramento River exceed 595 m3/s and the The Sacramento River Green Sturgeon, known as the south- Fremont Weir is inundated when flows increase above 200 m3/s ern Distinct Population Segment (DPS), is federally listed as (Feyrer et al. 2006). The flood recurrence interval in the Yolo threatened. The southern DPS Green Sturgeon is an anadro- Bypass averages every 2 years (Sommer et al. 2005). In most mous migratory species found throughout the bay and inlets of years, flooding of the Yolo Bypass occurs during the months of the Pacific Northwest as adults and subadults. Genetic analy- January through March. ses of spawning populations show signs of significant genetic Sturgeon species are vulnerable to water diversions during isolation between Green Sturgeon found in the northern rivers floods (USFWS 2009) principally due to their comparatively low compared with those found in the Sacramento River (Israel et al. swimming endurance in high water velocities and to their sub- 2004). Two similar northern spawning populations were identi- strate appression behavior (Boysen and Hoover 2009; Hoover fied in the Klamath and Rogue rivers and because of their genetic et al. 2011). For imperiled species, field monitoring is critical similarity were classified as the northern DPS. Alternatively, the for effective rescue efforts, quantifying “take,” and evaluating Sacramento River population was identified as genetically dif- long-term impacts of floods and artificial river structures on the ferent to those individuals from the Klamath and Rogue rivers, conservation status of individual populations. and was therefore classified as a separate population segment. GREEN STURGEON SPAWNING IN THE SACRAMENTO RIVER 289

There is currently no consensus on the abundance of southern through increased mortality and forgone reproduction has not DPS Green Sturgeon. However, those estimates, which have been explored for this population segment. The major features been provided in the literature or reports, are on the order of of the southern DPS—small estimated population size (e.g., thousands rather than tens of thousands. Currently our best sense Israel and May 2010), late maturity, episodic spawning, and a of population trajectory is based upon entrainment rates from the long-lived adult stage—all contribute not only to its overall vul- state and federal water export facilities located in the southern nerability but also to the possible impact of any threat targeting Sacramento–San Joaquin Delta and from incidental catch from reproductive adults (Heppell 2007). Therefore, the objectives the White Sturgeon A. transmontanus monitoring program in of this study were to (1) describe the reproductive condition of San Pablo Bay. In either case estimates from these two programs Green Sturgeon individuals and poststranding movements us- are either inherently exaggerated or biased and offer managers ing acoustic tags, (2) assess postrescue survival while in the little concrete evidence as to the true population trajectory (see river system, (3) determine which factors, if any, influenced Adams et al. 2007). movement rates and arrival of Green Sturgeon at the spawning Managers have already identified potential passage-related grounds, and (4) assess the population level impacts of stranding features that present potential population-level risk including and benefits of rescue activities for Green Sturgeon listed under the Freemont Weir (NMFS 2006). Potential risk from stranding the U.S. Endangered Species Act. in flood diversions was unknown until numerous Green Sturgeon were rescued from two diversions during the spring spawning migrations in 2011. METHODS Despite a mandate to preserve the Sacramento River pop- Capture and tagging.—In total, 25 Green Sturgeon were lo- ulation of Green Sturgeon, resource agencies have failed to cated within the two stranding sites. Sturgeon were captured adequately monitor several flood control structures that have using block nets at four separate sites at Fremont Weir and one been previously identified as potential barriers to fish migra- at Tisdale Weir (Figure 1). Upon capture, 23 Green Sturgeon tion (NMFS 2006). Sturgeon have previously been stranded at were found to be in “good” condition, based on the absences of Fremont Weir during high flood events. Informal sturgeon res- lacerations and abrasions from the concrete weir. One individual cues were performed at Fremont Weir in 2001, 2003, and 2004 was considered to be in “poor” condition due to excessive abra- (Zoltan Matica, California Department of Water Resources, sions and the presence of a spear tip near the back of the head. West Sacramento, personal communication). In 2006, a stur- One gravid female was found dead at the time of rescue and was geon rescue was conducted by the California Department of presumed to be a victim of poaching activities the night before. Fish and Game at Fremont Weir (California Department of Fish All 24 living Green Sturgeon were removed from the concrete and Game, unpublished data). Neither numbers nor species were channel behind each weir, turned upside down, and placed in a recorded for the years of 2001–2004; however, 26 unidentified hooded cloth stretcher. Gills were ventilated by pumping water sturgeon were removed from Fremont Weir in 2006. into the hood to maintain a continuous flow of fresh water. Fork Spawning migrations of Green Sturgeon typically occur dur- length (FL) and total length (TL) (±0.5 cm) were measured ing the months of March through June (Benson et al. 2007; with a retractable tape measure prior to surgery. Differences in Erickson and Webb 2007; Heublein et al. 2009). Spawning of lengths between males and females were tested at a later date us- Green Sturgeon in the Sacramento River has been documented ing a Student’s t-test implemented with JMP version 4 software from RKM 423.9 near Glenn–Colusa Irrigation District (GCID) (SAS Institute, Cary, North Carolina). Water temperatures dur- pumping facility to the confluence of Inks Creek at RKM 516.0 ing the rescue at Fremont Weir ranged from 14.2◦C to 16.6◦C. (Brown 2007; Poytress et al. 2009, 2010, 2011, 2012). No water temperatures were taken during the rescue at Tisdale Downloaded by [Department Of Fisheries] at 00:02 20 March 2013 During March and April 2011, Tisdale and Fremont weirs Weir. Average daily air temperatures during the days of rescue were flooded for approximately 26 and 24 d, respectively were 11.3◦C, with a high of 18.8◦C and a low of 4.4◦C. (California Department of Water Resources, California Data A uniquely coded beacon (V16-6L, VEMCO) with a fre- Exchange Center [CDEC], unpublished data). Personnel from quency of 69 kHz, a 75-s delay, and battery life of 3,650 d was the California Department of Fish and Game identified numer- implanted within the peritoneum of each Green Sturgeon (see ous stranded sturgeon on 11 April 2011 once water levels had Heublein et al. 2009 for surgical procedure). Prior to insertion receded. While it was unknown if any of the sturgeon stranded of the telemetry tag, the gonads were examined to determine sex were Green Sturgeon during the first sightings on 11 April 2011, and stage of maturity. For all males, except one that was releas- rescue operations were organized on 12 and 13 April 2011 to ing milt at capture, a small section of the testis was collected remove all fish stranded. During rescue operations both White for histological processing as described in Webb and Erickson Sturgeon and Green Sturgeon were found stranded at each of (2007). From each female approximately 15 eggs were collected the two diversions. for measurement of size and calculation of the polarization in- To our knowledge, the movements of Green Sturgeon have dex (PI) using the methods described by Van Eenennaam et al. yet to be described after releases from such an impediment. Fur- (2001, 2006). Surgery times ranged from 3 to 5 min, and to- thermore, the potential for population-level effects of stranding tal handling times averaged 10 min. Upon completion each 290 THOMAS ET AL.

sturgeon was transported approximately 0.1 km by foot or truck evaluated using Akaike’s information criterion (AIC; Stauffer and released into the Sacramento River. 2008). The eggs were examined to determine the reproductive state Population viability analysis.—In the absence of consen- of mature females. Egg length and width (n = 11 females) were sus on whether there is an upward or downward trend in the measured using a dark-field dissecting microscope equipped Sacramento River Green Sturgeon population, we constructed a with a camera lucida, an image-analyzing tablet, and a micro- baseline model in which the expected size of the population was computer interface (±0.001 mm). The eggs were then bisected neither growing nor declining. Modifications to the trendless with a razor blade along the animal–vegetal axis, and the dis- baseline model were then used to assess the marginal contribu- tance of the germinal vesicle from the inner border of the egg tion to ecological risk (Ginzburg et al. 1982) attending specific chorion and egg diameter were measured and used to calculate scenarios of stranding and rescue. To be conservative, all rates PI, a morphologic criterion of egg ripeness (Dettlaff et al. 1993; were assumed to be density independent (Ginzburg et al. 1990). Chapman and Van Eenennaam 2007). Histological slides were Projections of population growth were governed by a ma- examined under a compound scope and the germ cells scored for trix of transition rates describing the survival, development, stage of development using the developmental stages described and reproduction during distinct stages of the life history of in Webb and Erickson (2007). Green Sturgeon. Parameter values were compiled from pre- Monitoring movements.—An array of 47 stationary acoustic vious studies (Beamesderfer et al. 2007; Heppell 2007) and monitors (VR2 and VR2W, VEMCO) and a mobile tracking Green Sturgeon records of detections from monitors in the receiver (VR100, VEMCO) were used to describe postrescue Sacramento River. The life history of Green Sturgeon was di- movements and survival of tagged Green Sturgeon (Figure 1). vided into five stages: juveniles (ages 1–4), marine juveniles The lowermost-confirmed Green Sturgeon spawning site (GCID (ages 5–10), subadults (ages 11–15), young adults (ages 16– Hole) is located at RKM 424.0. Spawning was confirmed by 25), and adults (age 26+). No maximum age was imposed. Life egg collection at RKM 424.0 in 2010, by ongoing monitor- history parameters and justifications for choosing them are given ing of egg mats conducted by U.S. Fish and Wildlife Service in Table 1. (Poytress et al. 2011). Due to considerations of detectability The baseline model was implemented as a stochastic we selected the Irvine Finch monitor location (RKM 411.8) Monte Carlo simulation using RAMAS Metapop 5.0 software as our endpoint for determining migration success. A Green (Akçakaya and Root 2007). The Metapop software simulates Sturgeon detected at or above this location was judged to have population growth in a variable environment by sampling dis- reached the spawning grounds. A Green Sturgeon detected be- tributions of vital rates defined by user-provided means and low the Fremont Weir (RKM 225.5) was judged to have left standard deviations; the same means are used in all replicate the river system. Survival was defined by a successful summer simulations. Mean transition rates and fecundities are given in or winter out-migration past RKM 225.5. Detection probabili- Table 2. Deterministic analysis of the stage-based model pro- ties of monitors between Fremont Weir and the GCID monitor vided the expected stable stage structure indicated in Table 3. (RKM 225.5–424.0) were calculated using methods similar to Asymptotic elasticity analysis (Caswell 2001) performed by the Melnychuk et al. (2007). Despite only moderate detection prob- Metapop software provided a measure of the sensitivity of pop- abilities (by individual monitors; mean ± SD probability of ulation growth rate to proportional changes in elements of the detection, 0.71 ± 0.15), overall likelihood of detection was stage matrix. Only fecundity and adult survival yielded unam- high due to the extensive monitor array. Thus, we had high biguous elasticity; other entries jointly represented probabilities confidence in our ability to determine survival and movements. of survival and rates of development. Without information on Detections of tagged sturgeon were filtered to obtain the first and environmental stochasticity, we assumed 5% annual variability Downloaded by [Department Of Fisheries] at 00:02 20 March 2013 last detections at any particular monitoring site and movement of mortality and 10% annual variability of fecundity. Annual rates and residence times between monitors were calculated variation was lognormal. Under these conditions, the average using the last detection at monitor i and the first detection at population size declined by an average of fewer than six indi- monitor i + 1. Records at certain monitors were pooled due to viduals over 50 years. The model tracked only females under overlap in detection ranges. the assumption of a 1:1 sex ratio. All model projections used A generalized linear model (GLM; binary logistic regres- the same means and standard deviations for transition rates for sion) was used to identify which, if any, explanatory variables 10,000 replicated 50-year projections. All simulations started were associated with migratory success (defined as reaching with the population at its stable stage distribution. the spawning grounds at RKM 411.8) and a series of poten- The single recorded observation of stranding of Green tial explanatory variables (Stauffer 2008). Explanatory variables Sturgeon at the Fremont and Tisdale weirs provides only an used in the modeling included length, sex, surgery time, reach- anecdotal understanding of the frequency and magnitude of such specific migration rates, and reach-specific residence times. Six- events. We used a set of assumptions, described in the following teen candidate models (in two groupings: [1] all Green Sturgeon paragraphs, that link stranding with hydrological data to inform included, n = 24, and [2] Green Sturgeon from Fremont Weir, a population projection model. Within this set of assumptions, n = 12) using combinations of the explanatory variables were we used uncertainty regarding a subset of parameters to develop GREEN STURGEON SPAWNING IN THE SACRAMENTO RIVER 291

TABLE 1. Underlying parameters used to calculate survival and fecundity of Green Sturgeon in the Sacramento River under baseline (no entrainment) conditions.

Proportion Symbol for Value for Duration transitioning Parameter name model model (year) in duration Justiﬁcation Survival S 0.93 Beamesderfer et al. (2007) Transition to marine juvenile Mm 0.54 3 0.9 Transition rates chosen for appropriate residence time in stage according to Heppell (2007). Transition to subadult Ms 0.11 6 0.5 Transition to young adult My 0.13 5 0.5 Transition to adult Ma 0.07 10 0.5 Mature young adults Xy 0.29 Stage average based on linear interpretation of Heppell (2007). Mature adults Xa 0.93 Eggs young adult eggy 11,383 Egg production based on Beamesderfer et al. (2007) using stage average weighted by sexual maturity (females only). Eggs adult egga 73,465 Egg to age-1 survival s0 0.00002 First year survival and spawning success chosen to balance births and deaths in base model. Spawning success spwn 0.80 Reproduction frequency P 0.30 Telemetry data for Sacramento River.

objectively optimistic and pessimistic assessments of the impact gauge station (Figure 2) indicated a 0.43 annual probability of stranding on Green Sturgeon population dynamics. of a flood of concern occurring and lasting at least 1 d. Sus- We defined floods of concern as river stages that exceeded the tained periods of dry and wet years were implemented as con- Fremont Weir’s controlling elevation of 10.2 m by at least 0.3 m ditional flood probabilities. In the gauge data record, years with during the upstream migratory period from February through a flood were followed by flood years with a probability of 0.64, May. This criterion was chosen arbitrarily to help select flood whereas years without floods were followed by floods with a events in which appreciable volumes of water entered the di- probability of 0.31. While we included this autocorrelation in versions. We assumed that inundation at Fremont was a good simulations, it was not statistically significant (χ2 = 2.8, df = indicator of overall stranding risk and did not use weir-specific 25.26, P > 0.09) and had a negligible effect on results compared estimates of stranding. Data from 1984 to 2011 at the Fremont with additional simulations using the uniform flood probability of 0.43. Flood durations were measured in days. The 2011 in- TABLE 2. Transition and fecundity rates used in the stage-based model of undation of Fremont Weir was a flood of concern for 23 d. The Downloaded by [Department Of Fisheries] at 00:02 20 March 2013 Green Sturgeon in the Sacramento River. Formula parameters and their values average duration of floods of concern over the Fremont gauge are defined in Table 1. The asymptotic population growth rate is 1.0000. record was 28.25 d (SD = 17.34). Total rates Formula Value

Juvenile survival s·(1 − Mm) 0.432 TABLE 3. Stable stage distribution and abundance of Green Sturgeon in the Juvenile to marine s·Mm 0.498 southern Distinct Population Segment under baseline conditions assuming an Marine survival s·(1 − Ms) 0.829 average of 200 male and female spawners per year. s·Ms Marine to subadult 0.101 Fraction Female Total s· − My Subadult survival (1 ) 0.810 Stage Proportion mature spawners Females females Sub- to young adult s·My 0.120 Young adult survival s·(1 − Ma) 0.868 Juvenile 0.123 0 214 Young adult to adult s·Ma 0.062 Marine 0.359 0 625 Adult survival s 0.930 Subadult 0.191 0 333 Fecundity young adult eggy·s0·spwn·p 0.060 Young adult 0.173 0.29 26 301 Fecundity adult egga·s0·spwn·p 0.388 Adult 0.154 0.93 74 268 1,741 292 THOMAS ET AL.

The stranding rate used in the population model was developed using the total number of Green Sturgeon (n = 25) found stranded at Tisdale and Fremont weirs during 2011. We assumed that all individuals stranded were found and treated stranding as a system-wide risk rather than a site-specific rate. To estimate the proportion of migrants at risk of stranding during the 2011 flood, we developed a daily probability of migration during floods of concern. First, we used 2009–2011 records of tag detections in the array of monitors to find monthly probabilities of migration among tagged Green Sturgeon in the Sacramento River (M. J. Thomas, unpublished data). We then converted these to daily probabilities by dividing by the number of days in each month. Then we summed the daily probability of migration over the course of each flood of concern on record in the Fremont gauge data. Finally, we regressed these accumu- lated migration probabilities against flood length with an intercept constrained to zero. The resulting daily probability of 2 migration during floods of concern was 0.0123 (r = 0.98, F1, 27 = < FIGURE 2. The duration (in days) of floods of concern during the Green Stur- 2,746, P 0.0001). We assumed all migrants were at risk of geon migratory period at Fremont Weir for the period 1984–2011. The dashed stranding without respect to position in the river. Assuming 200 horizontal line indicates the average flood duration of 28.25 d (excluding years total migrants per year, the expected number of migrants dur- without floods). Floods of concern were defined as river stages that exceeded ing the 23 d of the 2011 flood was 56.6, with a Poisson 97.5% the weir’s controlling elevation by 0.3 m for at least 1 d during the migratory confidence interval (CI) of 41–76. This led to an estimated 0.44 period, February through May. per capita rate of stranding with a joint 95% CI of 0.216–0.780 (including uncertainty about both the number of migrants and To project possible demographic consequences of stranding, the true probability of stranding). Applied to the average flood we needed to establish a population size for Green Sturgeon in duration of 28.25 d, the per capita stranding rate yielded an ex- the Sacramento River. Population size estimates vary widely. pected stranding of 15.4 females with a 95% CI of 7.3–27.1 in For instance, Israel and May (2010) estimated that 5–14 fe- years with floods, assuming a 1:1 sex ratio. It is important to males contribute to juvenile production every year, translating note that similar stranding levels would result from any method to a range of 90–252 young adult and adult females accord- that assumes stranding risk is proportional to the duration of ing to the parameters in Table 3. Moyle (2002) gave a range floods. The translation of stranding levels into effects on adult of 70–800 female adults. Here, we used results of a river sur- survival and fecundity is most apparent for the fecundity multi- vey that estimated 220 total upstream migrants in 2010 (Ethan plier, assuming the population is at the stable stage distribution, Mora, University of California, Davis, unpublished data). We and rescued individuals survive stranding with 90% probability conservatively assumed this observation represented an above- and spawn with 50% probability (Table 4). average year and assumed the average spawning run was 200 To model the effect of the 2011 event in isolation, we ran adults. With a 1:1 sex ratio, stable stage distribution, and an replicated stochastic simulations starting with a single stranding event averaging 16 females (4.16 young adults and 11.84 adults). Downloaded by [Department Of Fisheries] at 00:02 20 March 2013 annual spawning probability of 0.3, the number of spawners corresponded to 569 young adult and adult females and a total Vital rate multipliers under 2011 stranding levels were 0.986 female population size of 1,741 (Table 3). for young adult survival, 0.956 for adult survival, and 0.840

TABLE 4. Mean and 95% CI around the impact of entrainment and rescue on Green Sturgeon demographic rates.

Lower estimate Mean Upper estimate Parameter No rescue Rescue No rescue Rescue No rescue Rescue Females stranded 7.3 15.4 27.1 Proportion of young adults 0.006 0.011 0.023 Proportion of adults 0.020 0.035 0.075 Proportion of spawners 0.073 0.154 0.271 Young adult survival multiplier 0.994 0.999 0.989 0.999 0.977 0.998 Adult survival multiplier 0.980 0.998 0.965 0.997 0.925 0.993 Fecundity multiplier 0.927 0.964 0.846 0.923 0.729 0.865 GREEN STURGEON SPAWNING IN THE SACRAMENTO RIVER 293

TABLE 5. Biological information for tagged Green Sturgeon (n = 24) at the Fremont (12 April 2011) and Tisdale (14 April 2011) weirs. M = male, F = female.

Tag identiﬁcation Location / site FL / TL (cm) Sex Maturation status Polar index GS1 Fremont / 1 182 / 200 M Mature GS2 Fremont / 1 144 / 161 M Mature GS3 Fremont / 1 183 / 202 F Eggs 0.048 GS4a Fremont / 2 175 / 190.5 M Mature GS5b Fremont / 2 156 / 175 F Eggs 0.038 GS6 Fremont / 2 188 / 207 F Eggs 0.056 GS7 Fremont / 2 164 / 179 M Mature GS8 Fremont / 2 173 / 158.5 M Mature GS9b Fremont / 3 207 / 219 F Eggs 0.067 GS10 Fremont / 3 163 / 177 M Mature GS11 Fremont / 3 193 / 213 F Eggs 0.066 GS12 Fremont / 3 166 / 185 F Eggs 0.054 GS13 Fremont / 3 150 / 167 M Mature GS14 Tisdale / 1 199 / 213 F Eggs 0.072 GS15 Tisdale / 1 150 / 165 M Mature GS16 Tisdale / 1 202 / 215 F Eggs 0.056 GS17 Tisdale / 1 190 / 201 F Eggs 0.040 GS18 Tisdale / 1 185 / 201 M Mature GS19b Tisdale / 1 171 / 184.5 F Eggs 0.031 GS20 Tisdale / 1 182 / 195 F Eggs 0.036 GS21 Tisdale / 1 183 / 200 M Mature GS22 Tisdale / 1 160 / 174 M Mature GS23 Tisdale / 1 174 / 182 M Mature GS24 Tisdale / 1 163 / 177 M Mature

aGS4 not included in movement analyses. bEggs in early stages of atresia.

for fecundity. Vital rate multipliers with rescue were 0.999 for of females was 0.051 ± 0.014, indicating spawning readiness young adult survival, 0.996 for adult survival, and 0.920 for (Table 5). However, three females had eggs showing early signs fecundity. of atresia, a form of degradation manifested by softening and marbling of the egg chorion. RESULTS Movement records showed a combined 71% migration success (reached RKM 411.8) for both males and females from the Reproductive State and Movements of Rescued Sturgeon Fremont and Tisdale rescue sites. Site-specific success was vari- Ultimately, 56 White Sturgeon and Green Sturgeon were capable depending on sex and location. Migration success from the Downloaded by [Department Of Fisheries] at 00:02 20 March 2013 tured and moved from behind the two weirs and returned to the Fremont Weir was three of six (50%) for females and five of six main-stem Sacramento River. In total, 11 female and 13 male (83%) for males. Success to the spawning ground was higher Green Sturgeon were rescued from Fremont and Tisdale weirs. for females stranded at Tisdale Weir with four of five (80%) Males were significantly smaller than the females (Student’s reaching the spawning grounds. Males stranded at the Tisdale t-test: P < 0.05). Males were 179.4 ± 14.8 cm TL (mean ± Weir had the same migratory success as those stranded at the SD) and females were 200.5 ± 13.8 cm. All Green Sturgeon Fremont Weir (five of six, 83%). Counting females only, 7 of were determined to be mature adults on their spawning migra- 11 (64%) rescued female Green Sturgeon reached the spawning tion (Table 5). Through examination of histological sections grounds, a rate higher than the 50% we used in our population from the testis, we verified that all testicular cysts contained models for spawning success. Individuals exhibited two differ- differentiated spermatozoa (stage 5; Webb and Erickson 2007), ent periodicities of occupation of the spawning grounds, either and one individual (GS4), although injured from a poaching at- spending weeks on the spawning grounds before out-migrating tempt, was spermiating upon capture. Rescued females were in in early summer or oversummering and out-migrating in early the late stages of final maturation with large, oval-shaped, olive- winter (Figure 3). brown-colored eggs that averaged 4.27 ± 0.13 mm in length A total of 16 candidate models were examined, and were and 3.75 ± 0.05 mm in width. The mean polarization index made up of eight candidate models that included all Green 294 THOMAS ET AL.

Population Viability Analysis Model projections over 50 years indicated that chronic stranding in flood control structures could have biologically significant impacts on the viability of the Sacramento River Green Sturgeon population (Figure 4). The estimated mean frequency and severity of stranding events reduced expected final adult female abundance by 33% compared with baseline conditions (Figure 4), from an average baseline population of 563 individuals to an average impacted population of 378 individuals after 50 years. As indicated by the lighter-weight series in Figure 4, individual replicates in each scenario led to higher or lower final abundance. In addition to this uncertainty, caused by environmental variability, there is uncertainty about the entrainment rate that stems from our lack of knowledge. Using high FIGURE 3. River kilometer plots for two individual Green Sturgeon, and low estimates of entrainment (Table 4), minimum abun- (a) GS13 and (b) GS17. Note period of out-migration for GS13 occurred in dance during the 50-year projections ranged from 246 to 435 late May compared with the out-migration of GS17, which did not occur until adult females (15–52% below the baseline expected minimum late January. of 513). The range of population projections stemming from uncertainty in the entrainment rate led to increases in the risk Sturgeon (Table 6) and eight that included only those rescued at of 20% decline greater than five times the baseline level. Under Fremont Weir (Table 7). Of the models that included all 24 Green baseline conditions, longer simulations yielded a median time Sturgeon, no strong candidate model was observed among the to 20% decline of 200 years; the probability of 20% decline in eight candidate models when evaluated by AIC values and asso- 50 years was 0.12 (Figure 5). With stranding, the median time ciated model likelihoods. In fact, the null model had the lowest to 20% decline was 13–39 years and the probability in 50 years AIC value and the strongest model likelihood (Table 6). Simi- was 0.68–1.0 (Figure 5). larly, of the models that only included Green Sturgeon rescued Simulations suggested that monitoring and rescue opera- at Fremont Weir, the null model was ranked as the second-best tions could greatly reduce the impact of stranding on population model (Table 7). Interestingly, the top model included water viability. Rescue with the estimated mean frequency and sever- temperature at the time of rescue, but the model did not provide ity of stranding events resulted in a final female adult abundance vast improvement in AIC values compared with other potential 7% below baseline (to 524 individuals, Figure 4). Expected min- models. Upon examination of this model, a slightly negative, imum adult abundance over 50 years was 466–502 (2–9% below but insignificant, relationship was observed (P = 0.140). baseline). Uncertainty about the entrainment rate resulted in a All but two individuals successfully out-migrated; male GS4, range of population projections that included a slight to moder- an individual that had a spear tip removed from its head, and ately elevated 20% decline risk after rescue (Figure 5). Rescue male GS1 that was repeatedly tracked throughout the spawning resulted in median times to 20% decline of 58–108 years; the grounds and was not detected after 16 June 2011. Assuming probability of 20% decline in 50 years was 0.19–0.45. Hence, these two individuals died, the observed survival rate of rescued rescue increased the time to 20% decline by a factor of 2.8–4.0 Green Sturgeon was 91.7% compared with the 90% used in our compared with the stranding scenario and decreased the 50-year

Downloaded by [Department Of Fisheries] at 00:02 20 March 2013 population models. risk of decline by 55–72%.

TABLE 6. Summary of binary logistic regression models ranked on Akaike’s information criterion (AIC) and AIC weights to explain migratory success (success deﬁned as reaching RKM 411.8) among individual Green Sturgeon (n = 24) rescued during spring 2012.

Model name AIC score Parameters Number of parameters Model likelihood AIC weights AIC Null 30.975 None 1 1.000 0.202 2 31.503 Total length 2 0.768 0.156 0.528 8 31.682 Group 2 0.702 0.142 0.707 5 32.466 Sex 2 0.474 0.096 1.491 3 32.683 Estimated weight 2 0.426 0.086 1.708 6 32.732 Surgery time 2 0.415 0.084 1.757 30 33.239 Surgery time, total length 3 0.322 0.065 2.265 GREEN STURGEON SPAWNING IN THE SACRAMENTO RIVER 295

TABLE 7. Summary of binary logistic regression models ranked on Akaike’s information criterion (AIC) and AIC weights to explain migratory success (success deﬁned as reaching RKM 411.8) among individual Green Sturgeon (n = 12) rescued from Fremont Weir during Spring 2012.

Model Number of Model AIC name AIC score Parameters parameters likelihood weights AIC 1 18.825 Water temperature 2 1.000 0.184 Null 19.320 None 1 0.781 0.144 0.495 10 19.768 Residence time (Knights Landing to China Bend) 2 0.624 0.115 0.943 25 19.865 Water temperature, sex 3 0.595 0.109 1.040 9 19.929 Migration rate (Knights Landing to China Bend) 2 0.576 0.106 1.104 12 19.951 Residence time (China Bend to Tisdale Weir) 2 0.570 0.105 1.125 11 20.016 Migration rate (China Bend to Tisdale Weir) 2 0.551 0.101 1.191 27 20.797 Water temperature, surgery time 3 0.373 0.069 1.972 26 20.822 Water temperature, total length 3 0.369 0.068 1.996

If the stranding event of 2011 was an isolated event, it posed DISCUSSION only a small risk to the population; rescue completely offset Our analysis of Green Sturgeon population viability in the the 50-year impact. One-time stranding decreased the expected Sacramento River suggests that stranding could have a biolog- minimum abundance of young adults and adults from 513 to ically signiﬁcant impact if it is a recurring event. Furthermore, 501 over 50 years and increased the probability of 20% decline it appears that monitoring and rescue will substantially reduce, from 0.10 to 0.17. When rescue was included in the simulations, though not completely offset, that impact. While our analysis both minimum abundance and decline risk returned to baseline indicated that the single event observed in 2011 was, by itself, a levels. small risk to population viability, it has since emerged that sim- Elasticity analysis of the baseline model provided a general ilar events have occurred at least four times over the last decade overview of the sensitivity of population growth rate to propor- (California Department of Fish and Game, unpublished data; tional changes in elements of the stage matrix. Elasticity of adult Zoltan Manteca, California Department of Water Resources, survival was 0.35 compared with 0.026 for adult fecundity and West Sacramento, personal communication). 0.0046 for young adult fecundity. The projected impact of stranding on the southern DPS can be attributed primarily to increased adult mortality rather than Downloaded by [Department Of Fisheries] at 00:02 20 March 2013

FIGURE 4. Projected adult female Green Sturgeon abundance in the Sacra- mento River population under estimated mean stranding rates with and without FIGURE 5. The effect of stranding and rescue on the probability of a 20% rescue. Both models assume that the expected population trajectory was ﬂat in decline in adult Green Sturgeon abundance over 50 years. The solid curve rep- the absence of stranding. Bold curves depict means of 10,000 replicates. Lighter resents background risk in the absence of stranding. Remaining curves describe series indicate minimum and maximum projected abundance in each year. the 95% CI developed around model assumptions, as described in the Methods. 296 THOMAS ET AL.

episodic reductions in fecundity. In long-lived species with Sturgeon in the southern DPS, it makes clear that the uncertainty delayed reproductive maturity, including sturgeon, population regarding that burden is larger than the interval represented in growth rate is most sensitive to adult survival (Heppell 2007). our chronic stranding scenarios. A monitoring program would Analysis of our baseline stage matrix demonstrated that popula- provide data critical to reducing that uncertainty. tion growth rate was more than an order of magnitude more sen- In the interest of being conservative with respect to popula- sitive to changes in adult survival than to proportional changes tion viability, we assumed Green Sturgeon population growth in fecundity. Rescued adults, if they survive, may return to the was density independent. This factor, in contrast to the three estuary rather than continuing their spawning migration, a phe- aforementioned assumptions, could overestimate effects of nomenon referred to as “dropback.” However, 1 year of foregone stranding. If we allowed the parameters contributing to fe- spawning only slightly lessens lifetime reproductive output. It cundity (sexual maturation, spawning interval, spawning suc- is worth noting that the same holds true for short-term research cess, egg production, or survival from egg to age 1) to increase activities that may lead to “dropback.” If handling fish provides with decreasing abundance, then the Green Sturgeon population greater understanding of population size and migration behavior, would exhibit some level of compensatory growth and therefore the gains in management efficacy easily exceed the maximum a smaller long-term response to stranding (Ginzburg et al. 1990). possible negative biological impact so long as adult survival is It is interesting to note that the use of a density-dependent pop- not reduced. ulation growth model would also be likely to make projections We made several consequential assumptions when modeling sensitive to the temporal autocorrelation of stranding events stranding and rescue. First, we assumed that searches behind (Jonzen´ et al. 2002). diversion structures located all stranded individuals. Second, we Future Green Sturgeon monitoring during stranding should assumed that stranding frequency is a simple function of flood attempt to detail the condition of fish at the time of rescue frequency at a single weir. Third, we assumed the magnitude of through blood collection and hormone testing. It should be stranding is proportional to flood duration. Finally, we used a noted that even prior to early ovarian atresia, observed at the density-independent model of population growth. histological level, plasma testosterone and estradiol-17β levels The assumption of 100% search efficiency has a two-fold decrease dramatically in female sturgeon when stressed by el- effect on our results. First, it minimizes our estimate of the evated temperatures (Talbott et al. 2011). Therefore, eggs may per capita stranding probability. Second, undetected fish would appear perfectly normal in the field or after histological pro- not be rescued, diminishing the degree to which rescue efforts cessing and examination, but in actuality the female has stopped would offset the impact of stranding. steroid production and will not spawn. Eggs of females that did The assumption that stranding occurs no more or less fre- not reach the spawning grounds may have already been atresic quently than inundations at Fremont Weir that exceed the con- due to stranding stresses. While results from the GLM did not trolling elevation by 0.3 m during the migratory season greatly find any strong candidate models or variables to explain migra- limits both the complexity of the model and the range of results tory success, there were many environmental and physiologi- it can project. Obviously, the model overestimates the impact of cal variables, including plasma testosterone and estradiol-17β stranding if the stranding event of 2011 was in reality a rare oc- levels, that could have been measured during the 2011 rescue currence. It is less obvious how a site-specific consideration of efforts. Furthermore, our definition of success (a fish made it the six flood-control structures on the Sacramento River would to the spawning grounds) does not necessarily mean that a fish affect projections. Gauge data near Tisdale Weir indicate that was successful in passing on its genetic material. Future re- it is inundated more often than Fremont Weir and that the du- search should be carried out to determine the relative genetic ration of inundation is generally longer. All else being equal, contributions of rescued Green Sturgeon to any larvae captured Downloaded by [Department Of Fisheries] at 00:02 20 March 2013 a model using the site-specific information on stranding and in ongoing juvenile Green Sturgeon monitoring. inundation at Fremont and Tisdale weirs would predict more To minimize the impacts of flood control diversions to ei- frequent stranding events. ther Green Sturgeon or White Sturgeon we believe there needs In addition to reducing the impact of stranding, monitoring to be more formal monitoring at Tisdale and Fremont weirs would result in a better understanding of its frequency, magni- and additional flood control diversions not evaluated in this tude, and causes. Our assumption that the magnitude of strand- study. Such monitoring should not only identify presence of stur- ing is proportional to the duration of inundation limited the range geon species within these structures but include a timely rescue of impact projected by the model. There is a chance that strand- response. ing is better predicted by other metrics, such as the volume of This study illustrates both the risks and the successes in water diverted, or by a multivariate suite of factors. Simulation managing for stranding in two flood diversions. We show that of the stranding event of 2011 in isolation, in which we not only removal of sturgeon from such diversions can allow a majority assumed a low frequency of stranding but were also freed of of individuals to both survive and continue natural migratory the need to make a duration–magnitude assumption, indicated behavior with low to moderate population level effect. How- a slightly elevated 50-year probability of decline. Though this ever, rescue efforts should only be considered as a short-term result bolsters the conclusion that stranding is probably a biolog- management strategy to reduce population-level risks of strand- ically meaningful burden on the population viability of Green ing. Ultimately, major modifications to flood control structures GREEN STURGEON SPAWNING IN THE SACRAMENTO RIVER 297

will be necessary to prevent stranding risks of sturgeon species Heublein, J. C., J. T. Kelly, C. E. Crocker, A. P. Klimley, and S. T. Lindley. during their spawning migration. 2009. Migration of Green Sturgeon, Acipenser medirostris, in the Sacramento River. Environmental Biology of Fishes 84:245–258. Hoover, J. J., K. A. Boysen, J. A. Beard, and H. Smith. 2011. Assessing the risk ACKNOWLEDGMENTS of entrainment by cutterhead dredges to juvenile Lake Sturgeon (Acipenser We are grateful to the many staff members and volunteers of fulvescens) and juvenile Pallid Sturgeon (Scaphirhynchus albus). Journal of the California Department of Fish and Game for their exhaustive Applied Ichthyology 27:369–375. Israel, J. A., J. F. Cordes, M. A. Blumberg, and B. May. 2004. Geographic efforts to capture and return sturgeon back to the river. Without patterns of genetic differentiation among collections of Green Sturgeon. North the foresight of agency personnel to include a telemetry com- American Journal of Fisheries Management 24:922–931. ponent to the rescue, this study would not have been possible. Israel, J. A., and B. May. 2010. Indirect genetic estimates of breeding population Tagging support was provided by Brian Mahardja and Emily size in the polyploid Green Sturgeon (Acipenser medirostris). Molecular Miller from University of California, Davis, and Robert Chase Ecology 19:1058–1070. Jaric,´ I., and J. Gessner. 2012. Analysis of publications on sturgeon research from the U.S. Bureau of Reclamation. Funding for population between 1996 and 2010. Scientometrics 90:715–735. modeling was provided by the U.S. Army Corps of Engineers, Jonzen,´ N., J. Ripa, and P. Lundberg. 2002. A theory of stochastic harvesting in Sacramento District. Permission to publish was provided by the stochastic environments. American Naturalist 159:427–437. Chief of Engineers. Melnychuk, M. C., D. W. Welch, C. J. Walters, and V. Christensen. 2007. River- ine and early ocean migration and mortality patterns of juvenile Steelhead Trout (Oncorhynchus mykiss) from the Cheakamus River, British Columbia. REFERENCES Hydrobiologia 582:55–65. Adams,P.B.,C.Grimes,J.E.Hightower,S.T.Lindley,M.L.Moser,and Moyle, P. B. 2002. Inland fishes of California, 2nd edition. University of M. J. Parsley. 2007. Population status of North American Green Sturgeon, California Press, Berkeley. Acipenser medirostris. Environmental Biology of Fishes 79:339–356. NMFS (National Marine Fisheries Service). 2006. Endangered and threatened Akçakaya, H. R., and W. Root. 2007. RAMAS metapop, version 5.0: viabil- wildlife and plants: threatened status for southern distinct population seg- ity analysis for stage-structured metapopulations. Applied Biomathematics, ment of North American Green Sturgeon. Federal Register 71:67(7 April Setauket, New York. 2006):17757–17766. Beamesderfer, R. C. P., M. L. Simpson, and G. J. Kopp. 2007. Use of life Poytress, W. R., J. J. Gruber, D. A. Trachtenbarg, and J. P. Van Eenennaam. history information in a population model for Sacramento Green Sturgeon. 2009. 2008 upper Sacramento River Green Sturgeon spawning habitat and Environmental Biology of Fishes 79:315–337. larval migration surveys. Final Annual Report to U.S. Bureau of Reclamation, Benson, R. L., S. Turo, and B. W. McCovey Jr. 2007. Migration and movement U.S. Fish and Wildlife Service, Red Bluff, California. patterns of Green Sturgeon (Acipenser medirostris) in the Klamath and Trinity Poytress, W. R., J. J. Gruber, and J. P. Van Eenennaam. 2010. 2009 upper rivers, California, USA. Environmental Biology of Fishes 79:269–279. Sacramento River Green Sturgeon spawning habitat and larval migration Boysen, K. A., and J. J. Hoover. 2009. Swimming performance of juvenile surveys. Final Annual Report to U.S. Bureau of Reclamation, U.S. Fish and White Sturgeon (Acipenser transmontanus): training and the probability of Wildlife Service, Red Bluff, California. entrainment due to dredging. Journal of Applied Ichthyology 25:54–59. Poytress, W. R., J. J. Gruber, and J. P. Van Eenennaam. 2011. 2010 upper Brown, K. 2007. Evidence of spawning by Green Sturgeon, Acipenser Sacramento River Green Sturgeon spawning habitat and larval migration medirostris, in the upper Sacramento River, California. Environmental Bi- surveys. Final Annual Report to U.S. Bureau of Reclamation, U.S. Fish and ology of Fishes 79:297–303. Wildlife Service, Red Bluff, California. Caswell, H. 2001. Matrix population models: construction, analysis, and inter- Poytress, W. R., J. J. Gruber, and J. P. Van Eenennaam. 2012. 2011 upper pretation, 2nd edition. Sinauer, Sunderland, Massachusetts. Sacramento River Green Sturgeon spawning habitat and larval migration Chapman, F. A., and J. P. Van Eenennaam. 2007. Determining the stage of surveys. Final Annual Report to U.S. Bureau of Reclamation, U.S. Fish and sexual maturity in female sturgeon for artificial spawning: the egg polarization Wildlife Service, Red Bluff, California. index or PI. Florida Cooperative Extension Service, Institute of Food and Sommer, T. R., W.C. Harrell, and M. L. Nobriga. 2005. Habitat use and stranding Agricultural Sciences, Document FA153, University of Florida, Gainesville. risk of juvenile Chinook Salmon on a seasonal floodplain. North American Available: edis.ifas.ufl.edu/pdffiles/FA/FA15300.pdf. (May 2012). Journal of Fisheries Management 25:1493–1504.

Downloaded by [Department Of Fisheries] at 00:02 20 March 2013 Dettlaff, T. A., A. S. Ginsburg, and O. I. Schmalhausen. 1993. Sturgeon fishes: Stauffer, H. B. 2008. Contemporary Bayesian and frequentist statistical research developmental biology and aquaculture. Springer-Verlag, New York. methods for natural resource scientists. Wiley, Hoboken, New Jersey. Erickson, D. L., and M. A. H. Webb. 2007. Spawning periodicity, spawning Talbott, M. J., J. P. Van Eenennaam, J. Linares-Casenave, S. I. Doroshov, C. S. migration, and size at maturity of Green Sturgeon, Acipenser medirostris, in Guy, P. Struffenegger, and M. A. H. Webb. 2011. Investigating the use of the Rogue River, Oregon. Environmental Biology of Fishes 79:255–268. plasma testosterone and estradiol-17β to detect ovarian follicular atresia in Feyrer, F., T. Sommer, and W. Harrell. 2006. Importance of flood dynamics ver- farmed White Sturgeon, Acipenser transmontanus. Aquaculture 315:283– sus intrinsic physical habitat in structuring fish communities: evidence from 289. two adjacent engineered floodplains on the Sacramento River, California. USFWS (U.S. Fish and Wildlife Service). 2009. Biological opinion on the 2008 North American Journal of Fisheries Management 26:408–417. operation of the Bonnet Carre´ Spillway. USFWS, Lafayette, Louisiana. Ginzburg, L. R., S. Ferson, and H. R. Akçakaya. 1990. Reconstructibility of Van Eenennaam, J. P., J. Linares, S. I. Doroshov, D. C. Hillemeier, T. E. density dependence and the conservative assessment of extinction risks. Con- Willson, and A. A. Nova. 2006. Reproductive conditions of the Klamath servation Biology 4:63–70. River Green Sturgeon. Transactions of the American Fisheries Society 135: Ginzburg, L. R., L. B. Slobodkin, K. Johnson, and A. G. Bindman. 1982. 151–163. Quasiextinction probabilities as a measure of impact on population growth. Van Eenennaam, J. P., M. A. H. Webb, X. Deng, S. I. Doroshov, R. B. Mayfield, Risk Analysis 2:171–181. J. J. Cech Jr., D. C. Hillemeier, and T. E. Willson. 2001. Artificial spawning Heppell, S. S. 2007. Elasticity analysis of Green Sturgeon life history. Environ- and larval rearing of Klamath River Green Sturgeon. Transactions of the mental Biology of Fishes 79:357–368. American Fisheries Society 130:159–165. This article was downloaded by: [Department Of Fisheries] On: 20 March 2013, At: 00:03 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 Escape Gaps as a Management Strategy for Reducing Bycatch in Net-Covered Traps for the Giant Mud Crab Scylla serrata Douglas Rotherham a , Daniel D. Johnson a , William G. Macbeth a & Charles A. Gray a a New South Wales Department of Primary Industries, Cronulla Fisheries Research Center of Excellence, Post Office Box 21, Cronulla, New South Wales, 2230, Australia Version of record first published: 06 Mar 2013.

To cite this article: Douglas Rotherham , Daniel D. Johnson , William G. Macbeth & Charles A. Gray (2013): Escape Gaps as a Management Strategy for Reducing Bycatch in Net-Covered Traps for the Giant Mud Crab Scylla serrata , North American Journal of Fisheries Management, 33:2, 307-317 To link to this article: http://dx.doi.org/10.1080/02755947.2012.760502

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 33:307–317, 2013 C American Fisheries Society 2013 ISSN: 0275-5947 print / 1548-8675 online DOI: 10.1080/02755947.2012.760502

ARTICLE

Escape Gaps as a Management Strategy for Reducing Bycatch in Net-Covered Traps for the Giant Mud Crab Scylla serrata

Douglas Rotherham,* Daniel D. Johnson, William G. Macbeth, and Charles A. Gray New South Wales Department of Primary Industries, Cronulla Fisheries Research Center of Excellence, Post Ofﬁce Box 21, Cronulla, New South Wales 2230, Australia

Abstract In response to experimental evidence and anecdotal concern about bycatch in the net-covered traps used increasingly by recreational and commercial fishers to target giant mud crabs Scylla serrata in New South Wales, Australia, experiments were done to examine the utility of escape gaps in reducing the bycatch of Yellowfin Bream Acanthopagrus australis (also known as Surf Bream) and undersized giant mud crabs. In each of two rivers, 10 different treatments comprising four different sizes of rectangular, horizontal escape gaps (85 × 45 mm, 85 × 55 mm, 95 × 45 mm, and 95 × 55 mm) and a control (no escape gaps) applied to both two-entrance and four-entrance traps were tested separately during the day and night. Traps fitted with escape gaps reduced the mean numbers of Yellowfin Bream and undersized giant mud crabs by 53–78% and 58–84%, respectively. Despite some inconsistent results, the effect of escape gaps was statistically significant in four of the seven analyses. There were, however, no significant differences in the mean numbers of Yellowfin Bream and undersized giant mud crabs among the different sizes of escape gap. By contrast, the largest escape gap treatment (i.e., 95 × 55 mm) reduced the mean number of legal-size giant mud crabs by 35–41%, which was significant in one of four analyses. Moreover, size-selectivity analyses indicated that the largest escape gap also allowed a comparatively greater proportion of undersized mud crabs to escape. Our findings show that implementing an 85-mm × 55-mm escape gap could substantially reduce bycatch without affecting catches of the target organism. Further reductions in catches of undersized giant mud crabs is possible with the 95-mm × 55-mm treatment, but at the cost of some reduction in the catch target.

Downloaded by [Department Of Fisheries] at 00:03 20 March 2013 The discarding of nontarget organisms and undersized indi- Although traps generally produce less bycatch than towed viduals of target species, or “bycatch,” by commercial fishing gears (Kennelly 2007), discarding of undersized individuals of gears is a problem that has received much attention in recent target organisms, particularly decapod crustaceans, is still of decades owing to global concern about the ecological impacts concern in trap fisheries worldwide (Miller 1990; Zhou and of such practices (Andrew and Pepperell 1992; Alverson et al. Shirley 1997; Guillory and Hein 1998). The negative effects 1994; Kelleher 2005). Consequently, strategies to reduce by- associated with discarding of undersized crustaceans (i.e., mor- catch have been the focus of extensive research in fisheries us- talities and sublethal injuries) have been well documented (re- ing towed gears (Hall 1996; Broadhurst 2000). The most widely viewed by Murphy and Kruse 1995). Thus, many studies have used strategy to minimize bycatch has been through physical examined modifications to trap design to reduce the capture and technological modifications of gear design, e.g., increasing of undersized lobsters (Nephropidae) and crabs (Cancridae, sizes of mesh and inserting bycatch-reduction devices (reviewed Oregoniidae and Portunidae), such as inserting escape gaps, by Broadhurst 2000). vents, or rings or increasing sizes of mesh (see reviews by Miller

*Corresponding author: [email protected] Received August 19, 2011; accepted December 11, 2012 307 308 ROTHERHAM ET AL.

1990; Guillory and Hein 1998). By comparison, much less atten- In this study, we tested the hypothesis that net-covered traps tion has been paid to reducing bycatch of nontarget organisms fitted with escape gaps would catch and retain fewer individuals in either commercial (e.g., Boutson et al. 2009) or recreational of Yellowfin Bream and undersized mud crabs (i.e., <85 mm (e.g., Rook et al. 2010) trap-based fisheries. CL) than traps without escape gaps. Examining the utility of In New South Wales, Australia, the giant mud crab Scylla escape gaps to solve problems of bycatch in traps is a logical serrata (hereafter “mud crab”) is caught by commercial (100– first choice because they are inexpensive, easy to insert and can 120 metric tons/year) and recreational (30–60 metric tons/year) be fitted to new or existing traps (Guillory and Hein 1998). An fishers in estuaries and rivers, primarily using traps (Rowling ideal escape gap would reduce catches of all Yellowfin Bream et al. 2010). Traditionally, such traps have been either square or and undersized mud crabs but not catches of legal-size crabs round and constructed of wire mesh. Although large proportions (sensu Guillory and Hein 1998; Rook et al. 2010). Thus, it was of mud crabs are estimated to be discarded by recreational fishers necessary to test the additional hypothesis that catches of legal- in Australia (about 68% of the total catch; Henry and Lyle size mud crabs (i.e., 85 mm CL) in traps would not be reduced 2003), studies directly examining bycatch in mud crab traps for by the addition of different sizes of escape gap. either recreational or commercial fisheries in New South Wales are lacking. Thus, there has been neither large concern about bycatch and ghost fishing nor any requirements for escape gaps METHODS to be fitted to wire mud crab traps in this region. Design of traps. —The circular traps used in this study Recent anecdotal evidence strongly indicates that recre- (Figure 1a, b) were 900 mm in diameter, 300 mm high, and ational and commercial fishers targeting mud crabs in New covered with 50-mm knotted PE mesh (48-ply, 2-mm-diameter South Wales have been switching from the traditional wire traps twine). This is the smallest size of mesh currently permitted to new types of collapsible, circular, steel-framed traps cov- in any type of recreational or commercial mud crab trap in ered with multifilament, polyethylene (PE) netting. This type of New South Wales. The two outer rings of the trap frame were net-covered trap has become an affordable off-the-shelf prod- made of galvanized steel rod (10-mm diameter) and held open uct, which has removed a barrier to entry for many recreational by four plastic stanchions, which could be unclipped from the fishers and subsequently increased the popularity of mud crab top ring, thus making the trap collapsible. Since recreational fishing. Collapsible, net-covered traps also offer practical (e.g., and commercial fishers are permitted to use traps with up to a less storage space required on boats and land) and economic ad- maximum of four entrances, we tested whether results of our vantages (e.g., improved durability and less maintenance) over experiments were consistent between traps configured with wire traps, particularly for commercial fishers who are permit- two or four entrances. The opening at the terminal end of each ted to use more traps (10 traps/fisher) than recreational fishers funnel-shaped entrance in a trap was approximately 50 mm (1 trap/fisher). high and 200 mm wide. In an unpublished study examining the utility of net-covered Determining shape, sizes and placement of escape gaps.— traps as a tool for sampling populations of mud crabs in New Rectangular, rounded corner, horizontal escape gaps (Figure 1c) South Wales, we found that large numbers of Yellowfin Bream were examined because we hypothesized that vertical gaps Acanthopagrus australis were incidentally caught (average of would restrict the escape of undersized mud crabs. Moreover, 5–9 fish/trap, maximum of 54 fish in a single trap), most of which Yellowfin Bream have been hypothesized to escape from hori- were under the legal size of 250 mm total length (TL) that applies zontal gaps by turning on their side (Stewart and Ferrell 2002). to both commercial and recreational fisheries. These observa- Existing data on morphometrics of bream and new data on mud tions were consistent with anecdotal evidence from compliance crabs were used to determine an appropriate range of minimum Downloaded by [Department Of Fisheries] at 00:03 20 March 2013 staff and stakeholders concerned about bycatch in net-covered and maximum dimensions of escape gaps. Most of the bream traps. While Yellowfin Bream are a key species in the recre- caught in our earlier study (i.e., 98%) were smaller than the min- ational and commercial sectors, fishers are only permitted to re- imum legal size of 250 mm TL. Thus, using the equations from tain legal-size mud crabs (85 mm carapace length [CL] for both Broadhurst et al. (2006), the maximum body width of Yellowfin males or females) in such traps; all other organisms, regardless Bream was 32.1 mm and the maximum body height (MBH) was of whether they are of legal size, must be returned to the water. 86.4 mm at the minimum legal size; for the largest bream caught If the catch rates of Yellowfin Bream observed in our earlier (280 mm fork length [FL] = 316 mm TL), the maximum body study are, indeed, typical of net-covered traps, the increased use width was 40.4 mm and the MBH was 96.6 mm. of such nets among recreational and commercial fishers may From a sample of 89 mud crabs collected from the Corindi have poor consequences for populations owing to mortality and River (29◦5830S, 153◦1335E), we measured CL (from the sublethal effects associated with capture and release (Murphy notch between the most protruding frontal teeth to the center of and Kruse 1995). Further, since PE netting is not biodegradable the posterior margin of the carapace), MBH, and the maximum (Macfadyen et al. 2009), lost or abandoned traps covered with heights of the right and left chelae separately for male and female this sort of netting may continue to “ghost-fish” over longer pe- crabs. Linear regressions between CL and the morphological riods than traditional wire traps (Campbell and Sumpton 2009). measurements showed that for legal-size male crabs (i.e., 85 mm ESCAPE GAPS IN GIANT MUD CRAB TRAPS 309

(a) CL), MBH was 47.6 mm and maximum height was 39.5 mm for the right and 38.7 mm for the left chelae. For legal-size female crabs, MBH was 46.9 mm and maximum height was 35.0 mm for the right and 33.0 mm for the left chelae (D.D.J. and coworkers, unpublished data). Based on the morphometric data for Yellowfin Bream and mud crabs and the minimum legal size of mud crabs, four different sizes of escape gap were constructed from galvanized steel (width × height): (1) 85 × 45 mm, (2) 85 × 55 mm, (3) 95 × 45 mm, and (4) 95 × 55 mm (Figure 1c). Traps with no escape gaps were the fifth trap type (control) tested. Each trap was fitted with two identical escape gaps (of one of the four different sizes) located opposite to one another in the bottom of the side panel (after Boutson et al. 2009; Figure 1b). Study area and design of experiment.—The experiment was done during the day and night in both the Kalang (30◦2935S, 153◦0127E) and Corindi (see above) rivers in northern New South Wales. This was done to test additional hypotheses that the effects of escape gaps on the numbers and sizes of Yellowfin Bream and mud crabs would be consistent between diel periods and different rivers. Nevertheless, because the two rivers were (b) sampled in different periods, differences in the effects of escape gaps between rivers were potentially confounded. Nine replicate days and nights were sampled in both the Kalang River (27 April to 7 May 2010) and the Corindi River (9–23 June 2010). At the start of each day (0700 hours) and night (1900 hours), four replicates of each of the 10 different trap treatments (i.e., two-entrance and four-entrance traps each fitted with a pair of the four different escape gaps or no escape gap) were deployed in a random order along each river in depths less than about 2 m. Following commercial and recreational practices, traps were set over predominantly bare substrates on either the right or left side of each river (selected at random) and within about 5–10 m of the shoreline. Traps were separated by 100 m to minimize nonindependence (Williams and Hill 1982). The design of sampling resulted in a total of 36 replicates of each of the 10 different treatments in each river (n = 360). (c) Each trap was baited with four individual striped mullet Mugil cephalus (each individual measured 30–40 cm FL and Downloaded by [Department Of Fisheries] at 00:03 20 March 2013 weighed about 1 kg) skewered on a wire attached to the center of the trap. After a soak time of 12 h, the catch in each trap was identified and counted. Individual bream and mud crabs were measured for FL and CL (as above), respectively. Analyses of data.—Initially, we planned to analyze the num- 85 mm bers of Yellowfin Bream, undersized mud crabs, and legal-size mud crabs caught in traps using a three-factor analysis of vari- 55 mm ance (ANOVA) model separately for each diel period (day and night) in each river (Kalang and Corindi). The three factors were entrance type (two-entrance versus four-entrance traps; fixed factor), treatment (the four different sizes of escape gap and the control; fixed factor), and trip (i.e., sampling day or FIGURE 1. Schematic diagrams of the collapsible, net-covered traps exam- night; random factor). For some variables, however, it was nec- ined in this study, together with an example of the general shape of escape gaps: essary to pool data across the four replicate traps for each trip (a) two-dimensional plan of a trap with four entrances (shaded areas) viewed (i.e., each trip became the replicate sample, resulting in n = from above, (b) three-dimensional view of a two-entrance trap with escape gaps, and (c) an 85-mm × 55-mm escape gap. 9) because we decided a priori that it was sensible to only 310 ROTHERHAM ET AL.

analyze variables containing catches in at least 25% of control in each treatment, so it was necessary to pool data from each traps (i.e., 18 out of 72 traps). In most cases, we could then replicate trap across day and night periods and the two rivers. analyze the pooled data using a two-factor model (i.e., entrance and treatment) separately for each diel period in each river. RESULTS Prior to analysis, data were tested for homogeneity of vari- In total, 3,514 Yellowfin Bream (2,111 in the Kalang River, ances using Cochran’s test and, where necessary, transformed to 1,403 in the Corindi River) and 1,196 mud crabs (620 in the log (x + 1) (Underwood 1997). To increase the power of tests e Kalang River, 576 in the Corindi River) were caught during the for the main effects, nonsignificant interaction terms were appro- experiment. Most bream caught in each river were undersized priately pooled when P > 0.25, which protected against making (94% of the total catch in the Kalang River and 99% in the type II errors (see Underwood, 1997). Our hypotheses specified Corindi River). By comparison, most crabs caught were of legal a priori that traps with escape gaps would generally catch fewer size (80% and 83%, respectively). Across all trap treatments in Yellowfin Bream and undersized mud crabs than traps with- both rivers, larger proportions of crabs were caught at night (64% out escape gaps (control). So, significant F-ratios for the factor of undersized crabs in Kalang River and 78% in the Corindi treatment were examined using a priori planned comparisons River and 78% and 91%, respectively, of legal-size crabs). (mean of all escape-gap treatments versus control). Significant differences among escape-gap treatments and between entrance Mean Numbers of Yellowfin Bream types were, however, examined using Student–Newman–Keuls The results of ANOVA of the mean numbers of Yellowfin (SNK) tests because we had no a priori expectations about the Bream revealed differing patterns between diel periods and patterns of differences among the different sizes of escape gap rivers (Table 1; Figure 2). In the Kalang River, two-entrance or between two-entrance and four-entrance traps. Similarly, for traps caught significantly more bream than four-entrance traps analyses of the mean number of legal-size crabs, we used SNK during the day (Figure 2a). Although there was a clear pattern tests to examine significant F-ratios for the factor treatment of smaller numbers of bream being caught in traps with escape (rather than a priori comparisons of the mean of all escape-gap gaps than in control traps (means reduced by 53–74%), ANOVA treatments versus control) because it was more appropriate undetected a significant trip × treatment interaction, with planned der our hypothesis that catches of legal-size mud crabs in traps comparisons revealing that the differences in the mean num- would not be reduced by different sizes of escape gap. We did bers of bream between traps with and without escape gaps were not investigate significant F-ratios for the trip factor because it inconsistent among trips (i.e., there was a significant trip × was a random factor. gap versus control interaction; Table 1). For bream caught in Analyses of size selectivity were limited to mud crabs the Kalang River at night, however, ANOVA detected a sig- because our hypothesis about the effects of escape gaps on nificant entrance × treatment interaction. Planned comparisons Yellowfin Bream did not specify any particular size-classes (entrance × gap versus control term) and subsequent SNK tests (i.e., we predicted that escape gaps would reduce the catches revealed that two-entrance traps with escape gaps caught sig- of all bream). For each entrance type (two and four) and each nificantly fewer bream than control traps (mean reduced by treatment (85 × 45 mm, 85 × 55 mm, 95 × 45 mm, 95 × 71%), but there were no differences in mean numbers of bream 55 mm, and control), the size frequencies of mud crabs caught between four-entrance traps with and without escape gaps (Ta- in each replicate trap were combined across all replicates in both ble 1; Figure 2b). rivers. Combined-hauls logistic and Richard’s size-selection By comparison, in the Corindi River, traps fitted with escape models were fitted to data for each treatment versus relevant gaps caught significantly fewer Yellowfin Bream than control control-trap pairing using maximum likelihood (ttfit; free R Downloaded by [Department Of Fisheries] at 00:03 20 March 2013 traps consistently both during the day (mean, 78% less) and at functions; available from www.stat.auckland.ac.nz/∼millar/ night (58% less; gap versus control term). There were no dif- selectware/code.html), with standard errors associated with ferences in the mean numbers of bream caught among the dif- estimated selectivity parameters corrected for overdispersion. ferent sizes of escape gap (among-gap term; Table 1; Figure 2c, (Millar et al. 2004). Successfully converged models were d). Further, while two-entrance traps caught significantly more assessed by comparing deviances and associated degrees of bream than four-entrance traps in the day (SNK test; Table 1; freedom against a chi-square distribution and by visual exam- Figure 2c), the number of entrances did not affect numbers of ination of residual plots. Where both logistic and Richard’s bream caught at night (Figure 2d). models converged, the most appropriate was determined via a likelihood ratio test. Where appropriate, the bivariate form of Mean Numbers of Undersized Mud Crabs Wald’s F-test was used to detect significant differences between In the Kalang River, the mean numbers of undersized mud selection curves (Kotz et al. 1982). crabs were significantly smaller (means reduced by 58 and 84%, Differences in the size-frequency distributions of mud crabs respectively) in traps fitted with escape gaps than in control traps among escape-gap treatments were examined separately for (gap versus control term; Table 2; Figure 3a, b). There were no two-entrance and four-entrance traps via Kolmogorov–Smirnov differences, however, among the different sizes of escape gap. (K–S) tests. These tests were only performed where n > 50 Although a similar pattern was observed for the Corindi River ESCAPE GAPS IN GIANT MUD CRAB TRAPS 311

TABLE 1. Results of ANOVA and a priori planned comparisons testing for differences in the mean numbers of Yellowfin Bream retained in mud crab traps, by period (day and night) and river (Kalang and Corindi). The factors in the model were entrance (En; two versus four entrances; fixed factor), treatment (Tr; four sizes of escape gap [Gp] and control trap with no escape gaps [Cn]; fixed factor), and trip (Tr; random factor). Prior to all analyses, data were tested for homogeneity of variances using Cochran’s test and in all cases were transformed to loge(x + 1). When the F-ratio for the interaction term was not significant at P = 0.25, the mean squares (MS) and degrees of freedom were pooled (Pld) with the residual; P < 0.05*, P < 0.01**, and P < 0.001***.

Kalang, day Kalang, night Corindi, day Corindi, night Source df MS F MS F MS F MS F En 1 5.75 27.54*** 13.04 32.83 3.54 4.94* 0.65 1.51 Tr 4 18.27 12.52 4.76 7.06 13.91 19.43*** 1.24 2.88* Gp vs. Cn 1 54.87 76.65*** 4.17 9.71** Among Gp 3 0.26 0.36 0.26 0.60 Tp 8 1.90 2.30 1.19 2.11* 1.06 1.48 1.16 2.71** En × Tr 4 2.43 2.36 1.84 3.26* 1.03 1.44 1.04 2.43 En × Gp vs. Cn 1 6.24 11.05** En × Among Gp 3 0.37 0.66 En × Tp 8 0.21 0.25 0.40 0.70 Pld Pld Tr × Tp 32 1.46 1.77** 0.67 1.19 Pld Pld Tp × Gp vs. Cn 8 2.49 3.01* Tp × Among Gp 24 1.12 1.35 En × Tr × Tp 32 1.03 1.25 Pld Pld Pld Residuals 270 0.83 Pooled residuals 0.56 (df = 302) 0.72 (df = 342) 0.43 (df = 342) Downloaded by [Department Of Fisheries] at 00:03 20 March 2013

FIGURE 2. Mean numbers of Yellowfin Bream retained in two-entrance and four-entrance mud crab traps fitted with five different treatments (four sizes of escape gap [see text] and a control with no escape gaps) in (a) day and (b) night sets in the Kalang River and (c) day and (d) night sets in the Corindi River. The whiskers represent SEs. 312 ROTHERHAM ET AL.

TABLE 2. Results of ANOVA and a priori planned comparisons (where appropriate) testing for differences in the mean numbers of undersized (<85-mm-CL) giant mud crabs retained in crab traps, by period (day and night) and river (Kalang and Corindi). See Table 1 for abbreviations. The Corindi River night samples were not analyzed owing to insufﬁcient data; the Kalang day data were analyzed with a two-factor model (En and Tr) using the number of crabs per trip as the replicate sample. Prior to all analyses, data were tested for homogeneity of variances using Cochran’s test and in all cases were transformed to loge(x + 1). Explanations of Pld and signiﬁcance are in Table 1.

Kalang, day Kalang, night Corindi, night Corindi, day Source df MS F MS F Not analyzed MS F En 1 0.06 0.41 0.30 3.13 0.08 0.62 Tr 4 0.69 4.86** 1.06 15.94*** 2.20 24.13*** Gp vs. Cn 1 1.71 12.02* 3.69 22.26*** 7.55 82.90*** Among Gp 3 0.35 2.47 0.19 1.12 0.41 4.53** Trip (Tp) 8 0.16 2.47* 0.07 0.75 En × Tr 4 0.21 1.46 0.05 0.72 0.27 2.97* En × Gp vs. Cn 1 0.02 0.21 En × Among Gp 3 0.35 3.89* En × Tp 8 0.09 1.42 0.13 1.45 Tr × Tp 32 Pld Pld Tp × Gp vs. Cn 8 Tp × Among Gp 24 En × Tr × Tp 32 Pld Pld Residuals 270 0.14 (df = 80) Final pooled residuals 0.17 (df = 334) 0.09 (df = 334)

(Figure 3c, d), data from the day were not analyzed because too Analyses of Size Selectivity few crabs were caught. Also, the ANOVA for undersized crabs No size-selectivity models successfully converged for caught at night revealed a significant entrance × treatment in- catches of mud crabs in two-entrance treatment traps. For mud teraction (Table 2; Figure 3d). Planned comparisons showed crabs caught in the four-entrance traps, only Richard’s models that there was no significant difference for the entrance × gap successfully converged for the 85-mm × 45-mm and 95-mm × versus control interaction (i.e., differences between traps with 55-mm treatments, while only logistic models converged for and without escape gaps were consistent between two-entrance 95-mm × 45-mm treatments (Table 4; Figure 5). The reduc- and four-entrance traps). However, there was a significant tion in model deviance achieved by applying Richard’s model entrance × gap interaction (i.e., inconsistent patterns of differ- to data for the 85-mm × 55-mm treatment was not significant ence among types of escape gap between two- and four-entrance (P > 0.05), so the simpler logistic model was accepted (Table 4; traps). Subsequent SNK tests revealed significant differences Figure 5). among escape-gap treatments for the two-entrance traps (85 × Despite the size of the standard errors associated with the 45 mm > 85 × 55 mm = 95 × 45 mm = 95 × 55 mm; Figure 3d), parameter estimates, a general trend of increasing CL at 50%

Downloaded by [Department Of Fisheries] at 00:03 20 March 2013 but not the four-entrance traps. The only significant difference retention (L50) for mud crabs was evident from the 85-mm × detected between the two- and four-entrance traps was for the 45-mm and 95-mm × 45-mm treatments (about 82 mm CL) 85-mm × 45-mm treatment (SNK tests, Figure 3d). to the 85-mm × 55-mm treatment (about 86.4-mm CL) and 95-mm × 55-mm treatment (about 91 mm CL; Figure 5; Ta- Mean Numbers of Legal-Sized Mud Crabs ble 4). However, Wald’s tests revealed no significant difference In all analyses, the mean numbers of legal-size mud crabs between the curves fitted for the 85-mm × 45-mm and 95-mm were significantly greater in two-entrance than in four-entrance × 45-mm treatments, the 95-mm × 45-mm and 85-mm × 55- traps in both the Kalang (61%) and Corindi (94%) rivers mm treatments, and the 85-mm × 55-mm and 95-mm × 55-mm (Table 3; Figure 4). The only significant difference detected treatments (P > 0.05). The only significant Wald’s test results among treatments was for the Corindi River at night (Table 3; were for the 95-mm × 55-mm treatment versus the 85-mm × Figure 4d). Subsequent SNK tests revealed that traps fitted with 45-mm and the 95-mm × 45-mm treatments (P < 0.05 in both 95-mm × 55-mm escape gaps caught significantly fewer legal- cases). size crabs; i.e., means were reduced by 35–41% relative to the For both the two-entrance and four-entrance traps and each other escape-gap treatments and control: 95 × 55 mm < 95 × of the four escape-gap treatments, the size-frequency distribu- 45 mm = 85 × 45 mm = 85 × 55 mm = control). tion of mud crabs caught was significantly different from that ESCAPE GAPS IN GIANT MUD CRAB TRAPS 313

FIGURE 3. Mean numbers of undersized (<85-mm-CL) giant mud crabs retained in two-entrance and four-entrance traps ﬁtted with ﬁve different treatments (four sizes of escape gap and a control with no escape gaps) in (a) day and (b) night sets in the Kalang River and (c) day and (d) night sets in the Corindi River. The whiskers represent SEs.

caught in the control trap (K–S tests; Table 5; Figure 6a, b). Although traps fitted with the 55-mm-high escape gaps (i.e., In most cases, catches from the control traps comprised larger 85 × 55 mm, 95 × 55 mm) caught smaller proportions proportions of crabs <75 mm CL and smaller proportions of of undersized crabs than the 45-mm treatments (i.e., 85 × crabs >75 mm CL than did traps fitted with escape gaps. 45 mm, 95 × 45 mm), significant differences among the various

TABLE 3. Results of ANOVA testing for differences in the mean numbers of legal-size (85-mm-CL) giant mud crabs retained in crab traps, by period (day and night) and river (Kalang and Corindi). See Table 1 for abbreviations. The Corindi River night samples were not analyzed owing to insufﬁcient data; the Kalang

Downloaded by [Department Of Fisheries] at 00:03 20 March 2013 day data were analyzed with a two-factor model (En and Tr) using the number of crabs per trip as the replicate sample. Prior to all analyses, data were tested for homogeneity of variances using Cochran’s test, but no transformations were necessary. Explanations of Pld and signiﬁcance are in Table 1.

Kalang, day Kalang, night Corindi, night Corindi, day Source df MS F MS F Not analyzed MS F En 1 2.43 9.45** 23.00 14.64** 43.40 49.10*** Tr 4 0.29 1.14 1.55 1.59 3.60 3.16* Tp 8 2.13 2.17 2.37 2.09* En × Tr 4 0.29 1.13 1.16 1.19 1.74 1.53 En × Tp 8 1.57 1.61 0.88 0.78 Tr × Tp 32 Pld Pld En × Tr × Tp 32 Pld Pld Residuals 270 0.26 (df = 80) 1.14 Pooled residuals 0.98 (df = 334) 314 ROTHERHAM ET AL.

FIGURE 4. Mean numbers of legal-size (≥85-mm-CL) giant mud crabs retained in two-and four-entrance traps ﬁtted with ﬁve different treatments (four sizes of escape gap and a control with no escape gaps) in (a) day and (b) night sets in the Kalang River and (c) day and (d) night sets in the Corindi River. The whiskers represent SEs.

escape-gap treatments were often inconsistent between two- net-covered traps while having little impact on the catch of legal- entrance and four-entrance traps (K–S tests; Table 5). Notably, size mud crabs. Further reductions in catch of undersized mud however, no significant differences in size-frequency distribu- crabs is also possible, but at the cost of some reduction in the tion of crabs in trap catches were detected between the two target organism. In most cases, the results were consistent with treatments involving the 55-mm-high escape gaps for either the our hypothesis that traps fitted with escape gaps would retain two-entrance or four-entrance configurations. significantly fewer Yellowfin Bream and undersized mud crabs than traps without escape gaps. Nevertheless, some analyses DISCUSSION indicated that the effects of escape gaps on mean numbers of Downloaded by [Department Of Fisheries] at 00:03 20 March 2013 The results of this study illustrate that it is possible to substan- Yellowfin Bream and mud crabs were inconsistent, as evidenced tially reduce the catches of nontarget organisms in collapsible, by the significant interactions in ANOVAs.

TABLE 4. Selectivity parameter estimates from combined trap-set model fits (logistic or Richard’s) to catches of giant mud crabs in four-entrance tapsforthe four different escape-gap treatments. The parameters are carapace length at 50% probability of retention (L50), selection range (SR; i.e., L75 – L25; mm), and relative fishing efficiency (p).

Treatment Combined trap sets model parameters 85 × 45 mm 85 × 55 mm 95 × 45 mm 95 × 55 mm Model type Richard’s Logistic Logistic Richard’s L50 82.34 (3.08) 86.41 (3.80) 82.02 (2.92) 91.02 (3.00) SR 7.42 (5.85) 8.83 (3.61) 6.56 (3.07) 6.39 (2.27) p 0.63 (0.06) 0.62 (0.06) 0.56 (0.06) 0.56 (0.06) ESCAPE GAPS IN GIANT MUD CRAB TRAPS 315

FIGURE 5. Selectivity curves for giant mud crabs caught in four-entrance taps with four different sizes of escape gap juxtaposed with size-frequency distribution for all mud crabs caught in the control traps; L50 is the carapace length at 50% probability of retention.

Although we did not test hypotheses explaining why the effects of escape gaps were inconsistent for some analyses, such FIGURE 6. Size-frequency distributions of giant mud crabs retained in results may be a consequence of the behavior of fish and crabs. (a) two-entrance and (b) four-entrance traps for each of the five different escape- Often, we observed Yellowfin Bream escaping (or attempting to gap treatments (four sizes of escape gap and a control with no escape gaps). escape) from a trap as it was being retrieved. In traps contain- Data are pooled across the Kalang and Corindi rivers and day and night sets. ing large numbers of bream, the effectiveness of escape gaps may have been reduced as individuals attempted to escape en highlights that escape gaps do not work if undersized crabs do masse upon retrieval of a trap. The formation of such bottle- not wish to leave a trap. necks at the escape gaps may have prevented some bream from Another consistent finding was that the four-entrance escaping before a trap reached the surface, a hypothesis that traps caught significantly fewer Yellowfin Bream than the could be tested by examining the effects of increasing the num- two-entrance traps. This result indicates that the additional ber of escape gaps in a trap. By comparison, mud crabs were entrances to a trap provided bream with additional avenues often still feeding on the bait after a trap was retrieved. Simi- of escape. Similarly, the four-entrance traps also consistently lar observations were also reported for another species of mud caught significantly fewer legal-size mud crabs than traps crab (Jirapunpipat et al. 2008), which together with our results, with two entrances. Although this finding suggests that larger

TABLE 5. Results of Kolmogorov–Smirnov tests examining differences in the size-frequency distributions of giant mud crabs between all combinationsof treatments for two-entrance versus four-entrance traps. Data were pooled across day and night periods and rivers. Levels of signiﬁcance are as in Table 1; ns = not signiﬁcant. Downloaded by [Department Of Fisheries] at 00:03 20 March 2013 Treatment (gap) Treatment (gap) 85 × 45 mm 85 × 55 mm 95 × 45 mm 95 × 55 mm Control Two-entrance traps 85 × 45 mm ** ** *** * 85 × 55 mm ns ns *** 95 × 45 mm ns *** 95 × 55 mm *** Four-entrance traps 85 × 45 mm ns ns ** *** 85 × 55 mm ns ns *** 95 × 45 mm ** *** 95 × 55 mm *** 316 ROTHERHAM ET AL.

numbers of legal-size crabs also escaped from four-entrance additional undersized mud crabs. Indeed, this option would be a traps, in an earlier laboratory experiment we found no differ- marked improvement over the status quo (i.e., no escape gaps). ences in the escapement of mud crabs between the two-entrance Although a recent experiment in New South Wales observed no and four-entrance traps (Johnson et al., unpublished data). mortalities of discarded mud crabs over a period of 3 d (Butcher Thus, the differences observed in the present study might also et al. 2012), the longer-term effects of discarding remain un- be explained by the reduced entry of legal-size crabs into known. four-entrance traps. While traps were identical, apart from the An ideal escape gap might be one that is more effective in number of entrances and size of escape gaps, the openings at the reducing the proportion of undersized mud crabs than the 85- terminal ends of the funnels in the four-entrance traps appeared mm × 55-mm escape gap but that does not reduce catches of to be less flexible than in the two-entrance traps owing to minor legal-size mud crabs, as we noted with the 95-mm × 55-mm (but necessary) differences in their construction. Reduced treatment. That ideal might lie in between those two sizes. The flexibility of the openings of the four-entrance traps may have 85-mm × 55-mm escape gap probably retained larger propor- restricted the entry of legal-size crabs, but not undersized crabs. tions of undersized mud crabs because the total length of crabs Further experiments are necessary to test this hypothesis and was larger than the width of this size of escape gap. As an also examine whether the number of entrances in rigid-wire example, mud crabs with a carapace length of 84 mm had esti- traps similarly affect catches of legal and undersized mud crabs. mated total lengths of 86 mm for females and 89 mm for males Although the number of entrances in a trap affected the catch (unpublished data). So, examining the effect of a 90-mm × of legal-size mud crabs more than the presence of escape gaps, 55-mm escape gap (compared with the 85-mm × 55-mm and there was a pattern of smaller mean numbers of legal-size crabs 95-mm × 55-mm treatments) would be logical. Preferably, such caught in traps fitted with the 95-mm × 55-mm escape gap; the an experiment should be done before examining other potential difference was significant in one analysis (Corindi, night). An modifications of crab traps, such as mesh size. Additional ex- escape gap that results in any significant or perceived reduc- periments are also necessary to test whether the findings we tions in catches of legal-size crabs is likely to be resisted by observed in the present study are consistent across other sizes fishers (Rook et al. 2010). Further, while traps fitted with 85- and types of traps, such rigid-wire traps, which are also used by mm × 55-mm escape gaps reduced catches of undersized crabs commercial and recreational fishers in New South Wales. without affecting catches of legal-size crabs in most cases, our Decisions about implementing appropriately sized escape size-selectivity analyses indicated that the probability of retain- gaps also need to consider the problem of ghost fishing. Lost ing undersized crabs was larger in this treatment (about 40%) or abandoned traps fitted with escape gaps will continue to re- than in traps with 95-mm × 55-mm escape gaps (about 16%). tain legal-size mud crabs. Sacrificial panels (300 × 300 mm) By comparison, the size-selectivity analyses showed that the constructed of galvanized wire are required to be fitted to com- 45-mm-high treatments (i.e., 85 × 45 mm and 95 × 45 mm) mercial fish traps used in estuaries of New South Wales, which both retained relatively large proportions of undersized crabs can also be used by some fishers to catch mud crabs. Such (>50%), indicating that the escapement of crabs was being lim- panels are not, however, required in commercial or recreational ited by their height, rather than width, which is consistent with crab traps. Although sacrificial wire panels corrode relatively studies of other species of crab elsewhere (Brown 1982; Zhou quickly (within months), traps constructed of PE mesh are likely and Shirley 1997). to fish for a period of between 5 and 10 years (Campbell and In light of our findings, managers now have the choice be- Sumpton 2009). While mandating the use of escape gaps across tween two different sizes of escape gap (i.e., 85 × 55 mm both fishery sectors would probably reduce the problem of ghost and 95 × 55 mm) for reducing bycatch in the net-covered traps. fishing of bream and undersized mud crabs, decisions about Downloaded by [Department Of Fisheries] at 00:03 20 March 2013 While both sizes of escape gap were similarly effective in reduc- whether to implement escape gaps together with (or as an altering catches of Yellowfin Bream, choosing one of these escape native to) sacrificial wire panels, are beyond the scope of this gaps for reducing catches of undersized mud crabs will depend paper. on the conservation and social objectives of the management of In addition to this or any future experimental work, larger- the fishery. Nevertheless, our study gives managers and stake- scale trials involving recreational and commercial fishers in New holders a basis for making decisions about the tradeoffs between South Wales should be considered. Although rarely done (but the two sizes of escape gap. see Macbeth et al. 2012), this sort of extension work is par- If minimizing the capture of undersized mud crabs was ticularly important when there are real or perceived reductions deemed to be more important than small losses of legal-size in catches of target organisms (as we demonstrated here) or mud crabs, the 95-mm × 55-mm escape gap would be appropri- when fishers are concerned about or suspicious of the general ate. By comparison, the 85-mm × 55-mm escape gap was less applicability of results obtained from highly controlled experi- effective in reducing catches of undersized mud crabs but did ments (Cox et al. 2007). Improved collaboration among fishers, not affect catches of legal-size crabs. So this escape gap might scientists, and managers is more likely to achieve acceptable be considered a more suitable option if retaining all legal-size outcomes that do not compromise the goals of sustainable man- mud crabs was a greater priority than catching and discarding agement and conservation. ESCAPE GAPS IN GIANT MUD CRAB TRAPS 317

ACKNOWLEDGMENTS Fisheries, and Forests, FRDC Project 99/158, New South Wales Fisheries We thank: the New South Wales Recreational Saltwater Trust Final Report, Canberra. for funding this research, Damian Young for assistance with Jirapunpipat, K., P. Phomikong, M. Yokota, and S. Watanabe. 2008. The effect of escape vents in collapsible pots on catch and size of the mud crab Scylla sampling, and Julian Hughes and Steve Kennelly for helpful olivacea. Fisheries Research 94:73–78. comments on the manuscript. Sampling was done under New Kelleher, K. 2005. Discards in the world’s marine fisheries: an update. FAO South Wales Agriculture Animal Care and Ethics Permit 02/15. (Food and Agriculture Organization of the United Nations) Fisheries Techni- cal Paper 470, Rome. Kennelly, S. J., editor. 2007. By-catch reduction in the world’s fisheries. REFERENCES Springer, Dordrecht, The Netherlands. Alverson, D. L., M. H. Freeburg, S. A. Murawski, and J. G. Pope. 1994. A global Kotz, S., N. L. Johnson, and C. B. Read, editors. 1982. Encyclopedia of statistical assessment of fisheries bycatch and discards. FAO (Food and Agriculture sciences. Wiley, New York. Organization of the United Nations) Fisheries Technical Paper 339, Rome. Macbeth, W. G., R. B. Millar, D. D. Johnson, C. A. Gray, R. S. Keech, and D. Andrew, N. L., and J. G. Pepperell. 1992. The by-catch of shrimp trawl fisheries. Collins. 2012. Assessment of relative performance of a square-mesh codend Oceanography and Marine Biology: An Annual Review 30:527–565. design across multiple vessels in a demersal trawl fishery. Fisheries Research Boutson, A., C. Mahasawasde, S. Mahasawasde, S. Tunkijjanukij, and T. 134–136:29–41. Arimoto. 2009. Use of escape vents to improve size and species selectivity Macfadyen, G., T. Huntington, and R. Cappell. 2009. Abandoned, lost or oth- of collapsible pot for blue swimming crab Portunus pelagicus in Thailand. erwise discarded fishing gear. United Nations Environment Programme Re- Fisheries Science 75:25–33. gional Seas Reports and Studies 185, FAO (Food and Agriculture Organiza- Broadhurst, M. K. 2000. Modifications to reduce bycatch in prawn trawls: tion of the United Nations) Fisheries and Aquaculture Technical Paper 523, a review and framework for development. Reviews in Fish Biology and Rome. Fisheries 10:27–60. Millar, R. B., M. K. Broadhurst, and W. G. Macbeth. 2004. Modelling between- Broadhurst, M. K., K. K. P. Dijkstra, D. D. Reid, and C. A. Gray. 2006. Utility haul variability in the size selectivity of trawls. Fisheries Research 67: of morphological data for key fish species in southeastern Australian beach- 171–181. seine and otter-trawl fisheries: predicting mesh size and configuration. New Miller, R. J. 1990. Effectiveness of crab and lobster traps. Canadian Journal of Zealand Journal of Marine and Freshwater Research 40:259–272. Fisheries and Aquatic Sciences 47:1228–1251. Brown, C. G. 1982. The effect of escape gaps on trap selectivity in the United Murphy, M. C., and G. H. Kruse. 1995. An annotated bibliography of capture Kingdom crab (Cancer pagurus L.) and lobster (Homarus gammarus (L.)) and handling effects on crabs and lobsters. Alaska Fishery Research Bulletin fisheries. ICES Journal of Marine Science 40:127–134. 2:23–75. Butcher, P. A., J. C. Leland, M. K. Broadhurst, B. D. Paterson, and D. G. Mayer. Rook, M. A., R. N. Lipcius, B. M. Bronner, and R. M. Chambers. 2010. Bycatch 2012. Giant mud crab (Scylla serrata): relative efficiencies of common baited reduction device conserves diamondback terrapin without affecting catch of traps and impacts on discards. ICES Journal of Marine Science 69:1511– blue crab. Marine Ecology Progress Series 409:171–179. 1522. Rowling, K., A. M. Hegarty, and M. Ives, editors. 2010. Status of fisheries Campbell, M. J., and W. D. Sumpton. 2009. Ghost fishing in the pot fishery for resources in NSW (New South Wales) 2008/09: industry and investment blue swimmer crabs Portunus pelagicus in Queensland, Australia. Fisheries NSW. Wild Fisheries Research Program, Cronulla, New South Wales. Research 95:246–253. Stewart, J., and D. J. Ferrell. 2002. Escape panels to reduce by-catch in the Cox, T. M., R. L. Lewison, R. Zydelis,ˇ L. B. Crowder, C. Safina, and A. J. Read. New South Wales demersal trap fishery. Marine and Freshwater Research 2007. Comparing effectiveness of experimental and implemented bycatch 53:1179–1188. reduction measures: the ideal and the real. Conservation Biology 21:1155– Underwood, A. J. 1997. Experiments in ecology: their logical design and 1164. interpretation using analysis of variance. Cambridge University Press, Guillory, V., and S. Hein. 1998. A review and evaluation of escape rings in blue Cambridge, UK. crab traps. Journal of Shellfish Research 17:551–559. Williams, M. J., and B. J. Hill. 1982. Factors influencing pot catches and pop- Hall, M. A. 1996. On bycatches. Reviews in Fish Biology and Fisheries 6: ulation estimates of the portunid crab Scylla serrata. Marine Biology 71: 319–352. 187–192. Henry, G. W., and J. M. Lyle, editors. 2003. The national recreational and in- Zhou, S., and T. C. Shirley. 1997. Performance of two red king crab pot designs. digenous fishing survey. Australian Government Department of Agriculture, Canadian Journal of Fisheries and Aquatic Sciences 54:1858–1864. Downloaded by [Department Of Fisheries] at 00:03 20 March 2013 This article was downloaded by: [Department Of Fisheries] On: 20 March 2013, At: 00:03 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 Age at Maturity, Fork Length, and Sex Ratio of Upper Willamette River Hatchery Spring Chinook Salmon Marc A. Johnson a & Thomas A. Friesen a a Oregon Department of Fish and Wildlife, Corvallis Research Laboratory, 28655 Highway 34, Corvallis, Oregon, 97333, USA Version of record first published: 06 Mar 2013.

To cite this article: Marc A. Johnson & Thomas A. Friesen (2013): Age at Maturity, Fork Length, and Sex Ratio of Upper Willamette River Hatchery Spring Chinook Salmon, North American Journal of Fisheries Management, 33:2, 318-328 To link to this article: http://dx.doi.org/10.1080/02755947.2012.760503

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 33:318–328, 2013 C American Fisheries Society 2013 ISSN: 0275-5947 print / 1548-8675 online DOI: 10.1080/02755947.2012.760503

ARTICLE

Age at Maturity, Fork Length, and Sex Ratio of Upper Willamette River Hatchery Spring Chinook Salmon

Marc A. Johnson* and Thomas A. Friesen Oregon Department of Fish and Wildlife, Corvallis Research Laboratory, 28655 Highway 34, Corvallis, Oregon 97333, USA

Abstract We used data from 17 brood years of coded-wire-tagged hatchery spring Chinook Salmon Oncorhynchus tshawytscha from the upper Willamette River to test for changes in mean age at maturity, fork length, and sex ratio. We found only limited evidence for any trend in age at maturity or sex ratio. However, Chinook Salmon sampled from tangle nets, recreational ﬁsheries, spawning grounds, and hatcheries all presented trends of decline in mean fork length. Rates of change in fork length ranged from 0 to 5 mm per year in most sample collections, though fork length declined more rapidly for samples from tangle nets. We also observed a positive relationship between adult fork lengths and the median monthly Paciﬁc Decadal Oscillation index in the year prior to juvenile liberation (the brood year). We suggest that future research should investigate the potential cause(s) for the decline in size of hatchery spring Chinook Salmon from the upper Willamette River, with attention to harvest, broad-scale environmental conditions, and hatchery spawning and rearing practices.

A variety of factors can affect the productivity of Pacific 2010). Large population size can intensify intraspecific compe- salmon Oncorhynchus spp. populations. Numerous studies have tition, while the probability of encountering mates decreases in demonstrated relationships between variable ocean conditions, small, structured populations (Frank and Brickman 2000). Infor- survivorship (Cole 2000; Mueter et al. 2002; Logerwell et al. mation on the stability of phenotypic traits and demographics is 2003), and growth (Bigler et al. 1996; Hobday and Boehlert therefore highly relevant to managers, who sometimes produce 2001; Wells et al. 2006) of diverse species. Size at maturity, hatchery salmon for both harvest and wild population supple- determined by growth rate and age at maturity, can in turn affect mentation purposes. the reproductive success of salmon through influence on egg Upper Willamette River (UWR) spring Chinook Salmon O. Downloaded by [Department Of Fisheries] at 00:03 20 March 2013 size (Quinn et al. 2004; Beacham 2010), egg deposition (Steen tshawytscha are listed as threatened under the U.S. Endangered and Quinn 1999), fecundity (Quinn et al. 2011), and competitive Species Act (NMFS 1999, 2005). Thirteen Willamette Project advantage for mates and redd sites (Foote 1990; Fleming and dams constructed in the upper watershed by the U.S. Army Gross 1994; Hendry et al. 2001; Ford 2012). However, reproduc- Corps of Engineers between 1941 and 1968 blocked 32% of tive advantages gained by larger size at maturity are sometimes historic spawning habitat (ODFW 2005), reducing natural pro- offset by the cost of increased vulnerability to size-selective duction within subbasins by up to 95% (ODFW and NMFS harvest (Hard et al. 2008; Kendall and Quinn 2009) or natural 2010). To mitigate for lost production and fishing opportunities, predation (Quinn et al. 2001). four state-operated hatcheries (Figure 1) release approximately In addition to individual fitness traits, demographics can in- 5.1 million juvenile Chinook Salmon into the upper Willamette fluence population productivity. For example, skewed sex ratios River annually (Johnson and Friesen 2010). Fish produced within a population can reduce the effective number of breeders by these facilities are harvested in recreational, commercial, (Waples 2002) and impact mate choice processes (Garner et al. and tribal fisheries and are used for reintroduction programs

*Corresponding author: [email protected] Received March 12, 2012; accepted December 11, 2012

318 AGE, LENGTH, AND SEX OF HATCHERY CHINOOK SALMON 319 Downloaded by [Department Of Fisheries] at 00:03 20 March 2013

FIGURE 1. The lower Columbia and Willamette rivers (major tributaries shown). Fish symbols denote upper Willamette River spring Chinook Salmon hatcheries and Z1–Z5 are lower Columbia River commercial ﬁshery management zones. 320 JOHNSON AND FRIESEN

intended to establish naturally reproducing populations in va- included both fall and spring juvenile releases (Figure 2, bottom cant or underutilized habitats (NMFS 2008). panel). Tags are typically recovered from salmon that return as Only limited information is currently available for demo- adults to hatcheries and spawning grounds or are harvested in graphic and life history traits of UWR spring Chinook Salmon. commercial and recreational fisheries. Coded wire tags pro- Length and sex data are routinely collected from hatcheries and vide a source of robust, publically available data (PSMFC various fisheries but are seldom analyzed across multiple gener- 2011) that have been used in other systems to address a wide ations. Counts of age-3 jacks and age-2 minijacks (Larsen et al. range of fisheries research and management questions, provid- 2010) are obtained through observations at the Willamette Falls ing information on marine distributions (Weitkamp and Neely fish passage facility (Figure 1). However, these precocious male 2002; Weitkamp 2010; Chamberlin et al. 2011), stock structure phenotypes typically comprise less than 5% of spring Chinook (Courtney et al. 2000; Tucker et al. 2011), and behavior Salmon that return to spawn in the UWR basin (ODFW 2011). (Mortensen et al. 2002; Parken et al. 2008). Other adult age-classes (ages 4, 5, and 6) cannot be observa- In this study, we used CWT data from 17 brood years of tionally discriminated and are pooled in counts. adult UWR hatchery spring Chinook Salmon sampled from the For several decades, the Oregon Department of Fish and Willamette and lower Columbia rivers to answer the following Wildlife has tagged large numbers of juvenile UWR hatchery questions: spring Chinook Salmon with coded wire tags (CWTs). Tagged Chinook Salmon have been released from all UWR hatcheries 1. Has the mean age at maturity of UWR spring Chinook in most recent years (Figure 2, top panel), and tag groups have Salmon changed over time? 2. Has the sex ratio of adult UWR spring Chinook Salmon changed over time? 2000 3. Has the mean fork length of adult UWR spring Chinook 1800 Marion Forks Hatchery McKenzie Hatchery Salmon changed over time? 1600 Willamette Hatchery South Santiam Hatchery 1400 By characterizing the nature or absence of trends for these 1200 population traits, we provide managers with basic phenotypic 1000 and demographic information, which may help to identify un- 800 desirable processes such as unintentional selection. 600 400 200 METHODS Chinook Released (x1000) Released Chinook 0 Data collection.—We examined CWT data from hatchery UWR spring Chinook Salmon sampled at (1) UWR hatcheries, 1990 1992 1994 1996 1998 2000 2002 2004 2006 (2) UWR spawning grounds, (3) lower Columbia River com- Release Year mercial net fisheries, and (4) combined lower Columbia and Willamette River recreational fisheries. We did not include data from tags collected in ocean fisheries, as our study focused on traits of mature UWR spring Chinook Salmon. From the Re- 300 gional Mark Information System (RMIS), we downloaded the following data for spring Chinook Salmon produced in brood years 1989–2005 and released as juveniles in the UWR basin: Downloaded by [Department Of Fisheries] at 00:03 20 March 2013 200 adult run year, fishery and gear used in collection, recovery location, recovery date, sex, fork length, tag code, release location, brood year, and release date. We then calculated the age of each 100

Day of the Year the Day of tagged fish at the time of collection (from the brood year and tag collection date). For CWTs collected in lower Columbia River commercial net 0 fisheries, we analyzed data only for Chinook Salmon released in the upper Willamette River and harvested in Columbia River 1990 1992 1994 1996 1998 2000 2002 2004 2006 management zones 1 through 5 (Figure 1). Specifically, we did Release Year not analyze data for UWR spring Chinook Salmon released from lower Columbia River net pens (see Claiborne et al. 2011) in FIGURE 2. Number of tagged juvenile spring Chinook Salmon (top panel) any sample collection. These fish are produced by Willamette released from upper Willamette River hatcheries into the upper Willamette River, 1990–2007, and day of the year (bottom panel) for release of tagged basin hatcheries and transferred to the lower Columbia River, groups of upper Willamette River spring Chinook Salmon, release years 1990– where they are acclimated in net pens to favor the local return 2007. January 1st is day 1 on the y-axis. of adult fish to support “select area” commercial fisheries, with AGE, LENGTH, AND SEX OF HATCHERY CHINOOK SALMON 321

a goal of 100% harvest (ODFW and NMFS 2010). Therefore, For each sample collection and the pooled dataset, we tested they are subject to release and harvest practices not common to for relationships between mean fork length and the brood year, other UWR hatchery spring Chinook Salmon. age, and sex of spring Chinook Salmon. We also considered We tested for effects from broad-scale ocean conditions on all first-order interaction variables. Samples that lacked data fork length at maturity. We first obtained monthly data for the for variables included in a given model were omitted from that Pacific Decadal Oscillation (PDO) index (JISAO 2012) and the analysis. We tested for the effects of ocean conditions on size at Multivariate El Nino–Southern˜ Oscillation index (MEI; NOAA maturity by evaluating relationships between mean fork length 2012), then tested for a relationship between fork length and the with median monthly PDO and MEI index values for all years median value for each index during the brood year and the five from juvenile production until adult collection. For these analy- subsequent years. ses, we used a mixed linear model approach and treated tag code, Age structure.—As different fish collection methods can pro- grouped by brood years, as a random effects variable. This ap- duce disparate sampling biases, we first tested for differences proach prevented pseudoreplication that would otherwise occur in age-class structure between fish harvested in commercial if individual observations (each fish) were treated as samples. gill nets and tangle nets to determine whether data from these We used Akaike’s Information Criterion (AIC) to assess model net types should be pooled or treated separately. In the lower fits and examined residual distributions to verify conformance Columbia River, gill nets are required to have minimum mesh with linear model assumptions. sizes of between 20.3 and 22.9 cm, whereas tangle nets have maximum mesh sizes of between 10.8 and 14.0 cm (WDFW RESULTS and ODFW 2011). Gill nets and tangle nets are not used simul- We obtained data from RMIS for 34,336 CWTs recovered taneously in the lower Columbia River, and we used the date of from UWR spring Chinook Salmon at hatcheries, spawning tag collection from these fisheries to determine which net type grounds, and terminal (freshwater) fisheries. Most tags were had been used during harvest. We then used a chi-square test to recovered from adults at UWR hatcheries (75%), followed by compare the age structures of fish collected by these two gear Willamette and Columbia River recreational fisheries (12%), types in years that employed both. Columbia River commercial net fisheries (7%), and UWR 1 Using CWT data from brood years 1989–2004 , we calcu- spawning grounds (5%). lated the mean age of each cohort in each sample collection On average, 74% of CWTs (interannual median and mean) (hatcheries, spawning grounds, net fisheries, and recreational were recovered from fish that had been released as juveniles fisheries). We then used linear regression analyses to test for during February–May. Although we observed no trend in the relationships between brood year and mean age at maturity, data for the weight of juvenile fish released in the spring (t = weighted by sample size for each brood year. We did not in- 0.43, df = 15, P = 0.6672), juvenile Chinook Salmon released clude brood year 2005 for this portion of our study, as data for in fall months were on average 0.4 g heavier each consecutive age-5 adults had not yet been uploaded to RMIS at the time of brood year (t = 2.72, df = 15, P = 0.0157, r2 = 0.29). our analyses. Sex ratio.—We examined sex ratios for spring Chinook Age Structure Salmon sampled in Columbia River commercial net fisheries, For all years and sample collections, most tags (52%) were UWR hatcheries, and UWR spawning grounds. Since these recovered from age-4 adults, followed by age-5 adults (43%) data are typically not recorded for Chinook Salmon taken in and age-3 jacks (3%). Age-2 minijacks and age-6 adults each Columbia or Willamette River recreational fisheries, we did not comprised only 1% of total tag collections (Table 1). analyze sex ratios in this sample collection. As with age struc- Despite the overall age structure of the stock, approximately Downloaded by [Department Of Fisheries] at 00:03 20 March 2013 ture analyses, we first used a chi-square test to compare the sex 86% of tagged UWR spring Chinook Salmon harvested in lower ratios of fish harvested by gill nets with those taken by tangle Columbia River net fisheries (n = 2,548) were from age-5 or nets. age-6 adults (Table 1). In years that both gill nets and tangle After plotting the number of females against the number nets were used (2002–2009) the age structure of fish taken in of males recovered from each cohort, we estimated linear re- gill nets differed significantly from that of tangle nets (χ2 = gression slopes that characterized the sex ratios in each sample 31.89, df = 3, P < 0.001), with proportionally more age-4 and collection, whereby b = 1 for 1:1 sex ratio. We calculated the age-6 fish harvested with gill nets. The greater proportion of 95% confidence intervals for each slope, then compared the age-4 fish harvested by gill nets, relative to tangle nets, could slopes for sex ratios of each sample collection to b = 1. be explained by May and June catches by gill nets, when 47% Fork length.—We performed our analyses of mean fork of harvested fish were age 4 and tangle nets were not used. length with data from age-4 and age-5 adults, since small sample We found no significant relationships between brood year sizes of other age-classes precluded formal statistical analyses. and mean age at maturity for Chinook Salmon sampled from gill nets (t = –1.18, df = 12, P = 0.260), tangle nets (t = –2.18, 1Gill net data were available for brood years 1989–1992 and 1995–2005. df = 7, P = 0.067), recreational fisheries (t = –0.92, df = 14, Tangle net data were available for brood years 1996–2005. P = 0.374), or hatcheries (t = –0.68, df = 14, P = 0.506). 322 JOHNSON AND FRIESEN

TABLE 1. Number of coded wire tags recovered from upper Willamette River hatchery spring Chinook Salmon, ages 2–6, by sample collection type. The percent of each sample collection total represented by various age-classes is in parentheses. Data from brood years 1989–2005.

Sample collection Age 2 Age 3 Age 4 Age 5 Age 6 Gill nets 0 (0%) 1 (0%) 172 (16%) 905 (83%) 17 (2%) Tangle nets 0 (0%) 1 (0%) 181 (12%) 1,256 (86%) 15 (1%) Recreational ﬁsheries 5 (0%) 62 (1%) 2,356 (54%) 1,891 (44%) 32 (1%) Spawning grounds 1 (0%) 20 (1%) 770 (43%) 996 (55%) 20 (1%) Hatcheries 353 (1%) 780 (3%) 14,519 (57%) 9,797 (38%) 186 (1%)

However, the mean age of Chinook Salmon sampled on UWR and age. Mean fork lengths of fish taken by gill nets were not, spawning grounds decreased by 4% each year (t = –2.89, df = however, different from those of all other sample collections 14, P = 0.012) for brood years 1989–2004. pooled (including tangle nets) after accounting for significant effects from age, sex, PDO, and brood year (t = 1.37, df = 12, = Sex Ratio P 0.195). Age-5 Chinook Salmon were significantly longer than age- We found no significant difference between the sex ratios of 4 fish in all sample collections (Table 2). We also found that fish taken with gill nets and tangle nets (χ2 = 0.80, df = 1, sex was associated with size differences for fish in all sample P = 0.371). The number of female Chinook Salmon recovered collections except recreational fisheries, where data for sex are from a given cohort was highly correlated with the number of typically not recorded (Table 2). In most sample collections, age- males recovered for the same cohort in lower Columbia River 4 females were longer than age-4 males. Age-4 males were only net fisheries (t = 21.23, df = 12,P< 0.001, r2 = 0.974), UWR longer than age-4 females for samples from spawning grounds. spawning grounds (t = 15.27, df = 14,P< 0.001, r2 = 0.943), Among age-5 fish, males were longer than females in all sample and UWR hatcheries (t = 17.40, df = 14,P< 0.001, r2 = collections (except where not detectable in recreational fisheries; 0.956) (Figure 3). Regression slopes (b) for these relationships Table 2). indicated that the ratio of males to females was not signifi- We found significant evidence for a decline in mean fork cantly different from 1:1 in lower Columbia River net fisheries length in age-5 Chinook Salmon in all sample collections ex- (b = 1.035 ± 0.087 for 95% CI) or UWR hatcheries (b = cept gill nets (Table 3). Although we detected no change in fork 1.111 ± 0.112 for 95% CI). Yet only half as many tags were length among samples collected with gill nets or age-4 females recovered from males than females on UWR spawning grounds collected at hatcheries, some significant change was detected (b = 0.450 ± 0.052 for 95% CI). The high r2 values that we among samples of both sexes and age-classes in all other collec- observed suggested stable sex ratios in each sample collection tions (Table 3; also see Appendix). Significant first order inter- among years. action terms suggested slightly different rates of change in fork length for males, females, and both dominant age-classes sam- Fork Length pled at hatcheries (Table 3), whereby fork length of age-5 males Chinook Salmon taken in gill nets were a mean 27 mm longer declined at the greatest rate (2 mm/year). Patterns observed in than those taken in tangle nets (t = 3.30, df = 6,P= 0.016), the pooled dataset largely reflected those of the hatchery sample

Downloaded by [Department Of Fisheries] at 00:03 20 March 2013 after accounting for differences explained by brood year, sex, collection (Table 3; Figure 4).

TABLE 2. Observed mean fork length (FL) for age-5 male spring Chinook Salmon from various sample collections, and mean relationships with FL for age-5 females, age-5 males, and age-4 females, as estimated from linear mixed models. Estimated mean FL differences for Hatcheries and Pooled represent the range of values for brood years 1989–2005.

Mean difference in FL from age-5 males Sample collection Observed mean FL for age-5 malesa Age-5 females Age-4 males Age-4 females Gill nets 847 (n = 88) –4 –125 –108 Tangle nets 814 (n = 110) –5 –86 –64 Recreational ﬁsheries 825 (n = 6) 0 –108 –108 Spawning grounds 865 (n = 39) –35 –124 –132 Hatcheries 831 (n = 647) –13 to –2 –110 to –96 –106 to –81 Pooled 832 (n = 890) –10 to + 1 –113 to –96 –106 to –79

aFor brood year 1999, which presented age and sex data in all collections. AGE, LENGTH, AND SEX OF HATCHERY CHINOOK SALMON 323

TABLE 3. Coefficients and degrees of freedom from linear mixed models 600 for the effects of median monthly PDO and brood year (as b in mm/year) on Net Fisheries mean fork length (mm) of Willamette River hatchery spring Chinook Salmon. 500 Data are for brood years 1989–2005, except gill nets (1989–1992 and 1995– 2005) and tangle nets (1996–2005). All values are rounded to the nearest whole 400 number and are significantly different from zero (P < 0.05), unless indicated as no relationship (NR) or zero. Additional model parameter values are provided 300 in the Appendix. 200 Slope (b) y=1.04x-5.79 Sample Age-5 Age-4 Age-5 Age-4 100 r2=0.97, P<0.001 collection PDO males males females females df 0 Gill nets 8 NR NR NR NR 11 0 100 200 300 400 500 600 Tangle nets NR –18 –18 –18 –18 5 Recreational 10 –1 –1 –1 –1 14 500 fisheries Spawning Grounds Recreational 10 –1 –1 –1 –1 14 400 fisheries Spawning grounds NR –5 –5 –5 –5 14 y=0.45x+3.73 Hatcheries 14 –2 –1 –1 0 14 300 Pooled 13 –2 –1 –1 0 14 r2=0.94, P<0.001 200

100 no relationship with the PDO index (Table 3). We found no relationship between fork length and the MEI index (in any year) or the PDO index in years other than the brood year. Number of Male Chinook 0 0 100 200 300 400 500 DISCUSSION 2500 By examining data from CWTs, we found evidence of mod- Hatcheries est but statistically significant declines in mean fork length for 2000 UWR hatchery spring Chinook Salmon. We also found that while sex ratios and mean ages of UWR hatchery spring Chi- nook Salmon differed significantly among sample collections, 1500 these demographic characteristics generally appeared stable. We believe our study to be the first to examine these important de- 1000 mographic and phenotypic traits across multiple generations of y=1.11x+31.64 this economically important stock. Although we did not perform extrapolations from CWT data Downloaded by [Department Of Fisheries] at 00:03 20 March 2013 500 2 r =0.96, P<0.001 to formally reconstruct UWR hatchery populations, inferences 0 from our data likely reflect patterns present in the entire hatchery population. According to WDFW and ODFW (2011), an 0 500 1000 1500 2000 2500 estimated 47% (mean; SE = 0.01) of all adult Willamette River Number of Female Chinook hatchery spring Chinook Salmon that entered the Columbia River from 1995 to 2010 were collected by hatcheries, har- FIGURE 3. Number of tagged female (x-axis) and male (y-axis) upper vested in Columbia and lower Willamette River recreational Willamette River (UWR) spring Chinook Salmon recovered from Columbia fisheries, or in lower Columbia River commercial net fisheries. River gill net fisheries (top panel), UWR spawning grounds (middle panel), and UWR hatcheries (bottom panel). Data are from age-4 and age-5 adults, brood In most years, the majority of these (collected) fish were recov- years 1989–2005. Note scale differences. ered by hatcheries (mean ± SE, 61% ± 13%), followed by lower river recreational fisheries (32% ± 12%) and commer- In most sample collections, we found that during the year cial net fisheries (5% ± 4%). Typically, another 5% of the total juvenile fish were produced (the brood year), single unit in- run was sampled through spawning ground surveys. Therefore, creases in the median monthly PDO index explained a signif- the proportions of CWTs present in each sample collection of icant increase in adult mean fork length (Table 3; Figure 5). our study (75% hatchery, 12% recreational fisheries, 7% com- Only samples from tangle nets and spawning grounds showed mercial net fisheries, and 5% spawning grounds) are similar to 324 JOHNSON AND FRIESEN

880 900 -1 860 b = -1.22 mm · yr 880 840 860 820 840 800 820 780 b = -0.14 mm · yr-1 800 Fork Length (mm) 760 780 740

Fork Length(mm) 760 720 1988 1992 1996 2000 2004 740 Brood Year 720

880 700 b = -1.87 mm · yr-1 860 -1.0 -0.5 0.0 0.5 1.0 1.5 840 PDO 820 1.5 800 780 1.0

Fork Length (mm) Fork 760 b = -0.79 mm · yr-1 740 0.5 720 1988 1992 1996 2000 2004 0.0 Brood Year

FIGURE 4. Relationships between brood year and mean fork length of mature PDO Index Value female (top panel) and male (bottom panel) upper Willamette River spring -0.5 Chinook Salmon for the dominant year-classes, age 4 (black dots) and age 5 (white dots). Mean fork lengths are for age–sex categories from the pooled dataset, each adjusted for the effect of Paciﬁc Decadal Oscillation in each brood year. Rates of change in fork length (b) are provided for each age–sex class. -1.0

Downloaded by [Department Of Fisheries] at 00:03 20 March 2013 1988 1992 1996 2000 2004 average harvest and collection rates, with some underrepresen- tation of recreational fisheries offset by overrepresentation of Year hatchery collections. Our estimates for the rates of decline in mean fork length of FIGURE 5. Relationship between median monthly Pacific Decadal Oscillation (PDO) index values for years 1989–2005 and the mean fork lengths of UWR UWR spring Chinook Salmon (Table 3) are similar to values hatchery spring Chinook Salmon produced in these brood years (top panel), reported for other stocks of Pacific salmon. Bigler et al. (1996) sampled as age-5 males (white triangles), age-5 females (white dots), age-4 found that mean fork lengths of Chinook Salmon from the Yukon males (black triangles), and age-4 females (black dots). Data are pooled from and Kenai rivers of Alaska declined by rates that ranged be- all sample collections. Median monthly PDO index values (bottom panel) for tween 1.74 and 6.52 mm per year, depending upon river and years 1989–2005. Dashed line is reference for zero PDO anomaly. age at maturity. Their meta-analysis also revealed that 45 of 47 Pacific salmon populations examined showed some evidence of decreasing body size, which they attributed to broad-scale male Chinook Salmon from Alaska’s Nushagak River declined environmental conditions and density-dependent competition by 2.6 mm per year and females declined by 3.1 mm per year. (Bigler et al. 1996). Using 29 years of catch and escapement They concluded that this decline could not be fully explained data, Kendall and Quinn (2011) found that mean fork length of by the intense fisheries on this river, citing possible effects from AGE, LENGTH, AND SEX OF HATCHERY CHINOOK SALMON 325

environmental conditions, ocean harvest, and competition with present, and evidence of a trend in age at maturity was absent hatchery-reared salmon. for samples from hatcheries and recreational fisheries. Simi- As these studies suggest, apparent trends in the mean size and larly, “fishing up effects” cannot explain the trend for declining age of Pacific salmon can be influenced by combinations of envi- size within age-classes. However, because data from spawning ronmental factors, selection-driven change, and sampling bias. grounds suggested a precipitous decline in mean age at matu- Ricker (1981) cited eight potential mechanisms that could lead rity and negative t-values were present in all other collections to observed declines in mean age or length of Chinook Salmon. (though not statistically significant), continued monitoring of These included (1) increased troll fishing on immature fish that age at maturity seems warranted for Willamette spring Chinook contribute to population estimates, (2) “fishing up” effects that Salmon populations. expose late-maturing fish to increased risk of mortality, (3) ex- A more direct mechanism than age selectivity might explain tirpation of innately larger or late-maturing populations within the trend in length of UWR hatchery spring Chinook Salmon. It the composite stock, (4) changes in ocean conditions that affect has long been recognized that size-selective fisheries can exert growth, (5) directional selection by fisheries that predominately strong evolutionary pressures on salmon populations, generat- harvest older individuals, (6) directional selection by fisheries ing detectable changes in size within ten or fewer generations that predominately harvest larger individuals, (7) changes in (reviewed by Hard et al. 2008). However, our data do not sug- regulations governing harvest practices, and (8) some aspect of gest that Chinook Salmon harvested in gill nets are significantly artificial propagation. larger than found in other sample collections. Moreover, selec- Two of the mechanisms proposed by Ricker (1981) cannot tion imposed by gill nets is limited by short seasons and strictly explain the trends in mean fork length that we found in UWR enforced quotas. Commercial net harvests of UWR spring Chi- hatchery spring Chinook Salmon. First, because we only used nook Salmon have been estimated at 0.1–7.4% (median 1.2%) data from fish collected during or after spawning migration, of the total run between 1995 and 2010 (WDFW and ODFW potential biases posed by sampling immature fish (mechanism 1) 2011). were largely avoided. Second, we focused our study on hatchery Both Ricker (1981) and, more recently, Hankin et al. (2009) populations, none of which were extirpated during the period proposed that hatchery practices could lead to declines in mean of sample collections (mechanism 3). Having excluded these size at maturity for Pacific salmon. Hankin et al. (2009) sug- two mechanisms, all others proposed by Ricker (1981) merit gested that random matings at hatcheries could unnaturally boost consideration. the fitness of younger, smaller individuals that would otherwise In their analysis of size trends, Bigler et al. (1996) cited the be outcompeted on spawning grounds by older, larger salmon. importance of competition and broad-scale ocean conditions on As age at maturity appears to be a heritable trait in Chinook growth of Pacific salmon. Though we found a strong positive Salmon (Hankin et al. 1993; Kinnison et al. 2011), random mat- relationship between mean fork length and PDO in most sample ings could then increase the proportion of early maturing indi- collections, this relationship did not explain the 17-year trend viduals in each generation. Yet jacks are relatively rare in UWR for fork length revealed by our data (Figure 5). Instead, the PDO Chinook Salmon populations, and we have shown that there is index explained short-term oscillations in mean fork length. In- little evidence for a trend toward earlier age at maturity in this terestingly, evidence of several recent “regime shifts” (Mantua stock. Nevertheless, random matings performed at hatcheries and Hare 2002) were present in the PDO index data during the could still increase the fitness of smaller individuals within year- period of our study. Such regime shifts are typically recorded classes, inadvertently relaxing natural selection on heritable size every 20–30 years (Mantua and Hare 2002). The positive rela- at age. Additional research is still needed to establish how hatch- tionship between PDO in the year prior to juvenile release and ery practices may influence life history and phenotypic traits of Downloaded by [Department Of Fisheries] at 00:03 20 March 2013 fork length of mature adults is consistent with the positive PDO– salmon. Given the trend toward smaller hatchery fish suggested ocean productivity relationship widely recognized to occur in by our data, we recommend that hatchery managers consider al- southeast Alaskan waters (Hare et al. 1999). In each year im- ternative spawning practices that favor the reproductive success mediately following positive PDO anomaly years, UWR spring of larger, perhaps older, individuals. Such practices may serve Chinook Salmon, which are thought to spend marine residence to compensate for selection against larger fish (regardless of the in southeast Alaskan waters (Weitkamp 2010), likely benefit mechanism) and more closely mimic mate selection processes from abundant forage fish and invertebrate prey species sup- that occur in nature (Hankin et al. 2009). ported by high primary productivity of the previous year. We Although we found sex ratios to differ among sample collec- suspect that this pattern was not detectable for fish from tangle tions, the high r2 values we obtained from our analyses (Figure 3) nets and spawning grounds because of the smaller sample sizes suggest that sex ratios within collections are stable among years. in these collections. The skewed sex ratios that we observed on spawning grounds Although commercial net fisheries appear to be highly se- were likely the consequence of male–female differences in be- lective for age-5 Chinook Salmon, the declines we observed in havior. Female salmon expend considerable energy searching mean fork length cannot simply be explained by harvest of these for suitable habitats to establish redd sites. After spawning, older fish, since size trends for both age-4 and age-5 fish were females then safeguard their reproductive investment through 326 JOHNSON AND FRIESEN

active parental care, remaining on redds to tend and fan eggs. In REFERENCES contrast, male salmon increase their fitness by successfully lo- Anderson, J. H., and T. P. Quinn. 2007. Movements of adult Coho Salmon cating additional mates as they become receptive (Esteve 2005; (Oncorhynchus kisutch) during colonization of newly accessible habitat. Quinn 2005; Anderson and Quinn 2007). These alternate behav- Canadian Journal of Fisheries and Aquatic Sciences 64:1143–1154. Beacham, T. D. 2010. Revisiting trends in the evolution of egg size in hatchery- iors place males at a higher risk to enter hatcheries and fish traps, enhanced populations of Chinook Salmon from British Columbia. Transac- often without the possibility to return to spawning grounds. In tions of the American Fisheries Society 139:579–585. agreement with this hypothesis, we found that sex ratios in the Bigler, B. S., D. W. Welch, and J. H. Helle. 1996. A review of size trends among large hatchery sample collections were slightly skewed in favor North Pacific salmon (Oncorhynchus spp.). Canadian Journal of Fisheries of males, though not significantly different from 1:1. In prac- and Aquatic Sciences 53:455–465. Chamberlin, J. W., T. E. Essington, J. W. Ferguson, and T. P. Quinn. 2011. tice, the skewed sex ratios of hatchery fish on spawning grounds The influence of hatchery rearing practices on salmon migratory behavior: suggests that, in the Willamette River, females may contribute is the tendency of Chinook Salmon to remain within Puget Sound affected more to the proportion of hatchery spawners on natural spawn- by size and date of release? Transactions of the American Fisheries Society ing grounds than males. A significant reduction in the propor- 140:1398–1408. tion of hatchery spawners on natural spawning grounds might Claiborne, A. M., J. P. Fisher, S. A. Hayes, and R. L. Emmett. 2011. Size at release, size-selective mortality, and age of maturity of Willamette River therefore be achieved by excluding females from programs that hatchery yearling Chinook Salmon. Transactions of the American Fisheries return hatchery fish to the river, as females may be less likely Society 140:1135–1144. than males to enter fish traps a second time. Cole, J. 2000. Coastal sea surface temperature and Coho Salmon production off In conclusion, we found that while sex ratios and age at ma- the north-west United States. Fisheries Oceanography 9:1–16. turity of UWR hatchery spring Chinook Salmon appeared to Courtney, D. L., D. G. Mortensen, J. A. Orsi, and K. M. Munk. 2000. Origin of juvenile Pacific salmon recovered from coastal southeastern Alaska identified be stable, this stock experienced a slow but significant decline by otolith thermal marks and coded wire tags. Fisheries Research 46:267–278. in mean fork length between brood years 1989 and 2005, with Dickerson, B. R., K. W. Brinck, M. F. Willson, P. Bentzen, and T. P. Quinn. males experiencing an approximate 2–3% decrease in mean 2005. Relative importance of salmon body size and arrival time at breeding body length over four generations. Positive relationships be- grounds to reproductive success. Ecology 86:347–352. tween body size with fitness traits and reproductive success have Dickerson, B. R., T. P. Quinn, and M. F. Willson. 2002. Body size, arrival date, and reproductive success of Pink Salmon, Oncorhynchus gorbuscha. been well documented in Pacific salmon (Fleming and Gross Ethology Ecology and Evolution 14:29–44. 1994; Dickerson et al. 2002; Knudsen et al. 2008; Williamson Esteve, M. 2005. Observations of spawning behaviour in Salmoninae: Salmo, et al. 2010; but see Dickerson et al. 2005). Thus, even slow Oncorhynchus and Salvelinus. Reviews in Fish Biology and Fisheries 15: declines in fork length, as we have described for UWR spring 1–21. Chinook Salmon, could have serious individual fitness and pop- Fleming, I. A., and M. R. Gross. 1994. Breeding competition in a Pacific salmon (Coho: Oncorhynchus kisutch): measures of natural and sexual selection. ulation productivity consequences if allowed to continue. Al- Evolution 48:637–657. though we have not identified the mechanism(s) responsible for Foote, C. J. 1990. An experimental comparison of male and female spawning the observed trend in fork length, we recommend that hatch- territoriality in a Pacific salmon. Behaviour 115:283–314. ery spawning practices be considered as a possible contributing Ford, M., A. Murdoch, and S. Howard. 2012. Early male maturity explains factor. Other candidate factors, such as climatic variables not a negative correlation in reproductive success between hatchery-spawned salmon and their naturally spawning progeny. Conservation Letters 5: considered in this study, should also receive due considera- 450–458. tion. Furthermore, we recommend that future research inves- Frank, K. T., and D. Brickman. 2000. Allee effects and compensatory population tigate the stability of fork length and other potentially impor- dynamics within a stock complex. Canadian Journal of Fisheries and Aquatic tant phenotypic traits in threatened wild UWR spring Chinook Sciences 57:513–517. Salmon. Garner, S. R., R. N. Bortoluzzi, D. D. Heath, and B. D. Neff. 2010. Sexual Downloaded by [Department Of Fisheries] at 00:03 20 March 2013 conflict inhibits female mate choice for major histocompatibility complex dissimilarity in Chinook Salmon. Proceedings of the Royal Society of London B 277:885–894. Hankin, D. G., J. Fitzgibbons, and Y. Chen. 2009. Unnatural random mating ACKNOWLEDGMENTS policies select for younger age at maturity in hatchery Chinook Salmon This work was partially funded by the U.S. Army Corps of (Oncorhynchus tshawytscha) populations. Canadian Journal of Fisheries and Engineers under Cooperative Agreement W9127N-10-02-0008, Aquatic Sciences 66:1505–1521. administered by David Leonhardt. Kirk Schroeder, Thomas Hankin, D. G., J. W. Nicholas, and T. W. Downey. 1993. Evidence for inheritance of age of maturity in Chinook Salmon (Oncorhynchus tshawytscha). Canadian Quinn, Chris Walker, and four anonymous reviewers provided Journal of Fisheries and Aquatic Sciences 50:347–358. helpful comments on an earlier draft. We thank Chris Kern of Hard, J. J., M. R. Gross, M. Heino, R. Hilborn, R. G. Kope, R. Law, and J. D. the Oregon Department of Fish and Wildlife for his insights into Reynolds. 2008. Evolutionary consequences of fishing and their implications lower Columbia River fisheries and Erin Gilbert of the Oregon for salmon. Evolutionary Applications 1:388–408. Department of Fish and Wildlife for assistance with producing Hare, S. R., N. J. Mantua, and R. C. Francis. 1999. Inverse production regimes: Alaska and West Coast Pacific salmon. Fisheries 24(1):6–14. the study area map. The conclusions presented here are those of Hendry, A. P., O. K. Berg, and T. P. Quinn. 2001. Breeding location choice in the authors and do not necessarily reflect the views of the state salmon: causes (habitat, competition, body size, energy stores) and conse- of Oregon. quences (life span, energy stores). Oikos 93:407–418. AGE, LENGTH, AND SEX OF HATCHERY CHINOOK SALMON 327

Hobday, A. J., and G. W. Boehlert. 2001. The role of coastal ocean variation tory, Physical Sciences Division, Boulder, Colorado. Available: www.esrl. in spatial and temporal patterns in survival and size of Coho Salmon (On- noaa.gov/psd/enso/mei./ (May 2012). corhynchus kisutch). Canadian Journal of Fisheries and Aquatic Sciences ODFW (Oregon Department of Fish and Wildlife). 2005. Oregon na- 58:2021–2036. tive fish status report. ODFW, Salem. Available: www.dfw.state.or.us/fish/ JISAO (Joint Institute for the Study of the Atmosphere and Ocean). 2012. Pacific ONFSR./ (September 2012). decadal oscillation (PDO) index. JISAO, University of Washington, Seattle. ODFW (Oregon Department of Fish and Wildlife) and NMFS (National Available: jisao.washington.edu/pdo./ (May 2012). Marine Fisheries Service). 2010. Upper Willamette conservation and re- Johnson, M. A., and T. A. Friesen. 2010. Spring Chinook Salmon hatcheries covery plan for Chinook Salmon and Steelhead. ODFW, Salem. Avail- in the Willamette basin: existing data, discernable patterns and information able: www.dfw.state.or.us/fish/CRP/upper willamette river plan.asp. (April gaps. Final report to U.S. Army Corps of Engineers, Task Order NWPPM- 2011). 09-FH-05, Oregon Department of Fish and Wildlife, Corvallis. ODFW (Oregon Department of Fish and Wildlife). 2011. Willamette Kendall, N. W., and T. P. Quinn. 2009. Effects of population-specific variation in Falls fish passage. ODFW, Salem. Available: www.dfw.state.or.us/fish/ age and length on fishery selection and exploitation rates of Sockeye Salmon fish counts/willamette falls.asp. (April 2011). (Oncorhynchus nerka). Canadian Journal of Fisheries and Aquatic Sciences Parken, C. K., J. R. Candy, J. R. Irvine, and T. D. Beacham. 2008. Genetic 66:896–908. and coded wire tag results combine to allow more-precise management of Kendall, N. W., and T. P.Quinn. 2011. Length and age trends of Chinook Salmon a complex Chinook Salmon aggregate. North American Journal of Fisheries in the Nushagak River, Alaska, related to commercial and recreational fishery Management 28:328–340. selection and exploitation. Transactions of the American Fisheries Society PSMFC (Pacific States Marine Fisheries Commission). 2011. Regional mark 140:611–622. processing center. PSMFC, Portland, Oregon. Available: www.rmpc.org. Kinnison, M. T., T. P. Quinn, and M. J. Unwin. 2011. Correlated contempo- (March 2011). rary evolution of life history traits in New Zealand Chinook Salmon, On- Quinn, T. P. 2005. The behavior and ecology of Pacific salmon and trout. corhynchus tshawytscha. Heredity 106:448–459. American Fisheries Society, Bethesda, Maryland. Knudsen, C. M., S. L. Schroder, C. Busack, M. V. Johnston, T. N. Pear- Quinn, T. P., T. R. Seamons, L. A. Vøllestad, and E. Duffy. 2011. Effects sons, and C. R. Strom. 2008. Comparison of female reproductive traits and of growth and reproductive history on the egg size–fecundity trade-off in progeny of first-generation hatchery and wild upper Yakima River spring Steelhead. Transactions of the American Fisheries Society 140:45–51. Chinook Salmon. Transactions of the American Fisheries Society 137:1433– Quinn, T. P., L. A. Vøllestad, J. Peterson, and V. Gallucci. 2004. Influences 1445. of freshwater and marine growth on the egg size–egg number tradeoff in Larsen, D. A., B. R. Beckman, and K. A. Cooper. 2010. Examining the con- Coho and Chinook Salmon. Transactions of the American Fisheries Society flict between smolting and precocious male maturation in spring (stream- 133:55–65. type) Chinook Salmon. Transactions of the American Fisheries Society 139: Ricker, W. E. 1981. Changes in the average size and average age of Pacific 564–578. salmon. Canadian Journal of Fisheries and Aquatic Sciences 38:1636–1656. Logerwell, E. A., N. Mantua, P. W. Lawson, R. C. Francis, and V. N. Agostini. Steen, R. P., and T. P. Quinn. 1999. Egg burial depth by Sockeye Salmon (On- 2003. Tracking environmental processes in the coastal zone for understand- corhynchus nerka): implications for survival of embryos and natural selection ing and predicting Oregon Coho (Oncorhynchus kisutch) marine survival. on female body size. Canadian Journal of Zoology 77:836–841. Fisheries Oceanography 12:554–568. Tucker, S., M. Trudel, D. W. Welch, J. R. Candy, J. F. T. Morris, M. E. Thiess, C. Mantua, N. J., and S. R. Hare. 2002. The Pacific decadal oscillation. Journal of Wallace, and T. D. Beacham. 2011. Life history and seasonal stock-specific Oceanography 58:35–44. ocean migration of juvenile Chinook Salmon. Transactions of the American Mortensen, D. G., A. C. Wertheimer, J. M. Maselko, and S. G. Taylor. 2002. Fisheries Society 140:1101–1119. Survival and straying of Auke Creek, Alaska, Pink Salmon marked with coded Waples, R. S. 2002. Effective size of fluctuating salmon populations. Genetics wire tags and thermally induced otolith marks. Transactions of the American 161:783–791. Fisheries Society 131:14–26. WDFW (Washington Department of Fish and Wildlife) and ODFW (Ore- Mueter, F. J., R. M. Peterman, and B. J. Pyper. 2002. Opposite effects of gon Department of Fish and Wildlife). 2011. Stock status and fisheries for ocean temperature on survival rates of 120 stocks of Pacific salmon (On- spring Chinook, summer Chinook, Sockeye, Steelhead, and other species, and corhynchus spp.) in northern and southern areas. Canadian Journal of Fish- miscellaneous regulations: 2011 joint staff report. WDFW, Joint Columbia eries and Aquatic Sciences 59:456–463. River Management Staff, Olympia. Available: wdfw.wa.gov/publications/ NMFS (National Marine Fisheries Service). 1999. Endangered and threatened 01161/wdfw01161.pdf. (February 2011).

Downloaded by [Department Of Fisheries] at 00:03 20 March 2013 species; threatened status for three Chinook Salmon evolutionarily signif- Weitkamp, L. 2010. Marine distributions of Chinook Salmon from the west coast icant units (ESUs) in Washington and Oregon, and endangered status for of North America determined by coded wire tag recoveries. Transactions of one Chinook Salmon ESU in Washington. Federal Register 64:56(24 March the American Fisheries Society 139:147–170. 1999):14308–14328. Weitkamp, L., and K. Neely. 2002. Coho Salmon (Oncorhynchus kisutch) ocean NMFS (National Marine Fisheries Service). 2005. Endangered and threatened migration patterns: insight from marine coded-wire tag recoveries. Canadian species: final listing determinations for 16 ESUs of West Coast salmon, Journal of Fisheries and Aquatic Sciences 59:1100–1115. and final 4(d) protective regulations for threatened salmonid ESUs. Federal Wells, B. K., C. B. Grimes, J. C. Field, and C. S. Reiss. 2006. Covariation Register 70:123(28 June 2005):37160–37204. between the average lengths of mature Coho (Oncorhynchus kisutch)and NMFS (National Marine Fisheries Service). 2008. Endangered species act sec- Chinook Salmon (O. tshawytscha) and the ocean environment. Fisheries tion 7(a)(2) consultation biological opinion and Magnuson-Stevens fishery Oceanography 15:67–79. conservation and management act essential fish habitat consultation: consulta- Williamson, K. S., A. R. Murdoch, T. N. Pearsons, E. J. Ward, and M. J. Ford. tion on the “Willamette River basin flood control project.” NMFS, Northwest 2010. Factors influencing the relative fitness of hatchery and wild spring Region, F/NWR/2000/02117, Seattle. Chinook Salmon (Oncorhynchus tshawytscha) in the Wenatchee River, Wash- NOAA (National Oceanic and Atmospheric Administration). 2012. Mul- ington, USA. Canadian Journal of Fisheries and Aquatic Sciences 67:1840– tivariate ENSO index (MEI). NOAA, Earth System Research Labora- 1851. 328 JOHNSON AND FRIESEN

APPENDIX: DETAILED DATA

TABLE A.1. Linear mixed model variable coefficients, standard errors (SE), degrees of freedom (df), t-values, and P-values. Variables are significant predictors of mean fork length for upper Willamette River hatchery spring Chinook Salmon derived from samples collected by various methods (see text), brood years 1989–2005. Age5 and Female are indicator variables (value = 0, 1); PDO and BroodYear are continuous. Coefficient units are in millimeters. For example, the mean fork length (FL) of an age-5 male collected from spawning grounds in 2002 is estimated as follows:

FL = 11436.46 + (2002 ·−5.34) + (1 · 123.80) + (0 ·−8.13) + (0 ·−26.86) = 871 (mm).

Sample collection Variables Value SE df t-value P-value Gill nets Intercept 731.67 5.95 928 122.90 <0.01 PDO 8.45 2.70 11 3.13 0.01 Age5 124.92 6.34 928 19.70 <0.01 Female 16.45 8.45 928 1.95 0.05 Female:Age5 –20.02 8.98 928 –2.23 0.03

Tangle nets Intercept 37,431.33 2,695.73 1,371 13.89 <0.01 BroodYear –18.36 1.35 5 –13.62 <0.01 Age5 86.17 5.66 1,371 15.22 <0.01 Female 22.25 7.43 1,371 3.00 <0.01 Female:Age5 –26.82 7.95 1,371 –3.37 <0.01

Recreational ﬁsheries Intercept 2,495.87 822.91 4,120 3.03 0.00 PDO 10.05 3.33 14 3.02 0.01 BroodYear –0.88 0.41 14 –2.13 0.05 Age5 108.33 2.16 4,120 50.23 <0.01

Spawning grounds Intercept 11,436.46 2,711.25 1,637 4.22 <0.01 BroodYear –5.34 1.36 14 –3.93 <0.01 Age5 123.80 6.06 1,637 20.43 <0.01 Female –8.13 5.35 1,637 –1.52 0.13 Female:Age5 –26.86 7.39 1,637 –3.63 <0.01

Hatcheries Intercept 2,590.32 685.66 24,245 3.78 <0.01 PDO 13.50 2.50 14 5.40 <0.01 BroodYear –0.93 0.34 14 –2.70 0.02 Age5 1,900.50 408.12 24,245 4.66 <0.01 Female –1,388.26 373.92 24,245 –3.71 <0.01 Female:Age5 –17.13 1.50 24,245 –11.44 <0.01 <

Downloaded by [Department Of Fisheries] at 00:03 20 March 2013 BroodYear:Female 0.70 0.19 24,245 3.74 0.01 BroodYear:Age5 –0.90 0.20 24,245 –4.38 <0.01

Pooled data Intercept 2,306.70 692.65 28,342 3.33 <0.01 PDO 13.04 2.53 14 5.16 <0.01 BroodYear –0.79 0.35 14 –2.26 0.04 Age5 2,281.65 401.15 28,342 5.69 <0.01 Female –1,285.62 369.79 28,342 –3.48 <0.01 Female:Age5 –17.03 1.39 28,342 –12.28 <0.01 BroodYear:Female 0.65 0.19 28,342 3.51 <0.01 BroodYear:Age5 –1.09 0.20 28,342 –5.41 <0.01 This article was downloaded by: [Department Of Fisheries] On: 20 March 2013, At: 00:04 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 Spawning Substrate Size, Shape, and Siltation Influence Walleye Egg Retention Derek P. Crane a & John M. Farrell a a Department of Environmental and Forest Biology, State University of New York–College of Environmental Science and Forestry, 246 Illick Hall, 1 Forestry Drive, Syracuse, New York, 13210, USA Version of record first published: 06 Mar 2013.

To cite this article: Derek P. Crane & John M. Farrell (2013): Spawning Substrate Size, Shape, and Siltation Influence Walleye Egg Retention, North American Journal of Fisheries Management, 33:2, 329-337 To link to this article: http://dx.doi.org/10.1080/02755947.2012.760504

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 33:329–337, 2013 C American Fisheries Society 2013 ISSN: 0275-5947 print / 1548-8675 online DOI: 10.1080/02755947.2012.760504

ARTICLE

Spawning Substrate Size, Shape, and Siltation Inﬂuence Walleye Egg Retention

Derek P. Crane* and John M. Farrell Department of Environmental and Forest Biology, State University of New York–College of Environmental Science and Forestry, 246 Illick Hall, 1 Forestry Drive, Syracuse, New York 13210, USA

Abstract We compared Walleye Sander vitreus egg retention among varying rock treatments placed in a hydraulic flume to test the influence of spawning substrate shape and size on egg retention and the influence of substrate siltation on egg adhesiveness. Egg loss from suboptimal spawning substrates has been hypothesized as a substantial source of mortality during the incubation period. To investigate the influence of substrate size and shape on Walleye egg retention, known numbers of Walleye eggs were pipetted onto rock substrate in a hydraulic flume and exposed to flowing water conditions. Eggs scoured from the substrate during each trial were collected and enumerated. Egg retention was higher in angular crushed limestone than in round glacial till and greater in coarse gravel than in larger size-classes, under all water velocity regimes. Angular coarse gravel had the highest egg retention rate (mean ± SD, 66.9 ± 3.7%), and round very coarse gravel had the lowest egg retention rate (47.0 ± 3.1%) of all the rock size and shape treatments. Substrate siltation significantly influenced egg adhesion. Clean rocks had a mean ± SD egg retention of 35.9 ± 36.6%, whereas fine-sediment-covered rocks did not retain any eggs throughout the course of the experiment. Differences in substrate size, shape, and siltation between Walleye spawning sites may contribute to variation in egg retention rates. Using angular gravel during the creation of Walleye spawning habitat and maintaining clean spawning substrates (through scouring effects and habitat restoration) may decrease mortality associated with egg entrainment and redistribution.

Spawning habitat loss and degradation, which have resulted sources, such as physical damage, predation, and deposition in from habitat fragmentation, watershed development, and pollu- lower quality habitat (Pitlo 1989; Newbury and Gaboury 1993; tion, have been recognized as major impediments to the natu- Roseman et al. 1996; Dustin and Jacobson 2003; Kelder and ral reproduction of Walleyes Sander vitreus (Kerr et al. 1997; Farrell 2009; Bozek et al. 2011b). For example, Roseman et al. Downloaded by [Department Of Fisheries] at 00:04 20 March 2013 Roseman et al. 2010; Bozek et al. 2011b). Inadequate interstitial (2001) documented an 80% reduction in Walleye eggs from space in spawning habitat (e.g., fine sediment clogging of inter- western Lake Erie reefs and subsequent poor larval catches af- stitial spaces in coarse substrates) can lead to egg loss from incu- ter a gale-force storm during the incubation period. The authors bation habitat, their subsequent entrainment in the water column, (Roseman et al. 2001) hypothesized that eggs lost from the and redistribution to low flow depositional areas. Fine sediment reefs were redistributed to deeper waters with silt substrates and deposition on Walleye spawning habitat may also inhibit egg consequently suffered high mortality rates. Alternatively, eggs adhesion to the substrate, potentially increasing their vulnera- scoured from suboptimal habitats may be trapped by higher bility to entrainment and redistribution. The adhesive properties quality incubating habitat downstream or adjacent to their orig- of eggs in many teleost species may be an ecological adaption inal deposition site. Dustin and Jacobson (2003) hypothesized to promote retention of eggs in the spawning habitat selected that eggs scoured from suboptimal Walleye spawning habitat in by the female (Riehl 1996). Movement of eggs away from their a Minnesota river were transported downstream and trapped by initial incubation sites may expose them to a variety of mortality coarse substrate in newly constructed spawning riffles.

*Corresponding author: [email protected] Received June 13, 2012; accepted December 12, 2012 329 330 CRANE AND FARRELL

Several studies have provided anecdotal evidence that egg re- TABLE 1. Description of rock size and shape treatments. The 50:50 mix rock distribution is likely a factor of incubation habitat quality (e.g., size-class was an equal mixture by volume of #2 and #3/#4 rock size-classes. substrate size, substrate shape, fine sediment accumulation) and A handheld one-half phi unit gravelometer was used to calculate the D50 and range of size-classes for each rock treatment. quantity. Johnson (1961) hypothesized that the higher survival of Walleye eggs incubated on gravel–rubble substrates than on sand Size class D50 (mm) Range of sizes (mm) or muck substrates was partly due to lower egg scour rates from gravel–rubble incubation habitat. Dustin and Jacobson (2003) Angular < < observed a higher catch-per-unit-effort of drifting Walleye eggs #2 18.8 16– 45 < < in an unmodified river reach with fine gravel and sand substrate #3/#4 44.3 32– 90 + < < than below a constructed spawning riffle with coarser substrate. #2 #3/#4 23.8 16– 90 The authors (Dustin and Jacobson 2003) suggested that newly Round placed coarse rock with clean interstitial spaces was better at re- #2 20.2 <16–<45 taining eggs than fine gravel and sand. In a recent study of Wall- #3/#4 43.9 <22.6–<90 eye spawning habitat in Kent’s Creek, New York, Kelder and #2 + #3/#4 26.5 <16–<90 Farrell (2009) documented redistribution of eggs to fine-grain, lower-quality incubating habitat during moderate and high flow events. Kelder and Farrell (2009) stated that high egg redistri- limestone, while round rock consisted of screened glacial till. A bution rates may be indicative of inadequate incubation habitat, handheld one-half phi unit gravelometer was used to calculate as suggested by Dustin and Jacobson (2003). However, Kelder the median particle size (D50) of each rock treatment (Potyondy and Farrell (2009) also recognized the possibility that egg re- and Bunte 2002). Each rock treatment was also assigned a cate- distribution may be common regardless of habitat conditions gorical class size according to the combined Udden–Wentworth and that the relationship between incubation habitat and egg re- grain size scale (Udden 1914; Wentworth 1922). The #2 rock tention is poorly understood. Assuming that eggs scoured from was primarily coarse gravel and the #3/#4 rock treatment was their original deposition site are at risk to predation, physical primarily very coarse gravel (Udden 1914; Wentworth 1922; damage, and deposition onto fine sediments, egg redistribution Table 1). The mixed rock treatment was a 50:50 mixture by vol- is a likely source of mortality (Pitlo 1989; Dustin and Jacobson ume of #2 rock and #3/#4 rock. Rock sizes were chosen based on 2003; Jones et al. 2003; Kelder and Farrell 2009), which may literature values for Walleye spawning substrates (Corbett and be addressed by optimizing substrates and hydrologic processes Powles 1986; Pitlo 1989; Kelder and Farrell 2009; Chalupnicki for egg retention during restoration projects. et al. 2010; Ivan et al. 2010) and size limitations of the flume. All Currently there is limited information that quantitatively de- substrates were washed prior to placement in the flume. Water scribes the relationship between spawning substrate character- velocities were based on a range of literature values for Wall- istics and Walleye egg retention. To address this knowledge gap, eye egg incubation sites in fluvial environments (Pitlo 1989; we used a hydraulic flume to investigate (1) the effects of rock Gillenwater et al. 2006; Chalupnicki et al. 2010). substrate size and shape on Walleye egg retention under three Ripe Walleyes were collected in April 2010 from the flow regimes and (2) the effect of fine sediment on the ability of Oswegatchie River, New York, and artificially spawned in the Walleye eggs to adhere to rock substrate in a fluvial environment. laboratory. The fertilized eggs were allowed to water-harden be- Results from this study will contribute to our understanding of fore being preserved in distilled white vinegar (5% acetic acid). Walleye reproductive ecology and provide managers with guid- Eggs were dyed using food coloring (Farrell 2001) to differen- ance for substrate selection during spawning habitat restoration tiate between replicates of a treatment and aid in observing the Downloaded by [Department Of Fisheries] at 00:04 20 March 2013 projects. eggs during the experiment. The diameter of dyed and preserved eggs did not deviate from literature values of fertilized, water- hardened Walleye eggs (McElman and Balon 1979). Preserved METHODS eggs lost the adhesive properties of freshly fertilized eggs, and Substrate size and shape.—Walleye egg retention among we assumed that there would be no interaction between rock substrate size, shape, and water velocity was examined in ex- shape or size and egg adhesion. Water-hardened preserved and perimental treatments in an Armfeild S6 MKII 7.5-m water– unpreserved eggs were dropped through a column of water to sediment recirculating flume (Armfield Limited, Ringwood, determine the influence of preservation on egg density based on England). A 2 × 3 × 3 factorial treatment design was specified, the settling rate of the eggs. Preserved, dyed eggs had a signifi- with two levels of rock shape (angular rock and round rock), cantly slower settling rate (mean = 1.85 cm/s; SD = 0.17; n = three levels of rock size (#2 rock, #3 or #4 rock, and a 50:50 30) than water-hardened, unpreserved eggs (mean = 2.23 cm/s; mixture of #2 and #3/#4), and three levels of water velocity SD = 0.06; n = 30; t-test (Satterthwaite): t = –11.6, df = 37.061, (0.55 m/s, 0.75 m/s, and 0.95 m/s). Each treatment was repli- P < 0.001). However, preserved eggs remained demersal and cated four times. Rock was obtained from a local quarry, and the settling rate difference of 0.38 cm/s, due to preservation, angular rock treatments consisted of crushed, screened would not affect treatment-level comparisons. WALLEYE EGG RETENTION 331

Preserved Walleye eggs were added to the substrate (n = 1,350 eggs per treatment) under no flow conditions. A no flow condition was used to ensure that all eggs settled on the substrate, so that an actual measure of retention could be obtained. The flume and bed slope were set to zero for all trials. Eggs were not added to the substrate within 1.5 cm of the flume walls in order to minimize interaction with this surface. The density of eggs used for each trial (5,000 eggs/m2) was based on values observed in riverine systems (Gibson and Hughes 1977; Corbett and Powles 1986; Eckersley 1986). Once eggs were allowed to settle on the substrate, the flume was activated and set to the predetermined velocity for 5 min. Eggs displaced from the substrate during each trial were collected in a mesh trap and counted. The total number of eggs displaced from the substrate was subtracted from the total number of eggs originally added to the substrate. The number of eggs retained in the substrate was then converted to a proportion retained for each treatment replicate. The substrate FIGURE 1. Walleye eggs adhered to a clean rock. Photo: Derek P. Crane. bed was agitated to remove any remaining eggs after each trial. [Figure available in color online.] Statistical analyses were conducted in SAS 9.2 (SAS 2009) using a significance level of 0.05. Factorial analysis of variance (ANOVA) was used to estimate treatment means and test amounts of milt were used to ensure fertilization (Malison and for significant treatment main effects and interactions. Planned Held 1996; Moore 2003). contrasts were used to estimate the simple effects of rock shape Immediately upon fertilization a pipette was used to place 10 at each level of rock size. The Waller–Duncan K-ratio t-test eggs on the surface of wetted clean and fine-sediment-covered for multiple comparisons was used to identify significant differ- rocks. Eggs were placed in a straight line and spaced about 6– ences between water velocity treatments, rock size treatments, 10 mm apart (Figure 1). The pair of rocks were placed side and the six combinations of rock shape and size treatments. by side in the flume so that the line of eggs on each rock Substrate siltation.—Clean rock substrate and rock substrate was oriented perpendicular to the water flow (Figure 2). The covered with a thin layer of fine sediment (i.e., clay – fine sand) were used to investigate the influence of substrate siltation on the adhesive capacity of freshly fertilized Walleye eggs. The experiment was conducted in the same water–sediment recirculating flume that was used to investigate the influence of rock size and shape on egg retention. Rocks were obtained from Kent’s Creek, a Lake Ontario tributary used by spawning Walleyes. Rocks in the fine-sediment-covered treatment were unaltered. Rocks in the clean treatment were scrubbed to remove any fine sediment. Because scrubbing was necessary to remove fine sediment, periphyton was likely removed as well. Thirty-two paired trials of Downloaded by [Department Of Fisheries] at 00:04 20 March 2013 clean and fine-sediment-covered rocks were conducted and new rocks were used for each trial. Clean and fine-sediment-covered rocks were paired so that their combined widths (b-axes) did not exceed the width of the flume. Walleyes were obtained from the New York State Depart- ment of Environmental Conservation’s Oneida Fish Hatchery. Ripe Walleyes were collected throughout the spawning run, and egg retention trials were conducted between 6 April 2011 and 14 April 2011 to account for any possible variability in egg adhesiveness due to run timing. To account for variability in the adhesive capacity of embryos from different female–male crosses, 32 unique pairs of Walleyes were used to conduct the trials. However, females were used to create unique crosses for FIGURE 2. Diagram of the experimental setup in the hydraulic flume. Ten two trials and males were used to create unique crosses for Walleye eggs were spaced evenly on each rock type and oriented perpendicular two or three trials. Fish were artificially spawned, and copious to the water flow. 332 CRANE AND FARRELL

flume pump was then activated, and the vertical mean water TABLE 2. Analysis of variance table for the F-test of the main effects of rock velocity in the center of the channel was maintained at approx- shape (angular, round), rock size (#2, #3/#4, 50:50 mix), and water velocity ± = ± (0.55m/s, 0.75m/s, 0.95m/s) treatments and interactions between treatments. imately 0.65 m/s (mean SD 0.653 0.015 m/s). Water Asterisks denote significant treatment main effects or significant interactions velocity was based on previous descriptions of incubating habi- between treatments (α = 0.05). tat for Walleye embryos in fluvial environments (Pitlo 1989; Gillenwater et al. 2006; Chalupnicki et al. 2010). Water temper- Treatment main effect or ature in the flume (range = 6.5–13.0◦C; mean ± SD = 10.09 ± interaction df FP 1.59◦C) was controlled to remain within the range of natural Rock shape 1, 54 97.01 <0.0001* Walleye spawning temperatures, but temperatures reached 12◦C Rock size 2, 54 67.41 <0.0001* in four trials and 13◦C in one trial, which is warmer than what Shape × size interaction 2, 54 8.44 0.0006* is often reported for Walleye spawning (Scott and Crossman Water velocity 2, 54 6.93 0.0021* 1973; Corbett and Powles 1986; Pitlo 1989; Chalupnicki et al. Shape × water velocity 2, 54 0.15 0.8579 2010). Simple linear regression was conducted using R (Fox interaction et al. 2010; Rcmdr: R Commander, R package version 1.6-1) to Size × water velocity 4, 54 1.76 0.1496 test the assumption that variation in water velocity and temper- interaction ature between trials did not significantly affect egg retention. A Shape × size × water 4, 54 0.83 0.5148 significant relationship was not observed between water velocity velocity interaction and egg retention (R2 = 0.11, df = 30, P = 0.063) or water temperature and egg retention (R2 = 0.05, df = 30, P = 0.220). A flume slope of zero was used for all trials. The water in the flume rock was 4.7% (SE, 1.46%) greater than round #2 rock. Angular was held at the same volume for each trial and dechlorinated mixed rock had a mean retention rate 7.2% (SE, 1.46%) greater using aquarium dechlorinator and activated charcoal. than round mixed rock, and angular #3/#4 rock had a mean At the end of each 1-h trial, the number of eggs remaining on retention rate 12.9% (SE, 1.46%) greater than round #3/#4 rock. each rock was enumerated prior to turning off the flow. For 30 There was also a significant main effect of rock shape. Angular of the 32 trials, egg loss was tracked over time by counting the rock treatments retained a greater proportion of eggs than round number of eggs remaining on each rock every 5 min. We did not rock treatments (ANOVA F-Test; F = 97.01; df = 1, 54; P anticipate being able to track egg loss over time; however, during < 0.0001; Table 3). We also observed significant main effects the initial two trials it became apparent that this was possible of rock size (ANOVA F-Test: F = 67.41; df = 2, 54; P = and it was conducted during the remainder of the experiment. 0.0001; Table 2) and water velocity (ANOVA F-Test: F = 6.93; Statistical analyses were conducted using R (Fox et al. 2010; df = 2, 54; P = 0.0021; Table 2). Rock treatments of size-class Rcmdr: R Commander, R package version 1.6-1). A McNemar chi-square test with continuity correction was used to test for a difference in egg adhesion between the two rock treatments after1h(α = 0.05). A paired t-test could not be used because no eggs remained on any of the 32 replicates of the fine-sediment- covered rock treatment (i.e., no variance structure existed for this treatment). To examine if differences in egg retention existed between the rock treatments on a shorter time scale, we compared the proportion of eggs remaining on the fine-sediment-covered Downloaded by [Department Of Fisheries] at 00:04 20 March 2013 and clean rocks after 5 min in the flume using a paired t-test (α = 0.05). Statistical comparisons between treatments at additional time periods (i.e., after 5 min) were not conducted due to the lack of eggs remaining on the fine-sediment-covered rocks.

RESULTS Substrate Size and Shape Rock shape, size-class, water velocity, and the interaction between rock shape and rock size significantly influenced egg retention (Table 2). The significant interaction between rock shape and rock size (ANOVA F-Test: F = 8.44; df = 2, 54; < FIGURE 3. A significant interaction was observed between rock shape and P 0.0006; Table 2) indicated increased differences between size (ANOVA F-Test: F = 8.44; df = 2, 54; P < 0.0006). As rock size increased, angular and round rock egg retention rates with increasing rock the difference in Walleye egg retention rates (mean ± SE) between angular rock size-class (Figure 3). The mean egg retention rate of angular #2 and round rock also increased. WALLEYE EGG RETENTION 333

TABLE 3. Estimated mean percentages of Walleye eggs retained based on rock shape, rock size, and water velocity.

Mean percentage of Treatment n eggs retained (SD) P Rock shape (df = 1, 54; F = 97.01)a Angular 36 61.9 (5.3) z <0.0001 Round 36 53.6 (7.4) y Rock size (k = 100; df = 2, 69; F = 24.17)b #2 24 64.6 (4.5) z <0.0001 50:50 mix 24 55.2 (5.2) y #3/#4 24 53.4 (7.6) y Water velocity (k = 100; df = 2, 69; F = 1.53)c 0.55 m/s 24 59.8 (7.5) z 0.2245 0.75 m/s 24 56.0 (7.7) z 0.95 m/s 24 57.5 (7.6) z

aMeans within the rock shape treatment groups that are not followed by a common letter are significantly different (ANOVA F-test, α = 0.05). bMeans within the rock size treatment groups that are not followed by a common letter are significantly different (Waller–Duncan K-ratio t-test for multiple comparisons, α = 0.05). cMeans within the velocity treatment groups that are not followed by a common letter are significantly different (Waller–Duncan K-ratio t-test for multiple comparisons, α = 0.05).

#2 had significantly higher egg retention rates than mixed rock and #3/#4 rock treatments (Table 3). Although the ANOVA FIGURE 4. Mean ± SD (n = 30) proportion of Walleye eggs retained on F-test identified a significant main effect of water velocity on clean and fine-sediment-covered rocks after being exposed to water velocities egg retention, post hoc analysis could not identify which water of 0.65 m/s for 5 min in a hydraulic flume. velocity treatment means differed (Table 3). Angular #2 rock had the highest mean egg retention averaged across all water velocity χ2 = 20.0455, df = 1, P < 0.001). No replicates of fine- treatments (mean = 66.9%; SD = 3.67%; n = 12; Table 4), while sediment-covered rocks retained any eggs for the entire dura- round #3/#4 rock had the lowest mean retention rate averaged tion of a trial, while clean rocks had a mean ± SD retention of across all water velocity treatments (mean = 47.0%; SD = 35.9 ± 36.6%. A difference in egg retention between rock treat- 3.10%; n = 12; Table 4). ments was also observed on a shorter time scale. All eggs were immediately lost from fine-sediment-covered rocks in 25 of the Substrate Siltation 30 trials when eggs loss was tracked over time. Furthermore, Fine sediment decreased the ability of Walleye eggs to adhere clean rocks had a greater mean egg retention rate after 5 min in to rock substrate. Eggs remained on clean rocks in a greater num- the flume than fine-sediment-covered rocks (t = 7.1974, df = ber of trials than on fine-sediment-covered rocks (McNemar’s Downloaded by [Department Of Fisheries] at 00:04 20 March 2013 29, P < 0.001). After 5 min, clean rocks had a mean ± SD egg retention rate of 58.7 ± 41.3%, while fine-sediment-covered TABLE 4. Estimated mean percentages of Walleye eggs retained based on rocks had a mean ± SD egg retention rate of 2.3 ± 7.7% (Fig- rock shape and rock size. ure 4). After pooling data within each treatment, only 7 of 300 Rock Mean percentage of eggs remained on the fine-sediment-covered rocks after 5 min, shape Size-class n eggs retained (SD)a P while 176 of the 300 eggs were retained on the clean rocks. Eggs were observed remaining on fine-sediment-covered rocks < Angular #2 12 66.9 (3.7) a 0.0001 after 20 min in only one trial. 50:50 mix 12 58.8 (4.0) c #3/#4 12 59.9 (4.5) b, c Round #2 12 62.2 (4.2) b DISCUSSION 50:50 mix 12 51.6 (3.8) d We found that rock shape and size and the interaction between #3/#4 12 47.0 (3.1) e shape and size influenced Walleye egg retention in a fluvial environment, with smaller angular rock having the highest egg aMeans that are not followed by a common letter are significantly different (Waller– Duncan K-ratio t-test for multiple comparisons: k = 100; df = 5, 66; α = 0.05; F = retention rates. Substrate siltation also affected egg retention. 41.8). Fine sediment covering the substrate prevented Walleye eggs 334 CRANE AND FARRELL

from adhering to the rocks, leading to lower retention rates than points in the angular rock matrix, untreated adhesive eggs may on clean rock. bond better to angular rock than to round rock. Therefore, our The underlying mechanism(s) resulting in significant differ- estimates for differences in egg retention rates between angu- ences in egg retention rates between angular and round rocks and lar and round rock may be conservative compared to natural rock size-class treatments was not quantitatively investigated. conditions with adhesive eggs. However, near-bed flow conditions were probably the dominant The adhesiveness of Walleye eggs typically lasts for 1–5 h factors controlling Walleye egg retention. Investigating the eco- (until water-hardened) (Priegel 1970; Waltemyer 1976; Bozek logical implications of near-bed flow conditions on aquatic biota et al. 2011a). However, Raabe and Bozek (2012) documented has proven to be challenging for ecologists due to the complex eggs remaining adhesive for 15–24 h in Big Crooked Lake, nature of near-bed flow and the small scale that it operates on Wisconsin, and Humphrey et al. (2012) observed Walleye eggs (Allan and Castillo 2007). Davis and Barmuta (1989) proposed remaining adhered to rock for at least 48 h postfertilization. D. P. that shear velocity, height of roughness elements (projection of Crane and J. M. Farrell (unpublished data) observed that Wall- rock into the water column), relative roughness (height of rough- eye eggs would not readhere to rock in a laboratory after water ness element divided by water depth), and distance between hardening, but eggs already adhered to rock remained adhered roughness elements are the primary variables dictating near-bed for several days if not removed. Walleye eggs are easily dam- flow conditions. The smaller #2 rock was tightly packed and had aged prior to water hardening and suffer high mortality rates relatively small interstitial spaces between each roughness ele- early in the incubation period (Bozek et al. 2011b). Heidinger ment, which may have resulted in quasi-smooth flow conditions. et al. (1997) documented the highest mortality rate of laboratory Quasi-smooth flow conditions are characterized by flow skim- incubated Walleye eggs occurring in the first 6 h after fertiliza- ming across the crests and spaces between roughness elements, tion, while in a similar study Latif et al. (1999) observed the creating microeddies between each rock (Davis and Barmuta highest mortality of eggs between 3 and 5 d postfertilization. 1989). Higher egg retention rates in the #2 rock may be related Therefore, it is important that eggs remain undisturbed in pro- to the trapping of eggs in lower-velocity microeddies associated tective habitat during the early incubation period, which the ad- with interstitial spaces. Larger rock treatments may have expe- hesive property of Walleye eggs may play an important role in. rienced wake-interference flow due to having relatively large We did not quantify the shear stress required to remove eggs interstitial spaces and greater distances between rock crests. from each rock treatment; however, our results indicate that Wake-interference flow, which is characterized by turbulence fine sediment decreases the critical shear stress needed to scour and high local velocities between rocks (Davis and Barmuta eggs from the substrate. Humphrey et al. (2012) also reported 1989), may have dislodged eggs from interstitial spaces. a significant influence of substrate condition on Walleye egg Although macroinvertebrates possess morphological adap- adhesion. They observed that Walleye eggs placed on rock sub- tations to exist in lotic conditions, studies of bed character- strates remained adhered at significantly higher water velocities istics on macroinvertebrate drift may serve as useful compar- than did eggs placed on sand, providing further evidence that isons to egg retention. Holomuzki and Biggs (2003) examined substrate condition can affect the critical shear stress needed to the influence of rock size on macroinvertebrate drift during scour Walleye eggs from incubation habitats (Humphrey et al. high water velocity conditions. Similar to our findings, they ob- 2012). Fine sediment in this experiment likely functioned in served higher macroinvertebrate drift in cobble substrates than a similar manner to the clay or diatomaceous earth used in in smaller gravel and mixed gravel and cobble substrates and hatchery production to bind Walleye egg membranes, thus re- hypothesized that gravel and mixed gravel and cobble rock treat- moving their adhesive properties (Malison and Held 1996). The ments created quasi-smooth flow conditions that provided mi- current study was conducted under stable bed conditions, and al- Downloaded by [Department Of Fisheries] at 00:04 20 March 2013 crorefugia. Macroinvertebrate drift in elevated water velocities though eggs were scoured from rocks covered in fine sediment, was also higher in round substrates than in angular substrates shear stress was not high enough to scour the fine sediment (Holomuzki and Biggs 2003). Holomuzki and Biggs (2003) at- covering the rocks. This suggests that the hydraulic forces act- tributed this to lower erosion potential, higher patch stability, ing directly on the fine sediment covering the rocks were not and less hydraulic stress in angular rock substrate than in round strong enough to scour fine sediment but were strong enough to rock substrate. Angular substrate tends to have more rock to scour eggs and any fine sediment bound to the eggs. However, rock contact points than spherical rock (Boggs 1987) and thus the hydraulic forces were not strong enough to break the egg– has smaller interstitial spaces and lower interstitial flow rates rock bonds in a large proportion of eggs used in the clean rock than does round rock. treatment. The interstitial pore matrix created by angular rock may also We clearly demonstrated that fine sediment covering coarse increase the number of egg to rock contact points and contribute rock decreased the ability of eggs to adhere to the substrate; to the effectiveness of angular rock at retaining eggs. We used however, our experiment was conducted in a laboratory setting preserved nonadhesive eggs and assumed that there would be and did not account for the variety of hydrodynamic, phys- no interaction between substrate shape and egg adhesiveness. ical, and biological conditions present at Walleye spawning However, because of the higher number of egg to rock contact sites. The current study was conducted under moderate water WALLEYE EGG RETENTION 335

velocity conditions compared to the maximum velocities ob- to wear relatively quickly its edges will soften, thus decreasing served at Walleye incubation sites in natural settings (see Kerr the likelihood of injury to spawning fish (Kerr et al. 1997). Kerr et al. 1997; Bozek et al. 2011a for a review). It is likely that shear (1996) also suggested the use of round rock because it provides stress exceeds the adhesive capacity of eggs under some natural sufficient interstitial space for incubating eggs and is less vul- circumstances, regardless of substrate condition. Additionally, nerable to packing than angular rock. We did not examine the high discharge events at Walleye spawning sites may cause effects of packing on egg retention and this may be a legitimate bed instability, leading to substrate and egg scour. Kelder and concern in fluvial ecosystems. However, our findings suggest Farrell (2009) observed scour of very fine gravel, clay, organic that angular rock provides sufficient interstitial space for in- material, and Walleye eggs after a heavy rain event and ele- cubating eggs. Future research that focuses on examining the vated discharge in Kent’s Creek, New York. Factors such as sustainability of restoration projects using various rock types rock texture, shape, size, and periphyton may also influence egg and addresses concerns about angular rock settling and causing adhesion. Because removing the fine sediment on rocks dur- abrasion to spawning fish is important given the discrepancy in ing this experiment likely removed periphyton, we were unable retention rates between angular and round rocks used in restora- to account for this possible source of variation in egg adher- tion projects. ence. Periphyton growth and species can vary substantially in Many Walleye spawning habitat restoration projects oper- fluvial environments depending on grazing, temperature, hydro- ate under the assumption that cobble and boulder size-class logic conditions, light availability, and nutrient concentrations substrates provide the best incubation conditions. Kerr et al. (Rosemond 1994; Cushing and Allan 2001). It is likely that dif- (1997) recommend glacial till 5–38 cm in diameter for spawn- ferent types of periphyton (i.e., bacteria, filamentous algae, and ing bed creation in rivers, with gabion limestone and angular diatoms) and their degree of growth have variable effects on granitic rock as second and third choices. Newburg (1975) rec- the ability of eggs to adhere to spawning substrates. Stoll et al. ommended a mixture of 7.6–12.7-cm rock (10%), 12.7–17.8- (2010) observed that the initial adherence of Bream Abramis cm rock (50%), and 17.8–22.9-cm rock (40%) for the creation brama eggs to substrate was 3.7 times greater on clean rock than of artificial spawning shoals in lakes. There is little scientific on rock covered in periphyton. Several studies have indicated evidence that suggests that very coarse gravel to boulder-size clean, periphyton-free substrate may be important for the repro- rocks provide the best incubating habitat for Walleye eggs, and ductive success of fishes with adhesive eggs (Gafny et al. 1992; our research provides evidence that smaller substrates (coarse Probst et al. 2009; Stoll et al. 2010). The effects of periphyton on gravel) may retain eggs better than more commonly used larger Walleye reproduction have not been investigated, and research substrate (very coarse gravel and small cobble). examining the relationship between periphyton, egg retention, Although this study provides valuable insight into how sub- and egg survival would further our understanding of Walleye re- strate shape and size influence egg retention, the data were productive ecology. However, Walleye spawning substrates may collected in a controlled laboratory setting under a specific set have limited or no periphyton growth due to cold water tem- of conditions that do not encompass the variety of conditions peratures, winter scour, and increased stream discharge during present at Walleye spawning sites. Given the physical habitat the early spring hydroperiod. Additionally, substrates inundated variability of Walleye spawning sites, site-specific conditions during the spring freshet (i.e., in close temporal proximity to the must be first considered during restoration planning. Although spawning run) are unlikely to have periphyton growth (Gafny this study showed smaller angular rock to have the highest egg et al. 1992; Probst et al. 2009) retention rate, this material is not suitable for all circumstances. Smaller material would be vulnerable to scour and downstream Implications for Habitat Restoration and Future Research displacement at high-energy sites. Smaller rock may also be Downloaded by [Department Of Fisheries] at 00:04 20 March 2013 The desire to address habitat degradation and increase natu- more vulnerable to sedimentation because it has less interstitial ral production of the ecologically and economically important flow to remove fine sediments. Walleye has triggered numerous restoration projects (Geiling Although maintaining or increasing clean coarse substrate in et al. 1996; Kerr 1996; Kapuscinski et al. 2010). A primary fo- order to provide adequate interstitial space for incubating eggs cus of these projects has been the creation of spawning habitat. is often the objective of Walleye spawning habitat restoration, it Of 406 Walleye restoration projects examined by Kerr (1996), may also inadvertently benefit Walleye eggs by increasing their 318 focused on creation of spawning habitat through substrate ability to adhere to the substrate and remain in place during the additions. Our findings contradict the limited guidance on Wall- water-hardening process. Our flume experiment demonstrated eye spawning habitat restoration that is available to resource that fine sediment decreases the adhesive capacity of eggs. Al- managers and engineers. Kerr (1996) suggested that round rock though the importance of adhesion to Walleye egg survival in is the best material for spawning habitat because it is less abra- natural settings is unknown, habitat management and restoration sive. However, because Walleye spawn close to the water surface plans should include measures to reduce fine sediment deposi- (Ellis and Giles 1965; Priegel 1970) and do not create redds or tion on spawning substrates because (1) Walleye eggs have a nests, the abrasiveness of the substrate may not be an issue for lower survival rate on fine sediment than on coarse substrate spawning fish. Additionally, because limestone has the tendency (Johnson 1961; Corbett and Powles 1986) and (2) increased 336 CRANE AND FARRELL

egg adhesion rates are unlikely to have a negative effect on egg Davis, J. A., and L. A. Barmuta. 1989. An ecologically useful classification of survival. mean and near-bed flows in streams and rivers. Freshwater Biology 21:271– Although anecdotal information suggests that egg redistribu- 282. Dustin, D. L., and P. C. Jacobson. 2003. Evaluation of Walleye spawning habi- tion is a likely source of Walleye egg mortality (Johnson 1961; tat improvement projects in streams. Minnesota Department of Natural Re- Pitlo 1989; Newbury and Gaboury 1993; Roseman et al. 1996; sources, Investigational Report 502, St. Paul. Dustin and Jacobson 2003; Kelder and Farrell 2009; Bozek Eckersley, M. 1986. Assessment of Walleye spawning grounds in Hoople Creek. et al. 2011b), empirical evidence from research focused on Pages 31–32 in 1986 report of the St. Lawrence subcommittee to the Lake this topic is lacking. Studies that quantify the links between Ontario committee. Great Lakes Fishery Commission, Ann Arbor, Michigan. Ellis, D. V., and M. A. Giles. 1965. The spawning behavior of the Walleye, Sti- Walleye spawning habitat characteristics, egg redistribution, zostedion vitreum (Mitchill). Transactions of the American Fisheries Society and egg mortality would increase our understanding of Wall- 94:358–362. eye population dynamics, allow for more effective management Farrell, J. M. 2001. Reproductive success of sympatric Northern Pike and of spawning habitat, and provide guidance so that spawning Muskellunge in an upper St. Lawrence River bay. Transactions of the Amer- habitat restoration projects maximize egg retention. Identifying ican Fisheries Society 130:796–808. Fox, J., L. Andronic, M. Ash, T. Boye, S. Calza, A. Chang, P. Grosjean, R. artificial spawning habitat structures that trap and retain eggs, Heiberger, G. J. Kerns, R. Lancelot, M. Lesnoff, U. Ligges, S. Messad, M. while also providing quality incubating habitat, would address Maechler, R. Muenchen, D. Murdoch, E. Neuwirth, D. Putler, B. Ripley, M. some of the current knowledge gaps and inform managers dur- Ristic, and P. Wolf. 2010. Rcmdr: R commander, R package version 1.6-1. ing project design. Available: CRAN.Rproject.org/package=Rcmdr. (March 2012). Gafny, S., A. Gasith, and M. Goren. 1992. Effect of water level fluctuation on shore spawning of Mirogrex terraesanctae (Steinitz), (Cyprinidae) in Lake Kinneret, Israel. Journal of Fish Biology 41:863–871. ACKNOWLEDGMENTS Geiling, W. D., J. R. M. Kelso, and E. Iwachewski. 1996. Benefits from incre- This project was funded by the Fish Enhancement, Mitiga- mental additions to Walleye spawning habitat in the Current River, with reference to habitat modification as a Walleye management tool in On- tion, and Research Fund, administered by the U.S. Fish and tario. Canadian Journal of Fisheries and Aquatic Sciences 53(Supplement 1): Wildlife Service, Cortland, New York. The authors would like 79–87. to thank Steve LaPan, Mark Babenzien, and Dave Gordon from Gibson, R. J., and C. E. Hughes. 1977. A Walleye stream spawning study on the New York State Department of Environmental Conserva- Hamilton’s Creek, 1969. Manitoba Department of Renewable Resources and tion for aiding in fish collection for the study. Thank you to Transportation Services, Technical Report 71-31, Winnipeg. Gillenwater, D., T. Granata, and U. Zika. 2006. GIS-based modeling of spawning Theodore Endreny for providing use of State University of habitat suitability for Walleye in the Sandusky River, Ohio, and implications New York–College of Environmental Science and Forestry’s for dam removal and river restoration. Ecological Engineering 28:311–323. hydraulic flume and technical support. We would also like to Heidinger, R. C., R. C. Brooks, D. Leitner, and I. Soderstrom. 1997. Prediction thank Stephen Stehman for his substantial help with statistical of Walleye egg and embryo survival at two stages of development. Progressive analysis, Kevin Kapuscinski and two anonymous readers for re- Fish-Culturist 59:64–67. Holomuzki, J. R., and B. J. F. Biggs. 2003. Sediment texture mediates high-flow view of a previous version of this manuscript, and the students effects on lotic macroinvertebrates. Journal of the North American Bentho- and staff of the Thousand Islands Biological Station for assisting logical Society 22:542–553. with data collection. Humphrey, S., Y. Zhao, and D. Higgs. 2012. The effects of water currents on Walleye (Sander vitreus) eggs and larvae and implications for the early survival of Walleye in Lake Erie. Canadian Journal of Fisheries and Aquatic Sciences 69:1959–1967. REFERENCES Ivan, L. N., E. S. Rutherford, C. Riseng, and J. A. Tyler. 2010. Density, produc- Allan, J. D., and M. M. Castillo. 2007. Stream ecology: structure and function tion, and survival of Walleye (Sander vitreus) eggs in the Muskegon River,

Downloaded by [Department Of Fisheries] at 00:04 20 March 2013 of running waters, 2nd edition. Springer, Dordrecht, The Netherlands. Michigan. Journal of Great Lakes Research 36:328–337. Boggs, S., Jr. 1987. Principles of sedimentology and stratigraphy. Macmillan, Johnson, F. H. 1961. Walleye egg survival during incubation on several types of New York. bottom in Lake Winnibigoshish, Minnesota, and connecting waters. Transac- Bozek, M. A., D. A. Baccante, and N. P. Lester. 2011a. Walleye and Sauger tions of the American Fisheries Society 90:312–322. life history. Pages 233–301 in B. A. Barton, editor. Biology, management, Jones, M. L., J. K. Netto, J. D. Stockwell, and J. B. Mion. 2003. Does the value of and culture of Walleye and Sauger. American Fisheries Society, Bethesda, newly accessible spawning habitat for Walleye (Stizostedion vitreum) depend Maryland. on its location relative to nursery habitats? Canadian Journal of Fisheries and Bozek, M. A., T. J. Haxton, and J. K. Raabe. 2011b. Walleye and Sauger habitat. Aquatic Sciences 60:1527–1538. Pages 133–197 in B. A. Barton, editor. Biology, management, and culture of Kapuscinski, K. L., T. G. Zorn, P. J. Schneeberger, R. P. O’Neal, and B. T. Walleye and Sauger. American Fisheries Society, Bethesda, Maryland. Eggold. 2010. The status of Lake Michigan Walleye stocks. Great Lakes Chalupnicki, M. A., J. H. Johnson, J. E. McKenna Jr., and D. E. Dittman. Fishery Commission Technical Report 69:15–70. 2010. Habitat selection and spawning success of Walleyes in a tributary to Kelder, B. F., and J. M. Farrell. 2009. A spatially explicit model to predict Owasco Lake, New York. North American Journal of Fisheries Management Walleye spawning in an eastern Lake Ontario tributary. North American 30:170–178. Journal of Fisheries Management 29:1686–1697. Corbett, B. W., and P. M. Powles. 1986. Spawning and larva drift of sym- Kerr, S. J. 1996. Walleye habitat creation and enhancement: an overview of patric Walleyes and White Suckers in an Ontario stream. Transactions of the selected projects. Ontario Ministry of Natural Resources, Percid Community American Fisheries Society 115:41–46. Synthesis, Peterborough. Cushing, C. E., and J. D. Allan. 2001. Streams: their ecology and life. Academic Kerr, S. J., B. W. Corbett, N. J. Hutchinson, D. Kinsman, J. H. Leach, D. Press, San Diego, California. Puddister, L. Stanﬁeld, and N. Ward. 1997. Walleye habitat: a synthesis WALLEYE EGG RETENTION 337

of current knowledge with guidelines for conservation. Ontario Ministry of Raabe, J. K., and M. A. Bozek. 2012. Quantity, structure, and habitat selection Natural Resources, Percid Community Synthesis, Walleye Habitat Working of natural spawning reefs by Walleyes in a north temperate lake: a multi- Group, Peterborough. scale analysis. Transactions of the American Fisheries Society 141:1097– Latif, M. A., R. A. Bodaly, T. A. Johnston, and R. J. P. Fudge. 1999. Critical 1108. stage in developing Walleye eggs. North American Journal of Aquaculture Riehl, R. 1996. The ecological significance of the egg envelope in teleosts with 61:34–37. special reference to limnic species. Limnologica 26:183–189. Malison, J. A., and J. A. Held. 1996. Reproduction and spawning in Walleye Roseman, E., P. Kocovsky, and C. Vandergoot, editors. 2010. Status of Walleye (Stizostedion vitreum). Journal of Applied Ichthyology 12:153–156. in the Great Lakes: proceedings of the 2006 symposium. Great Lakes Fishery McElman, J. F., and E. K. Balon. 1979. Early ontogeny of Walleye, Stizoste- Commission Technical Report 69. dion vitreum, with steps of saltatory development. Environmental Biology of Roseman, E., W. W. Taylor, D. B. Hayes, R. C. Haas, R. L. Knight, and K. O. Fishes 4:309–348. Paxton. 1996. Walleye egg deposition and survival on reefs in western Lake Moore, A. A. 2003. Manipulation of fertilization procedures to improve hatchery Erie (USA). Annales Zoologici Fennici 33:341–351. Walleye egg fertility and survival. North American Journal of Aquaculture Roseman, E., W. W. Taylor, D. B. Hayes, R. L. Knight, and R. C. Haas. 2001. 65:56–59. Removal of Walleye eggs from reefs in western Lake Erie by a catastrophic Newburg, H. J. 1975. Evaluation of an improved Walleye, Stizostedion vitreum, storm. Transactions of the American Fisheries Society 130:341–346. spawning shoal with criteria for design and placement. Minnesota Department Rosemond, A. D. 1994. Multiple factors limit seasonal variation in periphyton of Natural Resources Section of Fisheries Investigational Report 340. in a forest stream. Journal of the North American Benthological Society Newbury, R., and M. Gaboury. 1993. Exploration and rehabilitation of hydraulic 13:333–344. habitats in streams using principles of fluvial behaviour. Freshwater Biology SAS (Statistical Analysis Systems). 2009. SAS/STATR 9.2 user’s guide, 2nd 29:195–210. edition. SAS Institute, Cary, North Carolina. Pitlo, J., Jr. 1989. Walleye spawning habitat in Pool 13 of the upper Mississippi Scott, W. B., and E. J. Crossman. 1973. Freshwater fishes of Canada. Fisheries River. North American Journal of Fisheries Management 9:303–308. Research Board of Canada Bulletin 184. Potyondy, J., and K. Bunte. 2002. Sampling with the US SAH-97 hand-held Stoll, S., W. N. Probst, R. Eckmann, and P. Fischer. 2010. A mesocosm exper- particle size analyzer. Federal Interagency Sedimentation Project, Water- iment investigating the effects of substratum quality and wave exposure on ways Experiment Station, Vicksburg, Mississippi. Available: water.usgs.gov/ the survival of fish eggs. Aquatic Sciences 72:509–517. fisp/docs/Instructions US SAH-97 040412.pdf. (December 2010). Udden, J. A. 1914. Mechanical composition of clastic sediments. Bulletin of Priegel, G. R. 1970. Reproduction and early life history of the Walleye in the Geological Society of America 25:655–744. the Lake Winnebago region. Wisconsin Department of Natural Resources Waltemyer, D. L. 1976. Tannin as an agent to eliminate adhesiveness of Walleye Technical Bulletin 45. eggs during artificial propagation. Transactions of the American Fisheries Probst, W. N., S. Stoll, L. Peters, P. Fischer, and R. Eckmann. 2009. Lake water Society 105:731–736. level increase during spring affects the breeding success of bream Abramis Wentworth, C. K. 1922. A scale of grade and class terms for clastic sediments. brama (L.). Hydrobiologia 632:211–224. Journal of Geology 30:377–392. Downloaded by [Department Of Fisheries] at 00:04 20 March 2013 This article was downloaded by: [Department Of Fisheries] On: 20 March 2013, At: 00:05 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 Development of an Asian Carp Size Structure Index and Application through Demonstration Quinton E. Phelps a & David W. Willis b a Missouri Department of Conservation, 3815 East Jackson Boulevard, Jackson, Missouri, 63755, USA b Department of Natural Resource Management, South Dakota State University, Room 138, Northern Plains Biostress Building, Brookings, South Dakota, 57007, USA Version of record first published: 06 Mar 2013.

To cite this article: Quinton E. Phelps & David W. Willis (2013): Development of an Asian Carp Size Structure Index and Application through Demonstration, North American Journal of Fisheries Management, 33:2, 338-343 To link to this article: http://dx.doi.org/10.1080/02755947.2012.760506

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 33:338–343, 2013 C American Fisheries Society 2013 ISSN: 0275-5947 print / 1548-8675 online DOI: 10.1080/02755947.2012.760506

MANAGEMENT BRIEF

Development of an Asian Carp Size Structure Index and Application through Demonstration

Quinton E. Phelps* Missouri Department of Conservation, 3815 East Jackson Boulevard, Jackson, Missouri 63755, USA David W. Willis Department of Natural Resource Management, South Dakota State University, Room 138, Northern Plains Biostress Building, Brookings, South Dakota 57007, USA

range and now have established populations in many U.S. Abstract freshwater lentic and lotic environments. Asian carps persist Asian carp (Cyprinidae) are among the most invasive fishes in in these locations because of their rapid growth, dispersal North America, but assessment tools for evaluating stock structure capabilities, high reproductive potential, absence of natural have not been established. Thus, we used data from national and international published reports or manuscripts to develop stan- predators, and broad environmental tolerance (Ehrlich 1984; dard length categories for four species of Asian carp. Due to the Fuller et al. 1999). Because of these characteristics, Asian carps similarities in maximum reported size for Bighead Carp Hypoph- often achieve high abundance or biomass, can exert deleterious thalmichthys nobilis and Grass Carp Ctenopharyngodon idella,we effects on aquatic ecosystems (Kolar et al. 2005; Irons et al. propose the following standardized length categories for those two 2007), and are thus costly to control (Pimentel et al. 2000). species: stock = 30 cm, quality = 54 cm, preferred = 68 cm, memorable = 89 cm, and trophy = 111 cm. Proposed Black Carp My- To evaluate any fishery, including the Asian carps noted lopharyngodon piceus length categories are stock = 40 cm, qual- above, biologists must analyze stock structure and dynamics ity = 72 cm, preferred = 90 cm, memorable = 118 cm, and trophy = of the species of interest. The three dynamic rate functions 148 cm. Finally, we propose the following length categories for Sil- (recruitment, growth, and mortality) interact to determine = = = ver Carp H. molitrix:stock 25 cm, quality 45 cm, preferred stock structure (i.e., size structure, age structure). Of the many 56 cm, memorable = 74 cm, and trophy = 93 cm. We then used existing data collected in the Illinois River and Mississippi River fisheries assessment tools available, size structure is one of from 2003 to 2011 to evaluate Silver Carp proportional size distri- the most common, least expensive, and informative tools bution (PSD). Incremental PSD provided a temporal index of Silver that provide insight into population characteristics. However, Carp size structure in the Illinois River and strong cohorts could length-frequency distributions can be problematic to compare be followed through the incremental PSD values over time. Tra- among populations or temporally within a population. Thus, size ditional PSD provided a quantifiable, comparative index of Silver Downloaded by [Department Of Fisheries] at 00:05 20 March 2013 Carp size structure among two locations in the Mississippi River structure indices were developed to quantify and index length- and one location in the Illinois River. We believe this index provides frequency distributions (Weithman 1978; Gabelhouse 1984). an additional tool for fisheries managers to evaluate size structure Initial development of proportional size distribution (PSD; of Asian carp and help assess overall population status. formerly proportional stock density [Guy et al. 2007]) used a 2- cell (i.e., two length categories) model (Anderson 1976), which Bighead Carp Hypophthalmichthys nobilis, Black Carp My- was later converted to a 3-cell model (Wege and Anderson lopharyngodon piceus, Grass Carp Ctenopharyngodon idella, 1978), and then to a five-cell model (Gabelhouse 1984). and Silver Carp H. molitrix (Asian carps, family Cyprinidae) Despite the apparent benefits of using a size structure index are nonindigenous fishes that were introduced to the USA in to help assess the status of Asian carp populations, we are un- the 1960s and 1970s from Asia for human consumption and bi- aware of standardized length categories that have been proposed ological control (Fuller et al. 1999; Kolar et al. 2005; Schofield for Asian carps, other than for Common Carp Cyprinus carpio. et al. 2005). Since that time, Asian carps have expanded their Given the importance of standardization in fish sampling and

*Corresponding author: [email protected] Received September 19, 2012; accepted December 12, 2012

338 MANAGEMENT BRIEF 339

population analysis (Bonar et al. 2009), our primary objective RKM 128–254) and two reaches of the upper Mississippi River was to develop standard length categories for Bighead Carp, (Pool 26, Alton, Illinois, RKM 325–389 and Open River, Cape Black Carp, Grass Carp, and Silver Carp using the methods de- Girardeau, Missouri, RKM 47–129). Annual data (2003–2011) scribed by Gabelhouse (1984). To assess the utility of this index, from the La Grange reach of the Illinois River were used to cal- we then used long-term, resource-monitoring data previously culate incremental PSD. To calculate traditional PSD, we pooled collected in the upper Mississippi River basin to evaluate spa- data across years from each location to make comparisons. We tial and temporal trends of Silver Carp size structure using PSD. also employed 95% confidence intervals for each PSD value using the method developed by Gustafson (1988). METHODS Standard length categories.—To develop Asian carp stan- RESULTS dard length categories, we performed literature searches through The maximum lengths for each Asian carp species found in the university library (e.g., using Google Scholar, Ebsco, the literature were Bighead Carp = 150 cm (Kolar et al. 2005), Blackwell) for published reports and manuscripts for the longest Black Carp = 200 cm (Schofield et al. 2005), Grass Carp = (total length) specimen captured to date. We also spoke to the 150 cm (Schofield et al. 2005), and Silver Carp = 126 cm leading authorities on Asian carps in North America and their (Kamilov and Salikhov 1996). Because of the similarities colleagues overseas to assist in our literature search. As a fi- in reported maximum length for Bighead Carp and Grass nal check, we investigated the Long-Term Resource Monitoring Carp, we propose the same length categories for these two online database (United States Geological Survey, Upper Mid- species: stock = 30 cm, quality = 54 cm, preferred = 68 cm, west Environmental Science Center) to ensure that the values memorable = 89 cm, and trophy = 111 cm (Table 1). Proposed that we garnered from the literature were not exceeded by fish Black Carp standard length categories are stock = 40 cm, captured in the Mississippi and Illinois rivers. We then used quality = 72 cm, preferred = 90 cm, memorable = 118 cm, and an approach recommended by Gabelhouse (1984) to calculate trophy = 148 cm (Table 1). Finally, we propose the following standard length categories for each species. Specifically, mini- length categories for Silver Carp: stock = 25 cm, quality = mum lengths for stock, quality, preferred, memorable, and tro- 45 cm, preferred = 56 cm, memorable = 74 cm, and trophy = phy were calculated at 20, 36, 45, 59, and 74% of maximum 93 cm (37 in) (Table 1). observed lengths for each species (see Gabelhouse 1984 for specific methodology). TABLE 1. Recommended total length categories for four species of Asian Proportional size distribution.—Gabelhouse (1984) de- carp (Bighead Carp, Black Carp, Grass Carp, and Silver Carp) using the maxi- scribed two methods for using proportional size distribution: mum reported total length (Kamilov and Salikhov 1996; Kolar 2005; Schofield incremental and traditional. He recommended incremental pro- et al. 2005) and calculated using percentages with the five-cell model proposed portional size distribution to evaluate size structure using long- by Gabelhouse (1984). term data on an individual body of water over time. Thus, in- Length % Record Proposed cremental PSD can indicate variability in year-class strength Species category length standard (cm) and identify the relative influence of year-classes on size structure. Traditional PSD is recommended for comparisons between Bighead Carp Stock 20 30 aquatic systems by dampening the effects of year-class strength Quality 36 54 variability. Incremental and traditional PSD are calculated using Preferred 45 68 the following equations: Memorable 59 89

Downloaded by [Department Of Fisheries] at 00:05 20 March 2013 Trophy 74 111 Incremental PSD = (# of Fish in Specified Length Range) Black Carp Stock 20 40 /(# of Fish > Stock Length) × 100 Quality 36 72 Preferred 45 90 Traditional PSD = (# of Fish > Specified Length) Memorable 59 118 / > × (# of Fish Stock Length) 100 Trophy 74 148 Grass Carp Stock 20 30 Both incremental and traditional PSD were calculated as Quality 36 54 an example using the above equations with standard length Preferred 45 68 categories for Silver Carp (Table 2; Neumann et al 2012). Memorable 59 89 We used Silver Carp data from the Upper Mississippi River Trophy 74 111 Restoration–Environmental Management Program–Long-Term Silver Carp Stock 20 25 Resource Monitoring (UMRR-EMP-LTRM) collected by stan- Quality 36 45 dardized electrofishing from June 15 through October 31 from Preferred 45 56 2003 to 2011 (see Gutreuter et al. 1995 for a full description). Memorable 59 74 Specifically, we used Silver Carp total length data collected in a Trophy 74 93 reach of the Illinois River (La Grange Reach, Havana, Illinois, 340 PHELPS AND WILLIS

160 160 160 2003 2004 2005 140 N = 24 140 N = 377 140 N = 476 120 PSD = 100 120 PSD = 8 (+3) 120 PSD = 60 (+ 6) PSD S-Q = 0 PSD S-Q = 92 (+ 4) PSD S-Q = 40 (+ 5) 100 PSD Q-P = 13 (+19) 100 PSD Q-P = 1 (+ 1) 100 PSD Q-P = 56 (+ 5) 80 PSD P-M = 84 (+ 20) 80 PSD P-M = 7 (+ 3) 80 PSD P-M = 3 (+ 2) PSD M-T = 4 (+ 13) PSD M-T = 0 PSD M-T = 2 (+ 1) 60 60 60

Frequency 40 40 40

20 20 20

0 0 0 300400500600700800900 300 400 500 600 700 800 900 300 400 500 600 700 800 900 Total Length (mm) Total Length (mm) Total Length (mm)

160 160 160 2006 2007 2008 140 N = 258 140 N = 232 140 N = 150 120 PSD = 99 (+ 7) 120 PSD = 89 (+ 10) 120 PSD = 89 (+ 8) PSD S-Q = 1 (+ 1) PSD S-Q = 11 (+ 4) PSD S-Q = 26 (+ 7) 100 PSD Q-P = 71 (+ 6) 100 PSD Q-P = 19 (+ 5) 100 PSD Q-P = 19 (+ 7) 80 PSD P-M = 28 (+ 6) 80 PSD P-M = 66 (+ 7) 80 PSD P-M = 53 (+ 9) PSD M-T = 0 PSD M-T = 4 (+ 3) PSD M-T = 2 (+ 3) 60 60 60

Frequency 40 40 40

20 20 20

0 0 0 300 400 500 600 700 800 900 300 400 500 600 700 800 900 300 400 500 600 700 800 900 Total Length (mm) Total Length (mm) Total Length (mm)

160 160 160 2009 2010 2011 140 N = 1186 140 N = 525 140 N = 740 120 PSD = 7 (+ 1) 120 PSD = 97 (+ 5) 120 PSD=90(+ 8) PSD S-Q = 93 (+ 2) PSD S-Q = 40 (+ 4) PSD S-Q=9(+ 2) 100 PSD Q-P = 2 (+ 1) 100 PSD Q-P = 56 (+ 4) 100 PSD Q-P=86(+ 3) 80 PSD P-M = 5 (+ 1) 80 PSD P-M = 4 (+ 2) 80 PSD P-M=5(+ 2) PSD M-T = 0 PSD M-T = 0 PSD M-T=0 60 60 60 Frequency 40 40 40

20 20 20

0 0 0

Downloaded by [Department Of Fisheries] at 00:05 20 March 2013 300 400 500 600 700 800 900 300 400 500 600 700 800 900 300 400 500 600 700 800 900 Total Length (mm) Total Length (mm) Total Length (mm)

FIGURE 1. Silver Carp length-frequency distribution and incremental proportional size distribution (PSD) indices (± 95% conﬁdence intervals; Gustafson 1988) collected during 2003–2011 from the Illinois River (La Grange Reach, Havana, Illinois, RKM 128–254). All data were collected by the Upper MississippiRiver Restoration–Environmental Management Program–Long-Term Resource Monitoring (UMRR-EMP-LTRM) program using standardized electroﬁshing from 2003 to 2011 (see Gutreuter et al. 1995 for a full description). Abbreviations are as follows: S = stock, Q = quality, P = preferred, M = memorable, and T = trophy.

Incremental PSD provided a quantitative, temporal time- they pass through the incremental PSD values over time. To series representation of the Silver Carp size structure in the demonstrate the use of traditional PSD, we used data collected Illinois River (Figure 1). These indices allowed for following from the Illinois River and two reaches of the upper Mississippi cohorts over time (See PSD S-Q, Q-P over time in Figure 1; note River. Traditional PSD provided an index of Silver Carp size increases and decreases over time). For example, strong cohorts structure among two locations on the Mississippi River and the entered the population length frequency in 2004 and 2009. They Illinois River (Figure 2). Based on the use of similar sampling are visible in subsequent length frequencies as they grow, and methods and traditional PSD calculations, the Open River reach MANAGEMENT BRIEF 341

70 Pool 26, Mississippi River 60 N=845 50 PSD = 55 (+ 3) 40 PSD-P = 18 (+ 3)

30 PSD-M = 3 (+ 1) PSD-T = 0

Frequency 20

0 300 400 500 600 700 800 900 Total Length (mm)

20 Open River Reach, Mississippi River N=207 15 PSD = 66 (+ 7) PSD-P = 59 (+ 7) 10 PSD-M = 33 (+ 7) PSD-T = 1 (+ 2) Frequency 5

0 300 400 500 600 700 800 900 Total Length (mm)

250 LaGrange Reach, Illinois River 200 N=3968 PSD = 50 (+ 2) Downloaded by [Department Of Fisheries] at 00:05 20 March 2013 150 PSD-P = 13 (+ 1) PSD-M = 1 (+ 0) 100 PSD-T = 0 Frequency 50

0 300 400 500 600 700 800 900 Total Length (mm)

FIGURE 2. Silver Carp length-frequency distribution and traditional proportional size distribution (PSD) indices (± 95% conﬁdence intervals; Gustafson 1988) pooled from 2003 to 2011 and collected in the Illinois River (La Grange Reach, Havana, Illinois, RKM 128–254) and two reaches of the upper Mississippi River (Pool 26, Alton, Illinois, RKM 325–389; Open River, Cape Girardeau, Missouri, RKM 47–129). All data were collected by the Upper Mississippi River Restoration–Environmental Management Program–Long-Term Resource Monitoring (UMRR-EMP-LTRM) program using standardized electroﬁshing from 2003 to 2011 (see Gutreuter et al. 1995 for a full description). Abbreviations are as follows: S = stock, Q = quality, P = preferred, M = memorable, and T = trophy. 342 PHELPS AND WILLIS

of the upper Mississippi River has a size structure dominated Environmental Sciences Center. We also thank the Illinois River by large Silver Carp (PSD = 66 ± 7[± 95% confidence inter- Biological Field Station, the National Great Rivers Research vals]; PSD-P = 59 ± 7; PSD-M = 33 ± 7), the Illinois River and Education Center, and the Open Rivers and Wetlands Field has a size structure dominated by small Silver Carp (PSD = Station for data collection. 50 ± 2; PSD-P = 13 ± 2; PSD-M = 1 ± 0), while Pool 26 of the Upper Mississippi River exhibited intermediate size = ± = ± = ± REFERENCES structure (PSD 55 3; PSD-P 18 3; PSD-M 3 Anderson, R. O. 1976. Management of small warm water impoundments. Fish- 1). eries 1(6):5–7, 26–28. Bister, T. J., D. W. Willis, M. L. Brown, S. M. Jordan, R. M. Neumann, M. C. Quist, and C. S. Guy. 2000. Proposed standard weight (Ws) equa- DISCUSSION tions and standard length categories for 18 warmwater nongame and river- We developed standardized length categories that allow bi- ine fish species. North American Journal of Fisheries Management 20: ologists to employ a size structure index (i.e., proportional 570–574. size distribution) for four species of Asian carps. Size struc- Bonar, S. A., W. A. Hubert, and D. W. Willis, editors. 2009. Standard methods for sampling North American freshwater fishes. American Fisheries Society, ture indices were created to evaluate sport fish populations Bethesda, Maryland. but more recently have been applied to nongame fishes, river- Brown, M. L., and B. R. Murphy. 1993. Management evaluation of body con- ine fishes, and species of special concern (e.g., Brown and dition and population size structure for Paddlefish: a unique case. Prairie Murphy 1993; Quist et al. 1998; Bister et al. 2000; Shuman et al. Naturalist 25:93–108. 2006). Our study broadens the utililty of PSD to include Asian Ehrlich, P. R. 1984. Which animal will invade? Pages 79–95 in H. A. Mooney and J. A. Drake, editors. Ecology of biological invasions of North America carps. and Hawaii. Springer, New York. Proportional size distribution may at times provide insights Fuller, P. L., L. G. Nico, and J. D. Williams. 1999. Nonindigenous fishes intro- to population-level parameters such as recruitment, growth, and duced into inland waters of the United States. American Fisheries Society, mortality. For instance, fish populations exhibiting moderate Special Publication 27, Bethesda, Maryland. PSD values likely have moderate values for the dynamic rate Gabelhouse, D. W., Jr. 1984. A length-categorization system to assess fish stocks. North American Journal of Fisheries Management 4: functions (i.e., recruitment, growth, and mortality) (Willis et al. 273–285. 1993). Exceptions can occur, such as for newly expanding pop- Gustafson, K. A. 1988. Approximating confidence intervals for indices of fish ulations, and thus caution in interpretation is recommended population size structure. North American Journal of Fisheries Management (Willis et al. 1993). In contrast, fish populations exhibiting ex- 8:139–141. cessively low or high PSD values or temporally variable PSD Gutreuter, S., R. Burkhardt, and K. Lubinski. 1995. Long term resource monitoring program procedures: fish monitoring. National Biological Service, values may be indicative of population disruption (e.g., high Environmental Management Technical Center, Program Report 95-P002-1, commercial or recreational harvest, lack of recruitment, erratic Onalaska, Wisconsin. recruitment among years, slow or fast growth). Although PSD Guy, C. S., R. M. Neumann, D. W. Willis, and R. O. Anderson. 2007. Propor- indices will be useful for indexing size structure of Asian carps, tional size distribution (PSD): a further refinement of population size structure limitations likely exist. For example, PSD values typically vary index terminology. Fisheries 32:348. Irons, K. S., G. G. Sass, M. A. McClelland, and J. D. Stafford. 2007. Re- by sampling season for many fish species (Willis et al. 1993) duced condition factor of two native fish species coincident with invasion and such variability should be assessed for Asian carps. Further- of non-native Asian carps in the Illinois River, U.S.A.: is this evidence for more, in most cases PSD will not provide the only assessment competition and reduced fitness? Journal of Fish Biology 71(Supplement D): tool a fishery manager needs to evaluate Asian carps but can be 258–273. one of a suite of tools. We also suggest that developing PSD Kamilov, B. G., and T. V. Salikhov. 1996. Spawning and reproductive potential of the Silver Carp Hypophthalmichthys molitrix from the Syr Darya River. Downloaded by [Department Of Fisheries] at 00:05 20 March 2013 values for nongame or invasive species should not be actively Journal of Ichthyology 36(8):600–606. pursued without understanding the information provided by the Kolar, C. S., D. C. Chapman, W. R. Courtenay Jr., C. M. Housel, J. D. Williams, index and how it can be used to assist in management decisions. and D. P. Jennings. 2005. Asian carps of the genus Hypophthalmichthys Thus, we suggest that fishery managers develop a thorough un- (Pisces, Cyprinidae): a biological synopsis and environmental risk assess- derstanding of the strengths and weaknesses of the index, use ment. U.S. Fish and Wildlife Service, Washington, D.C. Neumann, R. M., C. S. Guy, and D. W. Willis. 2012. Length, weight, and an accepted standardized sampling approach, and couple PSD associated indices. Pages 637–676 in A. V. Zale, D. L. Parrish, and T. M. indices with other typical fish population indices (e.g., catch per Sutton, editors. Fisheries techniques, 3rd edition. American Fisheries Society, unit effort, condition, and growth) to provide a more complete Bethesda, Maryland. assessment of a population. Pimentel, D., L. Lach, R. Zuniga, and D. Morrison. 2000. Environmental and economic costs of nonindigenious species in the United States. BioScience 50:53–65. ACKNOWLEDGMENTS Quist, M. C., C. S. Guy, and P. J. Braaten. 1998. Standard weight (Ws) equation This study was funded by the U.S. Army Corps of Engineers’ and length categories for Shovelnose Sturgeon. North American Journal of Fisheries Management 18:992–997. Upper Mississippi River Restoration–Environmental Manage- Schofield, P. J., J. D. Williams, L. G. Nico, P. Fuller, and M. R. Thomas. ment Program’s Long-Term Resource Monitoring component 2005. Foreign nonindigenous carps and minnows (Cyprinidae) in the implemented by the U.S. Geological Survey, Upper Midwest United States—a guide to their identification, distribution, and biology. MANAGEMENT BRIEF 343

U.S. Geological Survey, Scientiﬁc Investigations Report 2005-5041, Reston, American Fisheries Society, North Central Division, Special Publication 5, Virgina. Bethesda, Maryland. Shuman, D. A., D. W. Willis, and S. C. Krentz. 2006. Application of a length- Weithman, A. S. 1978. A method of evaluating ﬁshing quality: develop- categorization system for Pallid Sturgeon (Scaphirhynchus albus). Journal of ment, testing, and application. Doctoral dissertation. University of Missouri, Freshwater Ecology 21:71–76. Columbia. Wege, G. J., and R. O. Anderson. 1978. Relative weight (Wr):anewindexof Willis, D. W., B. R. Murphy, and C. S. Guy. 1993. Stock density in- condition for Largemouth Bass. Pages 79–91 in G. D. Novinger and J. G. dices: development, use, and limitations. Reviews in Fisheries Science 1: Dillard, editors. New approaches to the management of small impoundments. 203–222. Downloaded by [Department Of Fisheries] at 00:05 20 March 2013 This article was downloaded by: [Department Of Fisheries] On: 20 March 2013, At: 00:06 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 Temporal and Regional Trends in Black Bass Release Rates in Minnesota Daniel A. Isermann a , Joshua B. Maxwell a c & Michael C. McInerny b a College of Natural Resources, University of Wisconsin–Stevens Point, 800 Reserve Street, Stevens Point, Wisconsin, 54481, USA b Minnesota Department of Natural Resources, 23070 North Lakeshore Drive, Glenwood, Minnesota, 56334, USA c Illinois Natural History Survey, Kaskaskia Biological Station, 1235 County Road 1000N, Sullivan, Illinois, 61951, USA Version of record first published: 06 Mar 2013.

To cite this article: Daniel A. Isermann , Joshua B. Maxwell & Michael C. McInerny (2013): Temporal and Regional Trends in Black Bass Release Rates in Minnesota, North American Journal of Fisheries Management, 33:2, 344-350 To link to this article: http://dx.doi.org/10.1080/02755947.2013.763877

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 33:344–350, 2013 C American Fisheries Society 2013 ISSN: 0275-5947 print / 1548-8675 online DOI: 10.1080/02755947.2013.763877

ARTICLE

Temporal and Regional Trends in Black Bass Release Rates in Minnesota

Daniel A. Isermann* and Joshua B. Maxwell1 College of Natural Resources, University of Wisconsin–Stevens Point, 800 Reserve Street, Stevens Point, Wisconsin 54481, USA Michael C. McInerny Minnesota Department of Natural Resources, 23070 North Lakeshore Drive, Glenwood, Minnesota 56334, USA

Abstract Determining the prevalence of voluntary catch and release occurring in recreational fisheries for black bass Micropterus spp. is important in selecting and evaluating harvest regulations. Voluntary release rates of black bass have increased in some southern U.S. fisheries, but release rates could vary spatially. We used angler interviews from 245 open-water creel surveys conducted on 167 Minnesota bodies of water during 1984–2006 to determine whether black bass release rates differed among anglers specifically targeting black bass and anglers that were not targeting black bass, among geographic regions within the state, and among time periods. Ranked values of release rate did not significantly differ between anglers targeting black bass and anglers not targeting them for any of the four time periods. Before 1990, median black bass release rates were similar in northern and southern regions of the state and did not exceed 80% in either region. During 1990–1994 and 1995–1999, median release rates in the southern region exceeded 85% and the mean ranked release rate observed for 1995–1999 (92%) was significantly higher than during 1984–1989 (76%). Median release rates of black bass also increased in the northern region of Minnesota, but rates did not exceed 85% in any time period and were not significantly different among time periods based on ranked values. Anglers in Minnesota released most of the black bass they caught, but angler harvest of black bass was more prevalent in the northern region of Minnesota; this pattern may reflect regional differences in angler attitudes regarding black bass harvest. Changes in harvest regulations may not affect black bass populations in many waters because anglers voluntarily release most of the black bass they catch, but regulation changes may be more effective in regions where harvest of black bass is more prevalent. Downloaded by [Department Of Fisheries] at 00:06 20 March 2013

Black bass Micropterus spp. represent one of the most pop- the premise that harvest is sufficiently high to negatively affect ular groups of freshwater fish sought by recreational anglers angler catch rates and population size structure (Redmond 1974; in North America (U.S. Fish and Wildlife Service [USFWS] Wilde 1997; Paukert et al. 2007). 2008a). Due to this popularity, state and provincial agencies Angler attitudes regarding harvest of black bass have changed have steadily increased the number and complexity of length- dramatically, and rates of voluntary release of black bass have based harvest regulations for Largemouth Bass Micropterus generally increased (Quinn 1996; Noble 2002; Myers et al. salmoides and Smallmouth Bass Micropterus dolomieu,thetwo 2008). Allen et al. (2008) demonstrated that fishing mortal- black bass species primarily sought by anglers across much of ity rates for Largemouth Bass populations in the southern USA North America (Paukert et al. 2007). These regulations are often have declined over the last several decades, and this decline implemented with the goal of reducing fishing mortality, under was largely attributed to increased rates of voluntary release by

*Corresponding author: [email protected] 1Present address: Illinois Natural History Survey, Kaskaskia Biological Station, 1235 County Road 1000N, Sullivan, Illinois 61951, USA. Received July 9, 2012; accepted December 26, 2012 344 BLACK BASS RELEASE RATES IN MINNESOTA 345

anglers. However, angler attitudes regarding black bass harvest Minnesota Department of Natural Resources (MNDNR) might vary spatially (Champeau and Thomas 1993; Myers et al. sampling records indicate that Largemouth Bass and Small- 2008) and in certain areas anglers may be less likely to voluntar- mouth Bass occur in more than 1,600 bodies of water in Min- ily release black bass, resulting in higher rates of exploitation. nesota and black bass represent one of the most popular groups For example, tournament fishing for black bass and black-bass- of fish sought by recreational anglers within the state (USFWS oriented fishing organizations that have promoted the release of 2008b). Sampling also indicates that the range of black bass black bass are more prevalent in the southern USA than in other has expanded in the state and the abundance of black bass has portions of North America, which could result in differences in generally increased (A. Carlson and N. Frohnauer, MNDNR, the prevalence of voluntary release of black bass among regions. unpublished data). In Minnesota, Largemouth Bass and Small- Additionally, while more than half of all angler trips in many mouth Bass are near the northern end of their native ranges southern states (e.g., Alabama, Florida, and Texas) target black (Lee et al. 1980; Becker 1983) and growth rates of black bass bass, in the Upper Midwest (e.g., Minnesota, Wisconsin, and are negatively correlated with latitude (Beamesderfer and North Michigan) less than half of all angling trips target black bass, 1995). Consequently, the effects of exploitation on black bass largely because Walleyes Sander vitreus and Northern Pike Esox size structure in Minnesota may be more dramatic (Schramm lucius are targeted by a larger percentage of anglers (USFWS et al. 1995; Beamesderfer and North 1995) due to slow growth 2008b). Regional differences in angler preferences for certain and harvest regulations may be effective in improving size struc- species could influence release rates if anglers specifically tar- ture. A recent evaluation of both mandatory catch-and-release geting black bass release them at a higher rate than anglers not regulations (i.e., all bass must be released regardless of length) targeting black bass. However, we found no study comparing and 305-mm maximum length limits for Largemouth Bass sug- release rates between these two angler categories. gested that exploitation was still an important factor regulating Increased voluntary release of black bass has important im- bass size structure in some Minnesota lakes (Carlson and Iser- plications for fisheries management. State and provincial re- mann 2010). This suggests that voluntary release rates of black source agencies continue to implement regulations designed to bass in some Minnesota fisheries may be lower than in other re- reduce angler harvest in an effort to improve angler catch rates gions of North America. Additionally, in 2006 a 305–508-mm and size structure of black bass (Wilde 1997). Alternatively, re- protected slot length limit for Smallmouth Bass was removed source agencies also implement harvest regulations designed to on Green Lake, Kandiyohi County, Minnesota, in response to promote harvest of smaller black bass to reduce densities and declines in the abundance of Walleyes and Yellow Perch Perca improve black bass growth rates and size structure (Neumann flavescens; removal of the slot length limit was intended to in- et al. 1994; Martin 1995; Wilde 1997). Recreational anglers crease angler harvest of Smallmouth Bass. are often supportive of these harvest regulations (Wilde and Understanding temporal and spatial trends in voluntary re- Ditton 1991; Quinn 1996), probably because they assume these lease rates is important for fishery managers when trying to changes will result in improved fishing quality. More stringent determine whether changes in black bass harvest regulations harvest regulations may not improve black bass fisheries if an- will achieve management objectives, but trends in release rates glers voluntarily release most of the black bass they catch. Ad- have not been analyzed for black bass fisheries outside of the ditionally, higher voluntary release rates may result in increased southern USA. Our primary objectives were to (1) determine black bass abundance, which could affect their population dy- if release rates differed between anglers specifically target- namics (Johnson and Hale 1977; Post et al. 1998; Pereira et al. ing black bass and anglers that were not targeting black bass, 2002) and community interactions (MacRae and Jackson 2001; (2) determine if angler release rates of black bass in Minnesota Fayram et al. 2005; Weidel et al. 2007). Specifically, higher den- have increased over time, and (3) determine if release rates dif- Downloaded by [Department Of Fisheries] at 00:06 20 March 2013 sities of black bass could result in density-dependent growth, fered among geographic regions within the state. smaller size structure, and fewer black bass attaining sizes desired by anglers (Neumann et al. 1994). Resource agencies may also want to reduce black bass abundance to limit potential neg- METHODS ative interactions with other popular sport fish or with native The angler release rates of black bass were calculated from species where black bass have been introduced (Jackson 2002; the estimates of the total number of black bass caught (TBC) Vander Zanden et al. 2003; Zipkin et al. 2008). For example, and the total number of black bass harvested (TBH), information increases in black bass abundance have become a concern for obtained from 245 open-water creel surveys conducted by the anglers and fishery managers in portions of the northern USA MNDNR on 167 bodies of water during 1984–2006. Creel clerks and Canada, due to potential interactions with other popular first began asking anglers about the number of fish they released sportfish such as Walleyes (Fayram et al. 2005; OMNR 2009; in 1984. We could not analyze more recent trends in black bass Wuellner et al. 2010). These concerns have led to liberalization release rates because few (<10 surveys) creel surveys have been of black bass harvest regulations on some lakes in Wisconsin (J. conducted since 2006. Roving (i.e., stratified-random sampling) Hansen, Wisconsin Department of Natural Resources, personal or access-based (i.e., nonuniform probability sampling) creel communication), Minnesota, and Ontario (OMNR 2009). survey designs were primarily used to obtain angler information. 346 ISERMANN ET AL.

All waters included in the analysis were regulated by a statewide daily bag limit of six black bass per angler and no minimum length limit when creel surveys were conducted. Angler release rates were calculated as follows:

TBC − TBH Release rate = . TBC

We did not calculate release rates for each species of black bass because in most bodies of water a single harvest regulation existed for both species and angler catch was generally dominated by Largemouth Bass. Release rates included both voluntary release (i.e., an angler could have legally kept the bass but released it) and release of black bass required by law (e.g., an angler had already harvested a legal limit of bass and was required to release additional fish). However, a previous analysis reported that only 0.1% of 15,150 angler parties interviewed during creel surveys conducted on Minnesota waters from 1980 to 1996 actually harvested a legal daily limit of six black bass (Cook et al. 2001), which suggests that the majority of black bass caught by Minnesota anglers are released voluntarily. Consequently, we assumed release rates largely represented voluntary release of black bass by anglers. For some creel surveys, individual estimates of TBC and TBH were available for anglers specifically targeting black bass and for anglers that were not specifically targeting them; for these surveys we calculated release rates for FIGURE 1. Regional designations (i.e., north and south) used to evaluate each angler type. General trends in black bass release rates were spatial trends in black bass release rates in Minnesota recreational fisheries summarized using median values and 95% confidence inter- during 1984–2006. vals based on a binomial distribution (Zar 1999); medians were used because release rate data were not normally distributed. Ranked values of release rate were used in all analyses because in the southern region because (1) they are generally similar in normal probability plots and Shapiro–Wilk tests indicated that black bass abundance and fertility to other waters in the southern errors within specific treatment or factor levels were not nor- region, (2) creel surveys were conducted on less than five bodies mally distributed. Release rates were ranked using the RANK of water in the MSP during 1984–1989 and 2000–2006, and (3) procedure available in SAS (version 9.2; SAS Institute, Cary, preliminary analysis showed that release rates for MSP anglers North Carolina); mean ranks were used for ties. were similar to those observed in the southern region. To describe temporal trends in release rates, we categorized We used Wilcoxon’s signed rank tests to determine whether estimates based on whether they were obtained within the fol- ranked values of release rate differed between the two angler lowing time period: 1984–1989, 1990–1994, 1995–1999, and types (i.e., targeting or not targeting black bass) within each Downloaded by [Department Of Fisheries] at 00:06 20 March 2013 2000–2006. We could not analyze temporal trends in black bass time period with no comparison among regions. We did not release rates for individual bodies of water because creel sur- conduct regional comparisons of release rates between angler veys were sporadically conducted on individual waters. In cases types because (1) rates for both angler types were not available where multiple creel surveys were conducted on an individual for all creel surveys, (2) few paired estimates were available for body of water during the same time period, release rates for each certain regions during some time periods, and (3) these rates angler type were averaged across creel surveys. were estimated simultaneously and were likely not independent To describe regional trends in release rates, the location of of one another. waters within the state was described as either north or south After evaluating differences in ranked release rates be- (Figure 1). The north/south designation was used to recognize tween angler types, we used two separate analyses of variance that (1) lakes are more common and generally less eutrophic (ANOVAs) to analyze trends in ranked black bass release rates. in the northern region than in the southern portion of the state The number of waters with information on black bass release (Heiskary and Wilson 1989), (2) black bass fisheries are more rates exceeded 10 for all combinations of region and time period, common in the northern portion of the state, and (3) the northern with the exception of the southern region during 2000–2006 region also supports more fishing-based tourism. Waters in the (Table 1). Therefore, the first ANOVA included both regions Minneapolis–St. Paul metropolitan areas (MSP) were included and the first three time periods (i.e., 2000–2006 not included) BLACK BASS RELEASE RATES IN MINNESOTA 347

TABLE 1. Numbers of waters with creel survey data that were used in comparing black bass release rates among time periods (i.e., 1984–1989, 1990–1994, 1995–1999, and 2000–2006) and geographic northern and southern regions within the state (Figure 1).

Region Period Number of waters North 1984–1989 21 1990–1994 17 1995–1999 48 2000–2006 12 South 1984–1989 23 1990–1994 37 1995–1999 25 2000–2006 1

FIGURE 2. Box-and-whisker plots for release rates (%) of black bass caught by anglers in the northern and southern regions of Minnesota (Figure 1) during as main factors and a second ANOVAwas used to analyze trends four designated time periods between 1984 and 2006. Boxes represent interquar- in ranked release rates for only the northern region over all four tile ranges, lines within each box represent medians, and error bars represent time periods. Separate analyses were conducted because only 10th and 90th percentiles for each distribution of release rate. An asterisk denotes the only time period (i.e., 1995–1999) when mean ranked release rates one creel survey was available for the southern region during significantly differed between regions (Tukey–Kramer tests; P < 0.05). Within 2000–2006. We used least-squares means and Tukey–Kramer each geographic region, box-and-whisker plots denoted with different letters (x tests to make multiple comparisons among factor or treatment or y) indicate significant differences in mean ranked release rate between time levels. Statistical significance was set at α = 0.05 for analyses. periods (Tukey–Kramer tests; P < 0.05).

RESULTS and 1990–1994 (q = –3.67; df = 165, 6; P = 0.10) and during Ranked values of release rate did not significantly differ be- 1990–1994 and 1995–1999 (q = –2.30; df = 165, 6; P = 0.58), tween anglers specifically targeting black bass and for anglers but mean ranked release rate was significantly lower during that were not specifically targeting black bass within any of the 1984–1989 than during 1995–1999 (q = –5.43; df = 165, 6; P < > four time periods (Wilcoxon’s signed rank tests, P 0.05). Con- 0.01; Figure 2). Mean ranked release rates did not significantly sequently, release rates were averaged across angler types for differ among time periods in the northern region of the state each body of water when analyzing spatial and temporal trends (F = 1.49; df = 3, 131; P = 0.22; Figure 2). in release rates. Median release rates ranged from 71% to 92% among regions and time periods (Figure 2). Median release rates were generally DISCUSSION higher in the southern region of Minnesota than in the north- The similarity in black bass release rates between anglers ern region of the state (Figure 2). Median release rate steadily specifically targeting black bass and those not targeting them increased over time in the southern region and reached an ob- indicates that voluntary catch and release of black bass was served median release rate of 92% during 1995–1999. Median a generalized practice among Minnesota anglers. The specific Downloaded by [Department Of Fisheries] at 00:06 20 March 2013 release rate also increased in the northern region of the state, but reasons for this similarity in release rates is not known at this median release rate of black bass in the northern region did not time; the similarity could reflect the fact that anglers would exceed 80% until 2000–2006 (i.e., 85%). rather harvest species other than black bass for consumption Analysis of variance indicated that there was a significant (OMNR 2009). interaction between region and time period (F = 3.89; df = 2, Median release rates of black bass caught by anglers in Min- 165; P = 0.02) in explaining variation among ranked values of nesota have increased since the 1980s, although this trend was black bass release rate. Tukey–Kramer pairwise comparisons less apparent in the northern region of the state, where release indicated that ranked values of black bass release rate were rates were not statistically different among time periods. These similar between regions during 1984–1989 (q = –0.61; df = trends suggest that voluntary release of black bass has increased 165, 6; P = 0.99) and 1990–1994 (q = –2.5; df = 165, 6; in Minnesota, because harvest regulations on the waters included P = 0.48), but ranked release rates in the northern region were our analysis did not change between 1984 and 2006. significantly lower than ranked values observed in the southern Increased voluntary release rates of black bass were also region during 1995–1999 (q = –6.79; df = 165, 6; P < 0.001; reported in a 47-state survey conducted by Quinn (1996) and Figure 2). For the southern region of the state, mean ranked for seven bodies of water in Texas and Florida (Myers et al. release rate was not significantly different during 1984–1989 2008). Median release rates during 1995–2006 suggest that 348 ISERMANN ET AL.

anglers in Minnesota release 85% or more of the black bass where resource agencies want to encourage harvest of black they catch; these release rates were equivalent to or higher than bass, angler education programs that describe the benefits of release rates reported by Myers et al. (2008). Voluntary release harvesting black bass or the use of incentives may be necessary of black bass has likely increased because of the promotion of to achieve desired exploitation rates. Our results suggest that this behavior by various fishing-based organizations and the out- the effectiveness of harvest regulations in relation to voluntary door media (Quinn 1996; Myers et al. 2008), but other factors, release rates is likely to vary among regions. Median release such as increases in fish consumption advisories (Jakus et al. rates indicate that harvest regulations will be more likely to im- 1997; Burger et al. 2001), may have also contributed to this prove black bass fisheries in the northern region of the state; trend. however, the relationships among release rate, exploitation, and Median release rate in the southern region of Minnesota dur- black bass populations in the northern USA and Canada are ing 1995–1999 (92%) was substantially higher than in the north- poorly understood. ern region of the state (78%). It is unclear if this difference is High rates of voluntary release also indicate that mortal- sufficient to influence the effectiveness of black bass harvest ity associated with the catch and subsequent release of black regulations between the two regions or whether this difference bass may represent an important component of black bass fish- has persisted since 2000. However, our results indicate that an- ing mortality. Mark–recapture population estimates and angler gler behavior can vary across a relatively small spatial scale. catch estimates from creel surveys indicate the Largemouth Bass Reasons for this difference in release rates between regions are in some northern Minnesota waters are caught and released by not specifically known at this time, but angler attitudes regarding anglers multiple times during an open-water fishing season (C. harvest of black bass likely vary among the northern and south- Shavlik, MNDNR, unpublished data). Mortality associated with ern regions of Minnesota, and many factors can influence angler the catch and eventual release of black bass following retention behavior. For example, consumption rates of fish by recreational in live wells during competitive tournaments has been rela- anglers can vary in relation to socioeconomic status or popula- tively well studied (Wilde 1997), but limited information exists tion demographics (Burger 2002; Burger and Campbell 2008; regarding the mortality of black bass caught and released by Shilling et al. 2010) and these metrics vary between the south- nontournament anglers, where release of fish is more often im- ern and northern regions of the state (McMurry 2006a, 2006b, mediate (Muoneke and Childress 1994). Moreover, most of the 2008). Additionally, fishing-based tourism is more prevalent in available studies on black bass release mortality outside of tour- the northern region of Minnesota. Anglers on vacation typi- nament events have included only black bass <350 mm total cally have a finite amount of time to fish on an individual trip length and the mortality of larger fish desired by many anglers (e.g., several days or a week). Vacationing anglers may be more (i.e., ≥381 mm) is virtually unknown (Muoneke and Childress likely to harvest black bass if obtaining fish for a meal is im- 1994). Most of these studies have suggested that black portant to trip satisfaction (e.g., an annual tradition) and species bass release mortality rates associated with nontournament another than black bass are not captured in sufficient numbers to glers are less than 10% (Muoneke and Childress 1994), but meet this goal during a trip. Conversely, local anglers may only if low levels of harvest can negatively influence abundance and harvest fish on trips where sufficient numbers of fish gener- size structure in northern black bass populations (Schramm et al. ally considered more palatable than black bass (e.g., Walleyes, 1995), then low levels of release mortality could have the same Bluegills Lepomis macrochirus) are caught or harvest these fish effect, especially if release mortality is positively related to bass over multiple trips. Regional differences in attitudes towards fish size. We suggest that further study is needed to determine the consumption in relation to contaminant loads or water quality true effects of high voluntary release rates on black bass popu- could also influence release rates if anglers perceive that fish lations because harvest does not represent the only component Downloaded by [Department Of Fisheries] at 00:06 20 March 2013 from one region are safer to eat. Because of the complexity of fishing-related mortality that should be considered when im- of factors affecting angler motivations and attitudes regarding plementing management strategies (Coggins et al. 2007; Reeves the harvest of fish (Burger 2002), angler attitude surveys sim- and Bruesewitz 2007). ilar to Allen and Miranda (1996) would be necessary to determine why release rates vary among regions and different angler types. ACKNOWLEDGMENTS We thank D. Dustin and J. Breeggemann for data acquisi- Management Implications tion and management and the numerous MNDNR personnel In general, the high voluntary release rates of black bass involved in collecting the information used in this evaluation. caught by Minnesota anglers make it less likely that changes We also thank P. Addison with the Ontario Ministry of Natural to harvest regulations will affect black bass abundance and size Resources for information regarding black bass harvest regula- structure. However, Carlson and Isermann (2010) demonstrated tions in Ontario. We also thank several anonymous reviewers that harvest regulations did improve Largemouth Bass size struc- who provided comments that greatly improved the manuscript ture in some Minnesota lakes, indicating that harvest remains an and the University of Wisconsin–Stevens Point for the funds to important source of mortality in some populations. In situations cover publication costs. BLACK BASS RELEASE RATES IN MINNESOTA 349

REFERENCES McMurry, M. 2006a. Minnesota personal income trends, 2000 to 2005. Min- Allen, M. S., and L. E. Miranda. 1996. A qualitative evaluation of special- nesota State Demographic Center, Report OSD-06-123, Minnesota Depart- ization among crappie anglers. Pages 145–151 in L. E. Miranda and D. R. ment of Administration, St. Paul. DeVries, editors. Multidimensional approaches to reservoir fisheries man- McMurry, M. 2006b. Minnesota age trends, 2000 to 2005. Minnesota State agement. American Fisheries Society, Symposium 16, Bethesda, Maryland. Demographic Center, Report OSD-06-120, Minnesota Department of Ad- Allen, M. S., C. J. Walters, and R. Myers. 2008. Temporal trends in Largemouth ministration, St. Paul. Bass mortality, with fishery implications. North American Journal of Fisheries McMurry, M. 2008. Minnesota’s nonwhite and Latino populations, 2007. Min- Management 28:418–427. nesota State Demographic Center, Report OSD-08-135, Minnesota Depart- Beamesderfer, R. C. P., and J. A. North. 1995. Growth, natural mortality, and ment of Administration, St. Paul. predicted response to fishing for Largemouth Bass and Smallmouth Bass pop- Muoneke, M. I., and W. M. Childress. 1994. Hooking mortality: a review for ulations in North America. North American Journal of Fisheries Management recreational fisheries. Reviews in Fisheries Science 2:123–156. 15:688–704. Myers, R., J. Taylor, M. Allen, and T. F. Bonvechio. 2008. Temporal trends in Becker, G. C. 1983. Fishes of Wisconsin. University of Wisconsin Press, voluntary release of Largemouth Bass. North American Journal of Fisheries Madison. Management 28:428–433. Burger, J. 2002. Consumption patterns and why people fish. Environmental Neumann, R. M., D. W. Willis, and D. D. Mann. 1994. Evaluation of Largemouth Research 90:125–135. Bass slot length limits in two small South Dakota impoundments. Prairie Burger, J., and K. R. Campbell. 2008. Fishing and consumption patterns of Naturalist 26:15–32. anglers adjacent to the Oak Ridge Reservation, Tennessee: higher income Noble, R. L. 2002. Reflections on 25 years of progress in black bass manage- anglers ate more fish and are more at risk. Journal of Risk Research 11:335– ment. Pages 419–431 in D. P. Philipp and M. S. Ridgway, editors. Black 350. bass: ecology, conservation, and management. American Fisheries Society, Burger, J., M. Gochfeld, C. W. Powers, L. Waishwell, C. Warren, and B. D. Symposium 31, Bethesda, Maryland. Goldstein. 2001. Science, policy, stakeholders, and fish consumption advi- OMNR (Ontario Ministry of Natural Resources). 2009. Fisheries management sories: developing a fish fact sheet for the Savannah River. Environmental plan for fisheries management zone 6. OMNR, Thunder Bay District, Ottawa. Management 27:501–514. Paukert, C., M. McInerny, and R. Schultz. 2007. Historical trends in creel Carlson, A. J., and D. A. Isermann. 2010. Mandatory catch and release and limits, length-based limits, and season restrictions for black basses in the maximum length limits for Largemouth Bass in Minnesota: is exploitation United States and Canada. Fisheries 32:62–72. still a relevant concern? North American Journal of Fisheries Management Pereira, D. L., M. G. Halverson, R. E. Ramsell, and M. T. Drake. 2002. Changes 30:209–220. in the Largemouth Bass fishery of Lake Minnetonka, Minnesota: evidence Champeau, T. R., and P. W. Thomas. 1993. Voluntary release of Largemouth from tournament monitoring. Pages 513–524 in D. P. Philipp and M. S. Bass by Florida anglers. Proceedings of the Annual Conference Southeastern Ridgway, editors. Black bass: ecology, conservation, and management. Association of Fish and Wildlife Agencies 45(1991):317–322. American Fisheries Society, Symposium 31, Bethesda, Maryland. Coggins, L. G., Jr., M. J. Catalano, M. S. Allen, W. E. Pine III, and C. J. Walters. Post, D. M., J. F. Kitchell, and J. R. Hodgson. 1998. Interactions among adult 2007. Effects of cryptic mortality and the hidden costs of using length limits demography, spawning date, growth rate, predation, overwinter mortality, and in fishery management. Fish and Fisheries 8:196–210. the recruitment of Largemouth Bass in a northern lake. Canadian Journal of Cook, M. F., T. J. Goeman, P.J. Radomski, J. A. Younk,and P.C. Jacobson. 2001. Fisheries and Aquatic Sciences 55:2588–2600. Creel limits in Minnesota: a proposal for change. Fisheries 26(5):19–26. Quinn, S. 1996. Trends in regulatory and voluntary catch-and-release fishing. Fayram, A. H., M. J. Hansen, and T. J. Ehlinger. 2005. Interactions between Pages 152–162 in L. E. Miranda and D. R. DeVries, editors. Multidimensional Walleyes and four fish species with implications for Walleye stocking. North approaches to reservoir fisheries management. American Fisheries Society, American Journal of Fisheries Management 25:1321–1330. Symposium 16, Bethesda, Maryland. Heiskary, S. A., and C. B. Wilson. 1989. The regional nature of lake water Redmond, L. C. 1974. Prevention of overharvest of Largemouth Bass in Missouri quality across Minnesota: an analysis for improving resource management. impoundments. Pages 54–68 in J. L. Funk, editor. Symposium on overharvest Journal of the Minnesota Academy of Science 55:71–77. and management of Largemouth Bass in small impoundments. American Jackson, D. A. 2002. Ecological effects of Micropterus introductions: the dark Fisheries Society, North Central Division, Special Publication 3, Bethesda, side of black bass. Pages 221–232 in D. P. Philipp and M. S. Ridgway, editors. Maryland. Black bass: ecology, conservation, and management. American Fisheries Reeves, K. A., and R. E. Bruesewitz. 2007. Factors influencing the hooking mor- Society, Symposium 31, Bethesda, Maryland. tality of Walleyes caught by recreational anglers on Mille Lacs, Minnesota.

Downloaded by [Department Of Fisheries] at 00:06 20 March 2013 Jakus, P. M., M. Downing, M. S. Bevelhimer, and J. M. Fly. 1997. Do sportfish North American Journal of Fisheries Management 27:443–452. consumption advisories affect reservoir anglers’ site choice? Agricultural and Schramm, H. L., Jr., P. E. McKeown, and D. M. Green. 1995. Managing black Resource Economics Review 26:196–204. bass in northern waters: summary of the workshop. North American Journal Johnson, F. H., and J. G. Hale. 1977. Interrelations between Walleye (Stizoste- of Fisheries Management 15:671–679. dion vitreum vitreum) and Smallmouth Bass (Micropterus dolomieui) in four Shilling, F., A. White, L. Lippert, and M. Lubell. 2010. Contaminated fish northeastern Minnesota lakes, 1948–69. Journal of the Fisheries Research consumption in California’s Central Valley delta. Environmental Research Board of Canada 34:1626–1632. 110:334–344. Lee, D. S., C. R. Gilbert, C. H. Hocutt, R. E. Jenkins, D. E. McAllister, and J. R. USFWS (U.S. Fish and Wildlife Service). 2008a. 2006 national survey of fishing, Stauffer Jr. 1980. Atlas of North American freshwater fishes. North Carolina hunting, and wildlife-associated recreation. USFWS, Report Series FHW/06- Biological Survey, Publication 1980-12, North Carolina Museum of Natural NAT, Washington, D.C. History, Raleigh. USFWS (U.S. Fish and Wildlife Service). 2008b. 2006 national survey of fish- MacRae, P. S. D., and D. A. Jackson. 2001. The influence of Smallmouth Bass ing, hunting, and wildlife-associated recreation: statewide summaries. US- (Micropterus dolomieu) predation and habitat complexity on the structure FWS, Report Series FHW/06, Washington, D.C. of littoral zone fish assemblages. Canadian Journal of Fisheries and Aquatic Vander Zanden, M. J., K. A. Wilson, J. M. Casselman, and N. D. Yan. 2003. Sciences 58:342–351. Species introductions and their impacts in North American shield lakes. Pages Martin, C. C. 1995. Evaluation of slot length limits for Largemouth Bass in 239–263 in J. M. Gunn, R. J. Steedman, and R. A. Ryder, editors. Boreal shield two Delaware ponds. North American Journal of Fisheries Management 15: watersheds: Lake Trout ecosystems in a changing environment. CRC Press, 713–719. Boca Raton, Florida. 350 ISERMANN ET AL.

Weidel, B. C., D. C. Josephson, and C. E. Kraft. 2007. Littoral fish Wuellner, M. R., S. R. Chipps, D. W. Willis, and W. E. Adams Jr. 2010. Interac- community response to Smallmouth Bass removal from an Adiron- tions between Walleyes and Smallmouth Bass in a Missouri River reservoir dack lake. Transactions of the American Fisheries Society 136:778– with consideration of the influence of temperature and prey. North American 789. Journal of Fisheries Management 30:445–463. Wilde, G. R. 1997. Largemouth Bass fishery responses to length limits. Fisheries Zar, J. H. 1999. Biostatistical analyses, 4th edition. Prentice-Hall, Upper Saddle 22(6):14–23. River, New Jersey. Wilde, G. R., and R. B. Ditton. 1991. Diversity among anglers in support for Zipkin, E. F., P. J. Sullivan, E. G. Cooch, C. E. Kraft, B. J. Shuter, and B. C. fishery management tools. Pages 329–335 in J. L. Cooper and R. H. Hamre, Weidel. 2008. Overcompensatory response of a Smallmouth Bass (Mi- editors. Warmwater fisheries symposium I. U.S. Forest Service General Tech- cropterus dolomieu) population to harvest: release from competition? nical Report RM-207. Canadian Journal of Fisheries and Aquatic Sciences 65:2279–2292. Downloaded by [Department Of Fisheries] at 00:06 20 March 2013 This article was downloaded by: [Department Of Fisheries] On: 20 March 2013, At: 00:06 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 Channel Unit Use by Smallmouth Bass: Do Land-Use Constraints or Quantity of Habitat Matter? Shannon K. Brewer a a U.S. Geological Survey, Oklahoma Cooperative Fish and Wildlife Research Unit, Oklahoma State University, 007 Agriculture Hall, Stillwater, Oklahoma, 74078, USA Version of record first published: 06 Mar 2013.

To cite this article: Shannon K. Brewer (2013): Channel Unit Use by Smallmouth Bass: Do Land-Use Constraints or Quantity of Habitat Matter?, North American Journal of Fisheries Management, 33:2, 351-358 To link to this article: http://dx.doi.org/10.1080/02755947.2013.763878

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 33:351–358, 2013 American Fisheries Society 2013 ISSN: 0275-5947 print / 1548-8675 online DOI: 10.1080/02755947.2013.763878

ARTICLE

Channel Unit Use by Smallmouth Bass: Do Land-Use Constraints or Quantity of Habitat Matter?

Shannon K. Brewer* U.S. Geological Survey, Oklahoma Cooperative Fish and Wildlife Research Unit, Oklahoma State University, 007 Agriculture Hall, Stillwater, Oklahoma 74078, USA

Abstract I examined how land use influenced the distribution of Smallmouth Bass Micropterus dolomieu in channel units (discrete morphological features—e.g., pools) of streams in the Midwestern USA. Stream segments (n = 36), from four clusters of different soil and runoff conditions, were identified that had the highest percent of forest (n = 12), pasture (n = 12), and urban land use (n = 12) within each cluster. Channel units within each stream were delineated and independently sampled once using multiple gears in summer 2006. Data were analyzed using a generalized linear mixed model procedure with a binomial distribution and odds ratio statistics. Land use and channel unit were strong predictors of age-0, age-1, and age->1 Smallmouth Bass presence. Each age-class was more likely to be present in streams within watersheds dominated by forest land use than in those with pasture or urban land uses. The interaction between land use and channel unit was not significant in any of the models, indicating channel unit use by Smallmouth Bass did not depend on watershed land use. Each of the three age-classes was more likely to use pools than other channel units. However, streams with high densities of Smallmouth Bass age >1 had lower proportions of pools suggesting a variety of channel units is important even though habitat needs exist at the channel- unit scale. Management may benefit from future research addressing the significance of channel-unit quality as a possible mechanism for how land use impacts Smallmouth Bass populations. Further, management efforts aimed at improving stream habitat would likely be more beneficial if focused at the stream segment or landscape scale, where a variety of quality habitats might be supported.

Streams are strongly influenced by their surroundings at mul- increases impervious land cover, soil compaction, and ripar- tiple spatial and temporal scales (Schlosser 1991; Allan et al. ian encroachment while decreasing vegetative cover, resulting 1997; Fausch et al. 2002). Ecological patterns observed at fine in modification to the stream drainage network (May et al. Downloaded by [Department Of Fisheries] at 00:06 20 March 2013 spatial scales are the product of local habitat structure and con- 1997) and the frequency and magnitude of storm flows (Arnold straints imposed by more coarse environmental filters (Frissell et al. 1982). Chemical contaminants are often prevalent in urban et al. 1986; Tonn 1990; Snelder and Biggs 2002; Brewer et al. streams (Brown et al. 2005). Streams located in watersheds with 2007). In this context, the influence of anthropogenic activities limited forest cover have higher bedloads, embeddedness, sus- in a watershed may manifest at multiple levels within the hier- pended sediment, and greater streambed instability than streams archical arrangement. For example, human activities that occur located in watersheds with more forest cover (Sutherland et al. across a watershed disrupt geomorphic processes that create 2002). Agricultural land use lowers the water table and provides and maintain habitat at fine spatial scales (Allan 2004). The re- excess nutrient enrichment in adjacent streams (Karr et al. 1985). lationship between hierarchical levels may influence the quality Richards et al. (1993) examined physical habitat in streams in- or quantity of habitat available to and used by fishes. fluenced by a range of agricultural conditions and found habitat Physicochemical degradation in streams may occur be- was most impacted in intense agricultural areas. Row crop agri- cause of land-use activities (Allan 2004). Urban development culture is associated with fine sediments and steep bank angles

*Email: [email protected] Received September 10, 2012; accepted December 23, 2012 351 352 BREWER

(Rowe et al. 2009a). In Midwestern agriculture-dominated wa- to be sensitive to changes in sedimentation and other anthro- tersheds, stream habitat is characterized by nonmeandering pogenic impacts (Sowa and Rabeni 1995; Brewer and Rabeni channels, eroding banks, fine substrates, and more homoge- 2011), which may result in a shift in habitat use. Regardless of nization of channel units (i.e., glides) (Rowe et al. 2009b). habitat used, I hypothesized that increased quantities of used Changes in the physicochemical condition of streams re- habitat would not relate to increased densities because coarse- sulting from land-use changes are linked to changes in aquatic scale features appear to structure riverine Smallmouth Bass pop- biota, including fishes. Urbanization relates to lower species di- ulations (Brewer et al. 2007) and the species is dependent on versity and richness (Walters et al. 2005) and higher abundance food sources produced in multiple channel units of streams of opportunistic species (Alexandre et al. 2010). Additionally, (Rabeni 1992; Zweifel et al. 1999). endemic species (Walters et al. 2005) and indices of biological integrity (Fitzpatrick et al. 2005) decline as urbanization increases. Converting land use from forest or prairie to agri- METHODS culture alters the integrity of fish populations (Richards et al. Study area.—Thirty-six study streams were selected from 1996; Wang et al. 1997; Waite and Carpenter 2000; Stewart et al. three biogeographic regions in Missouri (Central Dissected 2001). The specific type of agriculture (i.e., row crop, pasture) Till Plains, Osage Plains, and the Ozark Highlands; Nigh alters the detected changes or severity of change (Meador and and Schroeder 2002). The biogeographic regions were described Goldstein 2003; Marshall et al. 2008). Collectively, the relations in Brewer and Rabeni (2011) and are only briefly described here between anthropogenic land-use changes and associated biota to provide justification for the use of multiple sampling gears. illustrate the influence of human activities on aquatic systems Generally, streams transitioned across the three ecoregions from (Brown et al. 2005; Hughes et al. 2006). Unfortunately, no easily smooth plains with high sediment loads to karst topography with detected warning signs have been offered to managers to identify low sediments loads and a high frequency of springs. Each bio- possible declines in fish populations due to land-use changes. geographic region was historically a mixture of prairie, savanna, Responses of fishes to land-use changes might first be ad- woodland, and forest. The vegetative structure of the watersheds dressed by examining possible changes in habitat use related to has now largely been converted to secondary growth forest, land-use conditions in the watershed. Fish may alter their use of cropland, pasture, and urban areas, with small tracts of native habitat in response to land-use conversions before managers can vegetation remaining (Nigh and Schroeder 2002). detect fitness responses in the population. In this study, I deter- Study design and site selection.—Thirty-six study streams mined if land use in the watershed influenced Smallmouth Bass were selected across the range of Smallmouth Bass in Micropterus dolomieu use of particular channel units (morpho- Missouri to allow channel-unit habitat use to be examined un- logical features formed by landscape and stream interactions der varying land-use constraints (Table 1; Figure 1). These study at high discharges; Leopold et al. 1995). Secondarily, if par- streams were used for the current and related studies. A detailed ticular channel units were used differentially, I examined how description of the study design was presented in Brewer and quantity of the channel unit across the reach influenced Small- Rabeni (2011). Study sites were selected to allow assessment of mouth Bass densities. I hypothesized that adult and subadult the interaction between natural features thought to relate to the Smallmouth Bass (hereafter termed age >1) would use channel success of Smallmouth Bass populations (e.g., soils, landform; units differently depending on the land use in the watershed. Brewer et al. 2007), while choosing sites within these condi- Younger age-classes of Smallmouth Bass are expected to be tions that varied in their influence from various land-use classes somewhat plastic in their use of habitat (Pert et al. 2002; but (forest, pasture, and urban lands). Multivariate cluster analyses see Brewer 2011); however, adult Smallmouth Bass are thought were performed to create four distinct clusters of streams that Downloaded by [Department Of Fisheries] at 00:06 20 March 2013

TABLE 1. Mean (range in parentheses) conditions in forest, pasture, and urban streams. Descriptors include the following: percentages of land use in respective watersheds, watershed area (area), discharge (Q), residual pool depth (pool depth), and width:depth ratios (W:D).

Stream designation Descriptor Forest Pasture Urban Forest (%) 74 (62–89) 19 (7–25) 31 (10–60) Pasture (%) 22 (11–33) 69 (45–91) 55 (43–62) Urban (%) 0.12 (0.00–0.28) 0.30 (0.00–1.27) 6.00 (3.00–22.00) Area (km) 241 (96–497) 210 (97–499) 296 (66–795) Q(m3/s) 0.28 (0–0.90) 0.26 (0–1.25) 0.19 (0–0.61) Pool depth (cm) 158 (29–352) 117 (65–225) 91 (60–132) W:D (m) 36 (22–53) 33 (23–42) 41 (25–63) CHANNEL UNIT USE BY SMALLMOUTH BASS 353

100 and backwaters). Each site was mapped prior to sampling (1–2 d) so the relative percentages of channel units could be determined and so block nets could be placed at channel-unit boundaries 80 immediately prior to fish sampling. Most riffles were excluded from sampling because previous sampling indicated riffles were 60 only used by age-0 Smallmouth Bass in very small streams in Missouri or following swim-up when densities were very high (Brewer, unpublished data). However, as part of this study, 20 Pool (%) Pool 40 riffles were haphazardly selected for sampling (snorkeling) and no Smallmouth Bass were found in any of them. 20 Physical-habitat parameters were measured at each site. Dis- charge was measured using the velocity–area method at the time of fish sampling. Residual pool depth and width-to-depth ratios 0 were calculated using the Environmental Protection Agency’s Pasture Forest Urban Environmental Monitoring and Assessment Program protocol Land use (Kaufmann et al. 1999) as part of a related study using the same set of streams (Doisy et al. 2005). FIGURE 1. Box plots of percent pool in pasture, forest, and urban land-use categories. The horizontal line in each box represents the median and box Fish sampling.—Fish sampling was conducted from July to dimensions represent the 25th and 75th percentiles. The whiskers indicate the September 2006 at each of the 36 study streams. Each site was 10th and 90th percentiles and the black dots show the 5th and 95th percentiles. a stream reach with a thalweg length 40 times the average bankfull width of the stream channel. Sampling over a range of landscape conditions required the use of multiple sampling gears: contained varying amounts of rocky soils and soils with high snorkeling, above-water observation, and seining. Snorkeling runoff potential (Doisy et al. 2005). These clusters were created and above-water observation were combined because efficiency to investigate the interactive effects of natural environmental differences of individual methods varied by channel unit and features and land-use attributes (Brewer and Rabeni 2011) and, combining the gears resulted in similar detection probabilities for the current study, to investigate possible detectable changes and lower fright bias from observed fish (Brewer and Ellersieck (presence) in channel-unit use related to land use. All study 2011). Seining (4.57 m long and 1.22 m wide with 3-mm mesh) streams were perennial and had drainage areas that ranged from was used on a limited basis (<20% of any stream segment) 66 to 795 km2. After the initial clusters were created to account but was used in every land-use class and more than 40% of the for the natural conditions of soil composition and permeability, time in forest stream segments. Any bias associated with sein- stream segments within each cluster were sorted according to the ing would be expected to underestimate fish presence in forest highest percentages of forest, pasture, and urban lands in each stream segments. Seining followed standard protocols (Bonar respective category (Brewer and Rabeni 2011). The percent- et al. 2010) and above-water observation and snorkeling tech- ages of each land class were determined using a 1:100,000 dig- niques are detailed in Brewer and Ellersieck (2011) and Brewer ital stream network populated with land-cover attributes (Sowa and Rabeni (2011). Briefly, block nets were used to separate a et al. 2005). Final stream segments were chosen that had the series of channel units where necessary. Block nets were not highest percentages of the targeted land use (forest, pasture, or used to separate very shallow riffles from pools or edgewaters urban), but otherwise had minimal variation due to extraneous from main channel habitats. Each channel unit was sampled Downloaded by [Department Of Fisheries] at 00:06 20 March 2013 threats (e.g., known point sources). Land-use classes were not from downstream to upstream and fish that were physically cap- homogenous. The percentages of forest and pasture lands were tured via seine were released in the downstream channel unit to relatively high in those land-use classes, whereas the percent- prevent movements into unsampled channel units. All observed age of urban lands was generally quite low (<20%). However, fish were separated into different age-classes (age 0, age 1, and watersheds with low percentages of urban lands or impervious age >1; Brewer and Rabeni 2011). surfaces are known to disproportionately affect aquatic biota Analyses.—Data were analyzed as a split plot in space (Walsh et al. 2005), with degradation and species loss occurring using the generalized linear mixed-model procedure (PROC at very low levels of conversion (<10%; Meador et al. 2005; GLIMMIX, SAS 2000) to account for possible spatial auto- Moore and Palmer 2005; Utz et al. 2010). correlation. The general linear model contained the effects of Channel-unit designations and physical-habitat measures.— land use (forest, urban, or pasture), stream within land use, A simplified version of the channel-unit classification system channel unit, and the interaction between land use and channel proposed by Rabeni and Jacobson (1993) was used to designate unit. These models used a logit link and a binomial distribution. channel-unit boundaries at each site. Channel units were pools, The density of Smallmouth Bass in channel units was trans- riffles, runs, vegetated edges (water willow Justicia americana), formed to presence–absence data because of the high number nonvegetated edges, and side-channel habitats (i.e., forewaters of channel units containing zeros in the data set. The random 354 BREWER

effect of stream within land-use category was used as the de- RESULTS nominator of F (ratio of between and within treatment variance) Smallmouth Bass were found in 22 of the 36 streams for testing the fixed effect of land use, whereas the random sampled. Twelve of the streams where Smallmouth Bass were residual effect was used as the denominator of F for testing absent were classified as pasture or urban streams (six in each the fixed effects of stream and the interaction of land use and class), whereas only two were forest streams. In the 36 streams stream. sampled, 716 individual channel units were examined: 37% Odds ratio statistics were calculated for significant treatment pools, 27% runs, 10% vegetated edges, 25% nonvegetated differences (land use, channel unit, or the interaction between edges, and 1% off channels. In streams where Smallmouth land use and channel unit). The antilog of each mean logit Bass were found, 464 channel units were sampled: 51% pools, estimate was used to calculate the odds [(Pp /1 – Pp), where 27% runs, 3% vegetated edges, 18 nonvegetated edges, and 1% Pp is the probability of fish being present] of Smallmouth Bass off channels. being found versus not found in individual treatments (land- Age-0 fish were present in 58% of sampled pools, 26% of use categories or channel units). The antilog of the difference sampled runs, and 16% of sampled nonvegeated edgewaters. between logit estimates, or odds ratio [(Px /1–Px)/(Py /1–Py), The main effects of land use (F2, 33 = 3.26, P = 0.05) and where x and y are different treatments and P is the probability channel unit (F2, 51 = 3.59, P = 0.03) were strong predictors of of residing in that treatment], was used to calculate the odds of age-0 fish occurrence, but the interaction between land use and Smallmouth Bass being present in one treatment versus another. channel unit was not significant (F4, 51 = 1.22, P = 0.31). Age-0 For example, if the logit estimate for Smallmouth Bass in stream fish were equally likely to occur in urban streams as in pasture segments located in forested watersheds was 0.88 and in pasture streams (t33 = 0.25, P = 0.81), but were ≥ 10 times more likely watersheds was –1.47, then the odds of Smallmouth Bass would to occur in forested than in urban (e.g., from Table 2: 0.88 + be more than ten times greater in stream segments located in 1.77 = 2.66, ex [2.66] = 14.25) or pasture streams (Table 2: forested watersheds than in urban watersheds {e.g., ex [0.88 – ex [2.37] = 10.65). Regardless of land-use category, age-0 fish = } (–1.47)] 10.66 . were more likely to be present in pools than in runs (t51 = Some channel units had to be excluded from analyses for 1.77, P = 0.08) or vegetated edges (t51 = 2.52, P = 0.01). specific life stages of fish because of the limited availabil- This age-class was four times more likely to occur in pools than ity of particular channel units found in some reaches or because no Smallmouth Bass were encountered in those channel TABLE 2. Generalized linear model results for different life stages of Small- units. These exclusions were necessary as the algorithm will mouth Bass. Asterisks identify significant logit values (P < 0.10, standard error not converge when zeros exist in an entire unit (i.e., channel- in parentheses) indicating an unequal probability of presence versus absence. < unit type in any land-use group). Off-channel habitats were not Negative values indicate fish were more likely to be absent ( 50% probability of occurrence), and positive values indicate fish were more likely to be present included in any of the analyses because this channel-unit type (>50% probability). Odds represent the proportion of fish present divided by was not encountered in forested streams at the time of sam- absent. pling. Two off channels were sampled in urban streams and one from a pasture stream and at least one life stage of Small- Life Land Channel mouth Bass was present in all of them. We excluded vegetated stage use unit Logit Odds edgewaters from the age-0 analysis because fish were only de- Age 0 Pool 0.33 (0.57) 1.40 tected in this channel unit in one pasture stream. We excluded Run –1.06 (0.72) 0.34 edgewaters (vegetated and nonvegetated) from the analyses of Vegetated *–1.63 (0.69) 0.20 age-1 or older fish. We sampled a substantial number of edge- Downloaded by [Department Of Fisheries] at 00:06 20 March 2013 edgewater = = waters (n 93 vegetated; n 165 nonvegetated); however, Pasture *–1.48 (0.81) 0.23 no age-1 or older fish occurred in any of the nonvegetated Forest 0.89 (0.78) 2.43 channel units and only three vegetated egdwaters had age-1 or Urban *–1.77 (0.86) 0.17 older Smallmouth Bass present, one stream from each land-use Age 1 Pool 0.49 (0.46) 1.62 type. Run *–1.08 (0.58) 0.34 To provide insight into the relative importance of channel- Pasture –0.99 (0.67) 0.37 unit quantity necessary to support densities of Smallmouth Bass, Forest *1.31 (0.67) 3.71 simple scatter plots were developed for each life stage of Small- Urban –1.21 (0.74) 0.30 mouth Bass. Densities of Smallmouth Bass were plotted against Age >1 Pool 0.59 (0.44) 1.81 the percentage of major channel units (riffles, runs, and pools) Run *–1.48 (0.60) 0.23 in each stream. Densities of Smallmouth Bass within the stream Pasture *–1.20 (0.72) 0.30 reach were weighted by the relative proportion of channel units Forest *1.05 (0.63) 2.86 to account for changing densities in different habitats across the Urban *–1.18 (0.69) 0.31 study reach (Brewer and Rabeni 2011). CHANNEL UNIT USE BY SMALLMOUTH BASS 355

in runs and seven times more likely to occur in pools than in vegetated-edge habitats (Table 2). Age-1 Smallmouth Bass were present in the majority of pools sampled (62%), but only present in 25% of sampled runs. The main effects of land use (F2, 33 = 4.03, P = 0.02) and channel unit (F1, 26 = 5.37, P = 0.03) were strong predictors of age-1 presence, but the interaction between the main effects was not significant (F2, 26 = 0.41, P = 0.66). Age-1 fish were equally likely to occur in urban streams as in pasture streams (t33 = 0.22, P = 0.82), but were ≥9 times more likely to occur in forest than in urban (Table 2: ex [2.52] = 12.44) or pasture streams (Table 2: ex [2.29] = 9.87). Regardless of land-use category, age-1 fish were more likely to be present in pools than in runs (t26 = 2.32, P = 0.02). There was an equal probability of age-1 fish being present or absent in available pools, but they were 5 times more likely to be present in pools than in runs. > Age- 1 Smallmouth Bass were predominantly present in FIGURE 2. Scatter plot of density (age->1fish/m2, weighted by proportion pools (present in 65% of those sampled), but also occurred in of channel unit habitat in each reach) and percent pool within each of 36 streams 18% of sampled runs. Like other age-classes, the main effects sampled in Missouri. of channel unit (F1, 26 = 8.91, P < 0.01) and land use (F2, 33 = 3.77, P = 0.03) were strong predictors of age->1 fish pres- than in urban or pasture watersheds. The physical factors re- ence, but channel-unit use did not depend on land-use category lated to the patterns observed in this study are unclear, though (F2, 26 = 0.26, P = 0.77). Age->1 fish were equally likely to the physical and chemical consequences of urbanization have occur in urban streams as in pasture streams (t33 = –0.02, P = been well documented and include altered hydrology (Arnold 0.98), but were 9 times more likely to occur in forested than in et al. 1982; Paul and Meyer 2001; Walsh et al. 2005) and temper- urban (Table 2: ex [2.23] = 9.33) or pasture streams (Table 2: ature (Paul and Meyer 2001). Extreme discharges during critical ex [2.24] = 9.48). Regardless of land-use category, age->1fish periods increase Smallmouth Bass mortality and reduce recruit- were more likely to be present in pools than runs (t26 = 2.99, ment success (Peterson and Kwak 1999; Smith et al. 2005; P < 0.01). There was an equal probability of age->1 fish being Hafs et al. 2010) and high water temperatures restrict growth present in sampled pools, but they were 8 times more likely to and reduce Smallmouth Bass densities (Sowa and Rabeni 1995; be present in pools than in runs. Zweifel et al. 1999). The influence of pasture agriculture on The percentages of major channel units (riffles, run, and aquatic fauna may be less pronounced than that found un- pools) in each stream were highly correlated with one another. der well-studied, more intensive agriculture practices (e.g., row For example, the relationship between percent pool and densi- crop) (Meador and Goldstein 2003; Strayer et al. 2003). A pos- ties of smallmouth was essentially the reverse of the relationship itive benefit may be expected from having some agricultural between percent riffle or run and densities of Smallmouth Bass, lands (e.g., increase in invertebrate abundance) until some although more variation was shown using percent run. Further, threshold is reached (Quinn and Hickey 1990; Quinn 2000). The the patterns observed were similar for each life stage. The re- threshold is unknown for Midwestern USA pasture-dominated lationship between percent pool and densities of age->1 small- watersheds but would likely vary with management practices Downloaded by [Department Of Fisheries] at 00:06 20 March 2013 mouth is presented here because of the significance of pools (Marshall et al. 2008). relative to fish presence. Variation in percent pool was substan- Landscape-scale studies have limitations because of the diffi- tial in streams with low densities of age->1SmallmouthBass; culties in disentangling the effects of multiple land-use classes. however, in high-density streams, percent pool was much lower Land-use classes are not independent of one another, suggesting indicating a variety of channel units within a reach is important more than one land-use class may be acting on populations si- for these populations (Figure 2). multaneously (King et al. 2005). This is very likely the case for many of the streams I classified as “urban.” The highest percent of urban land use in any of the streams meeting the criteria of DISCUSSION the study was 22%. All of the streams classified as urban also A growing body of evidence suggests that the impacts of had a high proportion of pasture. Most of the streams (10 of 12) urbanization on stream biota are more severe than impacts from had more than 50% pasture in the watershed, in addition to the other land-use changes such as agriculture (e.g., Paul and Meyer urban percentage, suggesting that the influence of pasture land 2001; Brown et al. 2005; Walsh et al. 2005; Utz et al. 2010). use also relates to the results observed in this study. This study indicates native Smallmouth Bass are more likely to The relationships between Smallmouth Bass and land use be present in stream segments located in forested watersheds were expected given other analyses conducted as part of related 356 BREWER

studies on these streams (Brewer and Rabeni 2011); however, channel morphology could change the quality (e.g., depth, tem- land use did not impact habitat needs at the channel-unit scale perature, substrate size) of these channel units. Changes in the as I hypothesized. It is interesting that habitat needs exist at the quality of channel units could be linked to mechanisms (e.g., fine spatial scale (e.g., pools) but the coarse-scale factors over- increased metabolic demand) for population changes. Future ride Smallmouth Bass presence at the reach scale. This finding studies addressing habitat quality have the potential to increase supports stream hierarchical theory (Frissell et al. 1986) but is the management options available for improving Smallmouth a mismatch to the spatial scale at which stream management Bass populations. and restoration often takes place. In many instances, restoration Habitat-use results from one system are often not trans- targets are set at the channel-unit or reach scale with limited con- ferrable to other systems due to the relative availability of habitat sideration for the factors or processes at coarse spatial scales that features. However, when placed in the context of coarse-scale may be limiting to the population or habitat. features (in this case, land use) and taken over many stream Habitat needs for Smallmouth Bass extend beyond the segments, patterns may emerge that are useful for identifying channel-unit scale and likely include the interrelationship be- changes in habitat use in response to particular stressors. I have tween channel units within a stream reach. The hypothesis that identified the importance of particular channel-unit features to habitat use would change based on land-use characteristics in Smallmouth Bass but also emphasized the importance of land- the watershed was incorrect. Smallmouth Bass were present scape management in improving stream-fish populations. It is in pool channel units most of the time regardless of land-use clear that a variety of habitats is necessary to support Small- class. However, more pool habitat in a reach did not relate to mouth Bass populations even though the species disproportion- increases in Smallmouth Bass densities but rather the opposite. ately occupies only a few channel units. Monitoring changes Sowa and Rabeni (1995) found Smallmouth Bass population in physical-habitat features in a repeatable fashion, in addition densities to be negatively associated with percent pool area in to the fish populations, may provide more insight to possible streams of the Ozark border region in Missouri. Increased pool declines in the fitness of these populations. Data related to the quantity was associated with lower food production for Small- quality of physical features might be paired with changes in the mouth Bass (Zweifel et al. 1999). Secondary production by abundance of important prey species. Information such as this several crayfish species (a primary food source of Smallmouth should become readily available with the wide-ranging biologi- Bass: Rabeni 1992; Roell and Orth 1993; Roell and DiStefano cal monitoring programs initiated by many water resource agen- 2010) was higher in nonpool channel units even though they cies (e.g., Environmental Protection Agency’s Environmental represented only a small fraction of total stream area in Ozark Monitoring and Assessment Program). streams (Brewer et al. 2009). Additionally, Hafs et al. (2010) indicated survival of Smallmouth Bass was lower than expected ACKNOWLEDGMENTS when moving-water channel units dried during summer. Collec- This research is a contribution of the Oklahoma Coopera- tively, these studies suggest channel units that are rarely, if ever, tive Fish and Wildlife Research Unit (U.S. Geological Survey, used by Smallmouth Bass are still very important to the over- Oklahoma Department of Wildlife Conservation, Oklahoma all fitness of this species. In addition, this highlights a possible State University, and Wildlife Management Institute cooper- weakness associated with relating increased quantities of used ating), with support from the Missouri Cooperative Fish and habitat with instream flow needs (i.e., microhabitat or mesohab- Wildlife Research Unit. Funding was provided by the Missouri itat evaluations) or restoration targets (e.g., Parasiewicz 2008). Department of Conservation. The use of trade names or products Only if habitat quantity at these fine scales is limiting, would does not imply endorsement by the U.S. Government. Special strategies directed at increasing quantity of habitat be effective. Downloaded by [Department Of Fisheries] at 00:06 20 March 2013 thanks to Mark Ellersieck, Greg Wallace, Matt Mullins, Danny Understanding the mechanisms regulating populations, par- Moncheski, Allison McCluskey, and Brook Kruse for valuable ticularly under anthropogenic alterations, can help improve field and data assistance. This manuscript was improved by management decisions. Often managers assume the quantity comments from Phillip Bettoli, Mike Roell, Seth Wenger, and of habitat is limiting to populations (e.g., instream flows) when two anonymous reviewers. there may be several alternative explanations. Water quality (Brown et al. 2005), alterations in flow at critical times of the life cycle (Peterson and Kwak 1999), and sedimentation (Sowa and REFERENCES Alexandre, C. V., K. E. Esteves, and M. A. Marcondes de Moura e Mello. 2010. Rabeni 1995) have been suggested as factors that alter the poten- Analysis of fish communities along a rural–urban gradient in a neotropi- tial of Smallmouth Bass populations related to land use. I suggest cal stream (Piracicaba River basin, Sao˜ Paulo, Brazil). Hydrobiologia 641: the quality of the habitats occupied may also relate to changes 97–114. in densities of Smallmouth Bass. This attribute was not inves- Allan, J. D. 2004. Landscapes and riverscapes: the influence of land use on tigated here, but limited data collected to describe study sites stream ecosystems. Annual Review of Ecology, Evolution, and Systematics 35:257–284. indicate there may be differences in mean residual pool depth Allan, J. D., D. L. Erickson, and J. Fay. 1997. The influence of catchment land (pool depth independent of discharge) and width-to-depth ratios use on stream integrity across multiple spatial scales. Freshwater Biology among streams in different land-use categories. Alterations in 37:149–161. CHANNEL UNIT USE BY SMALLMOUTH BASS 357

Arnold, C. L., P. J. Boison, and P. C. Patton. 1982. Sawmill Brook: an example communities in southwest Wisconsin streams. North American Journal of of rapid geomorphic change related to urbanization. Journal of Geology Fisheries Management 28:944–953. 90:155–166. May, C. W., R. R. Horner, J. R. Karr, B. W. Mar, and E. B. Welch. 1997. Effects Bonar, S. A., W. A. Hubert, and D. W. Willis, editors. 2010. Standard methods of urbanization on small streams in the Puget Sound ecoregion. Watershed for sampling North American freshwater fishes. American Fisheries Society, Protection Techniques 2:483–494. Bethesda, Maryland. Meador, M. R., J. F. Coles, and H. Zappia. 2005. Fish assemblage responses Brewer, S. K. 2011. Patterns in young-of-year Smallmouth Bass microhabitat to urban intensity gradients in contrasting metropolitan areas: Birmingham, use in multiple stream segments with contrasting land uses. Fisheries Man- Alabama and Boston, Massachusetts. Pages 409–423 in L. R. Brown, R. H. agement and Ecology 18:506–512. Gray, R. M. Hughes, and M. R. Meador, editors. Effects of urbanization on Brewer, S. K., R. J. DiStefano, and C. F. Rabeni. 2009. The influence of age- stream ecosystems. American Fisheries Society, Symposium 47, Bethesda, specific habitat selection by a stream crayfish community (Orconectes spp.) Maryland. on secondary production. Hydrobiologia 619:1–10. Meador, M. R., and R. M. Goldstein. 2003. Assessing water quality at large Brewer, S. K., and M. R. Ellersieck. 2011. Evaluating two observational sam- geographic scales: relations among land use, water physicochemistry, ripar- pling techniques for determining the distribution and detection probability of ian condition, and fish community structure. Environmental Management age-0 Smallmouth Bass in clear, warmwater streams. North American Journal 31:504–517. of Fisheries Management 31:894–904. Moore, A. A., and M. A. Palmer. 2005. Invertebrate biodiversity in agricultural Brewer, S. K., and C. F. Rabeni. 2011. Interactions between natural-occurring and urban headwater streams: implications for conservation and management. landscape conditions and land use influencing the abundance of riverine Ecological Applications 15:1169–1177. Smallmouth Bass, Micropterus dolomieu. Canadian Journal of Fisheries and Nigh, T. A., and W. A. Schroeder. 2002. Atlas of Missouri ecoregions. Missouri Aquatic Sciences 68:1922–1933. Department of Conservation, Conservation Commission, Jefferson City. Brewer, S. K., C. F. Rabeni, S. P. Sowa, and G. Annis. 2007. Natural landscape Parasiewicz, P. 2008. Application of MesoHABSIM and target fish commu- and stream segment attributes influencing the distribution and relative abun- nity approaches to restoration of the Quinebaug River, Connecticut and dance of riverine Smallmouth Bass in Missouri. North American Journal of Massachusetts, USA. River Research and Applications 24:459–471. Fisheries Management 27:326–341. Paul, M. J., and J. L. Meyer. 2001. Streams in the urban landscape. Annual Brown, L. R., R. H. Gray, R. M. Hughes, and M. R. Meador, editors. 2005. Review of Ecology and Systematics 32:333–365. Effects of urbanization on stream ecosystems. American Fisheries Society, Pert, E. J., D. J. Orth, and M. J. Sabo. 2002. Lotic-dwelling age-0 Smallmouth Symposium 47, Bethesda, Maryland. Bass as both resource specialists and generalists: reconciling disparate litera- Doisy, K. E., C. F. Rabeni, S. P. Sowa, and M. A. Urban. 2005. Landscape ture reports. Pages 185–189 in D. P.Philipp and M. S. Ridgway, editors. Black factors associated with sediment in Missouri streams. Final Report to the Bass: ecology, conservation, and management. American Fisheries Society, U.S. Environmental Protection Agency, Report X7-98750001-0, Missouri Symposium 31, Bethesda, Maryland. Resource Assessment Partnership, University of Missouri, Kansas City. Peterson, J. T., and T. J. Kwak. 1999. Modeling the effects of land use and climate Fausch, K. D., C. E. Torgersen, C. V. Baxter, and H. W. Li. 2002. Landscapes change on riverine Smallmouth Bass. Ecological Applications 9:1391–1404. to riverscapes: bridging the gap between research and conservation of stream Quinn, J. M. 2000. Effects of pastoral development. Pages 208–229 in K. J. fishes. BioScience 52:483–498. Collier and M. J. Winterbourn, editor. New Zealand stream invertebrates: Fitzpatrick, F. A., M. W. Diebel, M. A. Harris, T. L. Arnold, M. A. Lutz, and ecology and implications for management. Caxton Press, Christchurch, New K. D. Richards. 2005. Effects of urbanization on the geomorphology, habi- Zealand. tat, hydrology, and fish index of biotic integrity of streams in the Chicago Quinn, J. M., and C. W. Hickey. 1990. Magnitude of effects of substrate particle area, Illinois and Wisconsin. Pages 87–115 in L. R. Brown, R. H. Gray, size, recent flooding, and catchment development on benthic invertebrates R. M. Hughes, and M. R. Meador, editors. Effects of urbanization on in 88 New Zealand rivers. New Zealand Journal of Marine and Freshwater stream ecosystems. American Fisheries Society, Symposium 47, Bethesda, Research 24:411–427. Maryland. Rabeni, C. F. 1992. Trophic linkage between stream centrarchids and their Frissell, C. A., W. J. Liss, C. E. Warren, and M. D. Hurley. 1986. A hierarchical crayfish prey. Canadian Journal of Fisheries and Aquatic Sciences 49:1714– framework for stream habitat classification: viewing streams in a watershed 1721. context. Environmental Management 10:199–214. Rabeni, C. F., and R. B. Jacobson. 1993. Geomorphic and hydraulic influences Hafs, A. W., C. J. Gagen, and J. K. Whalen. 2010. Smallmouth Bass summer on the abundance and distribution of stream centrarchids in Ozark U.S.A. habitat use, movement, and survival in response to low flow in the Illinois streams. Polskie Archiwum Hydrobiologii 40:87–99.

Downloaded by [Department Of Fisheries] at 00:06 20 March 2013 Bayou, Arkansas. North American Journal of Fisheries Management 30: Richards, C., G. E. Host, and J. W. Arthur. 1993. Identification of predomi- 604–612. nant environmental factors structuring stream macroinvertebrate communi- Hughes, R. M., L. Wang, and P.W.Seelbach, editors. 2006. Landscape influences ties within a large agricultural catchment. Freshwater Biology 29:285–294. on stream habitats and biological assemblages. American Fisheries Society, Richards, C., L. B. Johnson, and G. E. Host. 1996. Landscape-scale influences Symposium 48, Bethesda, Maryland. on stream habitats and biota. Canadian Journal of Fisheries and Aquatic Karr, J. R., L. A. Toth, and D. R. Dudley. 1985. Fish communities of Midwestern Sciences 53(Supplement 1):295–311. rivers: a history of degradation. BioScience 35:90–95. Roell, M. J., and R. J. DiStefano. 2010. Effects of a conservative Rock Bass Kaufmann, P. R., P. Levine, E. G. Robison, C. Seeliger, and D. V. Peck. 1999. length limit on angler participation, sport fish populations, and crayfish prey in Quantifying physical habitat in wadeable streams. U.S. Environmental Pro- a Missouri Ozark stream. North American Journal of Fisheries Management tection Agency, EPA/620/R-99/003, Washington, D.C. 30:552–564. King, R. S., M. E. Baker, D. F. Whigham, D. E. Weller, T. E. Jordan, P. F. Roell, M. J., and D. J. Orth. 1993. Trophic basis of production of stream-dwelling Kazyak, and M. K. Hurd. 2005. Spatial considerations for linking watershed Smallmouth Bass, Rock Bass, and Flathead Catfish in relation to invertebrate land cover to ecological indicators in streams. Ecological Applications 15: bait harvest. Transactions of the American Fisheries Society 122:46–62. 137–153. Rowe, D. C., C. L. Pierce, and T. F. Wilton. 2009a. Fish assemblage relationships Leopold, L. B., M. G. Wolman, and J. P. Miller. 1995. Fluvial processes in with physical habitat in wadeable Iowa streams. North American Journal of geomorphology. Dover, New York. Fisheries Management 29:1314–1332. Marshall, D. W., A. H. Fayram, J. C. Panuska, J. Baumann, and J. Hennessy. Rowe, D. C., C. L. Pierce, and T. F. Wilton. 2009b. Physical habitat and fish 2008. Positive effects of agricultural land use changes on coldwater fish assemblage relationships with landscape variables at multiple spatial scales 358 BREWER

in wadeable Iowa streams. North American Journal of Fisheries Management Sutherland, A. B., J. L. Meyer, and E. P. Gardiner. 2002. Effects of land cover on 29:1333–1351. sediment regime and fish assemblage structure in four southern Appalachian SAS (Statistical Analysis Systems). 2000. SAS/STAT user’s guide, version 8. streams. Freshwater Biology 47:1791–1805. SAS Institute, Cary, North Carolina. Tonn, W. M. 1990. Climate change and fish communities: a conceptual Schlosser, I. J. 1991. Stream fish ecology: a landscape perspective. BioScience framework. Transactions of the American Fisheries Society 119:337– 41:704–712. 352. Smith, S. M., J. S. Odenkirk, and S. J. Reeser. 2005. Smallmouth Bass recruit- Utz, R. M., R. H. Hilderbrand, and R. L. Raesly. 2010. Regional differences in ment variability and its relation to stream discharge in three Virginia rivers. patterns of fish species loss with changing land use. Biological Conservation North American Journal of Fisheries Management 25:1112–1121. 143:688–699. Snelder, T. H., and B. J. F. Biggs. 2002. Multiscale river environment clas- Waite, I. R., and K. D. Carpenter. 2000. Associations among fish as- sification for water resources management. Journal of the American Water semblage structure and environmental variables in Willamette basin Resources Association 38:1225–1239. streams, Oregon. Transactions of the American Fisheries Society 129: Sowa, S. P., D. D. Diamond, R. Abbitt, G. M. Annis, T. Gordon, M. E. Morey, 754–770. G. R. Sorensen, and D. True. 2005. A gap analysis for riverine ecosystems of Walsh, C. J., A. H. Roy, J. W. Feminella, P. D. Cottingham, P. M. Groffman, and Missouri. Final Report to the U.S. Geological Survey National Gap Analysis R. P. Morgan II. 2005. The urban stream syndrome: current knowledge and Program, Missouri Resource Assessment Partnership, University of Missouri, the search for a cure. Journal of the North American Benthological Society Columbia. 24:706–723. Sowa, S. P., and C. F. Rabeni. 1995. Regional evaluation of the relation of Walters, D. M., M. C. Freeman, D. S. Leigh, B. J. Freeman, and C. M. Pringle. habitat to distribution and abundance of Smallmouth Bass and Largemouth 2005. Urbanization effects on fishes and habitat quality in a southern Piedmont Bass in Missouri streams. Transactions of the American Fisheries Society river basin. Pages 69–85 in L. R. Brown, R. H. Gray, R. M. Hughes, and M. R. 124:240–251. Meador, editors. Effects of urbanization on stream ecosystems. American Stewart, J. S., L. Z. Wang, J. Lyons, J. A. Horwatich, and R. Bannerman. 2001. Fisheries Society, Symposium 47, Bethesda, Maryland. Influences of watershed, riparian-corridor, and reach-scale characteristics on Wang, L. Z., J. Lyons, P. Kanehl, and R. Gatti. 1997. Influences of watershed aquatic biota in agricultural watersheds. Journal of the American Water Re- land use on habitat quality and biotic integrity in Wisconsin streams. Fisheries sources Association 37:1475–1487. 22(6):6–12. Strayer, D. L., R. E. Beighley, L. C. Thompson, S. Brooks, C. Nilsson, G. Pinay, Zweifel, R. D., R. S. Hayward, and C. F. Rabeni. 1999. Bioenergetics insight and R. J. Naiman. 2003. Effects of land cover on stream ecosystems: roles of into Black Bass distribution shifts in Ozark border region streams. North empirical models and scaling issues. Ecosystems 6:407–423. American Journal of Fisheries Management 19:192–197. Downloaded by [Department Of Fisheries] at 00:06 20 March 2013 This article was downloaded by: [Department Of Fisheries] On: 20 March 2013, At: 00:07 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 Contrast of Degraded and Restored Stream Habitat Using an Individual-Based Salmon Model Steven F. Railsback a , Mark Gard b , Bret C. Harvey c , Jason L. White c & Julie K. H. Zimmerman d a Lang, Railsback and Associates, 250 California Avenue, Arcata, California, 95521, USA b U.S. Fish and Wildlife Service, Sacramento Fish and Wildlife Office, 2800 Cottage Way, Sacramento, California, 95825, USA c U.S. Forest Service, Pacific Southwest Research Station, 1700 Bayview Drive, Arcata, California, 95521, USA d U.S. Fish and Wildlife Service, Bay-Delta Fish and Wildlife Office, 650 Capitol Mall, Sacramento, California, 95814, USA Version of record first published: 19 Mar 2013.

To cite this article: Steven F. Railsback , Mark Gard , Bret C. Harvey , Jason L. White & Julie K. H. Zimmerman (2013): Contrast of Degraded and Restored Stream Habitat Using an Individual-Based Salmon Model, North American Journal of Fisheries Management, 33:2, 384-399 To link to this article: http://dx.doi.org/10.1080/02755947.2013.765527

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 33:384–399, 2013 C American Fisheries Society 2013 ISSN: 0275-5947 print / 1548-8675 online DOI: 10.1080/02755947.2013.765527

ARTICLE

Contrast of Degraded and Restored Stream Habitat Using an Individual-Based Salmon Model

Steven F. Railsback* Lang, Railsback and Associates, 250 California Avenue, Arcata, California 95521, USA Mark Gard U.S. Fish and Wildlife Service, Sacramento Fish and Wildlife Office, 2800 Cottage Way, Sacramento, California 95825, USA Bret C. Harvey and Jason L. White U.S. Forest Service, Pacific Southwest Research Station, 1700 Bayview Drive, Arcata, California 95521, USA Julie K. H. Zimmerman U.S. Fish and Wildlife Service, Bay-Delta Fish and Wildlife Office, 650 Capitol Mall, Sacramento, California 95814, USA

Abstract Stream habitat restoration projects are popular, but can be expensive and difficult to evaluate. We describe inSALMO, an individual-based model designed to predict habitat effects on freshwater life stages (spawning through juvenile out-migration) of salmon. We applied inSALMO to Clear Creek, California, simulating the production of total and large (>5 cm FL) Chinook Salmon Oncorhynchus tshawytscha out-migrants at a degraded and a restored site. The calibrated model reproduced observed redd locations and out-migrant timing and size. In simulations, the restored site had a much higher production of fry that established and grew before out-migration; it provided higher survival and positive growth due to moderate velocities, shallow depths, and cover for feeding and hiding. The restored site did not produce more total out-migrants because at both sites spawning gravel was sufficient and the vast majority of fry moved downstream soon after emergence. Simulations indicated that at both sites increasing food and cover availability could further increase production of large, but not total, out-migrants; spawning gravel, temperature, and flow appear nearly optimal already. Further gravel addition was predicted to increase total fry production but have little or even a negative effect on production of large out-migrants, illustrating that actions benefitting one

Downloaded by [Department Of Fisheries] at 00:07 20 March 2013 life stage can negatively affect others. The model predicted that further enhancements (e.g., in cover availability) would be more beneficial at the restored site than at the degraded site. Restoration efforts may be most effective when concentrated in “hot spots” with good habitat for growth and predator avoidance as well as for spawning. Contradicting the traditional notion of “limiting factors,” the model indicated that several factors each have strong effects. The model provided more understanding of restoration effects than would field studies alone and could be useful for designing projects to meet specific restoration objectives.

Designing and evaluating fish habitat restoration projects projects are either not evaluated or evaluations cannot clearly are important yet difficult tasks for freshwater fisheries man- document success (Bernhardt et al. 2005; Roni et al. 2008; agers. There is limited guidance available for designing suc- Jahnig¨ et al. 2011). Many restoration projects are designed to cessful river restoration projects, in part because most completed create a perception of “good” habitat without addressing the root

*Corresponding author: [email protected] Received June 23, 2012; accepted January 2, 2013 384 MODELING DEGRADED VS. RESTORED SALMON HABITAT 385

causes of degradation and ultimately fail to accomplish desired provide more potential for understanding the inherently com- objectives (Beechie et al. 2010). In response, recent literature plex effects of management on fish populations (e.g., Harvey has outlined best practices for achieving positive restoration and Railsback 2007; Railsback and Harvey 2011). outcomes (Beechie et al. 2010; Bernhardt and Palmer 2011). We introduce and illustrate the use of inSALMO, a new One of the fundamental principles is to develop a process-based IBM designed to support management of freshwater life stages approach to restoration focused on restoring hydrologic, geo- of salmon—spawning through rearing and out-migration. morphic, and biological processes that maintain river ecosys- The model is adapted from a family of IBMs that have been tem function. An important step in this approach is to explicitly applied to a variety of salmonid management and research link potential management actions to restoration objectives by issues (Railsback and Harvey 2001; Railsback et al. 2009; see predicting the effects of alternative management actions on eco- www.humboldt.edu/ecomodel/instream.htm). We describe the logical processes and fish population response. At the reach model and its application to two sites that represent restored scale, designing restoration actions that directly influence such and degraded conditions in Clear Creek, a productive but highly processes as habitat selection, predation, feeding, growth, and modified salmon spawning stream in California’s Central competition can achieve positive outcomes for fish populations Valley. We use the model and field observations to contrast (Beechie et al. 2010). Chinook Salmon Oncorhynchus tshawytscha spawning, incu- Modeling restoration actions allows managers to clarify their bation, and juvenile rearing between the two sites. The analyses objectives, define assumptions about the relationships between produced general conclusions about restoration and the model’s actions and ecological processes, explore sources of uncertainty, usefulness. and quantify predicted outcomes. Once restoration projects are completed, models can be used to evaluate their effects on fish populations by integrating field data on multiple processes. In STUDY SITE addition, models can be used as a decision-support tool by The location chosen for this demonstration application of evaluating alternative restoration scenarios to determine actions inSALMO is the Lower Clear Creek Flood Plain Restoration that best achieve the fundamental objectives of a project, such Project, Shasta County, California. Clear Creek flows generally as maximizing fish population response (Honea et al. 2009; east from the coastal mountain range into the Sacramento River Beechie et al. 2010; Stewart-Koster et al. 2010). Alternative and has a watershed area of approximately 650 km2.Baseflows scenarios can include type of management action as well as site at the site are mostly instream releases from Whiskeytown Dam, selection, recognizing that the spatial context of a restoration part of a large interbasin transfer system. Lower Clear Creek site may have a large effect on the project outcome (Bernhardt is a productive Chinook Salmon spawning stream (Yoshiyama and Palmer 2011). Use of models to evaluate completed projects et al. 2000); Whiskeytown Dam limits upstream passage but or potential future restoration actions can encourage projects to provides reliable base flows and moderate temperatures. The be process-based, have quantifiable objectives, and be designed creek supports runs of steelhead O. mykiss irideus and fall, for modification through adaptive management. late-fall, and spring runs of Chinook Salmon. Fall-run Chinook The use of models in monitoring and evaluation of restoration Salmon dominate in our study sites; adults arrive mostly in projects has long been advocated. In fact, modeling is an integral October and spawn from October through November, with most part of adaptive management, as originally envisioned (Walters juveniles migrating downstream as presmolts in their first spring 1986; Walters and Holling 1990). While many managers think and summer (Earley et al. 2010). of statistical models that are fit to field data as the natural frame- The lower end of Clear Creek was heavily disturbed by gravel work for adaptive management, simulation models have also mining such that much of the channel became narrow and steep Downloaded by [Department Of Fisheries] at 00:07 20 March 2013 been used in especially complex and prominent situations (e.g., with hard clay substrate instead of gravel. Several major actions Walters et al. 2000; Korman et al. 2011). Individual-based mod- have been taken to improve salmon and steelhead habitat. From els (IBMs) are especially attractive for evaluation and adaptive 1995 through 2001, flow releases from Whiskeytown Dam were management of restoration projects because they predict how gradually increased from a median of 1.9 m3/s to 6.1 m3/s, with meaningful population responses (e.g., number of successful a consequent decrease in mean annual water temperature of offspring) arise from habitat effects that are most clear at the 2.5◦C. Since 1996, spawning gravel has been injected in several individual level and can integrate the cumulative effects of the places for natural transport downstream. In 2000, a small dam multiple kinds of change made in restoration. Individual-based at river kilometer (RKM) 10.4 (upstream from the confluence models can combine models of (1) the physical environment with the Sacramento River) was removed. and how it is affected by restoration, (2) relevant physiological The Flood Plain Restoration Project (“restoration”) that and behavioral processes strongly affected by the environment we focused on was constructed in 2002 under the Central (e.g., feeding and growth, survival, spawning), and (3) key adap- Valley Project Improvement Act, which was intended to reduce tive behaviors such as selecting habitat and deciding when and effects of U.S. Bureau of Reclamation water projects on where to spawn. Individual-based models can be quite complex, Central Valley fisheries. Restoration design objectives included which makes them nontrivial to build and apply, but they also reestablishing a system of alternating riffles, exposed bars, and 386 RAILSBACK ET AL.

pools, establishing an active floodplain and coarse sediment MODEL DESCRIPTION transport, raising the bed to provide substrate of gravel and The model developed for this study is version 1.0 of in- cobble instead of the underlying hard clay, and increasing the SALMO, adapted from an unpublished full-life cycle salmon area of gravel suitable for spawning (NSR et al. 2000). This model and the inSTREAM family of stream trout models. The restoration project was an attractive study location because it inSTREAM models have been used for a variety of theoreti- offered contrasting sites with and without channel restoration cal and management applications (e.g., Railsback and Harvey and because the U.S. Fish and Wildlife Service (USFWS) (e.g., 2002, 2011; Railsback et al. 2003; Harvey and Railsback 2007, Earley et al. 2010) has extensively monitored variables such as 2009). We provide an overview of the model here, focusing spawner abundance, juvenile habitat use, and the redd location on processes that have not been published previously; a full and out-migrant data we describe below. description is provided by Railsback et al. (2012). We contrasted two study sites. The degraded site (designated “DEGRD”) was one of four used to represent prerestoration Model Purpose habitat (site 3C of Gard 2006). It is 460 m in length and has The model inSALMO is designed to help predict and explain a mean wetted width of 14 m at base flow. The restored site the interacting, cumulative effects of river management actions (designated “RESTO”; Figure 1) was the first in-channel phase on freshwater life stages of anadromous salmonids. Such actions of restoration (site 3A of USFWS 2006). Prior to its restoration include changes in flow, temperature, or turbidity regimes, alter- in 2002, the channel at RESTO had more bends and a lower ation of channel shape and availability of spawning gravel and gradient than DEGRD, but was similarly lacking in hydraulic cover for feeding or hiding, and manipulation of biological con- complexity and shallow habitat. The restoration work recon- ditions such as piscivorous fish densities and food availability. toured parts of the channel, added structures such as log root wads to stabilize banks and provide cover, and added spawning Entities, State Variables, and Scales gravel. The RESTO site is 490 m long and 27 m wide at base Version 1.0 of inSALMO represents habitat and fish in flow. streams where adult salmon arrive from the ocean and spawn, Downloaded by [Department Of Fisheries] at 00:07 20 March 2013

FIGURE 1. RESTO (Lower Clear Creek Flood Plain Restoration Project site 3A). This 2011 image (obtained from Google Earth) shows features of the 2002 restoration work: a wider and more sinuous channel, log structures for stability and habitat complexity in the bend at lower left, widespread spawning gravel, and a recontoured and replanted ﬂoodplain. [Figure available in color online.] MODELING DEGRADED VS. RESTORED SALMON HABITAT 387

eggs and embryos incubate, and juveniles rear until they migrate specified as a habitat reach parameter and has not changed more downstream. The model can represent one or several species or than 20% from the previous day (so spawning does not occur races (“runs”) of salmon, each with their own parameter values. when scour or dewatering are likely), and (5) a random number The model uses a daily time step. is less than a parameter we set to 0.2 (imposing a natural delay Habitat is represented at two scales: reaches and cells. A between arrival and spawning). However, adults always spawn “reach” represents a contiguous length of stream, typically sev- if it is the last day in their range of spawning dates. eral hundred meters in length. Simulations can represent one or When a female adult spawns, it first identifies the best cell more reaches, where fish move among them. Reach-scale habi- for spawning within a radius determined by its length (in this tat variables are daily flow, temperature, turbidity, and several study, this radius typically includes an entire reach). (Adults parameters that determine how high flows affect spawning and are not allowed to move to a different reach to spawn because the probability of redd scour. the number of spawners in each reach is an initial condition Habitats within each reach are depicted as two-dimensional of the model, not an outcome of simulations.) The “best cell” (depth-averaged) cells. Cells are irregular polygons, typically has the highest value of a spawning suitability measure that generated via GIS or hydraulic simulation software. Cells have considers depth, velocity, and area of spawning gravel not variables for depth and velocity, calculated from flow using currently guarded by another female. lookup tables generated using separate, standard, hydraulic The spawning female creates a redd in the selected cell, and models. The daily availability of drift food for fish is calculated the number of eggs are a power function of spawner length from depth, velocity, and cell area. Availability of a second type (Healey and Heard 1984). The female then identifies a male to of food (for “search feeding,” e.g., from the benthos) depends also spawn. Both spawners then have their weight reduced by on cell area. Availability of these two food types also depends 40% (Mesa and Magie 2006), which causes spawners to die on parameters estimated via calibration (see Model Calibration of poor condition (below) within a few days (median lifespan and Evaluation section below). Cells also have static habitat after spawning = 7 d). The female’s spawning status variable is variables that represent the amount of cell area that provides changed to “guarding.” velocity shelters for drift-feeding and spawning gravel, and a Adult habitat selection.—Female adults guarding a redd stay characteristic distance to hiding cover. These static variables in the cell containing the redd. Male and unspawned female are typically evaluated via field observation. The assumption adults move to the cell offering the highest probability of sur- that velocity shelter area (e.g., provided by boulders, habitat vival (below), typically one with high depth (to reduce predation structures, and banks) is independent of flow is an important risk) and low velocity (to reduce energy costs and hence starva- simplification made to avoid additional complexity and uncer- tion risk, as adults do not feed). tainty. Shelter area is typically observed during typical base Juvenile habitat selection and out-migration.—Juvenile flows, so this simplification should be reasonable most of the salmon select and move to the cell within the radius in which time, but perhaps not during high-flow events. they are assumed able to sense habitat conditions offering the Salmon are represented by three kinds of model entity. highest “expected fitness,” a tradeoff between expected future “Adults” have state variables for sex and spawning status growth and survival (Railsback et al. 1999; Railsback and Har- (whether they have spawned and, for females, whether they vey 2002). The radius increases with fish length and always are guarding a redd) and length, which determines fecundity of includes at least the adjacent cells. females. “Redds” represent the nests of eggs, with a variable Version 1.0 of inSALMO adds out-migration as a habitat for the number of eggs remaining alive each day. “Juveniles” selection alternative for juvenile salmon: if none of the available represent fish from emergence to out-migration, with variables cells offer expected fitness higher than the expected fitness from Downloaded by [Department Of Fisheries] at 00:07 20 March 2013 for length and weight. out-migration, the juvenile abandons its reach and moves into the next downstream reach. When a juvenile decides to move Process Overview and Schedule downstream from the downstream-most reach, it is recorded as The following actions are executed each simulated day. having migrated out of the system and removed from the model. Habitat updates.—Daily flow, temperature, and turbidity are A juvenile’s expected fitness from out-migration is represented read in for each reach. Cell depth, velocity, and food availability simply as “relative out-migration success,” a logistic function of variables are updated from flow. length (Figure 2) that reflects how the likelihood of successful Adult arrival.—Any adult salmon scheduled to arrive from migration to the ocean varies with length. Very small juveniles migration (see Initialization section below) are placed in their out-migrate only if expected fitness in their reach is very low, spawning reach. especially due to negative growth; as they grow, predation and Adult spawning.—Adults decide whether they spawn and, other risks become more important. Larger juveniles remain in if so, create a redd. An adult spawns if (1) it has not already their reach only if growth is positive and survival probability is spawned, (2) the date is within a range specified by parameters high (Figure 3). for each species or race, (3) the temperature is within a range Growth.—The daily change in weight is calculated from food specified by parameters, (4) the flow is below a maximum intake and energy costs using a standard bioenergetics approach, 388 RAILSBACK ET AL.

FIGURE 2. Relative out-migration success function for Chinook Salmon juveniles. The function’s shape is controlled by two parameters: the lengths at which a juvenile’s expected probability of surviving out-migration, relative to the maximum probability, is 0.1 (fishOutmigrateSuccessL1, 5.0 cm) and 0.9 (fishOutmigrateSuccessL9, 12.0 cm). FIGURE 3. Results of the out-migration decision method with fishOutmi- grateSuccessL1 equal to 5 cm and fishOutmigrateSuccessL9 at 12 cm. White then added to the fish’s previous weight. Because adults do not regions indicate combinations of growth and risk conditions under which juve- feed, their change in weight is always at least slightly negative. nile Chinook Salmon remain in their current reach, and grey regions indicate For juveniles, if the new weight exceeds the individual’s previ- when they migrate downstream. The x-axis is the daily probability of surviving ous maximum weight then an increase in length is based on an factors other than starvation (e.g., predation, high temperature). The y-axis is daily growth rate as grams of growth per gram of fish weight. The four panels observed length–weight relation. show how results depend on fish length. Fish were assumed to currently be in Survival.—Juveniles and adults are exposed to a variety of good condition; lower condition (weight at length) slightly increases the ten- mortality risks that depend on both habitat and fish variables. dency to migrate downstream. At lengths ≤5 cm, fish migrate downstream only Mortality due to poor condition (starvation and disease) be- if growth is negative. As length increases, out-migration becomes more depen- comes more likely as fish lose weight, which (for juveniles) dent on risk; by 9 cm, fish migrate downstream if growth is more than slightly negative or if survival is not greater than about 0.995. By 11 cm, juveniles often results from conditions such as low food intake and ex- out-migrate unless survival is very high and growth is positive. cessive velocities that cause high energetic costs. Juveniles are at risk of predation by other fish; this risk is lower if they use and Murray (1990). At constant temperatures, this model pre- habitat that is shallow or close to cover. All but the smallest ju- dicts eggs will develop fully in 94 d at 10◦C and in 57 d at veniles are also at risk of predation by terrestrial animals (e.g., 15◦C. birds, mammals), which can be reduced by selecting habitat that Redd emergence.—When a redd’s eggs are fully developed, is deeper or closer to hiding cover. Downloaded by [Department Of Fisheries] at 00:07 20 March 2013 they “emerge” as new juveniles over several days. Juvenile Redd survival.—Survival of redds is modeled by determining length is drawn from a uniform distribution ranging from 3.5 to how many eggs die each day due to dewatering (when a redd’s 4.1 cm fork length. cell is not submerged at the daily flow), scour and deposition (a function of flow, gravel size, and channel geometry), and Initialization of the Model temperatures either above or below optimal. Superimposition is At the start of an inSALMO simulation, no fish or redds modeled as a stochastic event that becomes possible when a redd are present. Instead, a number of adult salmon are created and is no longer guarded by its female and another female spawns given dates at which they “arrive from migration” by being in the same cell; the probability increases as the cell’s area of added to the model. Input for each species or race at each reach unguarded spawning gravel decreases. Females neither prefer specifies the number of arriving adults, their sex distribution, nor avoid superimposing redds. (inSALMO does not simulate the distribution from which individual lengths are drawn, and effects of fine sediment on redd survival; doing so would sub- the distribution from which individual arrival dates are drawn. stantially add to its complexity and uncertainty. Fine sediment is not believed to be a significant problem at Clear Creek.) Superindividuals Redd development.—The daily increment in egg develop- The “superindividual” technique (section 7.9 of Grimm and ment is calculated from temperature using model 4 of Beacham Railsback 2005) is used to make simulation of dense spawning MODELING DEGRADED VS. RESTORED SALMON HABITAT 389

populations computationally feasible. The model parameter vations (and hence predicted depths) were generally accurate juvenileSuperindividualRatio determines how many juveniles within 0.03 m. are represented by each juvenile object in the model. When Habitat cells.—We delineated habitat cells using GIS, start- redds emerge, the number of new juvenile objects is calculated ing with aerial photography of the site overlaid with the River2D by dividing the number of live eggs by juvenileSuperindividual mesh and habitat observations. Cell vertices were selected to Ratio. The consumption of food and velocity shelter area of capture important variation in habitat while producing no more each juvenile object is then multiplied by juvenileSuperindi- cells than necessary. Consequently, cells tended to be small vidualRatio. No other changes in juvenile rules or behavior where habitat varied sharply with distance (especially along were made to implement superindividuals in the model. When banks) and large in more homogeneous areas. This process pro- a juvenile object “dies,” all the fish it represents die. duced 552 cells at the more complex RESTO site and 162 cells To determine the extent to which using superindividuals af- at DEGRD. fects simulation results, we ran a sensitivity analysis on juvenile- Cell hydraulic and habitat variables.—To model depth and SuperindividualRatio, varying it from 1 (no superindividuals) to velocity from daily flow, inSALMO input includes lookup tables 100. Values of 10–20 were found to have negligible effects on of depth and velocity versus flow for each cell. The calibrated the kinds of outputs we analyze, while greatly increasing com- River2D models were used to simulate 24 flows at RESTO and putation speed. Here, we use a value of 20 individuals per object. 27 at DEGRD, ranging from below the lowest to above the highest recorded in Clear Creek during the period we modeled. Results of these hydraulic simulations were used to calculate METHODS depth and velocity in each inSALMO cell at each flow. Because Our application of inSALMO to fall-run Chinook Salmon the inSALMO cells are generally much bigger than the River2D in Clear Creek had three steps: (1) assembling input for two mesh elements and laid out so that hydraulics are relatively study sites representing habitat with and without restoration, (2) uniform within cells, we calculated depth and velocity of a cell calibrating and validating the model against field observations, as the simple mean (not weighted by area) of values from all and (3) designing analyzing simulation experiments to identify River2D nodes within the cell. the differences between sites predicted by the model and why Cell-specific input on velocity shelters for drift feeding, dis- they occur. tance to hiding cover, and area of spawning gravel was developed from field observations made March 23–25 and May 5, 2010. Input Data Assembly Velocity shelters and microhabitats providing hiding cover were Hydraulic simulation.—Hydraulic modeling methods for the relatively rare. We made point measurements of distance to DEGRD site are given in Gard (2006); similar methods were cover and water velocity throughout both reaches in locations used for RESTO. The River2D two-dimensional hydrodynamic that allowed us to capture sharp gradients in these variables. In model (Steffler and Blackburn 2002) was used to simulate how some areas, such as along stream margins where riparian veg- depth and velocity vary over space and with flow. Transects etation provided both hydraulic complexity and cover for fish, placed at the top and bottom of each site provided an upstream point measurements were spaced every few meters and each water-surface elevation for calibration of River2D, the down- point defined the velocity shelter and cover availability for a stream stage–discharge relationship input to River2D, and part specific cell. Points were less dense in midchannel areas, where of the site bed topography. Additional bed topography plus sub- we made measurements specifically where rare habitat features strate and cover data were collected using a total station, gen- provided velocity shelter or cover. We precisely located points erally in sets of points going across the channel. Points were with a Topcon Realtime Kinematic GPS (model R5). Spawning Downloaded by [Department Of Fisheries] at 00:07 20 March 2013 placed at changes in slope and to capture changes in substrate gravel was mapped onto aerial photos of the sites. Cover values type and cover-related habitat. Bed topography data were col- for each cell were assigned in GIS as the field-observed value lected in 2004 for RESTO and in 1999 for DEGRD. Bed to- nearest the cell center; spawning gravel values were assigned by pography, bed roughness, and substrate and cover distribution overlaying cells with the field-derived maps of gravel. Channel data were entered into River2D to create hydraulic models for boundaries were also located during the 2010 field observations each site. River2D was calibrated by adjusting bed roughness and compared with the River2D mesh in GIS to verify that heights until simulated water-surface elevations matched mea- channel shape had not changed substantially at either site. sured elevations at the upstream transect. Hydrodynamic model Adult numbers and characteristics.—Spawner input speci- predictions of stream velocity can be assessed by regressing fies the number, characteristics, and timing of adult salmon them against observed velocities (e.g., Booker et al. 2001). For arriving to spawn. Numbers of fall-run adults for each site this purpose, a minimum of 50 velocity measurements were were calculated from annual California Department of Fish collected per site in addition to those collected at the transects. and Game (CDFG) spawning escapement estimates for approx- Results (USFWS 2005, 2006) showed strong correlation, indi- imately 6.75 km of lower Clear Creek, which included both cating that velocity simulations are sufficient to represent the model sites. Adult numbers were produced by multiplying the differences between sites that we address. Water-surface ele- total escapement estimate by the proportion of the overall reach 390 RAILSBACK ET AL.

length represented by each model site. Input for adult sizes and The model was calibrated by systematically varying five pa- sex ratio were generated from unpublished carcass data provided rameters that were particularly uncertain and affected the timing by CDFG. and size of out-migrants: drift food concentration (habDrift- Conc), search food production (habSearchProd), relative survival of predation by fish (mortFishAqPredMin) and terrestrial Model Calibration and Evaluation animals (mortFishTerrPredMin), and the out-migration success After assembling input, we evaluated the model’s predic- function (fishOutmigrateSuccessL1). The model was run for wa- tions of where salmon spawn by comparing simulated versus ter years 2007–2009 using 360 combinations of these parame- observed redd locations, and calibrated its predictions of out- ters. The parameter combinations were evaluated by how often migrant timing and size by adjusting a few key parameters. (out of the 3 years) they met four criteria derived from the RST Redd locations.—Locations used for spawning have been data: (1) the number of out-migrants with length > 5 cm should observed at the study sites by mapping the extent of gravel that be above 10,000 per year from both sites, (2) out-migration appeared disturbed by spawners during and after fall Chinook should continue through at least June 1, (3) the date on which Salmon spawning in years including 2007–2010 (Giovannetti mean out-migrant length first exceeds 5 cm should be after April et al. 2008). These observed redd maps were overlain in GIS with 15, and (4) the maximum daily mean out-migrant length should the boundaries of inSALMO cells and then compared visually be between 6.5 and 8 cm. with redd locations in the inSALMO simulations. This approach allowed us to determine whether adults in the model spawned in Simulation Experiment Design and Analysis the same general areas and the kinds of habitat that real spawners We simulated spawning, incubation, juvenile rearing, and used, but has several uncertainties. First, the field observations out-migration of fall-run Chinook Salmon for five water years: are of a different type than the model results: the model predicts 2004–2008. The analysis looked at differences between the two how many redds are in each cell, not the exact location or extent sites throughout these life stages, examining the number of of redds, whereas the field observations attempt to delineate eggs produced, egg survival, and the fate of juveniles—survival, the extent of one or more redds. Second, there are potential growth, and timing of out-migration. When differences between errors in measuring redd extents and overlaying them with cell sites were found, differences in habitat availability and use were boundaries. Third, relatively high-quality spawning habitat is examined for explanations. widespread at the study sites (especially RESTO), so exactly In many analyses we distinguished between total out- which cell an adult spawns in is partially stochastic (in the model migrants and out-migrants >5 cm in length (referred to simply and probably in reality). Therefore, we did not expect simulated as “large out-migrants”). Because the model assigns juvenile redd locations to exactly match observations, but this analysis lengths of 3.5–4.1 cm upon emergence from redds, the 5-cm could illuminate any major errors in simulation of spawning threshold is useful for distinguishing between juveniles that location. moved downstream almost immediately after emergence, most- Out-migrant timing and size.—The main data-set available probably because they failed to find habitat providing positive for calibration of inSALMO to the Clear Creek sites are rotary growth (Figure 3, top left panel), and those that found productive screw trap (RST) data collected by the USFWS (e.g., Earley et al. habitat and grew at least a little before deciding to move down- 2010) downstream from the study sites. The RST at RKM 2.7 stream. Large numbers of Chinook Salmon fry moving down- is operated continuously during the juvenile Chinook Salmon stream (voluntarily or not) immediately after emergence have out-migration period. Combined with mark–recapture studies been observed at many sites (Healey 1991), but the relatively few conducted routinely to estimate trap efficiency, measurements fry that grow significantly before out-migration may contribute Downloaded by [Department Of Fisheries] at 00:07 20 March 2013 of individual juveniles captured in the RST were used to estimate disproportionately to adult returns. Miller et al. (2010) observed time series of numbers and size distributions of out-migrants. that 80% of the fish in a sample of Central Valley Chinook The RST data had important limitations. First, the data were Salmon adults were longer than 5.5 cm when they encountered collected from the capture of out-migrating fish from the entire salt water as juveniles. This evidence is not surprising in Cali- creek, not just our model sites; therefore, the RST data could fornia’s Central Valley: a few kilometers downstream from our not be used to distinguish between the study sites. Addition- sites, out-migrants encounter increasingly large and warm water ally, it was not clear how patterns in RST data were affected by bodies where productive feeding conditions are probably rarer differences between our two sites and the rest of the reach pro- and predation risks greater. ducing fry sampled by the trap (up to 4 km upstream from the model sites). Second, Clear Creek’s spring and late-fall runs of “Limiting Factors” Analysis Chinook Salmon could not be clearly distinguished in the RST The inSALMO model includes a “limiting factors tool” that data, making it difficult to discern a beginning or end of the automates analysis of how sensitive the simulated production fall-run out-migration that we simulated. Still, these data pro- of out-migrants is to a variety of factors that could be affected vided a useful view of how the number and size of out-migrating by habitat management. These factors are: base flow, food juveniles varied over time. availability, winter and summer temperature, spawning gravel MODELING DEGRADED VS. RESTORED SALMON HABITAT 391

availability, velocity shelter for drift feeding, hiding cover, 4. These values are: habDriftConc = 0.001 g/m3, habSearch- piscivory risk, redd scour, and the number of spawners. The tool Prod = 0.008 g/m2, mortFishAqPredMin = 0.94, mortFish sets up and executes simulations that vary these factors over TerrPredMin = 0.98, and fishOutmigrateSuccessL1 = 5.0 cm. ranges selected by the user and summarizes the response of Using these calibrated parameter values, the model’s out- key outputs such as the number of total and large out-migrants. migration timing was compared with the RST data by overlaying The simulation experiments consider parameter uncertainty graphs of weekly numbers of out-migrants in the (1) RST and by running each limiting factor scenario multiple times using (2) model results (Figure 5). Peaks in out-migration correspond combinations of values for a few especially uncertain parame- well, typically occurring in mid-February for both the model ters. We used the limiting factors tool as a way of identifying and the RST. However, out-migration from the model reaches and understanding effects of the habitat restoration and other begins later and ends earlier than out-migration at the RST. potential management actions. Model out-migration begins in mid-January while the RST data reports early December as the beginning of fall Chinook Salmon RESULTS out-migration. Model out-migration becomes rare by the end of June, while in some years the RST continues to catch small Model Calibration and Evaluation numbers of Chinook Salmon identified as fall-run fish as late as Redd locations.—The comparison of simulated versus ob- September. served redd locations indicates that the virtual adults in in- The size of simulated out-migrants was compared with the SALMO typically place redds in the same general locations RST data (Figure 6). The model generally predicts the start of as real salmon (Figure 4). The areas that real fish clearly avoid out-migration by large juveniles to be one to several weeks ear- (e.g., the large bend at the west end of RESTO, both ends of lier than observed in the screw traps. The model reproduces the DEGRD) are also avoided in the model. The areas of high model range of out-migrant sizes relatively well, which is not surpris- redd density, especially at RESTO, are similar to the large areas ing: the smallest out-migrant size is determined by parameters of observed redds. However, simulated redds at DEGRD are for fish size at emergence, which were based on observations more widely distributed than the observed redds. of real Chinook Salmon fry, and the largest out-migrant size Out-migrant timing and size.—Of the four criteria developed was considered in calibration. However, the model also closely from RST data for calibration of out-migration timing and size, reproduces patterns in out-migrant size that are not closely im- the first, second, and fourth were robustly met across wide ranges posed by model parameters: how out-migrant size (1) is constant of parameter values. However, the third criterion was rarely met; at a low value for a number of weeks, then (2) rises sharply for out-migrants > 5 cm length typically appeared one or several a few weeks, (3) levels off or even dips in April, and finally (4) weeks earlier in the model than in the RST data. One set of pa- continues to increase as the last few out-migrants leave. rameter values was selected as best meeting all four criteria and being typical of the many combinations that met criteria 1, 2, and Simulation Experiments This section presents results of simulation experiments contrasting the degraded and restored sites. All statistics and figures present simulation output, not field observations. Egg production and survival.—The inSALMO results indicated that spawning and incubation were equally successful at the two sites. At both sites and in all years, almost all (>94%) Downloaded by [Department Of Fisheries] at 00:07 20 March 2013 females produced redds. Survival to emergence was relatively constant among years and sites (across years, the mean and range at RESTO was 46% and 41–50%, respectively, and at DEGRD was 48% and 39–59%). Superimposition was the only major source of egg mortality, causing 50% loss (range among years, 44–54%) at RESTO and 48% loss (range, 41–50%) at DEGRD. Egg mortality due to excessively low or high temperature and associated disease was similar between sites because both were FIGURE 4. Comparison of observed and simulated redd locations for an exposed to the same temperatures. No eggs were lost to either de- example year, 2009. Observed redd extents are indicated by their curved bound- watering or scour; this was due in part to the upstream reservoir, aries. Simulated redd locations are indicated by the ovals placed randomly in so extreme high or low flows are rare. While our simulations did the redd’s cell, and cells are shaded progressively darker as depth increases not include scour-inducing flows, the redd scour function in in- (darker = deeper, white = not submerged). Redds extending outside the polyg- onal model cells are probably indicators of error in observing or overlaying SALMO (from Haschenburger 1999) predicted that widespread locations, or of changes in channel shape since the model’s topographic data scour should be much less frequent at RESTO than at DEGRD, were collected in 2004. [Figure available in color online.] because of RESTO’s wider and less-steep channel. 392 RAILSBACK ET AL.

FIGURE 5. Out-migration timing results for water years 2008 and 2009. The y-axis is the number of out-migrants per week either observed at the RST and determined to be fall Chinook Salmon (wide white bars), or in the simulation (grey bars). Simulated out-migrants are separated by their reach of origin, indicated by dark (RESTO) and light (DEGRD) grey.

The similarity between sites in superimposition rates may at DEGRD (20% of mortalities versus 9% at RESTO). The seem unexpected because one purpose of the restoration project number of juveniles that died at RESTO was much higher than at site RESTO was to increase the availability of high-quality at DEGRD (43,500 versus 14,500 over the entire simulation) spawning habitat. However, the model input indicated that such simply because far more fish remained at RESTO for more than habitat was almost equally available at the two sites at the time 1–2 d. it was observed in the field. At a typical winter base flow of We further addressed the question of why RESTO produced 7.5 m3/s, the simulated area of spawning gravel with suitability many more large out-migrants by turning on inSALMO’s op- Downloaded by [Department Of Fisheries] at 00:07 20 March 2013 > 0.5 for both depth and velocity was 2,900 m2 at RESTO and tional output file that reported the habitat characteristics and 2,500 m2 at DEGRD, or 5.8 and 5.5 m2/m reach length. Because fish fitness variables for each fish, each day, after the fish had the channel at DEGRD is much narrower than at RESTO, this selected the best cell available to it. We analyzed these results means a higher percentage of area is suitable spawning habitat for all juveniles; the vast majority of these were newly emerged at DEGRD. fry unsuccessfully seeking productive habitat near the cell con- Juvenile fates.—At the juvenile and out-migrant stages, the taining their redd and, for site DEGRD, fish entering the reach simulations produced large differences between sites. The two at its upstream end after migrating downstream from RESTO. sites produced approximately the same total number of out- This output allows examination of the habitat conditions fish migrants per spawner, but the number of large out-migrants was were able to find in their reach and the growth rates and sur- much higher at the restored site, RESTO (Figure 7). Of the vival probabilities they experienced as a consequence. These relatively few juveniles that died before migrating downstream, results emerged from (1) the model’s assumption that fish seek, >95% of mortality was due to three causes. Predation by fish over a distance limited by their size, habitat offering positive and terrestrial animals accounted for most mortality (88% of growth and high survival, and (2) simulated habitat charac- mortalities at RESTO and 75% at DEGRD). Poor condition teristics (depth, velocity, velocity shelter, hiding cover) near (due to rapid or persistent weight loss) was twice as common where juvenile salmon emerge from redds. For example, the MODELING DEGRADED VS. RESTORED SALMON HABITAT 393

FIGURE 6. Out-migrant size calibration and validation results. The x-axis is the start of the week over which data are averaged. The y-axis is mean length of out-migrating Chinook Salmon. For four example years (2000, 2003, 2004, 2008), RST data are compared with calibrated simulation results.

distribution of cell depths and velocities used by juveniles (Fig- Another potential explanation for the higher production of ure 8) indicate that many juveniles, especially at DEGRD, were large out-migrants at site RESTO is simply that RESTO is up- not able to find moderate velocities (where growth should be stream from DEGRD. Before being recorded as out-migrants, highest). At DEGRD, more juveniles used shallow depths where juveniles spawned in RESTO move downstream through DE- risk of being eaten by other fish was lower. GRD, where they could find productive habitat and grow; juve- Compared with fish in the degraded reach, juvenile salmon in niles spawned in DEGRD do not have another reach downstream RESTO occupied habitat with higher levels of velocity shelter to provide a second opportunity for growth prior to emigration. and hiding cover. At RESTO, juveniles used velocity shelter To test this explanation, we ran the experiment with sites re-

Downloaded by [Department Of Fisheries] at 00:07 20 March 2013 to reduce their swimming speed 20% of the time, compared versed; a major reduction in large out-migrants from RESTO with only 4% at DEGRD. Many more juveniles occupied cells when it is downstream would conﬁrm the explanation. Putting with distance to hiding cover low enough to reduce predation RESTO downstream from DEGRD did reduce the production of risk at RESTO. Median distance to hiding cover was 1.5 m at large out-migrants from RESTO but only slightly, from 0.95% RESTO, compared with 4.0 m at DEGRD; inSALMO’s survival to 0.83% of all out-migrants. Production of large out-migrants parameters assume this distance must be less than ∼1mto spawned at DEGRD more than tripled—from 0.14% to 0.48% reduce risk substantially. of all out-migrants—when they migrated through RESTO. This These habitat differences produced substantial differences result corresponds with the explanation that differences in juve- between sites in juvenile growth and survival potential (Fig- nile growth are due to habitat differences between the sites, not ure 9). Even though differences between sites in net energy their relative locations. intake (Figure 9, left panels) appear small, the percentage of simulated juveniles experiencing positive net energy intake (weight Limiting Factors Analysis gain instead of loss) was 38% at RESTO and only 18% at The limiting factor analysis indicated that the total number DEGRD. Juveniles more often obtained high survival probabil- of out-migrants at DEGRD could be affected by the availability ities (e.g., >0.98) at RESTO (Figure 9, right panels). of spawning gravel (Figure 10, upper panel). While addition of 394 RAILSBACK ET AL.

more than the baseline (actual) gravel area (x-axis > 0) had little benefit, reducing gravel area strongly reduced total out-migrant production at DEGRD. The mechanism for this effect was higher loss of eggs to superimposition when gravel is below baseline levels. The response of large-out-migrant production to gravel availability was small but opposite that of total out-migrants (Figure 10, lower panel). The number of large out-migrants was most affected by food availability. The simulated response to food availability typi- fies the response to factors expected to have unambiguously positive effects on juvenile survival and growth (availability of velocity shelter and hiding cover, in addition to food). These factors had essentially no effect on total numbers of out-migrants (Figure 11, upper panel) because the vast majority of juveniles left within a day or two after emergence. However, they clearly increased the numbers of large out-migrants at both sites (Fig- ure 11, lower panel). This positive effect (of velocity shelter and hiding cover, as well as food availability) was much more beneficial at RESTO because of the restored site’s higher productivity of large juveniles. For example, a 100% increase over baseline food availability increased the percentage of out-migrants that were large by about 100% at RESTO and 190% at DEGRD, but the number of large out-migrants increased by 6,800 per year at RESTO and only 1,400 at DEGRD. Even though the mass of additional food was the same between sites, the higher area of FIGURE 7. Number of total (top panel) and large (length > 5 cm; bottom good feeding habitat at RESTO allowed much more of the food panel) Chinook Salmon out-migrants in the habitat restoration analysis, by to be captured and turned into growth by juvenile salmon. water year and site. Downloaded by [Department Of Fisheries] at 00:07 20 March 2013

FIGURE 8. Distributions of cell depth (left panels) and velocity (right panels) for juvenile Chinook Salmon occupying the restored site RESTO (top panels) and degraded site DEGRD (bottom panels). The y-axis is the fraction of juveniles occupying cells with depth in the x-axis bin range (0–20, 20–40 . . . up to 180–200). The grey vertical line represents the median occupied depth or velocity. MODELING DEGRADED VS. RESTORED SALMON HABITAT 395

FIGURE 9. Differences between sites in Chinook Salmon juvenile fitness variables; format as in Figure 8. Left panels: histograms of daily net energy intake, i.e., the difference between energy intake from food and expenditure on metabolism and swimming; fish only grow if this variable remains above zero. Right panels: daily probability of surviving mortality sources other than starvation. Low survival probability values (e.g., <0.95, at which expected lifespan is less than 14 d) often resulted from occupying cells with a velocity above the fish’s maximum sustainable swimming speed (median survival was 0.013 at RESTO and 2.0 × 10−6 at DEGRD).

DISCUSSION and sizes of fish) that are of direct management relevance and testable against field observations. Mechanistic models such as Modeling in Habitat Restoration Design and Evaluation inSALMO are currently the only tools we have for forecasting Evaluation of river habitat restoration projects is important the benefits of alternative restoration designs (and even basin- but complex and difficult (Bernhardt et al. 2005). Conclusive wide restoration strategies) that affect multiple factors such evidence for the value of restoration projects, or relative value as physical habitat, food and cover availability, temperature, of alternative restoration designs, is very difficult to develop and flow.

Downloaded by [Department Of Fisheries] at 00:07 20 March 2013 from field studies alone because of the many interacting pro- While we studied a completed restoration project, models cesses and uncontrollable variables. Detailed, mechanistic mod- like inSALMO can also be used to support restoration planning els such as inSALMO contain many uncertainties that should decisions. Analyses like our “limiting factors” experiments can not be neglected, and the credibility of the model’s absolute be useful for prioritizing management actions; our analysis, for predictions is not yet clear. However, these models are, at the example, indicated that further modifying base flow or tem- least, useful for (1) providing a comprehensive framework for perature either up or down would have small or even negative identifying, documenting, and linking information and assump- effects. For channel reconstruction projects, simulation exper- tions, (2) exploring the consequences of such information and iments can both compare alternative designs and evaluate the assumptions, (3) producing hypotheses for what processes and relative benefits to juvenile salmon production of details such variables are important that can then be tested in the field, and as providing more gravel versus more hiding and feeding cover. (4) investigating relative differences among alternative manage- This is in contrast to other approaches and guidelines for design- ment actions. Habitat assessment approaches, whether based ing restoration projects (e.g., the spawning habitat restoration on habitat selection (e.g., PHABSIM; Bovee 1982) or feeding design process of Wheaton et al. 2004, and other approaches and energetics (e.g., Hayes et al. 2007), are not as useful for reviewed by Wheaton et al. 2004), which do not predict such such purposes, largely because they do not predict kinds of specific biological responses. Previous approaches predict, at results (e.g., timing of emergence and out-migration, numbers most, changes in “suitable” habitat area. 396 RAILSBACK ET AL.

FIGURE 10. Limiting factors analysis results for spawning gravel. The x-axis is the percentage by which spawning gravel availability was changed from the observed value for each sensitivity scenario. The y-axis values are the mean of nine simulations using all combinations of low, medium, and high values of two FIGURE 11. Limiting factors analysis results for availability of food for ju- uncertain parameters: food energy density and fishOutmigrateSuccessL1.The venile Chinook Salmon. The format is similar to that of Figure 10, except that upper panel reports total number of Chinook Salmon out-migrants. The lower both sites are represented on the left y-axis of the lower panel. panel reports large (length > 5 cm) out-migrants, with separate scales for sites RESTO (left y-axis) and DEGRD (right y-axis). input data collection in 2010 than it was in previous years, Our continuing research objectives include testing inSALMO perhaps due to the ongoing injection of gravel as part of the Clear against more kinds of data collected at Clear Creek and evalu- Creek restoration program, (2) spawners being less attracted to ating how results depend on the total area and number of sites. DEGRD, violating the assumption we used in preparing input, The goal of this research is to determine whether and how the that spawner density is constant over stream length, or (3) factors model can be used to support large-scale watershed restoration such as intragravel flow that affect spawning habitat quality but decisions. are not in the model. The model also predicted out-migration to occur over a nar- Evaluation of the inSALMO Application to Clear Creek rower range of dates than indicated by the RST data. In in- Downloaded by [Department Of Fisheries] at 00:07 20 March 2013 Our inSALMO simulations reproduced several key observa- SALMO, the timing of out-migration for the vast majority of tions that were not imposed by assumptions or parameter val- juveniles who leave as newly emerged fry is strongly deter- ues. Most important is the observation from RST data that the mined by the input defining when adults arrive and spawn and vast majority of juvenile Chinook Salmon move downstream as by the equation for how egg development varies with water newly emerged fry instead of staying and growing before out- temperature, which are all relatively certain. Hence, these dis- migration. Emergence of other patterns in out-migrant size and crepancies probably result from the RST collecting out-migrants timing in the model give us some confidence in its predictions from much more stream length and more kinds of habitat than of growth and out-migration. represented by the two inSALMO reaches, and from uncertain- Some model results did not closely match monitoring data. ties in distinguishing fall Chinook Salmon from other runs in Especially at site DEGRD, the model predicted redds to be the trap data. distributed over more area than that observed. This difference One seemingly counterintuitive result from the limiting fac- could be reduced by adjusting the parameters for redd defense tors analysis was that further addition of spawning gravel could area or adult mortality (so spawners die sooner and, hence, reduce the production of large out-migrants. The most likely defend redds for less time). However, the difference could also mechanism explaining this result is that more gravel produces result from (1) spawning gravel being more widespread during more emergent fry and, hence, greater competition and lower MODELING DEGRADED VS. RESTORED SALMON HABITAT 397

growth. We commonly see negative relations between abun- duction of small, newly emerged, juvenile salmon versus pro- dance and growth in inSALMO, due to competition for good duction of juveniles that survive and grow before migrating feeding sites. downstream. Our simulations indicated that widening and re- This application also identified several potential effects of shaping the channel and (from the limiting factors analysis) restoration that the current version of inSALMO does not ad- increasing the availability of food, velocity shelter, and hiding dress. One is the effect of gravel quality on both spawning cover all had little effect on production of small out-migrants habitat selection and redd survival. Defining useful, quantita- but strong positive effects on the number of large out-migrants. tive measures of gravel quality and modeling them is complex Increasing spawning gravel availability increased production of and uncertain, and we simply chose not to add these uncer- small out-migrants while having small and sometimes nega- tainties to the model. Second, we did not model how habitat tive effects on the number of large ones. If evidence that large differences among reaches affect which reach adults spawn in; out-migrants are more important (e.g., Miller et al. 2010) is instead, we simply specified how many adults used each reach. confirmed or accepted, it will be increasingly important for A straightforward approach for adding this process is to as- restoration project design to consider juvenile rearing habitat. sume no inherent “site fidelity” at the reach scale and let adults Jeffres and Moyle (2012) recently expressed concern about de- select the best available habitat among multiple reaches when signing habitat projects to enhance spawning success without they spawn. Third, many of our key results concern the ability giving adequate consideration to juvenile life stages. of newly emerged fry to find and occupy productive and safe A second general conclusion is that, in our simulation ex- habitat. The inSALMO model does not represent this process periments, a variety of factors affected restoration benefits— in detail; it neglects processes such as being swept downstream especially the number of large out-migrants—in a generally as passive drift and changes in swimming ability as fish com- multiplicative way, instead of there being a clear “limiting fac- plete their development from alevin to fry stages. The model’s tor.” We did not find just one process or variable that limited predicted number of large out-migrants is in fact sensitive to out-migrant production, but instead several factors—food avail- its parameters controlling the radius over which fish can find ability, velocity shelter, hiding cover—that each have strong habitat (Railsback et al. 2012, section 4.4.2). effects. Such results are typical of our individual-based fish models that include behaviors trading off feeding and preda- Contrast of Degraded and Restored Sites tion avoidance (e.g., Harvey and Railsback 2007; Railsback and This study did not directly evaluate the restoration project Harvey 2011). This conclusion implies it is not always nec- by simulating the RESTO site before and after restoration, but essary, or even possible, to identify a “limiting factor” before our comparison of RESTO to the degraded site provides useful implementing a successful habitat improvement project. inferences. The analysis indicates that RESTO was much more Finally, our limiting factors analysis supports the conclusion productive for larger juvenile Chinook Salmon than was the that total production of salmon may be improved the most if en- unrestored DEGRD. hancement efforts are concentrated in areas of especially good Why did the model predict that far more fry remain and habitat instead of dispersed among many unproductive areas. grow at RESTO than at DEGRD? One reason is simply that the This analysis indicated that further enhancements, such as in- RESTO site is much larger in area than DEGRD; the restora- creasing food production or cover availability, would produce tion project produced a channel nearly twice as wide as that independent and multiplicative increases in production of large at DEGRD at normal flows, with velocities 40–50% lower. A out-migrants. When multiple habitat factors have independent, wider, slower channel no doubt provides more area where newly multiplicative effects, the benefits of improving those factors are emerged fry can feed productively. However, higher availability highest where juvenile production is already high. For example, Downloaded by [Department Of Fisheries] at 00:07 20 March 2013 of velocity shelter and hiding cover at RESTO no doubt con- an addition of hiding cover that reduces predation mortality by tributed to higher retention, survival, and growth of juveniles. 10% will save more fish at a site producing 1,000 fish than at (We did not evaluate the potential for the wider, slower channel one producing 100 fish (assuming that there is not extreme com- to result in higher temperatures.) petition for escape cover during predation events). Creation of Differences between sites in total production of juvenile Chi- “hot spots” with high habitat quality for juvenile feeding and nook Salmon were small, in part because spawning gravel was predator avoidance as well as spawning may be an effective also abundant at DEGRD (perhaps due to upstream gravel in- salmon restoration technique. jection as part of the restoration program) and because most juveniles migrate downstream soon after emerging from the redd. The redd scour function in inSALMO indicates that the ACKNOWLEDGMENTS restored site is considerably less vulnerable to redd mortality This work was funded by the U.S. Bureau of Reclamation due to high flows. under the Central Valley Project Improvement Act, and by the USFWS. Claire Hsu was the initial project manager. Dan Cox, General Conclusions for Habitat Restoration USFWS, provided the initial motivation and direction. Soft- One important general conclusion of this study is that restora- ware and GIS support were provided by Steve Jackson, Charles tion actions can have different, even opposing, effects on pro- Sharpsteen, Colin Sheppard, and Diane Sutherland Montoya, 398 RAILSBACK ET AL.

and by Jeff Anderson, Corin Pilkington, and Bonnie Pryor of Jahnig,¨ S. C., A. W. Lorenz, D. Hering, C. Antons, A. Sundermann, E. Jedicke, Northern Hydrology and Engineering (McKinleyville, Califor- and P. Haase. 2011. River restoration success: a question of perception. Eco- nia). We especially thank Matt Brown, James Earley, and Sarah logical Applications 21:2007–2015. Jeffres, C., and P. Moyle. 2012. When good fish make bad decisions: Coho Giovannetti of the USFWS Red Bluff Office for the great deal Salmon in an ecological trap. North American Journal of Fisheries Manage- of time, data, and knowledge they contributed to this work. ment 32:87–92. Korman, J., S. J. D. Martell, and C. Walters. 2011. Describing population dynamics for early life stages of Rainbow Trout (Oncorhynchus mykiss) REFERENCES using a stock synthesis model. Canadian Journal of Fisheries and Aquatic Beacham, T. D., and C. B. Murray. 1990. Temperature, egg size, and develop- Sciences 68:1110–1123. ment of embryos and alevins of five species of Pacific salmon: a comparative Mesa, M. G., and C. D. Magie. 2006. Evaluation of energy expenditure in adult analysis. Transactions of the American Fisheries Society 119:927–945. spring Chinook Salmon migrating upstream in the Columbia River basin: Beechie, T. J., D. A. Sear, J. D. Olden, G. R. Pess, J. M. Buffington, H. Moir, P. an assessment based on sequential proximate analysis. River Research and Roni, and M. M. Pollock. 2010. Process-based principles for restoring river Applications 22:1085–1095. ecosystems. BioScience 60:209–222. Miller, J. A., A. Gray, and J. Merz. 2010. Quantifying the contribution of juvenile Bernhardt, E. S., and M. A. Palmer. 2011. River restoration: the fuzzy logic of migratory phenotypes in a population of Chinook Salmon Oncorhynchus repairing reaches to reverse catchment scale degradation. Ecological Appli- tshawytscha. Marine Ecology Progress Series 408:227–240. cations 21:1926–1931. NSR (North State Resources), McBain and Trush, and Matthews and Associates. Bernhardt, E. S., M. A. Palmer, J. D. Allan, G. Alexander, K. Barnas, S. Brooks, 2000. Lower Clear Creek floodway rehabilitation project: channel recon- J. Carr, S. Clayton, C. Dahm, J. Follstad-Shah, D. Galat, S. Gloss, P.Goodwin, struction, riparian vegetation, and wetland creation design document. Report D. Hart, B. Hassett, R. Jenkinson, S. Katz, G. M. Kondolf, P. S. Lake, R. Lave, to the Clear Creek Restoration Team, NSR, Redding, California, McBain J. L. Meyer, T. K. O’Donnell, L. Pagano, B. Powell, and E. Sudduth. 2005. and Trush, Arcata, California, and Matthews and Associates, Weaverville, Synthesizing U.S. river restoration efforts. Science 308:636–637. California. Available: www.srwp.org/documents/watershed/lowerclearcreek/ Booker, D. J., D. A. Sear, and A. J. Payne. 2001. Modelling three-dimensional restfish/LwrClearCreek FloodwayRehabProject08-30-00.pdf. (June 2012). flow structures and patterns of boundary shear stress in a natural pool–riffle Railsback, S. F., and B. C. Harvey. 2001. Individual-based model formulation for sequence. Earth Surface Processes and Landforms 26:553–576. Cutthroat Trout, Little Jones Creek, California. U.S. Forest Service General Bovee, K. D. 1982. A guide to stream habitat analysis using the instream flow Technical Report PSW-GTR-182. incremental methodology. U.S. Fish and Wildlife Service FWS/OBS-82/26. Railsback, S. F., and B. C. Harvey. 2002. Analysis of habitat-selection rules Earley, J. T., D. J. Colby, and M. R. Brown. 2010. Juvenile salmonid monitoring using an individual-based model. Ecology 83:1817–1830. in Clear Creek, California, from October 2008 through September 2009. U.S. Railsback, S. F., and B. C. Harvey. 2011. Importance of fish behaviour in Fish and Wildlife Service, Red Bluff Fish and Wildlife Office, Red Bluff, modelling conservation problems: food limitation as an example. Journal of California. Available: www.fws.gov/redbluff/cvpia.aspx. (June 2012). Fish Biology 79:1648–1662. Gard, M. 2006. Modeling changes in salmon spawning and rearing habitat Railsback, S. F., B. C. Harvey, S. K. Jackson, and R. H. Lamberson. 2009. associated with river channel restoration. International Journal of River Basin InSTREAM: the individual-based stream trout research and environmental Management 4:201–211. assessment model. U.S. Forest Service General Technical Report PSW-GTR- Giovannetti, S. L., M. R. Brown, and L. A. Stafford. 2008. Fall Chinook Salmon 218. spawning area mapping for the Clear Creek restoration program, 2007. U.S. Railsback, S. F., B. C. Harvey, and J. L. White. 2012. InSALMO version Fish and Wildlife Service, Red Bluff Fish and Wildlife Office, Red Bluff, 1.0: model improvements and demonstration application to Chinook Salmon California. Available: www.fws.gov/redbluff/cvpia.aspx. (June 2012). spawning, incubation, and rearing in Clear Creek, California. Final Report Grimm, V., and S. F. Railsback. 2005. Individual-based modeling and ecology. to U.S. Bureau of Reclamation, Lang, Railsback and Associates, and U.S. Princeton University Press, Princeton, New Jersey. Forest Service, Redwood Sciences Laboratory, Arcata, California. Avail- Harvey, B. C., and S. F. Railsback. 2007. Estimating multi-factor cumulative wa- able: www.fws.gov/sacramento/Fisheries/Instream-Flow/fisheries instream- tershed effects on fish populations with an individual-based model. Fisheries flow inSalmo.htm. (June 2012). 32:292–298. Railsback, S. F., R. H. Lamberson, B. C. Harvey, and W. E. Duffy. 1999. Move- Harvey, B. C., and S. F. Railsback. 2009. Exploring the persistence of stream- ment rules for individual-based models of stream fish. Ecological Modelling dwelling trout populations under alternative real-world turbidity regimes with 123:73–89.

Downloaded by [Department Of Fisheries] at 00:07 20 March 2013 an individual-based model. Transactions of the American Fisheries Society Railsback, S. F., H. B. Stauffer, and B. C. Harvey. 2003. What can habitat 138:348–360. preference models tell us? tests using a virtual trout population. Ecological Haschenburger, J. K. 1999. A probability model of scour and fill depths in Applications 13:1580–1594. gravel-bed channels. Water Resources Research 35:2857–2869. Roni, P., K. Hanson, and T. Beechie. 2008. Global review of the physical and Hayes, J. W., N. F. Hughes, and L. H. Kelly. 2007. Process-based modelling of biological effectiveness of stream habitat rehabilitation techniques. North invertebrate drift transport, net energy intake and reach carrying capacity for American Journal of Fisheries Management 28:856–890. drift-feeding salmonids. Ecological Modelling 207:171–188. Steffler, P., and J. Blackburn. 2002. River2D—two-dimensional depth averaged Healey, M. C. 1991. Life history of Chinook Salmon (Oncorhynchus model of river hydrodynamics and fish habitat: introduction to depth averaged tshawytscha). Pages 311–394 in C. Groot and L. Margolis, editors. Pacific modeling and user’s manual. University of Alberta, Edmonton. Available: salmon life histories. University of British Columbia Press, Vancouver. www.river2d.ualberta.ca/download.htm. (June 2012). Healey, M. C., and W. R. Heard. 1984. Inter- and intra-population variation Stewart-Koster, B., S. E. Bunn, S. J. MacKay, N. L. Poff, R. J. Naiman, and in the fecundity of Chinook Salmon (Oncorhynchus tshawytscha) and its P. S. Lake. 2010. The use of Bayesian networks to guide investments in flow relevance to life history theory. Canadian Journal of Fisheries and Aquatic and catchment restoration for impaired river ecosystems. Freshwater Biology Sciences 41:476–483. 55:243–260. Honea, J. M., J. C. Jorgensen, M. M. McClure, T. D. Cooney, K. Engie, D. USFWS (U.S. Fish and Wildlife Service). 2005. Monitoring of restora- Holzer, and R. Hilborn. 2009. Evaluating habitat effects on population status: tion projects in Clear Creek using 2-dimensional modeling method- influence of habitat restoration on spring-run Chinook Salmon. Freshwater ology. USFWS, Sacramento Fish and Wildlife Office, Sacramento, Biology 54:1576–1592. California. MODELING DEGRADED VS. RESTORED SALMON HABITAT 399

USFWS (U.S. Fish and Wildlife Service). 2006. Phase 3A Clear Creek restora- Walters, C. J., J. Korman, L. E. Stevens, and B. Gold. 2000. Ecosystem mod- tion: 2-d modeling. USFWS, Final Report, Sacramento Fish and Wildlife Of- eling for evaluation of adaptive management policies in the Grand Canyon. fice, Sacramento, California. Available: www.fws.gov/sacramento/Fisheries/ Conservation Ecology [online serial] 4(2):article 1. Instream-Flow/fisheries instream-flow reports.htm. (June 2012). Wheaton, J. M., G. B. Pasternack, and J. E. Merz. 2004. Spawning habitat Walters, C. J. 1986. Adaptive management of renewable resources. Macmillan, rehabilitation: I. conceptual approach and methods. International Journal of New York. River Basin Management 2:3–20. Walters, C. J., and C. S. Holling. 1990. Large-scale management experiments Yoshiyama, R. M., E. R. Gerstung, F. W. Fisher, and P. B. Moyle. 2000. Chinook and learning by doing. Ecology 71:2060–2068. Salmon in the California Central Valley: an assessment. Fisheries 25(2):6–20. Downloaded by [Department Of Fisheries] at 00:07 20 March 2013 This article was downloaded by: [Department Of Fisheries] On: 20 March 2013, At: 00:09 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 An Evaluation of Liquid Ammonia (Ammonium Hydroxide) as a Candidate Piscicide David L. Ward a , R. Morton-Starner a & Shaula J. Hedwall b a U.S. Geological Survey, Southwest Biological Science Center, Grand Canyon Monitoring and Research Center, 2255 North Gemini Drive, Flagstaff, Arizona, 86001, USA b U.S. Fish and Wildlife Service, Arizona Ecological Services, 2500 South Pine Knoll Drive, Flagstaff, Arizona, 86001, USA Version of record first published: 19 Mar 2013.

To cite this article: David L. Ward , R. Morton-Starner & Shaula J. Hedwall (2013): An Evaluation of Liquid Ammonia (Ammonium Hydroxide) as a Candidate Piscicide, North American Journal of Fisheries Management, 33:2, 400-405 To link to this article: http://dx.doi.org/10.1080/02755947.2013.765528

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 33:400–405, 2013 American Fisheries Society 2013 ISSN: 0275-5947 print / 1548-8675 online DOI: 10.1080/02755947.2013.765528

MANAGEMENT BRIEF

An Evaluation of Liquid Ammonia (Ammonium Hydroxide) as a Candidate Piscicide

David L. Ward* and R. Morton-Starner U.S. Geological Survey, Southwest Biological Science Center, Grand Canyon Monitoring and Research Center, 2255 North Gemini Drive, Flagstaff, Arizona 86001, USA Shaula J. Hedwall U.S. Fish and Wildlife Service, Arizona Ecological Services, 2500 South Pine Knoll Drive, Flagstaff, Arizona 86001, USA

The continued decline of native fish populations in the Abstract southwestern USA is largely attributed to interactions with in- Eradication of populations of nonnative aquatic species for the vasive aquatic species, and repeated studies have demonstrated purpose of reintroducing native fish is often difficult because very the inability of native fishes to persist in environments where few effective tools are available for removing aquatic organisms. This creates the need to evaluate new chemicals that could be used nonnative fish have become established (Marsh and Pacey as management tools for native fish conservation. Ammonia is a 2005). Warmwater nonnative species, such as Smallmouth Bass natural product of fish metabolism and is naturally present in the Micropterus dolomieu, Largemouth Bass M. salmoides,Com- environment at low levels, yet is known to be toxic to most aquatic mon Carp Cyprinus carpio, Green Sunfish Lepomis cyanellus, species. Our objective was to determine the feasibility of using Fathead Minnow Pimephales promelas, Red Shiner Cyprinella liquid ammonia as a fisheries management tool by evaluating its effectiveness at killing undesirable aquatic species and its persis- lutrensis, Mosquitofish Gambusia affinis, Black Bullhead tence in a pond environment. A suite of invasive aquatic species Ameiurus melas, Channel Catfish Ictalurus punctatus, Flathead commonly found in the southwestern USA were introduced into Catfish Pylodictis olivaris, tadpoles of the American bullfrog two experimental outdoor ponds located at the Rocky Mountain Lithobates catesbeianus, or northern crayfish Orconectes Research Station in Flagstaff, Arizona. Each pond was treated with virilis, dominate a majority of the ponds and streams of the ammonium hydroxide (29%) at 38 ppm. This target concentration was chosen because previous studies using anhydrous ammonia desert Southwest. Once established, these invasive species reported incomplete fish kills in ponds at concentrations less than flourish, ultimately jeopardizing the existence of native fish 30 ppm. Water quality was monitored for 49 d to determine how populations (Minckley and Marsh 2009).

Downloaded by [Department Of Fisheries] at 00:09 20 March 2013 quickly the natural bacteria in the environment converted the am- Attempts to eradicate nonnative aquatic species and rein- monia to nitrate. Ammonia levels remained above 8 ppm for 24 troduce native species are common, but success is often and 18 d, respectively, in ponds 1 and 2. Nitrite levels in each pond began to rise approximately 14 d after dosing with ammonia and limited because very few effective tools are available for manag- stayed above 5 ppm for an additional 21 d in pond 1 and 18 d in ing invasive aquatic species (Marking 1992; Dawson and Kolar pond 2. After 49 d all water in both ponds was drained and no ﬁsh, 2003). The only currently licensed and commercially available crayﬁsh, or tadpoles were found to have survived the treatment, general piscicide, rotenone, is undergoing increased scrutiny in but aquatic turtles remained alive and appeared unaffected. Liq- some states because of new data linking occupational rotenone uid ammonia appears to be an effective tool for removing many problematic invasive aquatic species and may warrant further in- use in agricultural settings to Parkinson’s disease (Tanner vestigation as a piscicide. et al. 2011). Even though there are substantial differences between the methods of application, formulation, exposures, and

*Corresponding author: [email protected] Received October 26, 2012; accepted January 2, 2013

400 MANAGEMENT BRIEF 401

dosages of rotenone used in agricultural versus fish management environment at low levels (ATSDR 2004). Naturally occurring applications (RRAC 2011), misunderstanding and fear about bacteria common in the environment detoxify ammonia through Parkinson’s disease may limit future use of rotenone as a fish a process commonly known as the nitrification cycle. Nitrify- management tool. This creates an urgent need to evaluate new ing bacteria (e.g., Nitrosomonas spp.) oxidizes ammonia (NH3 + − candidate piscicides that could be developed for use as addi- and NH4 ) into nitrite (NO2 ) and then nitric bacteria (e.g., − tional tools in native fish conservation. Nitrospira spp.) converts nitrite into nitrate (NO3 ), which is Ammonia is well known for its toxicity to fish (Randall subsequently utilized by plants. Ammonia can also volatilize to and Tsui 2002), and the toxic effects of ammonia to aquatic the atmosphere, be rapidly taken up directly by aquatic plants, organisms are well documented in the published literature or be adsorbed into sediments and organic materials (ATSDR (USEPA 1999, 2009). Ammonia is considered highly toxic 2004), and so, although the toxic effects of ammonia on aquatic on a scale of acute chemical toxicity to fish (96-h LC50 = organisms are well established from laboratory studies, the ef- 0.1–1.0 ppm) (Meyer and Barclay 1990). Wallen et al. (1957) fectiveness of applying liquid ammonia to ponds for the purpose rated ammonia seventh out of the 86 chemicals tested in of removing unwanted aquatic organisms is unknown. We eval- toxicity to mosquitofish. In aquaculture, ammonia commonly uated the use of liquid ammonia as a fish management tool for kills fish if biological filtration is inadequate (Piper et al. killing undesirable species and monitored its persistence in the 1986; Hargreaves and Tucker 2004). Ammonia has also been environment in a seminatural pond setting. responsible for a number of large-scale fish kills in natural rivers and lakes, as the result of chemical spills from tanker trucks or industrial processing plants (Brunson 1985; Meade METHODS 2004), yet it has largely been overlooked as a potential fisheries Two small, lined ponds located at the U.S. Forest Service management tool. Several researchers evaluated anhydrous Rocky Mountain Research Station in Flagstaff, Arizona, were ammonia for use in fisheries management in ponds in the used for this study. Pond 1 volume was 223,000 L (14.8 m wide, late 1960s and early 1970s and concluded that it was very 8.2 m long, and 2.2 m deep) and pond 2 volume was 141,000 L effective at removing unwanted organisms at concentrations (13.6 m wide, 11.4 m long, and 1.1 m deep). Ponds were filled above 30 ppm (Klussmann et al. 1970; Champ et al. 1973; with class A + reclaimed water from the city water treatment Prentice et al. 1978), but as rotenone became widely available, facility 14 d before the introduction of any fish. A suite of with more commercial formulations for aquatic applications, invasive species were captured from streams and ponds near investigations into alternative piscicides largely ceased. Flagstaff using trammel nets, seines, and angling (Table 1). All Ammonia is generally considered a contaminant and most fish were measured (TL, mm) and introduced into the ponds 30 studies have focused on reducing or preventing its harmful ef- d prior to dosing with ammonia to allow the newly introduced fects on the environment (ATSDR 2004), but ammonia also has fish and natural bacterial colonies to become established. Den- qualities that may make it a good tool for invasive species man- itrification bacteria will typically establish in new tanks within agement. At high concentrations (>2 ppm), ammonia can be 30 d at water temperatures above 20◦C (Timmons and Ebeling toxic to a wide variety of organisms (USEPA 1999, 2009); how- 2007). At the start of the experiment a total of 363 fish and 110 ever, ammonia is the natural byproduct of fish metabolism and crayfish were stocked into pond 1 and 372 fish and 110 crayfish decomposition of organic material and is naturally present in the were stocked into pond 2 (Table 1). Two hatchling striped

TABLE 1. Number and mean TL (mm; range in parentheses) of ﬁsh and other aquatic species stocked into each pond during the ammonia toxicity experiment.

Downloaded by [Department Of Fisheries] at 00:09 20 March 2013 Pond 1 Pond 2 Species Number Mean (range) TL (mm) Number Mean (range) TL (mm) Black Bullhead 20 233 (150–340) 20 230 (165–320) Channel Catfish 1 275 1 292 Flathead Catfish 1 371 1 290 Green Sunfish 97 90 (60–193) 79 96 (50–190) Largemouth Bass 4 261 (240–295) 5 264 (210–330) Smallmouth Bass 37 72 (50–110) 36 70 (50–95) Common Carp 11 329 (260–400) 13 339 (140–400) Fathead Minnow 72 65 (50–80) 92 65 (50–80) Red Shiner 55 65 (45–85) 55 64 (50–84) American bullfrog tadpoles 35 85 (20–176) 39 81 (20–157) Northern crayfish 110 55 (30–80) 110 54 (20–75) 402 WARD ET AL.

mud turtles Kinosternon baurii and two adult red-eared sliders Pond 1 Trachemys scripta elegans were also added to each pond to 40 2.5

allow an evaluation of the effects of ammonia on aquatic turtles. Ammonia Because of their small size, the hatchling mud turtles were Nitrate 2.0 placed in a mesh basket partially submerged in each pond so 30

they could easily be monitored. Red-eared sliders were allowed 1.5 to swim freely within each pond. Fish and turtles were fed ad 20 libitum every other day with Aquamax 6-mm ﬂoating pellets. 1.0

On September 15, 2011, ammonium hydroxide (29%, Uni- (mg/L) Nitrate Ammonia (mg/L) Ammonia

var Chemical) was added to both ponds at a dosage rate of 10 0.5 mL ammonium hydroxide per 3.78 L of water. A total of 0.5 29.5 L of ammonium hydroxide was added to pond 1 and 18.6 L of ammonium hydroxide was added to pond 2 to achieve ini- 0 0.0 tial calculated ammonia concentrations of 38 ppm. Ammonia 0 10203040 was poured directly from a bucket into each pond without mixing. Water quality was monitored daily for 5 d prior to dosing Pond 2 the ponds with ammonia until 45 d posttreatment. Dead ﬁsh 35 2.5

were collected daily and identiﬁed. Ammonia, nitrite, nitrate, 30 Ammonia and pH were tested with an API (Nessler reagent) aquarium Nitrate 2.0 test kit to obtain rapid visual assessments of water quality. Wa- 25 ter temperature was recorded every hour at the surface with a 1.5 20 Tidbit temperature logger. To obtain more precise measures of 15 ammonia and nitrate concentrations, a 60-mL water sample was 1.0 Nitrate (mg/L) Nitrate

also collected from each pond every other day, filtered through a (mg/L) Ammonia 45-µm nylon membrane and frozen for subsequent nutrient anal- 10 0.5 ysis at the Colorado Plateau Analytical Laboratory. A Lachat 5 QuikChem 8000 series spectrophotometer was used to analyze the ammonia and nitrate concentrations from each frozen sample 0 0.0 following standardized U.S. Environmental Protection Agency 0 10203040 (USEPA) methods (USEPA 1983). Analyses of eight samples of Days After Treatment ammonia calibration standard (0.20–45.0 ppm, Hach) and five FIGURE 1. Ammonia and nitrate concentrations measured with a spectropho- samples of nitrate standard (0.2–10.0 ppm, Hach) resulted in re- tometer for pond 1 (upper panel) and pond 2 (lower panel) for 40 d after treatment coveries between 93% and 100%, within the range specified for with ammonia. method accuracy, and five samples run in duplicate had relative deviations between 0% and 1%. which was lower than our calculated concentration of 38 ppm Nitrite is unstable and rapidly converts to nitrate when (Figure 1). Fish began dying within 20 min of dosing the ponds stored, so unfrozen water samples were also collected and with ammonia, although most fish were not visible until several analyzed immediately using spectrophotometry on September days later when they began to decompose and float to the water 30 and October 7, 2011, to obtain more precise estimates of surface. Dead individuals of all of the stocked species were Downloaded by [Department Of Fisheries] at 00:09 20 March 2013 maximum nitrite levels in each pond. After 45 d, 10 Green recovered with some dead specimens being recovered 15–18 d Sunfish were placed in a mesh basket in each pond and after treatment. Ammonia levels remained above 8 ppm (the monitored for 3 d to verify that the water was no longer toxic maximum our field test kit could measure) for 24 and 18 d in enough to kill Green Sunfish. On November 2, 2011, (49 ponds 1 and 2, respectively (Figure 2). Approximately 10–14 d after dosing the ponds with ammonia) both ponds were d after dosing each pond with ammonia, nitrite levels began to drained completely and all live organisms were recovered and rise and stayed above 5 ppm (also the maximum our field test kit counted. could measure) for an additional 21 and 18 d in ponds 1 and 2, respectively. On day 42 ammonia and nitrite levels had returned to pretreatment conditions in pond 2. Ammonia and nitrite had RESULTS still not returned to baseline levels in pond 1 at 45 d posttreat- Ammonia levels in both ponds before treatment (September ment and nitrate levels remained above 0.9 ppm (Figures 1, 2). 14, 2011) were less than 0.1 ppm (Figure 1). Water samples The pH in both ponds varied little during the study, averaging collected the day after treatment and analyzed using spectropho- 8.4 in pretreatment measurements and 8.7 after treatment. tometry revealed the actual initial concentrations of ammonia Water temperature in both ponds was similar, ranging from a in pond 1 and pond 2 were 29.8 and 33.9 ppm, respectively, daytime maximum of 25◦C at the beginning of the treatment MANAGEMENT BRIEF 403

Pond 1 DISCUSSION 10 Our study demonstrated that a single dose of ammonium Ammonia hydroxide (30–34 ppm) was sufficient to completely kill the Nitrite 8 aquatic organisms that we tested in experimental ponds, and ammonia and its degradation products persisted for less than 2 months. Our results are similar to other published studies con- 6 ducted in the 1960s and 1970s in which ammonia levels were artificially raised in ponds using anhydrous ammonia. Champ 4 et al. (1973) eradicated fish including Red Shiners, Black Bull- * heads, sunfishes, and Largemouth Bass from a pond of 0.72 sur-

Concentration (mg/L) Concentration * face hectares by dosing it with anhydrous ammonia at 37 ppm in 2 November. Klussmann et al. (1970) reported a complete kill of bullhead catﬁsh and Mosquitoﬁsh after dosing with anhydrous 0 ammonia at 20 and 40 ppm, respectively. In four ponds in Texas,

0 10203040 applications of 15 ppm anhydrous ammonia caused high mortality to catfish, sunfish, suckers, and minnows, but complete kills were not observed until concentrations reached 30 ppm Pond 2 10 (Prentice et al. 1978). In our study it took from 38 to 45 d for the nitrification Ammonia Nitrite cycle to complete and for ammonia, nitrite, and nitrate levels 8 to return to near pretreatment conditions. This is consistent with persistence times for ammonia when added to aquaculture

6 systems to establish and “cycle” bioﬁlters (Masser et al. * 1992; Timmons and Ebeling 2007). Klussmann et al. (1970) documented ammonia and nitrite concentrations reaching 4 nontoxic levels in ponds in less than 1 month during the summer in central Texas, while in winter ammonia persisted at

Concentration (mg/L) Concentration * 2 toxic levels for over 4 months (Champ et al. 1973). The bacteria that convert ammonia to nitrite and nitrate are temperature dependent. Average water temperatures remained between 0 10◦C and 15◦C during a majority of our study. These relatively 0 10203040cold water temperatures probably slowed down bacterial Days After Treatment growth and contributed to the persistence of ammonia. + The dissolution of ammonia in water creates ionized (NH4 ) FIGURE 2. Ammonia and nitrite concentrations measured with an API water and un-ionized (NH3) forms of ammonia. It is the un-ionized quality test kit in pond 1 (upper panel) and pond 2 (lower panel) for 45 d after ammonia that is most toxic to aquatic organisms because the treatment with ammonia. The API test kits cannot detect ammonia in excess of 8 ppm or nitrite in excess of 5 ppm. Asterisks indicate nitrite concentrations neutral molecule easily diffuses across gill membranes (USEPA measured using a spectrophotometer at 15 and 22 d after treatment. 1999). The proportion of un-ionized ammonia in water, and its

Downloaded by [Department Of Fisheries] at 00:09 20 March 2013 subsequent toxicity to organisms, increases with increasing pH and temperature (Emerson et al. 1975), but is determined primarily by pH. Toxicity results of studies conducted at different and dropping throughout the experiment to a nighttime low of temperatures and in ponds of different pH levels are therefore near 5◦C by October 31, with diurnal fluctuations of 5–8◦C. likely to vary. For example, an increase of one pH unit from 8 to An algae bloom began in each pond approximately 10 d 9 increases the amount of un-ionized ammonia approximately after dosing with ammonia and became very intense with a 10-fold (Piper et al. 1986). thick mat of algae completely covering both ponds by day 15. The USEPA has established criteria for maximum ammonia This algae bloom persisted in each pond for approximately 2 concentrations in surface water based on danger to fish. In most weeks. Pond water was no longer toxic enough to kill Green natural surface waters, ammonia concentrations greater than Sunfish in both ponds at 45 d as verified by sentinel sunfish, 2.0 ppm exceed the chronic exposure criteria for fish (Lewis which remained alive after being placed into each pond for 3 d and Morris 1986). Ammonia levels exceeded 2 ppm in each in a net-pen. After 49 d all water in both ponds was drained and pond for 18–24 d, and when ammonia levels began to decline to no fish, crayfish, or tadpoles were found to have survived the nonlethal levels, nitrite levels rose and stayed above 5 ppm for an treatment, but red-eared slider turtles and hatchling mud turtles additional 18–21 d (Figure 2). Nitrite is also highly toxic (Meyer remained alive and appeared unaffected. and Barclay 1990) to fish and oxidizes blood hemoglobin to 404 WARD ET AL.

methemoglobin, a form that is not capable of carrying oxygen, Management and removal of invasive aquatic species is often causing suffocation and death in fish (Piper et al. 1986). Un- difficult because of the large scale of the problem and the few ionized ammonia levels as low as 0.6 ppm can kill fish of some effective tools that are available. However, ammonia has many species in as little as a few days (Ogbonna and Chinomso 2010). qualities that may make it an effective piscicide. Ammonia may This combination of toxic chemicals that occur in succession as a be an effective tool for management of invasive aquatic species normal result of the nitrification process increases the likelihood in pond locations where rotenone treatments have previously that no aquatic species will survive an ammonia treatment. been attempted but were unsuccessful. Although licensing and Ammonia is highly soluble in water, reaching concentrations registration of new piscicides is very expensive (Marking 1992; of 1,000 ppm at 20◦C (Champ et al. 1973). This high solubility Dawson and Kolar 2003), the abundance of published informa- allows ammonia to spread throughout the water column without tion on toxicity of ammonia to humans (ATSDR 2004) as well spraying or mixing, which greatly facilitates application for as a wide variety of aquatic organisms (USEPA 1999) could fish management purposes. Liquid ammonia is relatively greatly speed up the licensing and registration process. Liquid inexpensive (US$200 for a 208-L [55 gal] drum) and readily ammonia therefore warrants further investigation as a candidate available, which makes its application on a large scale feasible, piscicide for use in fisheries management. although gaseous (anhydrous) ammonia may be even more cost effective for very large-scale applications (Prentice et al. ACKNOWLEDGMENTS 1978). Ammonia can kill snails and fish eggs (Wang et al. 2007; Oplinger and Wagner 2009), which could be advantageous We thank Alicia Burtner, Dave Foster, and Luke Avery for when a complete fish kill is the management goal, but also could assistance with collecting fish. Benjamin Moan performed water have disadvantages in locations with native snail or mussel sample analysis at the Colorado Plateau Analytical laboratory. species. Ammonia as an aquatic species management tool may Scott VanderKooi, Rob Clarkson, and two anonymous reviewers have additional management benefits such as control of aquatic provided helpful reviews of this manuscript. We thank the U.S. vegetation and pond fertilization to increase productivity Forest Service, Rocky Mountain Research Station in Flagstaff, (Ramachandran 1962; Klussmann et al. 1970). Ammonia is Arizona, for the use of their experimental ponds for these stud- already a USEPA-registered pesticide (Kegley et al. 2011), but ies, and the Southwest Biological Science Center for funding it is not labeled for aquatic use. However, extensive information support. The use of trade, firm, or corporation names in this is available on the toxicity of ammonia to aquatic organisms publication is for informational use only and does not constitute under a wide variety of water quality conditions (USEPA 1999), an official endorsement or approval by the U.S. Government. which could facilitate the licensing and registration process. Ammonia as a toxicant is unique because aquatic organisms have evolved mechanisms to passively diffuse ammonia from REFERENCES gill tissues. Most species of fish may not perceive ammonia ATSDR (Agency for Toxic Substances and Disease Registry). 2004. Toxicolog- ical profile for ammonia. U.S. Department of Health and Human Services, when these diffusion gradients are reversed (Richardson et al. ATSDR, Division of Toxicology/Toxicology Information Branch, Atlanta. 2001). When un-ionized ammonia is artificially spiked in the Available: www.atsdr.cdc.gov/toxprofiles/tp126.pdf. (October 2012). environment, reverse diffusive gradients cause the ammonia to Brunson, K. L. 1985. Summary of reported fish kills in Kansas during 1982. build up to toxic levels in gill tissue and blood (USEPA 1999; Transactions of the Kansas Academy of Science 88:71–83. Randall and Tsui 2002). We observed little erratic movement Champ, M. A., J. T. Lock, C. D. Bjork, W. G. Klussmann, and J. D. McCullough Jr. 1973. Effects of anhydrous ammonia on a central Texas pond, and a review and struggling of fish after treating ponds with ammonia, and of previous research with ammonia in fisheries management. Transactions of crayfish did not attempt to crawl out of the ponds, which is the American Fisheries Society 102:73–82. Downloaded by [Department Of Fisheries] at 00:09 20 March 2013 frequently a problem in attempts to control crayfish numbers. Dawson, V. K., and C. S. Kolar, editors. 2003. Integrated management tech- Use of ammonia as an aquatic species management tool also niques to control nonnative fishes. Completion Report to U.S. Bureau of has some disadvantages. Adding ammonia to natural systems Reclamation, U.S. Geological Survey, Upper Midwest Environmental Sci- ences Center, La Crosse, Wisconsin. provides additional nitrogen, which can stimulate algae blooms, Emerson, K., R. C. Russo, R. E. Lund, and R. V. Thurston. 1975. Aqueous as observed in this study. The lengthy natural detoxification pro- ammonia equilibrium calculations: effect of pH and temperature. Journal of cess may also be disadvantageous in circumstances where rapid the Fisheries Research Board of Canada 32:2379–2383. detoxification is desired. Because the proportion of toxic un- Hargreaves, J. A., and C. S. Tucker. 2004. Managing ammonia in fish ionized ammonia in water is highly dependent on pH it may ponds. Southern Regional Aquaculture Center, Publication 4603, Stoneville, < Mississippi. not be effective as a management tool under low pH ( 7) con- Kegley, S. E., B. R. Hill, S. Orme, and A. H. Choi. 2011. PAN pesticide ditions. Concentrated liquid ammonia and the fumes it creates database. Pesticide Action Network North America (PANNA), San Francisco. are dangerous to humans, and personal protective equipment Available: www.pesticideinfo.org. (October 2012). is needed for applicators. However, once dissolution occurs in Klussmann, W. G., M. A. Champ, and J. T. Lock. 1970. Utilization of anhy- water (at dosages necessary to act as a piscicide), there is little drous ammonia in fisheries management. Proceedings of the Southeastern Association of Game and Fish Commissioners 23(1969):512–519. health risk to humans or other mammals (Dawson and Kolar Lewis, W. M., Jr., and D. P. Morris. 1986. Toxicity of nitrite to fish: a review. 2003; ATSDR 2004). Transactions of the American Fisheries Society 115:183–195. MANAGEMENT BRIEF 405

Marking, L. L. 1992. Evaluation of toxicants for the control of carp and other Randall, D. J., and T. K. N. Tsui. 2002. Ammonia toxicity in fish. Marine nuisance fishes. Fisheries 17(6):6–13. Pollution Bulletin 45:17–23. Marsh, P. C., and C. A. Pacey. 2005. Immiscibility of native and non-native Richardson, J., E. K. Williams, and C.W. Hickey. 2001. Avoidance behaviour fishes. Pages 59–63 in C. L. Spring and S. Leon, editors. Proceedings of two of freshwater fish and shrimp exposed to ammonia and low dissolved oxygen symposia: Restoring native fish to the lower Colorado River: interactions of separately and in combination. New Zealand Journal of Marine and Fresh- native and non-native fishes; and Restoring natural function within a modi- water Research 35:625–633. fied riverine environment: the lower Colorado River. U.S. Fish and Wildlife RRAC (Rotenone Review Advisory Committee). 2011. Final report and Service, Southwestern Region, Albuquerque, New Mexico. recommendations to the Arizona Game and Fish Department. Arizona Game Masser, M. P., J. Rakocy, and T. M. Losordo. 1992. Recirculating aquaculture and Fish Department, Phoenix. Available: www.azgfd.gov/h f/documents/ tank production systems: management of recirculating systems. Southern Rotenone Review Advisory Committee Final Report 12 31 2011 version2. Regional Aquaculture Center, Publication 452, Stoneville, Mississippi. pdf. (October 2012). Meade, R. 2004. Fish and invertebrate recolonization in a Missouri prairie Tanner, C. M., F. Kamel, G. W. Ross, J. A. Hoppin, S. M. Goldman, M. Korell, stream after an acute pollution event. North American Journal of Fisheries C. Marras, G. S. Bhudhikanok, M. Kasten, A. R. Chade, K. Comyns, M. B. Management 24:7–19. Richards, C. Meng, B. Priestley, H. H. Fernandez, F. Cambi, D. M. Umbach, Meyer, F. P., and L. A. Barclay, editors. 1990. Field manual for the investiga- A. Blair, D. P. Sandler, and J. W. Langston. 2011. Rotenone, paraquat, and tion of fish kills. U.S. Fish and Wildlife Service, Resource Publication 177, Parkinson’s disease. Environmental Health Perspectives 119:866–872. Washington, D.C. Timmons, M. B., and J. M. Ebeling. 2007. Recirculating aquaculture, 2nd edi- Minckley, W. L., and P. C. Marsh. 2009. Inland fishes of the greater Southwest: tion. Northeastern Regional Aquaculture Center, Publication 01-007, Ithaca, chronicle of a vanishing biota. University of Arizona Press, Tucson. New York. Ogbonna, J. F., and A. A. Chinomso. 2010. Determination of the concentration USEPA (U.S. Environmental Protection Agency). 1983. Methods for chemical of ammonia that could have lethal effect on fish pond. ARPN Journal of analysis of water and wastes. USEPA, Office of Research and Development, Engineering and Applied Sciences 5(2):1–5. Report EPA-600/4-79/020, Washington, D.C. Oplinger, R. W., and E. J. Wagner. 2009. Toxicity of common aquacul- USEPA (U.S. Environmental Protection Agency). 1999. 1999 update of ambient ture disinfectants to New Zealand mud snails and mud snail toxicants water quality criteria for ammonia. USEPA, Office of Water, Report EPA- to Rainbow Trout eggs. North American Journal of Aquaculture 71:229– 822-R-99-014, Washington, D.C. 237. USEPA (U.S. Environmental Protection Agency). 2009. Draft 2009 update: Piper, R. G., I. B. McElwain, L. E. Orme, J. P. McCraren, L. G. Fowler, and J. R. aquatic life ambient water quality criteria for ammonia—freshwater. USEPA, Leonard. 1986. Fish hatchery management. U.S. Fish and Wildlife Service, Office of Water, Report EPA-822-D-09-001, Washington, D.C. Washington, D.C. Wallen, I. E., W. C. Greer, and R. Lasater. 1957. Toxicity to Gambusia affinis Prentice, J. A., P. P. Durocher, and D. L. Pritchard. 1978. Evaluation of anhy- of certain pure chemicals in turbid waters. Sewage and Industrial Wastes drous ammonia for fishery management uses. Proceedings of the Annual Con- 29:695–711. ference Southeastern Association of Fish and Wildlife Agencies 30(1976): Wang, N., C. G. Ingersoll, I. E. Greer, D. K. Hardesty, C. D. Ivey, J. L. Kunz, 88–98. W. G. Brumbaugh, F. J. Dwyer, A. D. Roberts, T. Augspurger, C. M. Kane, R. Ramachandran, V. 1962. The method and technique of using anhydrous am- J. Neves, and M. C. Barnhart. 2007. Chronic toxicity of copper and ammonia monia for aquatic weed control. Indo-Pacific Fisheries Council Proceedings to juvenile freshwater mussels (Unionidae). Environmental Toxicology and 10:146–153. Chemistry 26:2048–2056. Downloaded by [Department Of Fisheries] at 00:09 20 March 2013 This article was downloaded by: [Department Of Fisheries] On: 28 May 2013, At: 19:49 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 Fishery Management Complexes: An Impediment or Aid to Sustainable Harvest? A Discussion Based on the Northeast Skate Complex Fiona Hogan a , Steve Cadrin a & Alyssa Haygood b a University of Massachusetts, School of Marine Sciences, University of Massachusetts–Dartmouth, Department of Fisheries Oceanography, 200 Mill Road, Suite 325, Fairhaven, Massachusetts, 02719, USA b Arizona Western College, Environmental Science, 2020 South Avenue 8E, Yuma, Arizona, 85365, USA Published online: 29 Mar 2013.

To cite this article: Fiona Hogan , Steve Cadrin & Alyssa Haygood (2013): Fishery Management Complexes: An Impediment or Aid to Sustainable Harvest? A Discussion Based on the Northeast Skate Complex, North American Journal of Fisheries Management, 33:2, 406-421 To link to this article: http://dx.doi.org/10.1080/02755947.2013.763873

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 33:406–421, 2013 C American Fisheries Society 2013 ISSN: 0275-5947 print / 1548-8675 online DOI: 10.1080/02755947.2013.763873

ARTICLE

Fishery Management Complexes: An Impediment or Aid to Sustainable Harvest? A Discussion Based on the Northeast Skate Complex

Fiona Hogan* and Steve Cadrin University of Massachusetts, School of Marine Sciences, University of Massachusetts–Dartmouth, Department of Fisheries Oceanography, 200 Mill Road, Suite 325, Fairhaven, Massachusetts 02719, USA Alyssa Haygood Arizona Western College, Environmental Science, 2020 South Avenue 8E, Yuma, Arizona 85365, USA

Abstract There are seven species of skates (family Rajidae) found along the East Coast of the USA. All seven species are currently managed by the New England Fisheries Management Council as a single management complex extending from the Gulf of Maine to Cape Hatteras. The objective of the management plan is to ensure the long-term sustainability of fishing for each species via a trip limit approach. Two species are harvested in two distinct commercial fisheries. Northeast Fisheries Science Center trawl survey data and published literature were examined to investigate differences between the individual species in the skate complex. Each species exhibited a unique thermal and geographic range in addition to vital life history traits (e.g., age at maturity, longevity, and maximum size). Thorny Skate Amblyraja radiata and Smooth Skate Malacoraja senta have narrow thermal ranges and maintain a more northern distribution. Barndoor Skate Dipturus laevis have a moderate thermal habitat. Little Skate Leucoraja erinacea and Winter Skate L. ocellata have broad thermal ranges and are distributed throughout the management area. Limited inferences can be made about the thermal preferences of Clearnose Skate Raja eglanteria without data from south of Cape Hatteras, but they appear to have a broad thermal range within the management area. Rosette Skate L. garmani have a narrow thermal range and tend to be found in the deep offshore mid-Atlantic region. The validity of managing multiple distinct species in a complex is questioned. This example shows that a mixed-stock management strategy may be inadequate to meet the sustainability needs of each species and the associated fisheries. A management strategy focused on individual species may lead to a more efficient harvest of targeted species while allowing for the rebuilding of overfished species. Downloaded by [Department Of Fisheries] at 19:49 28 May 2013

Since the inception of the northeast skate complex fishery in accordance with 10 national standards. The national standards management plan (FMP) in 2003, the New England Fishery were designed to ensure a sustainable harvest from fisheries. The Management Council has managed all seven skate species as management plan is intended to conform to national standard 1, one stock complex. The management area covers the entire which defines a stock complex as northeast U.S. coast from the Gulf of Maine to Cape Hatteras a group of stocks in a FMP that are sufficiently similar in geographic (Figure 1). The objective of the plan is to ensure the long-term distribution, life history, and vulnerability to the fishery that the sustainability of each species via a trip limits approach. The impacts of management actions on the stocks in the complex is 2007 reauthorization of the Magnuson–Stevens Act requires similar. (DOC 2009) regional fisheries management councils to formulate and imple- The skate (family Rajidae) complex consists of seven species: ment annual catch limits and associated accountability measures Thorny Skate Amblyraja radiata, Smooth Skate Malacoraja

*Corresponding author: [email protected] Received November 8, 2011; accepted December 27, 2012

406 MANAGING THE U.S. NORTHEAST SKATE COMPLEX 407

obtained in the Northeast Fisheries Science Center (NEFSC) trawl surveys. Biological reference points are also based on NEFSC trawl survey biomass indices (NEFSC 2000) because fishery catch and biological data are uncertain and do not correspond to survey trends (NDPSWG 2009). The stock complex was assessed as part of a workshop on data-poor species during which several population models were applied to the available data, but all of the models were unsuccessful in describing the population dynamics of the skate complex (NDPSWG 2009). The skate FMP dictates the species composition of landings and discards, but the majority of landings remain reported as “unclassified” skates (NEFMC 2009). Species composition in fisheries is dependent on species size, geographic distribution, and management restrictions. Two of the seven skate species are targeted commercially, with the remaining five species being either discarded or outside the geographic range of the fisheries. FIGURE 1. Strata of the Northeast Fisheries Science Center’s bottom trawl The status of the multispecies resource varies latitudinally; land- survey, which extend from the Gulf of Maine to Cape Hatteras. The survey has ings of northern species (Thorny Skate and Smooth Skate) are a stratified random design. Strata 1–4 represent the mid-Atlantic region, strata 5–7 southern New England, strata 8–9 Georges Bank, and strata 10–11 the Gulf prohibited in the Gulf of Maine due to a low stock abundance of Maine. [Figure available in color online] proxy. Landings of Barndoor Skate are prohibited throughout their range. The allowable biological catch (ABC) that is set for senta, Barndoor Skate Dipturus laevis, Winter Skate Leucoraja the skate complex accounts for dead discards and landings and ocellata, Little Skate L. erinacea, Clearnose Skate Raja eglante- cannot be exceeded. The reliance on trip limits to reduce directed ria and Rosette Skate L. garmani. The skates in this complex fishing effort and discarding to reduce landings in all fisheries are oviparous and exhibit the same general life history strategy (direct and indirect) has resulted in the formation of a feedback (Collette and Klein-Macphee 2002; reviewed by Packer et al. loop for the landings; if the dead discards continue to comprise 2003a–g). However, differences exist in their biological char- a larger component of the ABC, the remaining portion for to- acteristics (e.g., age at maturity) and geographic distributions. tal allowable landings becomes smaller. This paper evaluates The spatial extent of skate distributions varies by species and whether the seven skate species conform to the stock complex may be determined by thermal preferences. Life history traits definition by inhabiting similar geographic and thermal regions vary by region because age and size at maturity tend to increase and having similar life history traits. Northeast Fisheries Science with increasing latitude and gestation period may be tempera- Center trawl survey data were used to construct cumulative den- ture dependent (Frisk and Miller 2006). sity functions of catch-weighted temperatures and a review of The skate wing fishery harvests Winter Skate that support a available life history traits to evaluate overlaps between species. frozen food export market. The bait fishery, which supplies the American lobster Homarus americanus trap industry, is dominated by Little Skate, with a small component of juvenile Winter METHODS Skate. The remaining five species contribute primarily to total Northeast Fisheries Science Center Trawl survey data from

Downloaded by [Department Of Fisheries] at 19:49 28 May 2013 discards. The status of skates as an underexploited, valueless 1963 to 2007 were examined to explore the appropriateness of species has changed in the last two decades. Landings of elas- managing the northeast skate complex as a whole under the mobranchs account for a large and increasing percentage of national standard 1 guidelines. The NEFSC trawl survey has a commercial fishery harvests in New England (NEFSC 2007). stratified random sampling design in which strata are defined Discards remain a large portion of the total catch, with an esti- by depth and location (Figure 1; Despres-Pantanjo et al. 1988). mated discard rate of 47.9% (NEFMC 2009). Only the Thorny The survey samples the defined skate management area from Skate is currently overfished and the Barndoor Skate is rebuild- the Gulf of Maine to Cape Hatteras. At each sample site the ing. The remaining five species are currently not overfished. species, bottom temperature, location, number of individuals, The reasons for these different population trends are not well weight, and month were recorded. These data were used to understood. Despite improvements in available life history data, describe the geographic distribution of each skate species. migration patterns are unknown and there are still large gaps in Catch-weighted temperatures (CWTs) were calculated by our understanding of what drives the skate populations. species for each season and compared with observed tempera- The northeast skate complex poses a series of challenges for tures using cumulative distribution functions (CDFs; Sokal and fishery management. The most significant impediment is the Rohlf 1995) to characterize each skate species’ seasonal envi- lack of reliable catch data and incomplete life history infor- ronment and thermal preferences. The CDFs were calculated mation for each species. Stock assessments are based on data as the cumulative relative frequencies of CWTs in 1-degree 408 HOGAN ET AL. Downloaded by [Department Of Fisheries] at 19:49 28 May 2013

FIGURE 2. Cumulative distribution functions (CDFs) showing the thermal range of the skate species in the northeast skate complex compared with all of the observed temperatures where the seven species were caught. The CDFs were calculated as the cumulative relative frequencies of CWTs in 1-degree temperature intervals in ascending order, ranging from 0 (for temperatures that are colder than those inhabited by skates) to 1 (for temperatures warmer than those inhabited by skates). MANAGING THE U.S. NORTHEAST SKATE COMPLEX 409

TABLE 1. Average catch-weighted temperatures (◦C) by season for the seven species in the northeast skate complex.

Skate species Winter Spring Summer Autumn Barndoor 5.99 6.22 8.78 10.05 Clearnose 10.39 9.47 11.63 19.31 Little 5.55 6.58 10.15 13.42 Rosette 10.19 11.72 10.56 12.58 Smooth 7.12 5.93 6.11 7.3 Thorny 6.25 5.3 6.16 7.74 Winter 5.16 6.97 10.06 13

all observed temperatures in each seasonal survey). Most tem- FIGURE 3. VonBertalanffy growth models fit to length and age for five species perature distributions were also significantly different between from the northeast skate complex using published von Bertalanffy parameters. species (Tables 2–5). The other five species have more restricted distributions along the northwest Atlantic coast (Table 6). temperature intervals in ascending order for each skate species, Thorny Skate have a narrow thermal habitat, with the largest ◦ ranging from 0 (for temperatures that are colder than those in- difference in average CWTs among seasons being only 2 C habited by skates) to 1 (for temperatures warmer than those (Table 1). Thorny Skate are distributed at the northern end of inhabited by skates). The CDF of CWTs for each species was the management area and are most abundant in the Gulf of compared with the CDF of all observed temperatures in the Maine and Georges Bank offshore strata, with very few fish survey. being caught inshore (<27 m depth) or in the southern New Northeast Fisheries Science Center trawl survey data were England or mid-Atlantic regions. This distribution is maintained further examined to detect any changes in thermal habitat among in all four seasons (Figure 4). Smooth Skate also have a narrow seasons by examining the differences between seasonal temper- thermal habitat, with the largest difference in average CWTs ◦ atures and geographic distributions by season. Seasons were de- among seasons again being only 2 C (Table 1). Smooth Skate fined as winter (December to February), spring (March to May), are distributed at the northern end of the management area and summer (June to August), and autumn (September to Novem- are most abundant in the Gulf of Maine and Georges Bank off- ber). The Kolmogorov–Smirnov test was used to test for temper- shore regions, with very few fish being caught inshore or in the ature preferences by comparing the distribution of all observed southern New England or mid-Atlantic regions; thus their distri- temperatures in each seasonal survey with the distributions of bution largely overlaps that of Thorny Skate. This distribution is catch-weighted temperatures for each species (Figure 2). maintained in all four seasons (Figure 5). Barndoor Skate have From the published literature, we derived life history traits for a moderate thermal habitat, with the largest difference in aver- ◦ all seven species in the northeast U.S. skate complex, including age CWTs among seasons being 5 C (Table 1). Barndoor Skate size and age at maturity and geographic range as well as the are abundant in the Gulf of Maine, Georges Bank, and southern growth parameters from the von Bertalanffy equation, New England offshore strata, with few fish being caught inshore Downloaded by [Department Of Fisheries] at 19:49 28 May 2013 or in the mid-Atlantic region; this distribution is maintained in −k(t−t0) Lt = L∞ 1 − e , (1) all four seasons (Figure 6). Winter Skate are found throughout the range of the NEFSC trawl survey in coastal and offshore waters. They are most abundant in the Georges Bank and south- where Lt is length at age, L∞ is the asymptotic maximum length, ern New England offshore strata, with few fish being caught in k is the von Bertalanffy growth coefficient, t is age, and t0 is the theoretical age at zero length (Figure 3). the Gulf of Maine and mid-Atlantic regions. This distribution is maintained year-round, but some movement offshore may occur in the summer, most likely due to a migration out of warmer RESULTS waters (Figure 7). Little Skate are abundant in the inshore and Distinct thermal habitats were identified for each of the seven offshore strata in all regions of the northeast U.S. coast but are skate species. Each species exhibited a unique tolerance range most abundant on Georges Bank and in southern New England. and seasonal preferences (Table 1). There was some overlap in This distribution is maintained year-round and largely overlaps geographic distribution, especially for Winter and Little skates, with that of Winter Skate (Figure 8). Limited inferences can the two commercially important species (Figure 2). Temperature be made on the thermal preferences of Clearnose Skate with- preference was significant for each species (i.e., the distributions out data from south of Cape Hatteras. Within the survey area, of the CWTs were significantly different from the distribution of this species exhibits a broad thermal habitat. However, they 410 HOGAN ET AL.

TABLE 2. Kolmogorov–Smirnov D-statistics for the temperature distributions for the seven species in the northeast skate complex during the winter season. Values in bold italics are signiﬁcant at P ≤ 0.05.

All observed Skate species temperatures Barndoor Clearnose Little Rosette Smooth Thorny Barndoor 0.248 Clearnose 0.535 0.446 Little 0.116 0.132 0.485 Rosette 0.686 0.596 0.273 0.653 Smooth 0.24 0.127 0.492 0.151 0.706 Thorny 0.175 0.163 0.555 0.107 0.737 0.12 Winter 0.094 0.204 0.60.115 0.758 0.233 0.124

TABLE 3. Kolmogorov–Smirnov D-statistics for the temperature distributions for the seven species in the northeast skate complex during the spring season. Values in bold italics are signiﬁcant at P ≤ 0.05.

All observed Skate species temperatures Barndoor Clearnose Little Rosette Smooth Thorny Barndoor 0.15 Clearnose 0.489 0.489 Little 0.178 0.109 0.413 Rosette 0.703 0.785 0.325 0.677 Smooth 0.203 0.098 0.552 0.187 0.864 Thorny 0.159 0.110 0.573 0.198 0.862 0.136 Winter 0.165 0.103 0.443 0.034 0.682 0.182 0.185

TABLE 4. Kolmogorov–Smirnov D-statistics for the temperature distributions for the seven species in the northeast skate complex during the summer season. Values in bold italics are signiﬁcant at P ≤ 0.05.

All observed Skate species temperatures Barndoor Clearnose Little Rosette Smooth Thorny Barndoor 0.242 Clearnose 0.576 0.604 Little 0.361 0.154 0.514 Rosette 0.466 0.276 0.511 0.134 Smooth 0.332 0.574 0.732 0.693 0.759 Thorny 0.236 0.476 0.685 0.595 0.703 0.105 Downloaded by [Department Of Fisheries] at 19:49 28 May 2013 Winter 0.344 0.151 0.514 0.027 0.139 0.676 0.578

TABLE 5. Kolmogorov–Smirnov D-statistics for the temperature distributions for the seven species in the northeast skate complex during the autumn season. Values in bold italics are signiﬁcant at P ≤ 0.05.

All observed Skate species temperatures Barndoor Clearnose Little Rosette Smooth Thorny Barndoor 0.204 Clearnose 0.681 0.885 Little 0.249 0.38 0.657 Rosette 0.294 0.404 0.844 0.231 Smooth 0.452 0.368 0.957 0.696 0.741 Thorny 0.401 0.281 0.943 0.645 0.685 0.103 Winter 0.236 0.384 0.675 0.018 0.242 0.688 0.638 MANAGING THE U.S. NORTHEAST SKATE COMPLEX 411

TABLE 6. Summary of life histories and geographical distributions of the seven species of the northeast skate complex. Abbreviations are as follows: GB = Georges Bank, SNE = southern New England, GOM = Gulf of Maine, MA = mid-Atlantic, and fm = fathom.

Maximum Maximum estimated age Species General distribution length (cm) (years) Reference(s) Thorny Skate Inshore and offshore GOM, Female: 105 Female: 16 Sulikowski et al. (2005b) along the 100-fm edge of GB Male: 103 Male: 16 (very few in SNE or MA) Smooth Skate Inshore and offshore GOM, 71 Female: 14 Packer et al. (2003e); along the 100-fm edge of GB Male: 15 Natanson et al. (2007) (very few in SNE or MA) Barndoor Skate Offshore GOM (Canadian 152 11 Bigelow and Schroeder waters), offshore GB and SNE (1953); Gedamke et al. (very few inshore or in the (2005) MA region) Winter Skate Inshore and offshore GB and Female: 121.8 Female: 18 Sulikowski et al. (2003) SNE with lesser amounts in Male: 137.4 Male: 19 GOM and MA Little Skate Inshore and offshore GB, SNE, 62 12.5 Richards et al. (1963); and MA (very few in GOM) Waring (1984); Packer et al. (2003c); Frisk and Miller (2006) Clearnose Skate Inshore and offshore MA 94–95 Female: 7 Gelsleichter (1998); Packer Male: 5 et al. (2003b) Rosette Skate Offshore MA 57 Packer et al. (2003d)

inhabit warmer temperatures than the other six species contained for Smooth Skate is currently unavailable. Little Skate mature in this complex. Clearnose Skate are not well represented in the at a younger age than Winter Skate; estimates of age at matu- NEFSC trawl survey because the majority of the population rity are unavailable for Clearnose and Rosette skates. Growth is found south of Cape Hatteras (NEFSC 2007); they are most parameters are not available for all species, but the available abundant in the mid-Atlantic offshore and inshore strata regions, von Bertalanffy parameters differ among species (Table 8). Al- with very few fish being caught in southern New England and no though all the skates are characteristically slow-growing, the fish in other survey regions (Figure 9). Some seasonal northward von Bertalanffy growth parameters (k, L∞, and t0)differby movements are suggested in the autumn. Rosette Skate are most species. abundant in the mid-Atlantic offshore strata region, with very few fish being caught in southern New England and Georges

Downloaded by [Department Of Fisheries] at 19:49 28 May 2013 Bank and no fish in the Gulf of Maine or inshore (Figure 10). DISCUSSION There is no overlap between this species and Thorny, Smooth, Given the differences in geographic distribution and life his- or Barndoor skates. tory among skate species, the current management strategy chal- Life history parameters differ by species regardless of simi- lenges the stock complex definition as outlined by the national larities in thermal habitats or geographical distribution (Table 6). standard 1 guidelines. The seven species in the skate complex do Of the northerly species, Thorny and Barndoor skates reach a not have a common geographic distribution, nor do they mature maximum size over 100 cm while the maximum size of Smooth at the same size. They are seven distinct species that must be Skate is less than 100 cm; estimated maximum age is over managed sustainably. 10 years for all three species. The maximum size of Winter Skate The distribution of each species influences the landings com- is estimated to be approximately double that of Little Skate, but position (NDPSWG 2009). The wing fishery is focused in the the estimated maximum age is over 10 years for both species. Of Gulf of Maine and Georges Bank, with New Bedford, Mas- the southerly species, Clearnose Skate reach a larger maximum sachusetts, being the center of processing for skate products. size than Rosette Skate; no comparisons of estimated maximum The fishery is restricted by prohibitions on the harvest of some ages can be done due to lack of data. Age and size at maturity species. Under this constraint, landings consist primarily of are different for each species in the complex (Table 7). Thorny Winter Skate. Clearnose Skate do reach a large enough size to Skate mature at an older age than Barndoor Skate; an estimate be harvested in the wing fishery, but their southern distribution 412 HOGAN ET AL. Downloaded by [Department Of Fisheries] at 19:49 28 May 2013

FIGURE 4. Seasonal distributions of Thorny Skate from Northeast Fisheries Science Center trawl survey data compared with all observed bottom temperatures where the seven skate species were caught. Seasonal distributions were deﬁned by the range of temperatures at the locations where Thorny Skate were observed in each season. MANAGING THE U.S. NORTHEAST SKATE COMPLEX 413 Downloaded by [Department Of Fisheries] at 19:49 28 May 2013

FIGURE 5. Seasonal distributions of Smooth Skate from Northeast Fisheries Science Center trawl survey data. See Figure 4 for additional information.

limits their exposure to the fishery. The contribution of Rosette reduce the harvest of a more abundant species to allow rebuild- Skate to the Winter Skate quota may provide a short-term ben- ing of a co-occurring species (Punt et al. 2005). In a regional efit to the wing fishery, but inflating the allowable Winter Skate management strategy, this may require the reduced catch of Lit- catch may have unforeseen consequences, e.g., overfishing tle and Winter skates in the Gulf of Maine to allow an increase Winter Skate. In mixed-species fisheries, it may be necessary to in the Thorny Skate stock to occur with the long-term goal of 414 HOGAN ET AL. Downloaded by [Department Of Fisheries] at 19:49 28 May 2013

FIGURE 6. Seasonal distributions of Barndoor Skate from Northeast Fisheries Science Center trawl survey data. See Figure 4 for more information.

harvesting Thorny and Winter skates together in the wing fish- The ability of skates to sustain heavy fishing pressure has ery. The bait fishery targets smaller skates, primarily in the been questioned because of local extirpations of other rajid southern New England area, with the majority of process- species (Brander 1981; Walker and Heessen 1996; Dulvy et al. ing occurring in Point Judith, Rhode Island. Rosette Skate 2000; Dulvy and Reynolds 2002). Varying life history traits are small enough to be used in the bait fishery, but their and thermal distributions may contribute to species’ differential distribution is restricted to deep waters in the mid-Atlantic responses to fishing pressure; slow-growing, late-maturing fish region. are less likely to sustain an intensive fishery than fast-growing, MANAGING THE U.S. NORTHEAST SKATE COMPLEX 415 Downloaded by [Department Of Fisheries] at 19:49 28 May 2013

FIGURE 7. Seasonal distributions of Winter Skate from Northeast Fisheries Science Center trawl survey data. See Figure 4 for more information.

early-maturing fish, but a broad thermal range may allow any over the entire management area and has had large changes in species to utilize areas experiencing the least fishing pressure. abundance. Although it had a low survey index in 2007, its in- At the northern extent of the management area, Thorny Skate dex has since increased threefold (the cause of which remains are overfished (NEFSC 2007) and the population has continued unknown). to decline despite a prohibition on landings since 2003. Smooth No direct inferences on seasonal migration patterns can be Skate have fluctuated with respect to being overfished but appear made from seasonal distributions. This study is able to identify to have responded positively to regional possession prohibitions the seasonal temperatures at which species are most likely to be (NDPSWG 2009). The Winter Skate population is distributed found, but it does not track individual skates. The migration 416 HOGAN ET AL. Downloaded by [Department Of Fisheries] at 19:49 28 May 2013

FIGURE 8. Seasonal distributions of Little Skate from Northeast Fisheries Science Center trawl survey data. See Figure 4 for more information.

patterns of the skate species remain largely unknown. Rays R. clavata tagged in the Thames Estuary, UK. U.S. and Templeman (1984) tagged Thorny Skate off Newfoundland us- Canadian survey data suggest some connectivity between the ing Peterson discs. The movement of individuals from the three Winter Skate populations on Georges Bank and the Scotian tagging sites varied, with some being recaptured near the orig- Shelf (Frisk et al. 2008). Winter Skate populations have shown inal site and others over 160 km away. Hunter et al. (2005a, large ﬂuctuations in survey abundance that have not appeared in 2005b) found evidence of seasonal migration by Thornback the other species trends. The statistical analysis suggests some MANAGING THE U.S. NORTHEAST SKATE COMPLEX 417 Downloaded by [Department Of Fisheries] at 19:49 28 May 2013

FIGURE 9. Seasonal distributions of Clearnose Skate from Northeast Fisheries Science Center survey data. See Figure 4 for more information.

level of thermal habitat mixing may be occurring between a few since its implementation, but multiple species have started to species on a seasonal scale. A tagging study targeting each skate rebuild. An amendment to this plan was initiated when the Win- species would help fill the gaps in our understanding of skates’ ter Skate was designated as overfished, but before the analysis seasonal movement patterns. and drafting of the amendment document was completed Win- The goal of the Northeast Skate Complex FMP is the sustain- ter Skate were no longer designated as overfished (following the able harvest of skates in all fisheries. The success of this plan is inclusion of additional survey years in the reference point calcu- difficult to determine. Several species have become overfished lations; NDPSWG 2009). Prohibiting the possession of Thorny 418 HOGAN ET AL. Downloaded by [Department Of Fisheries] at 19:49 28 May 2013

FIGURE 10. Seasonal distributions of Rosette Skate from Northeast Fisheries Science Center trawl survey data. See Figure 4 for more information.

and Barndoor skates throughout their range has had contrasting servation efforts for each species instead of relying solely on results. Barndoor Skate have been steadily rebuilding since the discarding to reduce fishing mortality on overfished species. inception of the FMP; Thorny Skate have shown a steady de- Skates are caught mainly as bycatch in many fisheries. The cline. This indicates that a blanket strategy is be appropriate for only mechanism currently used to reduce fishing mortality and all seven species. increase abundance proxies is to prohibit possession, which Species-specific management would improve the ability of increases skate discards. Under the reauthorized Magnuson– the management council to enhance the fishery and the con- Stevens Fishery Conservation and Management Act, dead MANAGING THE U.S. NORTHEAST SKATE COMPLEX 419

TABLE 7. Summary of the reproductive life history characteristics of the seven species of the northeast skate complex. See Table 6 for abbreviations.

Age at 50% Fecundity Size at 50% maturity Reproductive (eggs per Species maturity (cm) (years) cycle year) Reference(s) Thorny Skate Females (GOM): 87.5 Females: 11 Year-round 40.5 Sulikowski et al. (2005a, Males (GOM): 86.5 Males: 10.9 2006); Parent et al. 2008) Smooth Skate Females: 56 Year-round Kneebone et al. (2007); Males: 61 Sulikowski et al. (2007) Barndoor Skate Females (GB): 116 Females: 6.5 47; 69–85 Packer et al. (2003a); Parent Males (GB): 108 Males: 5.8 et al. (2008) Winter Skate Females: 76 Females: Year-round Sulikowski et al. (2004, Males: 73 11–12 (one peak 2005c); Frisk and Miller 85 Males: 11 during (2006) summer) Little Skate Females: 48 Females: 9.5 Cicia et al. (2009) Males: 46 Males: 7.7 Clearnose Skate Female: 59–65a 35 Packer et al. (2003b); Males: 56a Sosebee (2005) Rosette Skate Females: 33–35a Sosebee (2005) Males: 33a aSize at maturity.

discards must be accounted for; if these become too high, the Little Skate, effectively separating the wing and bait fisheries. allowable catch is decreased, potentially decreasing fishing ac- New England fisheries rarely have the opportunity to harvest tivity. Under separate management plans, Barndoor Skate could Rosette Skates, but that species remains vulnerable to southern remain prohibited throughout its range pending a reevaluation deepwater fisheries. The majority of Clearnose Skate landings of the NEFSC survey years used in the reference points. Thorny occur south of New England, so this species may benefit from Skate could remain prohibited and regional closures imple- being managed as one stock throughout its range. mented to protect dense aggregations in an effort to increase The inclusion of Clearnose and Rosette skates may affect the population. The regulations for Smooth Skate could remain the level of catch allotted to more economically important as they are. The status quo would pertain to Winter Skate and species in New England. The maximum sustainable yield of a

TABLE 8. Summary of the von Bertalanffy growth parameters of ﬁve of the seven species of the northeast skate complex. Values are not available for the Clearnose and Rosette skates.

Combined Species Males Females sexes Reference(s)

Downloaded by [Department Of Fisheries] at 19:49 28 May 2013 Winter Skate k = 0.074 k = 0.059 Sulikowski et al. (2003) L∞ = 121.8 cm L∞ = 137.4 cm t0 = –1.418 t0 = –1.609 Barndoor Skate k = 0.1414 Gedamke et al. (2005) L∞ = 166.3 cm t0 = –1.2912 Thorny Skate k = 0.11 k = 0.13 Sulikowski et al. (2005b) L∞ = 127 cm L∞ = 120 cm t0 = –0.37 t0 = –0.4 Smooth Skate k = 0.12 k = 0.12 Natanson et al. (2007) L∞ = 75.4 cm L∞ = 69.6 cm a a t0 = 1.96 t0 = 1.96 Little Skate k = 0.19 Frisk and Miller (2006) L∞ = 56.1 cm t0 = –1.77 a Converted to t0 from l0. 420 HOGAN ET AL.

single species decreases when it is harvested and managed in a species. The difficulties with species identification and the multispecies fishery (Ricker 1958). The maximum sustainable ability to catch a large aggregation in a single haul present yield for each species differs based on its reproductive char- challenges. It may be necessary to use regions or fisheries as a acteristics and initial abundance proxy; such a yield may not proxy for species. A general assumption is that total allowable be achievable for each species without the risk of reducing its landings of the two targeted species would increase if they were biomass to a level below the threshold or driving the species managed individually. Under the current management paradigm, to extinction. The Northeast Skate Complex FMP does not es- all seven species contribute to the annual catch limit. A shift to timate the maximum sustainable yield because of unreliable a species-specific management strategy may result in decreased identification but uses a proxy (the median exploitation ratio) total allowable landings for Winter and Little skates if the con- instead. The removal of Clearnose and Rosette skates may re- tribution of the five other species is removed. duce the noise in stock assessments, potentially increasing the The efficiency of skate fisheries management would be chances that a maximum sustainable yield can be estimated. greatly improved by the separation of the skate complex The issue of skate complex management is not solely a U.S. plan into seven separate plans. The two southernmost species, problem. As in the northeast U.S. skate fishery, the majority of Clearnose and Rosette skates, are not overfished. Clearnose landings in Europe are reported as unspecified. Approximately a Skate contribute to the fishery mostly in areas south of Rhode dozen skate and ray species are found in the waters of the British Island. Management of this species might be greatly improved Isles. The sustainability of recent harvests has been questioned by including its entire geographic range into the management because of the local disappearance of three skate species (Dulvy plan. The contribution of Rosette Skate to landings is negligible. et al. 2000). A port sampling program was initiated to speciate The survey catches few Rosette Skates (NEFSC 2007); this may the catch and collect length composition data on skates landed indicate that the population is small or that the survey is ineffi- whole. Current management strategies in UK waters include a cient in sampling this species. Although the seven skate species minimum landing size in certain areas and a total allowable catch occupy similar ecological niches, the life history and habitat for skates and rays in the North Sea (Ellis et al. 2005). The skate differences between the species may be substantial enough to fishery is largely a bycatch fishery in the UK, and high discard warrant a shift to species-specific or regional management. survival is essential to reduce the negative impact of fishing practices on the population. The short-term survival of skates ACKNOWLEDGMENTS was 55% for demersal trawlers in the Bristol Channel (Enever Trawl survey data were from the Northeast Fisheries Science et al. 2009). Survival was found to increase by modifying the Center. Dan Georgiana and Emily Keiley of the University of cod end of the net, a measure designed to reduce skate bycatch Massachusetts–Dartmouth assisted in a challenging discussion (Enever et al. 2010). of Skate management. We also thank Ken Oliveira, Francis By contrast, the Northwest Pacific Fisheries Management Juanes, and Greg Skomal for their guidance. Council has had the opportunity to regulate the Alaskan skate fishery from its inception. A targeted fishery began in 2003 that focused on two large skates, the Big Skate Raja binoculata and REFERENCES the Longnose Skate R. rhina. In an effort to achieve a sustainable Bigelow, H. B., and W. C. Schroeder. 1953. Fishes of the Gulf of Maine. U.S. harvest, the fishery management plan was amended to remove Fish and Wildlife Service Fishery Bulletin 74:577. Raia batis these two species from the “other species” category, which had Brander, K. 1981. Disappearance of Common Skate from Irish Sea. Nature 290:48–49. previously included all skate species (DOC 2004). A separate Cicia, A. M., W. B. Driggers III, G. W. Ingram Jr., J. Kneebone, P. C. W. Tsang,

Downloaded by [Department Of Fisheries] at 19:49 28 May 2013 total allowable catch was established for these two skate species D. M. Koester, and J. A. Sulikowski. 2009. Size and age estimates at sexual while allowing the remainder of the skate species (which have maturity for the Little Skate Leucoraja erinacea from the western Gulf of no commercial value and are difficult to identify) to be managed Maine, USA. Journal of Fish Biology 75:1648–1666. as a complex. The Alaskan forage fish and octopus (dominated Collette, B. B., and G. Klein-Macphee. 2002. Bigelow and Schroeder’s Fishes of the Gulf of Maine. Smithsonian Books, Washington, D.C. by the giant Pacific octopus Enteroctopus dofleini) fisheries are DOC (Department of Commerce). 2004. Fisheries of the Exclusive Economic also managed in complexes. These complexes contain mostly Zone off Alaska; skates management in the groundfish fisheries of the Gulf nontargeted species for which long-term catch and life history of Alaska. Federal Register 69:92(12 May 2004):26313–26320. data may not be available and have each posed a challenge to DOC (Department of Commerce). 2009. Magnuson–Stevens Act provisions; compliance with the catch limit strategy (Reuter et al. 2010). annual catch limits; national standard guidelines. Federal Register 74:11(16 January 2009):3178–3213. Catch limit management structures have been put into place Despres-Patanjo, L. I., T. R. Azarovitz, and C. J. Byrne. 1988. Twenty-five years for each of these Alaskan complexes, but the risk to individual of fish surveys in the Northwest Atlantic: the NMFS Northeast Fisheries population sizes remains unknown (Reuter et al. 2010). These Center’s bottom trawl survey program. Marine Fisheries Review 69:50–53. examples illustrate the need for detailed, species-specific plans Dulvy, N. K., J. D. Metcalfe, J. Glanville, M. G. Pawson, and J. D. Reynolds. to achieve sustainable harvests. 2000. Fishery stability, local extinctions, and shifts in community structure in Skates. Conservation Biology 14:283–293. The implementation of a species-specific strategy would Dulvy, N. K., and J. D. Reynolds. 2002. Predicting extinction vulnerability in require improvements to reported landings and discards by skates. Conservation Biology 16:440–450. MANAGING THE U.S. NORTHEAST SKATE COMPLEX 421

Ellis, J. R., N. K. Dulvy, S. Jennings, M. Parker-Humphreys, and S. I. Rogers. and habitat characteristics. NOAA Technical Memorandum NMFS-NE- 2005. Assessing the status of demersal elasmobranchs in UK waters: a re- 177. view. Journal of the Marine Biological Association of the United Kingdom Packer, D. B., C. A. Zetlin, and J. J. Vitaliano. 2003f. Essential fish habitat 85:1025–1047. source document: Thorny Skate, Amblyraja radiata, life history and habitat Enever, R., T. Catchpole, J. Ellis, and A. Grant. 2009. The survival of skates (Ra- characteristics. NOAA Technical Memorandum NMFS-NE-178. jidae) caught by demersal trawlers fishing in UK waters. Fisheries Research Packer, D. B., C. A. Zetlin, and J. J. Vitaliano. 2003g. Essential fish habitat 97:72–76. source document: Winter Skate, Leucoraja ocellata, life history and habitat Enever, R., A. S. Revill, R. Caslake, and A. Grant. 2010. Discard mitigation characteristics. NOAA Technical Memorandum NMFS-NE-179. increases skate survival in the Bristol Channel. Fisheries Research 102:9–15. Parent, S., S. Pepin, J. Genet, L. Misserey, and S. Rojas. 2008. Captive breed- Frisk, M. G., and T. J. Miller. 2006. Age, growth, and latitudinal patterns of ing of the Barndoor Skate (Dipturus laevis) at the Montreal Biodome, with two Rajidae species in the northwestern Atlantic: Little Skate (Leucoraja eri- comparison notes on two other captive-bred Skate species. Zoo Biology nacea) and Winter Skate (Leucoraja ocellata). Canadian Journal of Fisheries 27:145–153. and Aquatic Sciences 63:1078–1091. Punt, A. E., F. Pribac, B. L. Taylor, and T. I. Walker. 2005. Harvest strategy Frisk, M. G., T. J. Miller, S. J. D. Martell, and K. Sosebee. 2008. New hypoth- evaluation for School and Gummy Shark. Journal of Northwest Atlantic esis helps explain elasmobranch “outburst” on Georges Bank in the 1980s. Fishery Science 35:387–406. Ecological Applications 18:234–245. Ricker, W. E. 1958. Maximum sustained yields from fluctuating environments Gedamke, T., W. D. DuPaul, and J. A. Musick. 2005. Observations on the life and mixed stocks. Journal of the Fisheries Research Board of Canada 15:991– history of the Barndoor Skate, Dipturus laevis, on Georges Bank (western 1006. North Atlantic). Journal of Northwest Atlantic Fishery Science 35:67–78. Richards, S. W., D. Merriman, and L. H. Calhoun. 1963. Studies on the marine Gelsleichter, J. 1998. Vertebral cartilage of the Clearnose Skate, Raja eglanteria: resources of southern New England, IX. The biology of the Little Skate, development, structure, ageing, and hormonal regulation of growth. Doctoral Raja erinacea Mitchell. Bulletin of the Bingham Oceanographic Collection dissertation. College of William and Mary, Williamsburg, Virginia. 18:1–67. Hunter, E., A. A. Buckley, C. Stewart, and J. D. Metcalfe. 2005a. Repeated Reuter, R. F., M. E. Conners, J. Dicosimo, S. Gaichas, O. Ormseth, and seasonal migration by a Thornback Ray in the southern North Sea. Journal T. T. Tenbrink. 2010. Managing nontarget, data-poor species using catch of the Marine Biological Association of the United Kingdom 85:1199–1200. limits: lessons from the Alaskan groundfish fishery. Fisheries Management Hunter, E., A. A. Buckley, C. Stewart, and J. D. Metcalfe. 2005b. Migra- and Ecology 17:323–335. tory behaviour of the Thornback Ray, Raja clavata, in the southern North Sokal, R. R., and F. J. Rohlf. 1995. Biometry: the principles and practice of Sea. Journal of the Marine Biological Association of the United Kingdom statistics in biological research, 3rd edition. Freeman, New York. 85:1095–1105. Sosebee, K. A. 2005. Maturity of Skates in northeast United States waters. E- Kneebone, J., D. E. Ferguson, J. A. Sulikowski, and P. C. W. Tsang. 2007. Journal of Northwest Atlantic Fishery Science [online serial] 35(9):141–153. Endocrinological investigation into the reproductive cycles of two sympatric Sulikowski, J. A., S. Elzey, J. Kneebone, J. Jurek, W. Huntting Howell, and skate species, Malacoraja senta and Amblyraja radiata, in the western Gulf P.C. W.Tsang. 2007. The reproductive cycle of the Smooth Skate, Malacoraja of Maine. Environmental Biology of Fishes 80:257–265. senta, in the Gulf of Maine. Marine and Freshwater Research 58:98–103. Natanson, L. J., J. A. Sulikowski, J. R. Kneebone, and P. C. Tsang. 2007. Age Sulikowski, J. A., J. Kneebone, S. Elzey, J. Jurek, P. D. Danley, W. Huntting and growth estimates for the Smooth Skate, Malacoraja senta, in the Gulf of Howell, and P. C. W. Tsang. 2005a. Age and growth estimates of the Thorny Maine. Environmental Biology of Fishes 80:293–308. Skate (Amblyraja radiata) in the western Gulf of Maine. U.S. National Marine NDPSWG (Northeast Data Poor Stocks Working Group). 2009. Northeast Data Fisheries Service Fishery Bulletin 103:161–168. Poor Stocks Working Group Report, part A. Skate species complex, deep Sulikowski, J. A., J. Kneebone, S. Elzey, J. Jurek, P. D. Danley, W. Huntting sea red crab, Atlantic Wolfish, Scup, and Black Seabass. Northeast Fisheries Howell, and P. C. W. Tsang. 2005b. The reproductive cycle of the Thorny Science Center, Reference Document 09-02a, Woods Hole, Massachusetts. Skate (Amblyraja radiata) in the western Gulf of Maine. U.S. National Marine NEFMC (New England Fisheries Management Council). 2009. Final Amend- Fisheries Service Fishery Bulletin 103:536–543. ment 3 to the Fishery Management Plan (FMP) for the northeast skate com- Sulikowski, J. A., J. Kneebone, S. Elzey, J. Jurek, W. H. Howell, and P. C. W. plex and final environmental impact statement (FEIS) with an initial Regula- Tsang. 2006. Using the composite variables of reproductive morphology, tory Flexibility Act analysis. NEFMC, Newburyport, Massachusetts. histology, and steroid hormones to determine age and size at sexual maturity NEFSC (Northeast Fisheries Science Center). 2000. 30th Northeast Regional for the Thorny Skate Amblyraja radiata in the western Gulf of Maine. Journal Stock Assessment Workshop (30th SAW), Stock Assessment Review Com- of Fish Biology 69:1449–1465. Downloaded by [Department Of Fisheries] at 19:49 28 May 2013 mittee (SARC), consensus summary of assessments. NEFSC, Reference Doc- Sulikowski, J. A., M. D. Morin, S. H. Suk, and W. Huntting Howell. 2003. Age ument 00-03, Woods Hole, Massachusetts. and growth estimates of the Winter Skate (Leucoraja ocellata)inthewest- NEFSC (Northeast Fisheries Science Center). 2007. 44th Northeast Regional ern Gulf of Maine. U.S. National Marine Fisheries Service Fishery Bulletin Stock Assessment Workshop (44th SAW). 44th SAW assessment summary 101:405–413. report. NEFSC, Reference Document 07-03, Woods Hole, Massachusetts. Sulikowski, J. A., P. C. W. Tsang, and W. Huntting Howell. 2004. An annual Packer, D. B., C. A. Zetlin, and J. J. Vitaliano. 2003a. Essential fish habitat cycle of steroid hormone concentrations and gonad development in the Winter source document: Barndoor Skate, Dipturus laevis, life history and habitat Skate, Leucoraja ocellata, from the western Gulf of Maine. Marine Biology characteristics. NOAA Technical Memorandum NMFS-NE-173. 144:845–853. Packer, D. B., C. A. Zetlin, and J. J. Vitaliano. 2003b. Essential fish habitat Sulikowski, J. A., P. C. W. Tsang, and W. Huntting Howell. 2005c. Age and size source document: Clearnose Skate, Raja eglanteria, life history and habitat at sexual maturity for the Winter Skate, Leucoraja ocellata, in the western characteristics. NOAA Technical Memorandum NMFS-NE-174. Gulf of Maine based on morphological, histological, and steroid hormone Packer, D. B., C. A. Zetlin, and J. J. Vitaliano. 2003c. Essential fish habitat analyses. Environmental Biology of Fishes 72:429–441. source document: Little Skate, Leucoraja erinacea, life history and habitat Templeman, W. 1984. Migrations of Thorny Skate, Raja radiata, tagged in the characteristics. NOAA Technical Memorandum NMFS-NE-175. Newfoundland area. Journal of Northwest Atlantic Fishery Science 5:55–63. Packer, D. B., C. A. Zetlin, and J. J. Vitaliano. 2003d. Essential fish habitat Walker, P. A., and H. J. L. Heessen. 1996. Long-term changes in ray populations source document: Rosette Skate, Leucoraja garmani virginica, life history in the North Sea. ICES Journal of Marine Science 53:1085–1093. and habitat characteristics. NOAA Technical Memorandum NMFS-NE-176. Waring, G. T. 1984. Age, growth, and mortality of the Little Skate off the Packer, D. B., C. A. Zetlin, and J. J. Vitaliano. 2003e. Essential fish northeast coast of the United States. Transactions of the American Fisheries habitat source document: Smooth Skate, Malacoraja senta, life history Society 113:314–321. This article was downloaded by: [Department Of Fisheries] On: 28 May 2013, At: 19:52 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 Ghost Fishing in the Southeast Alaska Commercial Dungeness Crab Fishery Jacek Maselko a , Gretchen Bishop b & Peter Murphy c a National Marine Fisheries Service, Alaska Fisheries Science Center, Auke Bay Laboratories, 17109 Point Lena Loop Road, Juneau, Alaska, 99801, USA b Alaska Department of Fish and Game, Division of Commercial Fisheries, Post Office Box 115526, Juneau, Alaska, 99811-5526, USA c National Oceanic and Atmospheric Administration Marine Debris Program, Office of Response and Restoration, 7600 Sand Point Way, Seattle, Washington, 98115, USA Published online: 29 Mar 2013.

To cite this article: Jacek Maselko , Gretchen Bishop & Peter Murphy (2013): Ghost Fishing in the Southeast Alaska Commercial Dungeness Crab Fishery, North American Journal of Fisheries Management, 33:2, 422-431 To link to this article: http://dx.doi.org/10.1080/02755947.2013.763875

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 33:422–431, 2013 C American Fisheries Society 2013 ISSN: 0275-5947 print / 1548-8675 online DOI: 10.1080/02755947.2013.763875

ARTICLE

Ghost Fishing in the Southeast Alaska Commercial Dungeness Crab Fishery

Jacek Maselko* National Marine Fisheries Service, Alaska Fisheries Science Center, Auke Bay Laboratories, 17109 Point Lena Loop Road, Juneau, Alaska 99801, USA Gretchen Bishop Alaska Department of Fish and Game, Division of Commercial Fisheries, Post Ofﬁce Box 115526, Juneau, Alaska 99811-5526, USA Peter Murphy National Oceanic and Atmospheric Administration Marine Debris Program, Ofﬁce of Response and Restoration, 7600 Sand Point Way, Seattle, Washington 98115, USA

Abstract Entrapment of crabs by derelict crab pots (also known as ghost fishing) can be a significant consequence of commercial fishing. The prevalence of lost commercial pots and ghost-fishing entrapments was estimated for the commercial Dungeness crab Cancer magister fishery in southeastern Alaska during the 2009 and 2010 summer closures of the commercial season (16 August through 30 September). Teams of divers retrieved a random subsample of the derelict crab pots located using side-scan sonar. Altogether, we retrieved 123 derelict crab pots containing 215 entrapped Dungeness crabs. The densities of derelict crab pots varied from 1.5 to 10.1/km2, while the densities of entrapped Dungeness crabs ranged from 0 to 54.5/km2, depending on the area surveyed. Derelict crab pots were discovered to effectively ghost-fish for at least 7 years, indicating that there are long-term cumulative impacts on Dungeness crab populations. The number of derelict crab pots and entrapped Dungeness crabs at each of the surveyed areas was highly correlated with the number of fishermen, the number of pot lifts, and annual harvest in numbers, allowing for extrapolation to a regionwide estimate of crab entrapment and derelict crab pot abundance. Overall, our findings show instantaneous entrapment of less than 1% of the commercial crab harvest with a cumulative annual loss of less than 3% of the regional commercial crab harvest. We challenge the efficacy of the biodegradable escape mechanism currently employed in commercial Dungeness crab pots in southeastern Alaska and present alternatives which may require further in situ or laboratory verification of their effectiveness. Downloaded by [Department Of Fisheries] at 19:52 28 May 2013

Ghost fishing is defined as the continued entrapment of ma- depends largely upon pot design, especially the pot entrance rine animals by derelict fishing gear, including longlines, gill configuration and the effectiveness of escape (cull) rings and nets, and fish and invertebrate pots (Matsuoka et al. 2005). The biodegradable escape mechanism. High fishery intensity can be magnitude of the problem of ghost fishing in invertebrate pot caused by short seasons or a high pot density. Gear conflicts fisheries is a function of the annual proportion of pots lost (pot may occur when more than one fishery is being prosecuted in loss rate), the proportion of pots ghost fishing (ghost-fishing the same area during the same time period and can result in pots rate), and the length of time over which ghost fishing occurs. being dragged into deep waters or in buoy lines being purpose- The pot loss rate is a function of fishery intensity, gear con- fully or accidentally cut. Environmental conditions which may flicts, and environmental conditions, while the ghost-fishing rate result in pot loss include storms, sedimentation, and ice cover.

*Corresponding author: [email protected] Received February 23, 2012; accepted December 27, 2012 422 GHOST FISHING IN THE DUNGENESS CRAB FISHERY 423

Unattended pots with long soak periods are more likely to be (Hebert et al. 2008). The summer closure protects crabs during lost due to challenging environmental conditions. The length of the primary female molt and mating period (Stone and O’Clair time over which ghost fishing occurs is a function of the rate of 2001). degradation of the pot. Alaska state regulations are in place to provide for a sus- Derelict crab pots ghost fishing for long periods of time can tainable Dungeness crab harvest and prevent ghost fishing. The result in lethal and sublethal effects on both crabs and noncommercial harvest is prosecuted by a pot fishery restricted targeted species (High 1976; Breen 1990; Kruse and Kimker to males of at least 165 mm carapace width excluding spines. 1993; Stevens 1996). The lethal effects of ghost fishing on tar- Commercial crab pots cannot exceed 127 cm in diameter and geted invertebrates can include increased predation by species are required to have two 111-mm diameter cull rings to allow such as octopus (Octopodidae; High 1976), increased cannibal- for egress of undersized crabs. In addition to these cull rings, ism (Sheldon and Dow 1975; Paul et al. 1993), and starvation commercial pots must have a biodegradable escape mechanism (Breen 1987). Sublethal effects can include carapace damage, in the form of a loop of no greater than 60-thread cotton twine appendage loss (Durkin et al. 1984; Muir et al. 1984; Kimker (rot cord) or a galvanic timed-release device attached to the 1992; Barber and Cobb 2007), decrease in weight (Muir et al. hook that secures the pot lid (Escape mechanism for shellfish 1984), and increased biofouling of the exoskeleton (Dahlstrom and bottomfish pots 1978). Although Alaska wildlife troopers 1975). Leg loss is known in some Cancer species to reduce molt regularly examine a subset of actively fishing crab pots for com- increment (Bennett 1973). pliance, no reliable estimates of regulatory compliance rates are Dungeness crabs Cancer magister inhabit muddy and sandy available. bottom areas of southeastern Alaska waters, generally in depths The effects of ghost fishing on Dungeness crab populations less than 100 m, and support valuable commercial and noncom- in southeastern Alaska are unknown, but estimates of gear loss mercial fisheries. They live 8–10 years and reach the legal size and ghost-fishing rates as well as crab mortality from other pot of 165 mm (6.5 in) at 4–5 years of age (Butler 1961; Bishop fisheries suggest they are substantial. Barry (1981) estimated et al. 2007). Southeast Alaska includes an archipelago but also that 20% of the Dungeness crab pots fished in the 1978–1979 nonisland terrain. Dungeness crabs are distributed throughout its coastal Washington state commercial season were lost. In the inland waters, often near coastal communities, providing easy Fraser River District in British Columbia, Breen (1987) esti- access to both commercial and noncommercial users. The size mated an 11% annual Dungeness crab pot loss rate with annual of vessels used varies from about 6 to 27.5 m and fishermen ghost-fishing mortality of 7% of the annual commercial har- are allowed 75–300 pots depending upon the provisions of their vest. Similarly, an average pot loss of 10% was reported for permit class (Hebert et al. 2008). As in other pot fisheries, some the Alaskan red king Paralithodes camtschaticus fishery in the pots are lost and continue to ghost-fish. This ghost fishing can 1970s (High and Worlund 1979) and a loss of 20% for the Bering continue for multiple years and could account for significant Sea and Aleutian Islands crab (red and blue Paralithodes platy- mortality due to continuous rebaiting by entrapped animals that pus king crab, Tanner crab Chionoecetes bairdi, and snow crab die (Breen 1990). C. opilio) fisheries in 1991 (Kruse and Kimker 1993); Kruse The Dungeness crab fisheries in southeastern Alaska have and Kimker attributed the high pot loss rate to ice movement long provided important economic and social benefits to the re- during the snow crab fishery. In the Puget Sound (Washington gion. The first recorded commercial harvest occurred in the State) Dungeness crab fishery, 37% of recovered derelict crab 1930s (Hebert et al. 2008). The fisheries are managed by pots were ghost fishing for at least 1 year even if the rot cord the Alaska Department of Fish and Game (ADFG), which re- was degraded (June 2007). The annual ghost-fishing catch rate

Downloaded by [Department Of Fisheries] at 19:52 28 May 2013 quires that harvest and effort be reported on fish tickets and was estimated at 74 crabs per pot in Puget Sound, for an an- assigned to 1 of 254 statistical areas (Figure 1). Statistical ar- nual mortality of 372,000 Dungeness crabs, which represents eas are mutually exclusive geographical areas of varying sizes 30% to 40% (337,000 kg) of the mean annual harvest (June where commercial fishing may be permitted. The commercial 2007). However, a more recent study (Antonelis et al. 2011) fishery is under limited entry, with 308 permit holders currently estimated the ghost-fishing mortality rate in Puget Sound at a eligible to participate in southeastern Alaska, which is where more modest 4.5% of the mean annual harvest. Unlike com- the majority of the Dungeness crab harvest in the State occurs. mercial and noncommercial harvests, which target legal males Detailed commercial harvest records indicate a mean annual only, ghost fishing removes both males and females from the harvest of 1.1 million kg since the 1960s, with an exvessel value population, which may have a greater effect on reproductive of US$11.3 million for the 2007–2008 season (Hebert et al. potential. 2008). In the 2000–2001 through 2007–2008 seasons, an av- The goal of this study was to determine the number of erage of 192 permit holders participated in the fishery (Hebert derelict crab pots and the magnitude of crab entrapment et al. 2008). Commercial Dungeness crab fishing seasons for resulting from pot loss by the commercial Dungeness crab most of northern southeastern Alaska consist of a 15 June–15 fishery in southeastern Alaska. In 2009 and 2010, we conducted August summer season, followed by a 16 August–30 September systematic side-scan sonar surveys of fishing grounds to late summer closure, and a 1 October–30 November fall season identify derelict pots, then retrieved pots by scuba diving 424 MASELKO ET AL. Downloaded by [Department Of Fisheries] at 19:52 28 May 2013

FIGURE 1. Locations of the 11 areas surveyed (purple shading) with side-scan sonar for ghost fishing by derelict Dungeness crab pots (arrows). Each study was within a specific ADFG statistical area; boundaries are indicated with dashed lines. A map of all ADFG statistical areas is available at www.adfg.alaska.gov/ static/fishing/PDFs/commercial/maps/chart05 salm shell all.pdf. [Figure available in color online] GHOST FISHING IN THE DUNGENESS CRAB FISHERY 425

and inventoried their contents. The known distribution of the 30 August in 2010. Due to the relatively small size of commer- commercial fishery harvest and effort provided a basis for cial Dungeness crab pots (∼1 m in diameter), a frequency of expansion of derelict crab pot and entrapped crab estimates to 900 kHz was used; this limited the effective survey swath width the entire southeastern Alaska region. to 80 m, 40 m on either side of the vessel survey track. The side- scan sonar topside processing unit was linked to a real-time METHODS differential Global Positioning System (GPS); together with ca- Field methods.—Over the two study years, derelict crab pot ble offset calculations, this allowed accurate geopositioning of surveys were conducted in eleven ADFG statistical areas, cho- the towfish and subsequently of sonar targets. The sonar towfish sen based on their proximity to Juneau as well as on their large was towed between 5 and 10 m above the bottom at a speed of annual harvest and fishing effort. Nine areas near Juneau were 5.5–9 km/h in order to obtain the best imagery for differentiating surveyed in 2009 and two different areas near Petersburg in 2010 targets. The 11 areas surveyed contained a total of 93.71 km2 of (Figure 1). The areas surveyed produced 19.8% of the commer- fishing grounds. cial Dungeness crab harvest in southeastern Alaska in 2009 and After the surveys, targets (potential derelict crab pots) 28.3% in 2010. The surveys were conducted during the sum- were identified using bottom imagery acquired with Sonar- mer commercial closure in order to minimize interactions with Wiz (Chesapeake Technologies) software in conjunction with actively fishing crab pots. the Klein 3900 side-scan sonar. The factors used to identify Within each area Dungeness crab habitat was identified and potential derelict crab pots included target dimensions, shape, mapped using ArcGIS. Three criteria were used: bottom depth and sonar image clarity. Since we did not want to bias the less than 30 m; muddy, sandy, or silty substrate; and 10 years sampling to favor newer, more intact crab pots having more of ADFG aerial survey data showing the distribution of com- unambiguous imagery, all sonar objects that even vaguely re- mercial crab pot buoys. These habitat maps were used to select sembled a Dungeness crab pot were selected. Subsequently, a the side-scanning sonar transects and to estimate the area of report was produced for each of the 11 surveyed areas detail- fishing grounds within each area. It is important to note that ing the GPS location, dimensions, and side-scan sonar image our expanded estimates of ghost-fishing loss to the fishery de- of each target. Within each area, a subsample of all targets was pend heavily on accurate available fishing ground area estimates. randomly selected for retrieval (Table 1). The number of tar- However, these estimates would be more accurate with better gets selected was based on available dive time at each location habitat mapping, which was not available at the time of this and varied from 17% (n = 24; Central Lynn Canal) to 54.5% study. (n = 6; Auke Bay) of the total identified targets at each area To locate and identify derelict crab pots, Klein 3900 dual- (Table 1). frequency side-scan sonar was towed from a 10-m research Pot retrieval was conducted from a larger chartered vessel vessel from 17 August to 15 September in 2009 and from 16 to with the capability of hauling potentially heavy, sediment-laden

TABLE 1. Density and number of derelict crab pots from 11 commercial Dungeness crab harvest areas as determined through side-scan sonar and scuba pot retrieval surveys in southeastern Alaska. Duncan Canal and Wrangell were surveyed in 2010 and all other areas in 2009. Detection probabilities for the three areas that were not dive-sampled were computed as the mean of the 2009 surveys.

Derelict Sonar Dive Veriﬁed Detection Survey pots/km2

Downloaded by [Department Of Fisheries] at 19:52 28 May 2013 targets targets targets probability Derelict pots area; (95% CI) di 2 tî Area (Statistical Area) (Ni)(ni)(di)(ρî = )(tî = ρî Ni )km(ai)(Tî = ) ni ai Auke Bay (111-50) 11 6 2 0.333 3.7 2.44 1.5 (0.0– 3.0) Berners Bay (115-20) 46 12 11 0.917 42.2 4.58 9.2 (7.5–10.0) Limestone Inlet (111-90) 7 3 1 0.333 2.3 0.52 4.5 (0.0–13.5) Stephens Passage (111-40) 29 6 5 0.833 24.2 7.29 3.3 (2.0– 4.0) Central Lynn Canal (115-10) 142 24 24 1.000 142.0 5.45 26.1 Young Bay (111-41) 7 3 2 0.667 4.7 2.23 2.1 (0.0– 3.1) Duncan Canal (106-43) 204 41 31 0.756 154.2 24.04 6.4 (5.4– 7.5) Wrangell (108-40) 222 52 52 1.000 222.0 9.47 23.4 Howard Bay (112-61)a 9 0.681 6.1 0.97 6.3 (4.9– 7.6) Funter Bay (112-63)a 4 0.681 2.7 0.27 10.1 (7.8–12.1) Hawk Inlet (112-65)a 13 0.681 8.9 1.65 5.4 (4.2– 6.5) Total 694 147 128 0.871 604.3 58.91 10.3 (9.7–10.7)

aLocations at which side-scan sonar surveys were performed without dive-sampling pot retrieval surveys. 426 MASELKO ET AL.

derelict crab pots. The derelict crab pots were located by plac- bility, ρî at area i is calculated as ing a buoyed anchor on the target GPS coordinates. Two divers then descended, attached a 10-m rope to the anchor and swam di ρî = , (2) a 10-m radius circular search-swath around it. If visibility was ni poor, the divers conducted two search circles, with one diver positioned at 5 or 10 m and the other at 2.5 or 7.5 m from where di is the number of verified targets found by divers and the anchor, to increase the probability of encountering the pot ni is the number of dive-sampled targets at each area i.Note by physical contact. Pots were located in zero visibility using that only targets found (either crab pots or other objects) were this technique. If multiple pots were found by divers at a tar- used for the calculations, as not finding a target was attributed get, a separate retrieval line was attached to each pot. Multiple to poor visibility or misplacement of the anchor and not to pots were also sometimes accidentally retrieved when tangled misidentification of the sonar target. The density of derelict with a single pot attached to the retrieval line. In all cases of crab pots per square kilometer may then be calculated as both types of multiple pot retrieval, sonar images confirmed the location of multiple targets in close proximity to the se- tî Tî = , (3) lected pot. When a derelict crab pot was located and visibility ai permitted, a video of the area surrounding the in situ crab pot and its content was recorded. The divers then attached a line where ai is the bottom area surveyed by side-scan sonar within to the crab pot, threading it around the pot lid in order to pre- area i. The estimated number of Dungeness crabs entrapped in vent accidental opening during retrieval. Finally, a hydraulic derelict commercial crab pots at area i is crab block on the chartered fishing vessel was used to retrieve = γ δˆ , the pot. eî tî ˆ i i (4) After retrieval, each pot was photographed and inspected, and γ data pertaining to the condition, content, and probable cause of where ˆ i is the proportion of all retrieved derelict pots that were δˆ pot loss and crab entrapment were recorded. ADFG regulatory ghost fishing and i is the mean number of entrapped crabs per γ buoy tags labeled with the fishing year were found on 12% (15 pot at area i. ˆ i can be calculated as pots) of all retrieved pots. These were used to assign derelict gi crab pots to age-classes and to establish age-class standards. γˆ = , (5) i d Pots without buoy tags were then assigned to age-classes by i qualitative assessment of the degree of biofouling and metal where gi is the number of retrieved pots containing crabs at degradation relative to these known-age pots. area i. Note that d is the number of retrieved derelict crab pots Any Dungeness crabs contained in the retrieved pots were i and is not necessarily equal to d . This is due to the fact that measured, sexed, and weighed and their shell and appendage i some verified crab pots were lost during retrieval and some were conditions were recorded; after this they were released. Shells tangled with other derelict crab pots, resulting in multiple pots and appendages of dead Dungeness crabs were present in eight being retrieved from a single target. In addition, some derelict pots but were not included in the analysis because we could not crab pots were retrieved although they were not sonar targets. definitely account for the number of crabs the parts represented. Subsequently, δˆ can be calculated as No dead crabs were found in derelict pots and no fish or other i invertebrates were entrapped, live or dead. All retrieved crab gi h=1 chi Downloaded by [Department Of Fisheries] at 19:52 28 May 2013 pots were inventoried and subsequently brought back to port for δˆ = , (6) i g recycling. i Analysis.—For each area sampled, the subsequent derelict where chi is the number of entrapped crabs in ghost-fishing crab crab pot and entrapped crab counts were converted to densi- pot h at area i. ties based on the area swept by side-scan sonar. These density The density of entrapped Dungeness crabs per square kilo- estimates were expanded by the area of fishing grounds to ob- meter at area i is calculated as tain estimates of the total abundance of derelict crab pots and entrapped crabs for each area. ˆ = eî . More specifically, the estimated number of derelict commer- Ei (7) ai cial crab pots in area i is Finally, the total numbers of derelict crab pots and entrapped crabs at each area i were calculated by multiplying their esti- tî = ρî Ni , (1) mated densities, Tî and Eˆ i ,byAi, the area of fishing grounds in area i. where ρî is the detection probability and Ni is the total number The 95% confidence intervals for the number of lost commer- of targets identified with side-scan sonar. The detection proba- cial Dungeness crab pots per square kilometer at each area i (Tî ) GHOST FISHING IN THE DUNGENESS CRAB FISHERY 427

was obtained using a bootstrap distribution of 1,000 samples was forced to be zero since we were not taking into account (Efron and Tibrshirani 1993) by subsampling with replacement noncommercial harvest. from the dive-sampled targets (ni) at each of the 11 areas and recalculating ρî . The 95% confidence interval for the number of entrapped Dungeness crabs was obtained in the same fash- RESULTS 2 ion, except that γˆ i and δî were recalculated for each ρî based An overall total of 63% of the 93.71 km of fishing grounds on sampling with replacement from the distribution of retrieved were surveyed; within areas, this ranged from 32% to 99% 2 crab pots (di ) at each area i. of 0.55–49.61 km (Table 1). A total of 694 targets (potential In 2009, the three areas most distant from Juneau—Howard derelict commercial crab pots) were identified using side-scan Bay, Funter Bay, and Hawk Inlet—were not sampled by divers sonar, of which 147 were sampled. Of the sampled targets, to verify sonar targets due to lack of time. For these areas, the 87% (n = 128) were verified to be derelict crab pots, with 123 mean of the detection probabilities from the other six areas was being successfully retrieved and 5 being located but not retrieved used to calculate derelict crab pot and entrapped Dungeness crab due to their being either snagged or mudded in on the bottom densities. (Tables 1, 2). In order to calculate the proportion of the mean annual har- Detection probability increased between the two survey vest lost due to ghost fishing, the total number of entrapped years, from 0.681 in 2009 to 0.880 in 2010 (Table 1). This large Dungeness crabs was divided by the mean annual commercial increase was likely due partly to surveyors’ having to learn to harvest in numbers (obtained from the ADFG) for each area. identify targets in 2009 and partly to the facts that different areas The mean of the 1991–1992 through 2010–2011 seasons, rather were surveyed and there were separate dive teams in the 2 years. than the 2009–2010 and 2010–2011 values, was used to pre- We were able to examine ghost fishing and estimate age for all serve confidentiality by grouping information, as required by of the derelict crab pots retrieved, but were only able to positively Alaska law (Confidential nature of certain reports and records determine legal rot cord compliance for a portion of them. Of 1970). This was considered valid as the standard deviation of the 123 pots retrieved, 32.5% (n = 40) were determined to be harvest in numbers is greater among areas (σ = 7.43 × 105) ghost fishing (Table 2). Of the 106 pots for which we were able to than between years (σ = 1.75 × 105). determine whether legal rot cord was used, 91% (n = 96) were or Data by area on annual Dungeness crab commercial harvest had been legally rigged. Only 1 of 18 pots in Stephens Passage, in numbers and effort in terms of permit holders and pot pulls 4 of 33 pots in Duncan Canal, and 5 of 40 pots in Wrangell were obtained from the ADFG. Three linear regressions of the were noncompliant. There was no difference in ghost-fishing measures of harvest, permit holders, and pot pulls against each rates between overall noncompliant and compliant derelict crab other and six regressions of the these factors against total lost pots; 40% of noncompliant (n = 4) and 35% of compliant (n = pots and total entrapped crabs were conducted. The intercept 34) derelict crab pots were ghost fishing. However, there was

TABLE 2. Density and number of entrapped Dungeness crabs in ghost-ﬁshing derelict crab pots from 11 commercial harvest areas as determined through side-scan sonar and scuba pot retrieval surveys in southeastern Alaska. Note that di is the actual number of pots retrieved after adjusting di for the pots lost or found during retrieval of derelict crab pots originally detected by side-scan sonar.

Ghost- Entrapped Percent Estimated Entrapped Pots ﬁshing crabs ghost Entrapped entrapped crabs/km2 retrieved pots captured ﬁshing crabs/pot crabs (95% CI)

Downloaded by [Department Of Fisheries] at 19:52 28 May 2013 gi gi gi h=1 chi eî Area (d )(gi )(= chi)(γˆ i = )(δî = )(eî = tî γˆ i δî )(Eˆ i = ) i h 1 di gi ai Auke Bay 2 1 6 50.0% 6.0 11.0 4.5 (0.0–18.0) Berners Bay 13 5 26 38.5% 5.2 84.3 18.4 (2.1–39.7) Limestone Inlet 1 0 0 0.0% 0.0 0.0 0.0 Stephens Passage 5 0 0 0.0% 0.0 0.0 0.0 Central Lynn Canal 27 8 25 29.6% 3.1 131.5 24.1 (6.8–47.3) Young Bay 2 1 2 50.0% 2.0 4.7 2.1 (0.0–6.3) Duncan Canal 33 10 63 30.3% 6.3 294.5 12.3 (4.9–21.1) Wrangell 40 15 93 37.5% 6.2 516.2 54.5 (26.9–87.3) Howard Baya 28.0% 2.7 4.1 4.8 (0.8– 8.6) Funter Baya 28.0% 2.7 1.8 7.7 (1.2–13.7) Hawk Inleta 28.0% 2.7 5.9 4.1 (0.7–7.3) Total 123 40 215 32.5% 3.6 1,055.1 12.0 (9.2–14.8)

aLocations at which side-scan sonar surveys were performed without dive-sampling pot retrieval surveys. 428 MASELKO ET AL.

a difference in the ghost-fishing rate between compliant and and crab entrapment were all highly significant (P < 0.001). noncompliant pots that had recently been lost (i.e., less than There was one derelict crab pot for every 570 Dungeness crabs 2 years previously); 100% of recently lost noncompliant pots harvested (derelict crab pots = 0.00175·annual harvest; r2 = (n = 3) were ghost fishing but only 41% of compliant pots (n = 0.89), 209 pot pulls (derelict crab pots = 0.00480·annual pot 29). Ghost-fishing rates generally declined with the length of pulls; r2 = 0.88), or three permit holders (derelict crab pots = time that pots had been lost. The age of the 123 derelict crab 0.333·annual permit holders; r2 = 0.87). Likewise, the number pots retrieved ranged from less than 1 year to more than 6 years; of entrapped Dungeness crabs was estimated as 1 for every 268 of those, 19% (n = 23) had been lost for less than 2 years, crabs harvested (entrapped crabs = 0.00372·annual harvest; r2 = 41% (n = 50) for 3–4 years, 37% (n = 46) for 5–6 years, and 0.98), 98 pot pulls (entrapped crabs = 0.0102·annual pot pulls; 3% (n = 4) for more than 6 years. Those lost for 4 years or r2 = 0.98), or permit holder (entrapped crabs = 0.705·annual less had significantly higher ghost-fishing rates (43.8%) than permit holders; r2 = 0.96). The mean annual Dungeness crab the 5–6-year group (17.4%) and the older than 6 years group harvest in all of southeastern Alaska for the 1991–1992 through (0.0%). However, the two oldest groups did not differ in their 2010–2011 commercial seasons was 1.95 million crabs. Thus, ghost-fishing rates. we estimate that there were 3,072 derelict crab pots containing A number of factors were responsible for crab entrapment. 6,525 Dungeness crabs at the time of the survey. This amounts to Overall, 32.5% (n = 40) of the derelict crab pots retrieved 0.37% of the commercial Dungeness crab harvest in southeast- contained Dungeness crabs. In 50% (n = 20) of the ghost-fishing ern Alaska being unavailable to the fishery due to ghost fishing. pots, crabs were entrapped with no apparent lid obstruction Among areas, the highest entrapment was 1.28% for Auke Bay except that the lid was lying flush with the rim; in 10% (n = 4), (Table 3). the pots were illegally rigged; for 15% (n = 6) marine growth restricted the opening of the lid, whereas in 13% (n = 5) the rubber band used to keep the lid closed was obstructing the lid and preventing it from opening, even though the rot cord had DISCUSSION disintegrated. The cause of crab entrapment was undetermined Although derelict crab pots eventually get mudded in, cor- for the remaining 13% (n = 5) of the ghost-fishing pots. roded, or otherwise mechanically disabled, they ghost-fish for By area, the percentage of ghost-fishing pots ranged from much longer in southeastern Alaska than previously reported for 0% to 50%, and the average catch rate was 3.6 crabs per pot other areas. Our estimated 6-year minimum ghost-fishing life (Table 2). The highest number of crabs found in any one pot for derelict crab pots contrasts sharply with the 2-year ghost- was 19; this pot was from the Wrangell area. The 123 pots fishing life previously estimated for derelict crab pots (Breen retrieved in 2009 and 2010 contained a total of 215 entrapped 1987; June 2007). However, the experiments of Breen (1987) crabs (Table 2). Of all entrapped crabs, 75% (n = 162) were and June (2007) were not continued past 2 years, and therefore malesoflegalsize(>165 mm), 5% (n = 10) were sublegal the actual efficacy of an older derelict crab pot was never phys- males, and 20% (n = 43) were females. Furthermore, 70% (n = ically evaluated. The fact that buoy tags labeled with fishing 30) of entrapped females were larger than 165 mm. year are required for the southeastern Alaska commercial Dun- The density of entrapped Dungeness crabs was highly vari- geness crab fishery also assisted us in arriving at this finding. able. Six of the areas had an entrapment density of less than Stevens et al. (2000) recovered derelict commercial Tanner crab 6 crabs/km2 (Table 2). The highest density of entrapped Dun- pots that had been lost for at least 2 years since the closure of geness crabs was in the Wrangell area (54.5 crabs/km2), fol- the commercial fishery and found no difference in ghost-fishing 2 Downloaded by [Department Of Fisheries] at 19:52 28 May 2013 lowed by the Central Lynn Canal (24.1 crabs/km ) and Berners rates between recently lost and older (presumably more than Bay (18.4 crabs/km2) (Table 2). 2-year-old) pots recovered from Chiniak Bay. Therefore, this Correlations were calculated for derelict crab pots and en- discrepancy in the estimated ghost-fishing life span could also trapped crabs, with the recent 20-year history of commercial be due to unique characteristics of the Alaskan marine environ- harvest and effort compiled by the ADFG. All three mea- ment that retard the corrosion of pots. sures of Dungeness crab harvest and effort were significantly Corrosion is a complex multiphase process that changes over (P < 0.001) correlated. Harvest was highly correlated with time and is most likely the primary source of derelict crab pot the number of permit holders (r2 = 0.998) and the number degradation. Corrosion rates are affected by various physical of pot pulls (r2 = 0.999), while the number of pot pulls was and biological factors, including the activity of sulfate-reducing highly correlated (r2 = 0.997) with the number of permit bacteria, which require rich nutrient sources for metabolism. holders. Each of these measures provided useful predictions Increased water temperature and nutrient pollution also sub- of the number of derelict crab pots and entrapped Dungeness stantially increase long-term corrosion rates (Melchers 2008). crabs. Nutrient pollution from runoff and water temperatures are gen- The six regressions of the number of crabs harvested and erally higher at the more developed lower latitudes of Pacific fishing effort in terms of pot lifts and vessels (from the 1991– Northwest coastal waters. In the cold, relatively unpolluted ma- 1992 through 2010–2011 season data) on the estimated pot loss rine waters of southeastern Alaska, we therefore expect derelict GHOST FISHING IN THE DUNGENESS CRAB FISHERY 429

TABLE 3. Total derelict crab pots, total entrapped crabs, mean annual commercial harvest for the 1991–1992 through 2010–2011 seasons, and entrapped Dungeness crabs determined through side-scan sonar and scuba pot retrieval surveys as a proportion of the harvest from the 11 commercial harvest areasin southeastern Alaska. See Tables 1 and 2 for deﬁnitions of unexplained parameters.

Entrapped crabs as Area of Derelict Total Total Mean proportion of mean fishing pots/km2 derelict entrapped annual annual harvest grounds, (95% CI) pots crabs harvested (95% CI) 2 tî Ai Eˆ i Area km (Ai)(Tî = )(Ai Tî )(Ai Eˆ i ) crabs (Hˆ i )() ai Hˆ i Auke Bay 2.77 1.5 (0.0–3.0) 4.2 12.5 978 1.28% (0.00–5.11%) Berners Bay 5.29 9.2 (7.5–10.0) 48.7 97.4 10,749 0.91% (0.10–1.95%) Limestone Inlet 0.55 4.5 (0.0–13.5) 2.5 0.0 338 0.00% Stephens Passage 7.30 3.3 (2.0–4.0) 24.2 0.0 11,331 0.00% Central Lynn Canal 5.99 26.1 156.1 144.5 27,022 0.53% (0.15–1.05%) Young Bay 2.89 2.1 (0.0–3.1) 6.1 6.1 955 0.63% (0.00–1.90%) Duncan Canal 49.61 6.4 (5.4–7.5) 318.3 607.7 139,943 0.43% (0.17–0.75%) Wrangell 14.42 23.4 338.0 785.9 226,972 0.35% (0.17–0.55%) Howard Baya 1.54 6.3 (4.9–7.6) 9.7 7.4 3,503 0.21% (0.03–0.38%) Funter Baya 0.83 10.1 (7.8–12.1) 8.4 6.4 2,904 0.22% (0.04–0.39%) Hawk Inleta 2.52 5.4 (4.2–6.5) 13.5 10.3 5,416 0.19% (0.03–0.34%) Total 93.71 10.3 (9.7–10.7) 961.3 1124.5 430,111 0.26% (0.20–0.32%)

aLocations where side-scan sonar surveys were performed without dive-sampling pot retrieval surveys.

crab pots to ghost-fish for at least 7 years, resulting in a substan- penses incurred by Dungeness crab fishermen. The relatively tial cumulative loss of crabs due to ghost fishing. large numbers of nonlegal crabs that we found retained by ghost- The long-term ghost-fishing efficacy of derelict crab pots fishing derelict crab pots suggest that consideration should also would be reduced by an effective biodegradable escape be given to increasing the size of the cull rings to enable all mechanism. We found that 40% of the pots that met regulatory sublegal male crabs to escape. requirements for the biodegradable escape mechanism were Although reducing the ghost-fishing rate of derelict crab pots still ghost-fishing. We identified two main factors, aside from is important, reducing the pot loss rate is just as important. noncompliant biodegradable escape mechanisms, that affected While this is largely outside the scope of the current work, the long-term derelict crab pot ghost-fishing rates: marine growth effectiveness of such measures as reducing pot density, regulat- and metal fatigue. Marine growth, specifically barnacles ing maximum soak time, reducing gear conflicts, and adjusting Semibalanus spp. and algae Saccharina spp., was observed to fishing seasons to avoid pots being iced into the heads of inlets disable crab pot lid openings. Metal fatigue resulted in released should be investigated. lids lying flat on top of the crab pots, whereas in new pots the The proportion of derelict crab pots that we found actively lids spring open when released. The configuration of older crab ghost fishing is consistent with previously reported rates. June

Downloaded by [Department Of Fisheries] at 19:52 28 May 2013 pots resulted in there being no gap between the lids and the (2007) estimated that 37% of derelict crab pots were ghost fish- frames and appeared to prevent crabs from escaping. ing in Puget Sound; our overall estimate for Southeast Alaska is The biodegradable escape mechanisms and cull rings of Dun- comparable at 32.5% (95% CI = 14.9–44.1%). Given that crab geness crab pots should be redesigned to reduce crab mortality pots of the same design are used in both fisheries, this relatively caused by ghost fishing. Amending the regulations to require small discrepancy is likely due to the higher crab densities found a cut in the pot lid or a pot side laced with biodegradable rot in Puget Sound. With higher crab densities, the probability of cord, as provided for in the Washington Dungeness crab fishery crabs encountering a derelict crab pot would be higher, resulting (Commercial shellfish pot gear: escape mechanism required. in more pots being populated by crabs. Additionally, while we 1987), should be investigated. This mechanism would be re- found over 90% of derelict crab pots to be legally rigged, only silient to mudding in, algal growth, and metal fatigue, while 76% of those retrieved in Puget Sound were legally rigged (June still allowing crabs to escape as the derelict crab pot aged. An 2007). These two differences likely account for the discrepancy even better solution would be to use biodegradable cull panes, in annual ghost-fishing estimates between Southeast Alaska and as is currently being investigated for the Chesapeake Bay blue Puget Sound. crab Callinectes sapidus fishery (Havens et al. 2008). A newly Our estimates of ghost-fishing rates reflect the number of designed biodegradable escape mechanism would have to be crabs entrapped at the time of the survey; however, Breen (1987), tested for its effectiveness and designed to minimize the ex- June (2007), and Antonelis et al. (2011) calculate a cumulative 430 MASELKO ET AL.

annual mortality. June (2007) suggests that the number of en- ACKNOWLEDGMENTS trapped crabs at any one time is representative of one-twelfth of This study was funded by the NOAA Marine Debris Division. the annual entrapment. The use of 12 as a multiplier to convert Special thanks to Eric Brown, Andrew Eller, and Haley Poole our ghost-fishing estimates to annual losses assumes that 100% for help with field work and data processing and to the divers: of entrapped crabs die within 30 d, that no crabs escape, and that Manuel Cruz, Peter Fischel, David Francksen, Bill Heard, Justin the catch rate is constant. However, a number of studies suggest Keese, Pat Malecha, Jennifer Mondragon, Matthew Nardi, Kalei that these assumptions are rather liberal. First, the mortality rate Shotwell, Elizabeth Siddon, Bob Stone, Ryan Wattam, Brad is not 100%. Breen (1987) used an in situ experiment to calculate Weinlaeder, and Alex Wertheimer. We also thank the review- an annual mortality rate of about 50% for entrapped Dungeness ers for their invaluable contribution. Reference to trade names crabs, while Antonelis et al. (2011) found a similar annual mor- does not imply endorsement by the U.S. Government. The tality rate of 46%. Second, some entrapped crabs do escape. findings and conclusions in this report are those of the au- In an experiment in Puget Sound, High (1976) demonstrated thor(s) and do not necessarily represent the views of the funding that 45% of legal-sized crabs in pots without rot cord escaped agency. through functioning cull rings and/or triggered tunnel openings; the escapement rate was 77% for sublegal males and 65% for females. In a Salish Sea study, Antonelis et al. (2011) found a REFERENCES Antonelis, K., D. Huppert, D. Velasquez, and J. June. 2011. Dungeness crab smaller entrapped crab escapement rate of 37% for legal-sized mortality due to lost traps and a cost-benefit analysis of trap removal in males, 50% for sublegal males, and 41% for females. In a sim- Washington State waters of the Salish Sea. North American Journal of Fish- ilar experiment with Dungeness crabs in the Columbia River eries Management 31:880–893. estuary, Muir et al. (1984) found that 49% of legal males and Barber, J. S., and J. S. Cobb. 2007. Injury in trapped Dungeness crabs (Cancer 66% of sublegal males escaped from pots during a 22-d study. magister). ICES Journal of Marine Science 64:1–9. Barry, S. T. 1981. Coastal Dungeness crab study. State of Washington, Depart- Finally, there is seasonal variability in catch rates. Larger catch ment of Fisheries, Project Progress Report 1–35-R, Olympia. rates would be expected during late summer and early fall (when Bennett, D. B. 1973. The effect of limb loss and regeneration on the growth of our surveys were conducted) than in late winter (after the fall the edible crab Cancer pagurus L. Journal of Experimental Marine Biology fishery) and spring when water temperatures are cold and crabs and Ecology 13:45–53. are inactive (Breen 1987; Antonelis et al. 2011). Bishop, G. H., M. S. M. Siddeek, and J. M. Rumble. 2007. Growth of male Dungeness crabs in southeastern Alaska. Pages 1339–1342 in J. Nielsen, It appears therefore, that based on our entrapment estimate J. J. Dodson, K. Friedland, T. R. Hamon, J. Musick, and E. Verspoor, editors. of 0.37% of harvest in the Southeast Alaska commercial Dun- Reconciling fisheries with conservation: proceedings of the fourth world geness crab fishery at the time of the survey, the cumulative fisheries congress. American Fisheries Society, Symposium 49, Bethesda, annual entrapment is at most 4.47% of the annual harvest. The Maryland. resulting cumulative annual mortality is half that: 2.23% of Breen, P. A. 1987. Mortality of Dungeness crabs caused by lost traps in the Fraser River Estuary, British Columbia. North American Journal of Fisheries the annual harvest. In contrast, the estimated mortality for the Management 7:429–435. Fraser River district of British Columbia was 7% (Breen 1987) Breen, P. A. 1990. A review of ghost fishing by traps and gill nets. NOAA and 30–40% for Puget Sound (June 2007). However, a more Technical Memorandum NMFS-SWFSC-154:571–599. recent Puget Sound study (Antonelis et al. 2011) yielded a more Butler, T. H. 1961. Growth and age determination of the Pacific edible crab, similar ghost-fishing mortality estimate of 4.5% of the annual Cancer magister Dana. Journal of the Fisheries Research Board of Canada 18:873–889. harvest (95% CI, 2.6–6.4%). Perplexingly, the estimate most Commercial shellfish pot gear: escape mechanism required. 1987. Washington different from ours (June 2007) came from a study which used Administrative Code, Title 220-52-035.

Downloaded by [Department Of Fisheries] at 19:52 28 May 2013 similar side-scan sonar methods, while the estimate closest to Confidential nature of certain reports and records. 1970. Alaska Statutes, section ours (Antonelis et al. 2011) used very different methods, em- 16.05.815. ploying fish ticket data, harvest reporting cards, interviews, and Dahlstrom, W. A. 1975. Report of lost crab trap recovery. State–Federal Dun- geness Crab Scientific Committee meeting, Bellevue, Washington. surveys to estimate the number of commercial and recreational Durkin, J. T., K. D. Buchanan, and T. H. Blahm. 1984. Dungeness crab leg loss traps lost annually. in the Columbia River Estuary. Marine Fisheries Review 46:22–24. The magnitude of ghost fishing in the southeastern Alaska Efron, B., and R. J. Tibrshirani. 1993. An introduction to the bootstrap. Chapman commercial Dungeness crab fishery was smaller than expected and Hall, New York. based on prior studies. Lower crab density may explain the Escape mechanism for shellfish and bottomfish pots. 1978. Alaska Administra- tive Code, Title 5, section 39.145. smaller catch rates, while the greater age of derelict pots could Havens, K. J., D. M. Bilkovic, D. M. Stanhope, and K. T. Angstadt. 2008. Crab account for their decreased catchability. Another factor may be trap with degradable cull ring panel. U.S. patent application 12/394,917. predation by octopuses, as eight of the retrieved pots contained October 8, 2009. Dungeness crab carapaces and appendages with markings in- Hebert, K., J. Stratman, K. Bush, G. Bishop, C. Siddon, J. Bednarski, and A. dicative of octopus predation (D. Scheel, Alaska Pacific Uni- Messmer. 2008. 2009 Report to the Board of Fisheries on Region 1 shrimp, crab, and scallop fisheries. Alaska Department of Fish and Game, Fisheries versity, personal communication). However, the magnitude of Management Report 08-62, Douglas. the effect of octopus predation on ghost-fishing estimates is still High, W. L. 1976. Escape of Dungeness crabs from pots. Marine Fisheries unknown. Review 38:19–23. GHOST FISHING IN THE DUNGENESS CRAB FISHERY 431

High, W. L., and D. D. Worlund. 1979. Escape of king crab Paralithodes River Estuary. North American Journal of Fisheries Management 4:552– camtschatica from derelict pots. NOAA Technical Report NMFS SSRF-734. 555. June, J. 2007. A cost-benefit analysis of derelict fishing gear removal in Puget Paul, A. J., J. M. Paul, and A. T. Kimker. 1993. Starvation resistance in Alaskan Sound, Washington. Natural Resources Consultants, Report prepared for the crabs. Alaska Department of Fish and Game, Division of Commercial Fish- Northwest Straits Foundation, Seattle. eries, Regional Information Report 2A93–03, Anchorage. Kimker, A. T. 1992. Tanner crab survival in closed pots. Alaska Department Sheldon, W. W., and R. L. Dow. 1975. Trap contributions to losses in the of Fish and Game, Division of Commercial Fisheries, Regional Information American lobster fishery. U.S. National Marine Fisheries Service Fishery Report 2H90-05, Anchorage. Bulletin 73:449–451. Kruse, G. H., and A. T. Kimker. 1993. Degradable escape mechanisms for pot Stevens, B. G. 1996. Crab bycatch in pot fisheries: causes and solutions. Pages gear. Alaska Department of Fish and Game, Summary Report to the Alaska 151–158 in T. Wray, editor. Proceedings of the Solving Bycatch Workshop. Board of Fisheries 5J93-01, Juneau. Alaska Sea Grant College Program 96–03, Seattle. Matsuoka, T., T. Nakashima, and N. Nagasawa. 2005. A review of ghost fishing: Stevens, B. G., I. Vining, S. C. Byersdorfer, and W. E. Donaldson. 2000. Ghost scientific approaches to evaluation and solutions. Fisheries Science Research fishing by Tanner crab (Chionocetes bairdi) pots off Kodiak, Alaska: pot 71:691–702. density and catch per trap as determined from sidescan sonar and pot recov- Melchers, M. E. 2008. Development of new applied models for steel corrosion ery data. U.S. National Marine Fisheries Service Fishery Bulletin 98:389– in marine applications, including shipping. Ships and Offshore Structures 399. 3:135–144. Stone, R. P., and C. E. O’Clair. 2001. Seasonal movements and distribution Muir, W. D., J. T. Durkin, T. C. Coley, and G. T. McCabe. 1984. Escape of Dungeness crabs, Cancer magister, in a glacial Southeast Alaska estuary. of captured Dungeness crabs from commercial crab pots in the Columbia Marine Ecology Progress Series 214:167–176. Downloaded by [Department Of Fisheries] at 19:52 28 May 2013 This article was downloaded by: [Department Of Fisheries] On: 28 May 2013, At: 19:52 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 Genetic Population Structure of Bull Trout in the East Fork Bitterroot River Drainage, Montana Leslie G. Nyce a b , Lisa Eby a , Christopher G. Clancy b , Sally Painter c & Robb F. Leary d a Wildlife Biology Program, University of Montana, 32 Campus Drive, Missoula, Montana, 59812, USA b Montana Fish, Wildlife and Parks, 1801 North First Street, Hamilton, Montana, 59840, USA c Conservation Genetics Laboratory, Division of Biological Sciences, University of Montana, 32 Campus Drive, Missoula, Montana, 59812, USA d Montana Fish, Wildlife and Parks, Conservation Genetics Laboratory, Division of Biological Sciences, University of Montana, 32 Campus Drive, Missoula, Montana, 59812, USA Published online: 02 Apr 2013.

To cite this article: Leslie G. Nyce , Lisa Eby , Christopher G. Clancy , Sally Painter & Robb F. Leary (2013): Genetic Population Structure of Bull Trout in the East Fork Bitterroot River Drainage, Montana, North American Journal of Fisheries Management, 33:2, 432-445 To link to this article: http://dx.doi.org/10.1080/02755947.2013.768565

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 33:432–445, 2013 C American Fisheries Society 2013 ISSN: 0275-5947 print / 1548-8675 online DOI: 10.1080/02755947.2013.768565

ARTICLE

Genetic Population Structure of Bull Trout in the East Fork Bitterroot River Drainage, Montana

Leslie G. Nyce* Wildlife Biology Program, University of Montana, 32 Campus Drive, Missoula, Montana 59812, USA; and Montana Fish, Wildlife and Parks, 1801 North First Street, Hamilton, Montana 59840, USA Lisa Eby Wildlife Biology Program, University of Montana, 32 Campus Drive, Missoula, Montana 59812, USA Christopher G. Clancy Montana Fish, Wildlife and Parks, 1801 North First Street, Hamilton, Montana 59840, USA Sally Painter Conservation Genetics Laboratory, Division of Biological Sciences, University of Montana, 32 Campus Drive, Missoula, Montana 59812, USA Robb F. Leary Montana Fish, Wildlife and Parks, Conservation Genetics Laboratory, Division of Biological Sciences, University of Montana, 32 Campus Drive, Missoula, Montana 59812, USA

Abstract Investigation of the genetic population structure of Bull Trout Salvelinus confluentus is useful for developing biologically sound conservation and management strategies. We focused on the East Fork Bitterroot River (hereafter, East Fork), Montana, because it is a relatively undisturbed, connected watershed that contains a migratory life history form of Bull Trout. Fin clips were collected from 17 sites: nine East Fork tributaries, the main-stem East Fork, and seven other tributaries across the Bitterroot River drainage. Considering all the population samples, principal components analysis of allele frequencies at 15 microsatellite loci indicated that the East Fork samples formed a cluster that was distinct from the other Bitterroot River tributary samples. Within the East Fork, there was significant divergence among population samples, with pairwise FST ranging from 0.016 to 0.188 and a global FST of Downloaded by [Department Of Fisheries] at 19:52 28 May 2013 0.063. Relative to other Bull Trout studies, levels of genetic variation within our samples were typically higher, while the overall level of differentiation among samples was lower. Based on ONCOR analyses of multiple-locus genotypes, most individuals in the East Fork basin were assigned to their water body of capture, with an average probability of 88%. Within the East Fork, 16 fish that were collected in the main stem were assigned to tributary populations; 26 individuals sampled from seven tributaries were assigned to the main-stem population. In addition, there were five tributaries in which sampled individuals were assigned to tributary populations other than the water body of capture. Based on GeneClass2 analysis, 76 individuals were identified as first-generation migrants. These observations suggest movement and potential gene flow between the main-stem East Fork and its tributaries and between tributaries via the main stem. The main-stem East Fork therefore appears to be an integral component for maintaining the migratory form of Bull Trout in the drainage and may serve as an important vehicle for genetic exchange among tributary populations.

*Corresponding author: [email protected] Received May 18, 2012; accepted January 9, 2013

432 BULL TROUT GENETIC POPULATION STRUCTURE 433

An understanding of the patterns of gene flow across, 2008a, 2011a, 2011b; Warnock et al. 2010). To ensure a broad within, and among river basins provides information about the understanding of Bull Trout genetic population structure, it is degree of isolation among populations of freshwater fishes. important to compare population structure across an array of This information can be valuable for formulating biologically systems and scales, as appropriate restoration and conservation sound management and conservation programs (Slatkin 1987; goals may differ at these levels. Harrison and Hastings 1996; Rieman and Dunham 2000; Lowe In addition to an understanding of genetic population and Allendorf 2010). For example, dispersal that leads to structure, the assessment of Bull Trout life history diversity gene flow among populations, thus resulting in metapopula- is a conservation priority (Montana Bull Trout Restoration tion structure, may be particularly important for buffering pop- Team 2000). Bull Trout exhibit a variety of life histories, ulations against stochastic environmental risk, supporting sink including both resident and migratory individuals that spawn populations, and refounding extirpated populations (Harrison in tributary systems (Rieman and McIntyre 1993; McPhail and and Hastings 1996; Hanski 1998; Dunham and Rieman 1999; Baxter 1996; Northcote 1997). Resident Bull Trout spend their Rieman and Dunham 2000). In addition, gene flow among entire lives within their natal stream or tributary, moving only spawning populations helps to maintain long-term genetic di- short distances (e.g., <2 km; Jakober et al. 1998). Migratory versity within populations. Bull Trout spend 1–4 years in their natal tributaries before The genetic population structure of the Bull Trout Salvelinus migrating—often tens to hundreds of kilometers—to larger confluentus is of particular conservation interest because this habitats (e.g., rivers, lakes, and estuaries), where they forage and species is currently listed as threatened under the Endangered overwinter for several years before returning to the tributaries to Species Act in the USA (USOFR 1998) and is classified as “at spawn (Fraley and Shepard 1989; Rieman and McIntyre 1993; risk” in western Canada (McCart 1997). Most published studies McPhail and Baxter 1996; Swanberg 1997). Migratory Bull of Bull Trout population structure have found relatively high Trout are typically over twice the size of resident adults; this genetic divergence (FST; typical range of 0.100–0.250, with larger size greatly increases fecundity and possibly reproductive differences up to 0.400) not only among populations within success (Fraley and Shepard 1989; Jonsson and Jonsson 1993; large river basins (e.g., Columbia River and Clark Fork River McPhail and Baxter 1996; Lockard 2006). Populations may be basins) but also among populations in adjacent second- and composed of a single life history type, or resident and migratory third-order tributaries (Leary et al. 1993; Kanda and Allendorf fish may coexist and breed as a single panmictic population 2001; Costello et al. 2003; Spruell et al. 2003; Whiteley et al. (Jonsson and Jonsson 1993; McPhail and Baxter 1996; Jakober 2006; DeHaan et al. 2007, 2010, 2011b; Ardren et al. 2011). et al. 1998; Nelson et al. 2002; Homel et al. 2008). Conser- Explanations for the high genetic divergence that is commonly vation of life history diversity, especially migratory forms, is observed among proximate Bull Trout populations include their considered important for population persistence in unstable tendency to spawn in headwater streams, their strong site fi- environments. The existence of migratory Bull Trout can lead delity to natal spawning areas, and their small effective pop- to gene flow among populations and allow for the possibility of ulation sizes (Ne), which lead to increased effects of genetic recolonization after catastrophic events (Thorpe 1994; Rieman drift (McPhail and Baxter 1996; Swanberg 1997; Neraas and et al. 1997; Rieman and Dunham 2000; Bahr and Shrimpton Spruell 2001; Whiteley et al. 2004). Genetic divergence may 2004; Burton 2005). Despite the importance of maintaining life also be affected by environmental stochasticity, such as de- history diversity (e.g., Schindler et al. 2010), the known threats bris flows caused by severe wildfires. Such stochasticity may to Bull Trout population persistence (e.g., habitat degradation, severely reduce the population size and/or isolate the popula- habitat fragmentation, and interactions with exotics) have had

Downloaded by [Department Of Fisheries] at 19:52 28 May 2013 tion, thus resulting in the loss of genetic variation, especially disproportionately greater influences on the migratory form rare alleles (Hallerman 2003; Allendorf and Luikart 2007). An- than on the resident form throughout the Bull Trout’s range thropogenic activities that modify the landscape (e.g., habitat (Swanberg 1997; Neraas and Spruell 2001). fragmentation and altered hydrology) can also reduce gene flow Resident and migratory Bull Trout were once common and Ne and consequently increase genetic divergence (Neraas throughout western Montana, and migratory fish were docu- and Spruell 2001; Morita and Yamamoto 2002; Yamamoto et al. mented regularly in the Bitterroot River drainage throughout the 2004; DeHaan et al. 2007; Morita et al. 2009). 19th century and early 20th century (M. H. Williams, Bitterroot Much of the existing knowledge about the genetic population National Forest, unpublished data). Although Montana Fish, structure of Bull Trout has been obtained from studies com- Wildlife, and Parks (MFWP) and the Bitterroot National For- paring populations across multiple watersheds or populations est monitor trout populations in the Bitterroot River drainage, within disturbed watersheds, including areas where populations little is known about the genetic population structure of Bull have been fragmented by barriers or habitat degradation (Leary Trout and the status of the migratory life history form (C. G. et al. 1993; Kanda and Allendorf 2001; Costello et al. 2003; Clancy, personal communication). Only two areas within the Spruell et al. 2003; Whiteley et al. 2006; DeHaan et al. 2007, Bitterroot River drainage are still known to support migratory 2008b, 2010). Few published studies have been conducted in rel- Bull Trout: the West Fork Bitterroot River (hereafter, West Fork) atively undisturbed, connected systems (but see DeHaan et al. and the East Fork Bitterroot River (hereafter, East Fork). We 434 NYCEETAL.

focused on the East Fork because it is a relatively undisturbed, is estimated to be roughly 5,000–10,000 individuals (MFWP, connected, core conservation area for Bull Trout (Montana Bull unpublished data). Tributaries occupied by Bull Trout contain Trout Scientific Group 1995). Despite these conditions, the Bull both resident and migratory individuals, but the relative propor- Trout migratory life history component has likely been expe- tions of these life history forms are unknown (MFWP, unpub- riencing a decline over the last few decades, as evidenced by lished data). Other native species that are present in the drainage decreases in catches of large fish in the main-stem East Fork include the Westslope Cutthroat Trout, Mountain Whitefish, (Montana Bull Trout Restoration Team 2000; Nelson et al. 2002; Largescale Sucker Catostomus macrocheilus, Longnose Sucker MFWP, unpublished data). Our objectives for this study were Catostomus catostomus, Longnose Dace Rhinichthys catarac- to (1) describe the genetic population structure of Bull Trout in tae, and Slimy Sculpin Cottus cognatus. Nonnative fishes that the East Fork and (2) identify the potential sources of migratory are present in the East Fork include Rainbow Trout O. mykiss, Bull Trout in the drainage. Given results from previous studies, Brown Trout Salmo trutta, and Brook Trout Salvelinus fonti- we predicted that neighboring tributaries would be relatively nalis. Hybrid fishes found in the East Fork include Westslope isolated and would have high FST. We also predicted that fish Cutthroat Trout × Rainbow Trout, Westslope Cutthroat Trout × sampled from the main-stem East Fork would be migratory in- Yellowstone Cutthroat Trout O. clarkii bouvieri × Rainbow dividuals that would be assigned to several different tributaries. Trout, and Bull Trout × Brook Trout. We used an individual assignment test (ONCOR; Kalinowski 2008) and detection of first-generation migrants (GeneClass2; Piry et al. 2004) to help identify which tributaries were produc- METHODS ing migratory fish. Sample collection.—During the summers of 2008 and 2009, we electrofished 17 East Fork tributaries where Bull Trout were previously encountered to obtain a population sample (hereafter, STUDY AREA “sample” refers to a collection of individuals from a population) The Bitterroot River basin is located in southwest Montana from putative spawning populations within the basin (Figure 2). and includes the main-stem Bitterroot River, the West Fork, and To capture fish, we used either a bank electrofishing system the East Fork (Figure 1); the total drainage area is 7,288 km2.The or backpack electrofisher, and we sampled 304.8-m (1,000-ft) majority of the lower drainage is private land, while higher in the sections in an attempt to obtain 50 Bull Trout tissue samples drainage the major landowner is the U.S. Forest Service (USFS). (hereafter, fin clips) representing multiple size-classes. Reaches The Bitterroot River basin supports three native salmonids: were added if 50 fin clips were not collected from the initial Westslope Cutthroat Trout Oncorhynchus clarkii lewisi,Bull 304.8-m section. In most tributaries, we were unable to obtain Trout, and Mountain Whitefish Prosopium williamsoni.Al- 50 fin clips despite our efforts and the addition of extra sam- though the East Fork was our focal area, population samples ple reaches. We also electrofished approximately 26 km of the were included from four tributaries of the main-stem Bitterroot main-stem East Fork by using a johnboat mobile anode elec- River (Daly, Skalkaho, Burnt Fork, and Willow creeks) and three trofishing system to capture fish that were residing in the river. tributaries of the West Fork (North Fork Sheephead, Sheephead, The MFWP and USFS provided additional fin clips collected and Slate creeks) for comparative purposes. from four tributaries of the main-stem Bitterroot River and three The East Fork headwaters are approximately 73 river kilo- tributaries of the West Fork; these samples allowed us to exam- meters (rkm) from the confluence with the West Fork and ine how East Fork genetic diversity fits in a broader geographic main-stem Bitterroot River, with roughly 166 km2 lying within scale. Each captured Bull Trout was measured and weighed; a

Downloaded by [Department Of Fisheries] at 19:52 28 May 2013 the Anaconda-Pintler Wilderness (managed jointly by the caudal fin clip was collected; and the fish was then released. All Beaverhead-Deerlodge and Bitterroot national forests). The wa- fin clips were preserved in vials with a 95% solution of ethanol. tershed encompasses 1,057 km2. Much of the upper drainage is Laboratory methods.—All laboratory work was performed managed by the USFS, with some private lands being present in in the Conservation Genetics Laboratory at the University of the valley bottom. Historical anthropogenic impacts in the wa- Montana, Missoula. Analyses of the fin clips included DNA tershed include timber harvest, forest roads, agriculture, water extraction, PCR, and fragment analysis using 15 variable mi- diversions, and wildfire. Star Falls (rkm 64), a natural barrier crosatellite loci: Omm1128 and Omm1130 (Rexroad et al. 2001); to upstream movement, is the upper limit of known Bull Trout Sco102, Sco105, Sco106, and Sco107 (Washington Depart- occurrence in the East Fork. There are no known barriers to ment of Fish and Wildlife, unpublished data); Sco200, Sco202, movement between the tributaries within the watershed below Sco212, Sco215, Sco216, Sco218, and Sco220 (DeHaan and Star Falls, thereby allowing for the expression of both resident Ardren 2005); Sfo18 (Angers et al. 1995); and Smm22 (Crane and migratory life histories. Biologists from MFWP and USFS et al. 2004). Seven of the 15 loci are diagnostic between Bull have surveyed every tributary of the East Fork, and Bull Trout Trout and Brook Trout, thus allowing examination of hybridiza- occupancy has been documented in 17 of the 23 main tributaries. tion between these fishes (Table 1). We extracted DNA from each Based on electrofishing data associated with fish monitoring fin clip by using a cell lysis buffer and ammonium acetate pro- surveys, the number of Bull Trout in the East Fork drainage tein precipitation followed by isopropanol DNA precipitation. BULL TROUT GENETIC POPULATION STRUCTURE 435 Downloaded by [Department Of Fisheries] at 19:52 28 May 2013

FIGURE 1. Location of Bull Trout sampling sites within the Bitterroot River drainage, Montana: East Fork Bitterroot River (site numbers 1–10), West Fork Bitterroot River (11–13), and Bitterroot River (14–17). Numbers correspond to the numbered populations listed in Table 2. 436 NYCEETAL.

FIGURE 2. Focal study area of the East Fork Bitterroot River drainage, depicting Bull Trout presence and tributaries where tissue samples (fin clips) were collected (highlighted in black). The main-stem East Fork was also sampled to collect fin clips from fluvial fish (highlighted in dark gray). Light-gray highlighting indicates creeks in which no Bull Trout have been captured during any of the Montana Fish, Wildlife, and Parks or U.S. Forest Service sampling efforts to date. Numbers correspond to the numbered populations listed in Table 2.

The PCR reactions were conducted via the QIAGEN microsatel- ment (Rice 1989). We used Fisher’s exact test in GENEPOP lite protocol using the QIAGEN Multiplex PCR Kit (QIAGEN, to test for allele frequency differences between all pairs of Valencia, California). We used three different PCR profiles: samples and thus to determine whether it would be appro- multiplexes 1 and 2 used a touchdown profile with an initial priate to consider each sample independently in subsequent annealing temperature of 63◦C stepping down to 53◦C; multi- analyses. After we established whether samples could be in- Downloaded by [Department Of Fisheries] at 19:52 28 May 2013 plex 3 used a typical PCR profile with an annealing tempera- dependently considered, we explored the within-sample genetic ture of 54◦C; and multiplex 4 used a typical PCR profile with diversity as well as the magnitude and pattern of among-sample an annealing temperature of 55.4◦C (Table 1). The DNA from differences. each fin clip was amplified in a PTC-200 thermocycler (MJ Re- To examine within-sample genetic diversity, we calculated search, Waltham, Massachusetts). After PCR, fragments were the mean number of alleles per locus (A), number of private al- separated in an ABI 3130xl Genetic Analyzer (Applied Biosys- leles, observed heterozygosity (Ho), and expected heterozygos- tems, Inc. [ABI], Foster City, California) at the Murdock DNA ity (He) by using GENALEX version 6.4 (Peakall and Smouse Sequencing Facility, University of Montana, Missoula. Allele 2006). The program HP Rare version 1.1 (Kalinowski 2005) sizes were determined by using the ABI GS600LIZ ladder, and was used to calculate allelic richness (Ar). This program stan- chromatogram output was viewed and analyzed using ABI Gen- dardizes the number of alleles detected in each sample to the eMapper version 3.7. smallest sample size by using rarefaction. We looked for evi- Genetic analyses.—We performed exact tests of Hardy– dence of recent genetic bottlenecks in the samples by using the Weinberg equilibrium (HWE) for all samples and loci by us- program Bottleneck (Cornuet and Luikart 1996), assuming the ing the program GENEPOP version 4.0 (Rousset 2008). Sig- two-phased model of mutation with a variance of 12.0. Tests for nificant values for HWE in each sample were adjusted for gametic disequilibrium between all pairs of loci in each sample multiple comparisons by using a sequential Bonferroni adjust- were conducted with Fisher’s exact test in GENEPOP. BULL TROUT GENETIC POPULATION STRUCTURE 437

TABLE 1. Microsatellite loci analyzed in Bull Trout from the Bitterroot River analysis) was applied to the entire data set to assess the ability basin (asterisks indicate loci that are diagnostic for Bull Trout × Brook Trout of our baseline data set to correctly assign fish to their water hybridization); PCR multiplex annealing temperatures (TA), PCR multiplex final primer concentrations, and associated references (WDFW = Washington body of capture. This procedure involves removal of an individ- Department of Fish and Wildlife). ual from the data set and use of a maximum likelihood algorithm to estimate the probability that the individual came from each Final of the samples; thus, the method was used to ensure that tribu- concentration tary populations in our baseline data set were different enough Locus (µM) Reference from each other to permit an informative genetic assignment Multiplex 1 (T = 55◦C) of fish collected from the main-stem East Fork. After the jack- A knife analysis, we used the individual genetic assignment test Sco106 0.10 WDFW, unpublished in ONCOR to determine the tributary of origin for each indi- Sfo18* 0.16 Angers et al. 1995 vidual fish collected from the main-stem East Fork, as we were Smm22 0.15 Crane et al. 2004 interested in identifying sources of migratory fish. Initially, we Sco216* 0.15 DeHaan and Ardren 2005 planned to treat the main-stem East Fork fish as unknowns in the = ◦ Multiplex 2 (TA 56 C) individual assignment analysis because we suspected that they Sco218* 0.10 DeHaan and Ardren 2005 would mainly represent fish that had dispersed from the tribu- Sco202 0.10 DeHaan and Ardren 2005 taries. Results of the above analyses, however, suggested that Sco200 0.15 DeHaan and Ardren 2005 this was unlikely to be the case. Therefore, all 17 samples were Sco220 0.12 DeHaan and Ardren 2005 included in the individual assignment analysis. Next, we used ◦ Multiplex 3 (TA = 54 C) GeneClass2 (Piry et al. 2004) to search for evidence of potential Sco215* 0.075 DeHaan and Ardren 2005 first-generation migrants among the samples. This analysis used Omm1128* 0.10 Rexroad et al. 2001 the Bayesian approach of Rannala and Mountain (1997), with Sco105 0.10 WDFW, unpublished 10,000 Monte Carlo-simulated individuals per sample. Desig- ◦ Multiplex 4 (TA = 55.4 C) nation of an individual as a “migrant” was based on the ratio Sco102* 0.10 WDFW, unpublished Lhome/Lmax, where Lhome is the likelihood computed from the Omm1130 0.10 Rexroad et al. 2001 sample in which the individual was collected and Lmax is the Sco107* 0.10 WDFW, unpublished highest likelihood among all samples (including the sample in Sco212 0.10 DeHaan and Ardren 2005 which the individual was collected). Individuals were considered to be first-generation migrants (and therefore migratory fish) if the probability of Lhome/Lmax was less than 0.010. Even though our focus was on the East Fork, we included all 17 sam- In addition to allele frequency comparisons, among-sample ples in this analysis to identify potential migrants from other differences were examined with multivariate analyses and FST. locations. We used Minitab 15 (Minitab 2007) to perform a principal components analysis (PCA) of the allele frequencies among the sam- RESULTS ples. Overall genetic divergence among samples (global FST) Of the 17 East Fork tributaries sampled, Bull Trout were not was estimated in FSTAT version 2.9.3 (Goudet 2001); the as- detected in five tributaries, and three tributaries (Bertie Lord, sociated 95% confidence level was estimated based on 1,000 Carmine, and Camp creeks) were not included in the data set Downloaded by [Department Of Fisheries] at 19:52 28 May 2013 bootstrap replicates. To further examine genetic divergence, due to very small sample sizes (≤3 fin clips). Thus, samples we computed FST estimates for all pairs of samples by using from nine East Fork tributaries were examined. Each sample GENEPOP following standard ANOVA as described by Weir included in the data set had at least 15 fin clips (Table 2). Sizes and Cockerham (1984). Tests of statistical significance were of individual fish in the East Fork tributary samples ranged from conducted using Arlequin version 3.1 (Excoffier et al. 2005). 25 to 422 mm TL. In contrast, we captured 73 fish in the main- A sequential Bonferroni correction was used to adjust the sig- stem East Fork, with sizes ranging from 102 to 610 mm TL. nificance level for multiple comparisons. To test for a potential Individual fish from the three West Fork tributaries and the four isolation-by-distance relationship between FST and fluvial dis- Bitterroot River tributaries ranged in size from 76 to 432 mm tance (i.e., rkm between sampling locations), we used Mantel’s TL (Table 2). Overall, the final data set included 17 samples test in GENALEX (999 permutations). across the Bitterroot River basin. We used two procedures (individual assignment and detection of migrants) to examine the likelihood of dispersal of indi- Within-Sample Variation viduals among the populations represented by our samples and Our data set was composed of 591 individuals. Among them, to identify potential sources of migratory fish. First, we used 41 possessed both Bull Trout and Brook Trout alleles at the the leave-one-out and individual assignment options in ON- diagnostic loci, indicating that they were hybrids (Table 2). All COR (Kalinowski 2008). The leave-one-out test (a jackknife hybrid individuals occurred in samples from the West Fork and 438 NYCEETAL.

TABLE 2. Bull Trout sample information, including the sample size after hybrid ﬁsh were excluded (N), the mean number of alleles per locus (A), allelic richness (Ar), expected heterozygosity (He), observed heterozygosity (Ho), the number of private alleles, and the number of hybrid individuals. Sample Private number Location NA Ar He Ho alleles Hybrids East Fork Bitterroot River 1 Clifford Creek 23 7.133 6.594 0.674 0.704 6 0 2 Martin Creek 35 7.267 5.887 0.624 0.710 0 0 3 Meadow Creek 69 8.267 6.535 0.696 0.718 1 0 4 Moose Creek 38 7.867 6.569 0.660 0.681 0 0 5 Orphan Creek 22 5.267 4.872 0.586 0.685 0 0 6 Star Creek 15 6.133 6.133 0.689 0.756 0 0 7 Swift Creek 50 7.600 5.802 0.628 0.640 0 0 8 Tolan Creek 28 5.467 4.898 0.622 0.650 0 0 9 Warm Springs Creek 27 7.800 6.809 0.680 0.699 1 0 10 East Fork main stem 73 9.467 7.099 0.703 0.700 2 0 Mean 38 7.227 6.120 0.656 0.694 1 0 West Fork Bitterroot River 11 North Fork Sheephead Creek 15 6.867 6.867 0.704 0.747 4 4 12 Sheephead Creek 18 7.267 6.922 0.684 0.763 2 1 13 Slate Creek 15 6.933 6.932 0.688 0.756 2 1 Mean 16 7.022 6.907 0.692 0.755 3 2 Main-Stem Bitterroot River 14 Daly Creek 51 11.067 8.463 0.734 0.741 3 1 15 Skalkaho Creek 53 10.867 8.142 0.727 0.724 3 6 16 Burnt Fork Creek 32 9.667 8.123 0.751 0.792 10 2 17 Willow Creek 27 6.267 5.352 0.586 0.602 3 26 Mean 41 9.467 7.515 0.700 0.715 5 9

Bitterroot River tributaries, and the majority of these individuals test indicated that allele frequencies statistically differed be- were collected in Willow Creek, a Bitterroot River tributary tween all pairs of samples at one or more loci. Furthermore, (Table 2). Hybrids were removed before further analyses, as the all pairwise FST estimates were significantly greater than zero presence of nonnative alleles can distort population structure (P < 0.010; see below and Table 3). Therefore, each sample was estimates (Forbes and Allendorf 1991; Cegelski et al. 2006; treated separately for subsequent analyses. Allendorf and Luikart 2007). Comparing all samples in the data set, A ranged from 5.267 For the 15 microsatellite loci analyzed, A ranged from 2 to 11.067 with a mean of 7.717 (Table 2). Values of Ar ranged Downloaded by [Department Of Fisheries] at 19:52 28 May 2013 to 34, and all loci were polymorphic in all 17 samples. Af- from 4.872 to 8.463, with a mean of 6.587 (Table 2). Values of A ter a sequential Bonferroni correction (15 comparisons/sample; and Ar were lowest in Orphan Creek and highest in Daly Creek. α = 0.003), there was little evidence of deviations from HWE Thirty-seven private alleles occurred in 11 different samples, expectations in the samples. Only Orphan Creek (a tributary and the frequency ranged from 0.007 to 0.156. Averaged over all of the East Fork) showed a significant departure from HWE, samples, He and Ho were 0.673 and 0.710, respectively, and the and this was due to an excess of heterozygotes at Smm22 and average within-population expected heterozygosity (Hs) ranged Omm1130. Although most samples did not show a significant from 0.586 to 0.751. Orphan and Willow creeks had the lowest departure from HWE, there was a significant tendency for our He, while Burnt Fork Creek had the highest He. Willow Creek East Fork samples to have an excess of heterozygotes (94 of had the lowest Ho, and Burnt Fork Creek had the highest Ho. 150 comparisons; Statistical Package for the Social Sciences, Considering all samples (East Fork, West Fork, and main-stem sign test: P = 0.002). Interestingly, these results include the Bitterroot River), values of A, Ar, and He did not significantly main-stem East Fork sample, which we initially assumed would differ among populations from the different drainages (Kruskal– consist of a mixture of individuals from different tributaries and Wallis analysis: P > 0.050). thus would possess a deficit of heterozygotes based on HWE. Within the East Fork basin, A ranged from 5.267 to This, however, was not the case. After correction, Fisher’s exact 9.467 (mean = 7.227), and Ar ranged from 4.872 to 7.099 BULL TROUT GENETIC POPULATION STRUCTURE 439

TABLE 3. Estimates of genetic divergence (FST) between all possible pairs of Bull Trout samples in the data set (East Fork: sample numbers 1–10; West Fork: 11–13; main-stem Bitterroot River: 14–17). All FST values were signiﬁcantly greater than zero (P < 0.01).

Sample number Sample number andlocation 1 2345678910111213141516

1. Clifford Creek 2. Martin Creek 0.071 3. Meadow Creek 0.045 0.082 4. Moose Creek 0.041 0.059 0.032 5. Orphan Creek 0.103 0.120 0.073 0.073 6. Star Creek 0.043 0.067 0.047 0.048 0.095 7. Swift Creek 0.068 0.098 0.043 0.060 0.085 0.065 8. Tolan Creek 0.124 0.165 0.105 0.116 0.188 0.147 0.148 9. Warm Springs Creek 0.045 0.090 0.041 0.038 0.106 0.041 0.066 0.113 10. East Fork main stem 0.016 0.056 0.019 0.018 0.071 0.023 0.030 0.106 0.020 11. North Fork 0.083 0.140 0.076 0.098 0.156 0.089 0.106 0.131 0.074 0.070 Sheephead Creek 12. Sheephead Creek 0.087 0.122 0.074 0.102 0.161 0.097 0.091 0.126 0.083 0.071 0.041 13. Slate Creek 0.088 0.114 0.078 0.089 0.139 0.089 0.100 0.122 0.071 0.068 0.046 0.050 14. Daly Creek 0.080 0.111 0.066 0.072 0.119 0.086 0.092 0.105 0.063 0.060 0.042 0.066 0.050 15. Skalkaho Creek 0.073 0.109 0.059 0.075 0.119 0.082 0.078 0.093 0.061 0.054 0.042 0.054 0.046 0.009 16. Burnt Fork Creek 0.097 0.133 0.095 0.096 0.136 0.102 0.130 0.143 0.093 0.081 0.063 0.098 0.079 0.030 0.057 17. Willow Creek 0.240 0.284 0.212 0.244 0.300 0.227 0.262 0.212 0.239 0.224 0.215 0.208 0.221 0.195 0.179 0.209

(mean = 6.120; Table 2). Orphan Creek had the lowest A and Ar, Among-Sample Variation whereas the East Fork main-stem sample had the highest A and Comparisons among all samples indicated the presence of Ar. A total of 10 private alleles occurred in four different sam- geographic genetic structure at the basin level, with the East ples, and the frequency ranged from 0.007 to 0.130. Averaged Fork samples being distinguishable from West Fork and Bitter- over all East Fork samples, He and Ho were 0.656 and 0.694, re- root River tributary samples (Figure 3). The ﬁrst two axes of the spectively, while Hs ranged from 0.586 to 0.703. Orphan Creek PCA explained 43% of the variation in allele frequencies. All had the lowest He, and the main-stem East Fork sample had the highest He. Swift Creek had the lowest Ho, and Star Creek had the highest Ho. We observed evidence of gametic disequilibrium in 8 of the 17 samples. Out of 105 comparisons per sample, the following samples showed signiﬁcant evidence of disequilibrium: Martin Creek at two pairs of loci, Orphan Creek at three pairs, Star Creek at one pair, Swift Creek at four pairs, Tolan Creek at one

Downloaded by [Department Of Fisheries] at 19:52 28 May 2013 pair, North Fork Sheephead Creek at one pair, Sheephead Creek at one pair, and Willow Creek at two pairs. The specific locus pairs that showed evidence of disequilibrium differed among the samples; the exception was observed for Sco200 and Sfo18, which were out of equilibrium in samples from Tolan and Wil- low creeks. Therefore, disequilibrium was probably not due to linkage but rather to some random factor, such as low Ne or a recent bottleneck. The latter scenario seems unlikely because results of the genetic bottleneck test suggested that only the Star Creek sample had a signal indicating a recent bottleneck, as evidenced by a significant allele frequency mode shift (P = 0.018), but the Wilcoxon test was not significant (P = 0.679). However, FIGURE 3. Principal components (PC) analysis based on allele frequencies for Bull Trout population samples collected in the Bitterroot River basin. The first our power to detect a signal indicative of a recent bottleneck two axes (PC1 and PC2) collectively explained 43% of the variation; Tolan and was weak due to the relatively small number of individuals and Willow creeks are identified as outliers. Numbers correspond to the numbered loci analyzed in each sample. populations listed in Table 2. 440 NYCEETAL.

but one of the East Fork samples clustered together, and all but to be between Tolan Creek and the other East Fork samples, and one of the samples outside of the East Fork (i.e., the West Fork the smallest differences were between the main-stem East Fork and Bitterroot River tributaries) clustered together. Both Willow and the tributaries. We did not detect a significant isolation-by- Creek (a Bitterroot River tributary) and Tolan Creek (an East distance relationship between genetic distance and fluvial dis- Fork tributary) showed up as outliers from the two clusters (Fig- tance among tributaries of the East Fork (r2 = 0.017; P = 0.285). ure 3). This genetic structure across the basin was also evidenced Results of the leave-one-out test suggested that our baseline by pairwise FST estimates, which ranged from 0.009 to 0.300 had high potential for successful assignment of individual Bull (Table 3), with a global FST value of 0.089. The highest pairwise Trout to the correct water body of capture (range = 47–100%; FST estimates were observed between Willow Creek and all of mean = 75%). All individuals that were captured in tributaries the other samples, and pairwise values with Tolan Creek were of the East Fork were assigned to populations within that basin. generally above 0.100. As expected, there was more genetic di- Fish from the West Fork and main-stem Bitterroot River trib- vergence between tributaries among the three basins (East Fork, utaries also had high assignment success within each specific West Fork, and Bitterroot River) than between tributaries within basin, with the exception of five individuals. The results from the East Fork basin. Specifically, FST values for paired East Fork ONCOR (individual assignment test) and GeneClass2 (first- tributary samples were generally lower than FST values from generation migrant detection) analyses suggested that based on comparisons of East Fork tributary samples with West Fork or our samples, movement of individuals among the East Fork Bitterroot River tributary samples. We did not detect a signifi- main stem and tributaries may be common (Tables 4, 5). Within cant isolation-by-distance relationship between genetic distance the East Fork basin, 76 of 380 individuals were identified and fluvial distance for the 17 samples distributed across the Bit- as first-generation migrants (P < 0.010). Individuals in the terroot River basin (r2 = 0.009, P = 0.250). main-stem East Fork sample were assigned to or identified as Focusing on the East Fork, we found that all nine tribu- first-generation migrants from all East Fork tributaries except tary samples and the main-stem East Fork sample significantly Martin and Tolan creeks. Furthermore, individuals in all East differed from each other (P < 0.010); pairwise FST estimates Fork tributary samples except Tolan Creek were assigned to or ranged from 0.016 to 0.188 (Table 3), with a global FST value of identified as first-generation migrants from the main-stem East 0.063. As previously indicated, the greatest differences tended Fork. Finally, a number of individuals sampled in the East Fork

TABLE 4. ONCOR results for individual genetic assignment of Bull Trout samples from the East Fork (sample numbers 1–10), West Fork (11–13), and main-stem Bitterroot River (14–17). Sample location is the water body of capture (where the sample was collected); N is the number of fish in the sample; underlined values represent the number of individuals that were assigned to their water body of capture; values that are not underlined represent the number of fish that were assigned to each population other than the location of capture; and “Total” is the total number of fish that were assigned to locations other than the capture location. Empty cells indicate a value of zero.

Assigned population number Sample number and location N 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 Total 1. Clifford Creek 23 19 1 1 2 4 2. Martin Creek 35 31 1 1 2 4 3. Meadow Creek 69 59 1 9 10

Downloaded by [Department Of Fisheries] at 19:52 28 May 2013 4. Moose Creek 38 35 3 3 5. Orphan Creek 22 21 1 1 6. Star Creek 15 1 12 2 3 7. Swift Creek 50 1 1 41 7 9 8. Tolan Creek 28 28 0 9. Warm Springs Creek 27 27 0 10. East Fork main stem 73 3 5 1 2 3 2 57 16 11. North Fork Sheephead 15 15 0 Creek 12. Sheephead Creek 18 18 0 13. Slate Creek 15 15 0 14. Daly Creek 51 45 6 6 15. Skalkaho Creek 53 4 49 4 16. Burnt Fork Creek 32 32 0 17. Willow Creek 27 27 0 BULL TROUT GENETIC POPULATION STRUCTURE 441

TABLE 5. GeneClass2 results for detection of first-generation migrants in Bull Trout samples from the East Fork (sample numbers 1–10), West Fork (11–13), and main-stem Bitterroot River (14–17). Sample location is the water body of capture (where the sample was collected); N is the number of fish in the sample; underlined values represent the number of individuals that originated from the water body of capture (i.e., were not considered migrants); values that are not underlined represent the number of migrants, listed under the most likely population of origin; and “Migrants” is the total number of migrants in the sample. All migrant values are significantly greater than zero (P < 0.01). Empty cells indicate a value of zero.

Assigned population number Sample number and location N 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 Migrants 1. Clifford Creek 23 16 1 1 1 4 7 2. Martin Creek 35 27 2 1 5 8 3. Meadow Creek 69 60 1 1 7 9 4. Moose Creek 38 1 29 1 1 6 9 5. Orphan Creek 22 1 19 2 3 6. Star Creek 15 1 81 5 7 7. Swift Creek 50 4 1 36 9 14 8. Tolan Creek 28 28 0 9. Warm Springs Creek 27 4 20 3 7 10. East Fork main stem 73 4 2 2 2 2 61 12 11. North Fork Sheephead 15 11 3 1 4 Creek 12. Sheephead Creek 18 1 3 13 1 5 13. Slate Creek 15 12 1 2 3 14. Daly Creek 51 43 8 8 15. Skalkaho Creek 53 12 41 12 16. Burnt Fork Creek 32 2 30 2 17. Willow Creek 27 27 0

tributaries were assigned to or identified as first-generation mi- undisturbed, open system. Such conditions have not been com- grants from other East Fork tributaries. Thus, there appears to be mon in many previous studies on Bull Trout genetic popula- movement (and potential gene flow) from the main-stem East tion structure. We found geographic genetic structure among Fork to most tributaries and vice versa as well as movement the East Fork, West Fork, and Bitterroot River samples. The between the tributaries (Tables 4, 5). East Fork samples demonstrated high within-population diver- No individuals from the West Fork or main-stem Bitterroot sity and lower among-population diversity. Movement occurred tributaries were assigned to or identified as potential migrants from the main-stem East Fork to its tributaries, from the tribu- in the East Fork samples. However, one individual that was cap- taries to the main-stem East Fork, and between the tributaries, tured in Sheephead Creek (a West Fork tributary) was identified all of which are indicative of metapopulation structure. We de- as a first-generation migrant (GeneClass2) from the main-stem tected no conclusive evidence of fish movement between the Downloaded by [Department Of Fisheries] at 19:52 28 May 2013 East Fork. Results from ONCOR indicated that movement be- East Fork basin and tributaries of the West Fork, or main-stem tween Daly and Skalkaho creeks (main-stem Bitterroot River Bitterroot River. Therefore, the main-stem East Fork is vital for samples) may be common. GeneClass2 results also indicated maintaining the migratory form of Bull Trout in this basin. that there was movement of fish between West Fork tributaries Although it has been suggested that Bull Trout may func- and between main-stem Bitterroot River tributaries. Likewise, tion as a metapopulation (Dunham and Rieman 1999; Rieman five individuals from tributaries of the main-stem Bitterroot were and Dunham 2000), a few genetic population studies have in- identified as migrants in the West Fork tributaries (Table 5). dicated that there may be little if any metapopulation structure A few of these very likely represent spurious results, as they (Spruell et al. 1999; Kanda and Allendorf 2001). Studies have would have required the movement of fish above the impassable also shown that typically, local populations in close geographic Painted Rocks Dam. proximity (e.g., adjacent tributaries) are very different genetically (Leary et al. 1993; Spruell et al. 1999; Kanda and Allendorf 2001; Whiteley et al. 2006; DeHaan et al. 2010). Many of these DISCUSSION studies compared populations across watersheds or within wa- Our study of the East Fork offers a unique view of Bull tersheds characterized by degraded or fragmented habitat. Thus, Trout genetic population structure. This single watershed is es- the results of each study may have been greatly influenced by the timated to contain 5,000–10,000 Bull Trout and is a relatively condition of the watershed in which it was conducted (Costello 442 NYCEETAL.

2 TABLE 6. Summary of comparable studies of Bull Trout genetic population structure, including watershed size (km ), global FST, pairwise FST range, presence of known barriers, number of populations (N), total number of individuals (Ni), mean number of alleles per locus (A), mean allelic richness (Ar), mean expected heterozygosity (He), and mean observed heterozygosity (Ho). Watershed Global Pairwise Known 2 Location size (km ) FST FST range barriers NNi AAr He Ho Reference East Fork, Montana 1,057 0.063 0.016–0.188 No 10 380 7.23 6.12 0.66 0.70 This study West Fork and 6,231 0.089 0.009–0.300 Yes 7 211 8.42 7.25 0.70 0.73 This study Bitterroot River, Montana Warm Springs 405 0.206 0.121–0.245 Yes 4 123 3.95 3.82 0.52 0.53 DeHaan et al. Creek, Montana 2010 Metolius River, 1,160 0.054 0.005–0.148 No 7 332 6.28 5.65 0.58 0.59 DeHaan et al. Oregon 2008a Swan River, 1,732 0.037 0.008–0.060 No 7 277 8.41 7.4 0.69 0.69 DeHaan et al. Montana 2011a

et al. 2003; Yamamoto et al. 2004; Meeuwig et al. 2010). The than that of Bull Trout in other undisturbed, connected systems evaluation of population structure in different watersheds across of similar size, although there are few studies available for com- scales and levels of impact is critical as we apply our understand- parison. Comparable studies (Table 6) that used the same 15 ing of Bull Trout population structure to the establishment of microsatellite loci within a single watershed were conducted restoration goals and conservation guidelines. in Warm Springs Creek, Montana (DeHaan et al. 2010); the Our data indicated that populations in tributaries were ge- Metolius River, Oregon (DeHaan et al. 2008a); and the Swan netically different from each other, but the pairwise divergence River, Montana (DeHaan et al. 2011a). Estimates of mean A, (estimated by FST) between samples within the East Fork, West Ar, He, and Ho were generally higher in the East Fork, West Fork, and main-stem Bitterroot River basins generally was not Fork, and Bitterroot River samples than in samples from these large. For example, the global FST among the East Fork sam- other systems, with the exception of the Swan River (Table 6). ples was 0.063. This is lower than the global FST observed in We can only hypothesize why these measures of genetic diver- many other studies of Bull Trout (typical range of 0.100–0.250, sity were generally greater in the Bitterroot River watershed. If with differences up to 0.400; Whiteley et al. 2006; DeHaan tributary populations only have a few breeders, then an overes- et al. 2010; Ardren et al. 2011) except studies that have been timation of population divergence may be caused by sampling conducted in relatively undisturbed, connected habitats of ap- the progeny of relatively few breeders (Allendorf and Phelps proximately the same size as the East Fork (see Table 6). The 1981; Hansen et al. 1997; Hansen and Mensberg 1998). Sam- lower FST values for the East Fork and similar basins suggests pling methods, population age structure, and breeding ecology that connected, relatively undisturbed watersheds containing a could vary among studies, thereby inﬂuencing the estimates migratory life history component allow a greater opportunity of divergence. A more likely explanation is that the Bitterroot for gene ﬂow among Bull Trout populations. This is supported River watershed may have higher population sizes and/or greater

Downloaded by [Department Of Fisheries] at 19:52 28 May 2013 by our findings from the assignment test and the detection of connectivity (lack of barriers and number of first-generation mi- first-generation migrants; those results indicated that within the grants) than the other systems. Higher proportions of migratory East Fork basin, movement from the main-stem East Fork to its fish and increased habitat size and quality in the East Fork may tributaries, from the tributaries to the main stem, and among the also contribute to the observed differences (Hansen and Mens- tributaries was common, likely resulting in gene flow. Overall, berg 1998; Knutsen et al. 2001; Neville et al. 2006). We suggest these results suggest the existence of metapopulation structure that the relatively high within-population genetic diversity is in the East Fork basin (e.g., “metapopulations . . . [are] any as- also a reflection of the higher levels of gene flow among pop- semblage of discrete local populations with migration among ulations in such undisturbed, connected systems. This idea is them . . . regardless of the rate of population turnover”: Hanski supported by studies that have demonstrated a reduction in ge- and Simberloff 1997). This metapopulation structure may be netic variation within Bull Trout populations that are isolated more typical of smaller, connected watersheds with a diversity above barriers (e.g., Whiteley et al. 2006; DeHaan et al. 2007, of life history types; however, few Bull Trout studies have been 2011b). The authors of these prior studies suggested that the size conducted in such systems (but see DeHaan et al. 2008a, 2011a; of Bull Trout populations and the distance between populations Warnock et al. 2010). with migratory fish could be a factor influencing the levels of ge- Overall, the genetic diversity we observed within Bull Trout netic variation. Similarly, in Brown Trout, resident populations populations in the Bitterroot River drainage tended to be higher show greater divergence and lower diversity than populations BULL TROUT GENETIC POPULATION STRUCTURE 443

with migratory individuals, thus highlighting the role of life The presence of migratory fish and dispersal among popu- history variation in patterns of genetic variation (Hansen and lations may be important for the persistence of Bull Trout in Mensberg 1998; Knutsen et al. 2001). the East Fork and throughout the Bitterroot River system. Dis- In comparing the allele frequencies among all samples from tribution of adults into multiple habitats (rivers and tributaries) the East Fork, West Fork, and Bitterroot River, we found two can buffer populations from disturbances. For example, Rieman major groups of samples: (1) the East Fork, with Tolan Creek et al. (1997) and Sestrich et al. (2011) determined that Bull as an outlier; and (2) the West Fork and the Bitterroot River, Trout populations recovered rapidly after extensive wildfires. with Willow Creek as an outlier (Figure 3). Even though there Rieman et al. (1997) concluded that two important mechanisms are no known barriers between Tolan and Willow creeks and contributing to Bull Trout recovery were dispersal and life his- their respective main-stem rivers, these populations appeared tory (migratory fish). Even though wildfire burned extensively to be more isolated than the other populations in the region. throughout the East Fork basin in 2000, with several creeks ex- It is possible that Tolan and Willow creeks are more isolated periencing high-severity riparian burns and subsequent debris because of watershed disturbances, including (but not limited flows (Sestrich et al. 2011; USFS, unpublished data), only the to) hydrology and land use. For example, the lower end (3.2 km) Star Creek sample showed any signs of a recent population bot- of Tolan Creek is heavily impacted by agricultural operations tleneck. However, this drainage experienced very little wildfire, (grazing and water diversion). The Willow Creek watershed is and where it did experience wildfire the severity of the wildfire affected by roads running along the creek, channelization and was low (USFS, unpublished data). private development, and grazing and diversion structures that One of our goals was to identify the potential sources of likely reduce dispersal (R. Brassfield and M. Jakober, USFS, migratory Bull Trout. Even though some fish sampled in the personal communication). In addition, Willow Creek had the main-stem East Fork were assigned to the tributaries, 78% of largest proportion of Bull Trout × Brook Trout hybrids (about fish in the main-stem sample were assigned to that same popu- 50% of samples), suggesting that there is an abundance of Brook lation and grouped together in HWE. Historical redd count data Trout, more suitable habitat for Brook Trout, or fewer Bull Trout. and radiotelemetry studies have identified some spawning in the Regardless of the reason, isolation of the two creeks appears to upper main-stem East Fork, thus indicating that this area is a have resulted in lower within-population genetic diversity and potential source of migratory fish (Nyce 2011; MFWP, unpub- unusually high divergence for these populations. lished data). In the tributary samples, many of the identified first- Populations with less genetic diversity, such as those in Tolan generation migrants were assigned to the main-stem East Fork. and Willow creeks, should not necessarily be treated as having Likewise, in the main-stem East Fork sample, many of the iden- a lower conservation value than the more diverse populations. tified first-generation migrants were assigned to the tributaries. These populations constitute part of the overall genetic diversity These results imply that the main-stem spawning population of the system (Rieman and Allendorf 2001). It is also possible produces and receives a large proportion of migrants and that that Bull Trout in these two tributaries contain important adap- the East Fork main stem currently plays an important role for tive differences resulting from the lack of gene flow (Leary migratory fish. In addition, many of the East Fork tributaries, et al. 1993; Giles and Goudet 1997). Due to their genetic differ- especially Clifford, Meadow, Star, Swift, and Warm Springs ences and despite their isolation, Bull Trout in Tolan and Wil- creeks, were identified as sources of migratory fish (Tables 4, low creeks may represent important components of the among- 5). Given that we detected first-generation migrants collected population genetic diversity of Bull Trout in the Bitterroot River over a short time period (2 years), additional sampling across basin. a longer time period could eventually identify other tributaries

Downloaded by [Department Of Fisheries] at 19:52 28 May 2013 In 8 of the 17 samples we examined, there was significant that are important and that may contribute to the persistence of gametic disequilibrium between one or more pairs of loci. Im- the migratory life history form. migrants that differ genetically from local individuals can create Although continued protection of Bull Trout populations in gametic disequilibrium, especially if migration rates are low and the tributaries is essential, the main-stem East Fork appears to pulsed or if populations are small (Waples and England 2011). be an integral component for maintaining the migratory form Waples and England (2011) also noted that the same effect could of Bull Trout in the drainage. The main-stem East Fork not occur if individuals from more than one population are included only appears to produce and receive migratory fish but also in the sample. The gametic disequilibrium and the observed serves as a dispersal corridor for potential genetic exchange tendency toward an excess of heterozygotes could indicate that among tributary populations. Conservation of migratory Bull Ne is relatively low. In contrast, the relatively low estimates of Trout in the East Fork is critical (USFWS 2008), and although FST, especially in the East Fork, and the results from the as- the management unit is the entire East Fork basin, a concerted signment test and migrant detection suggest that there is gene monitoring effort should be focused on the main-stem East Fork. flow among populations. Gene flow among populations with To help ensure Bull Trout persistence, we would prioritize both a breeding ecology that promotes a relatively low Ne may be the identification of spawning locations in the main-stem East very important for maintaining genetic diversity (Rieman and Fork and the preservation of habitat as necessary for migratory Allendorf 2001). Bull Trout. 444 NYCEETAL.

ACKNOWLEDGMENTS DeHaan, P. W., M. Diggs, and W. R. Ardren. 2008a. Analysis of genetic varia- We are grateful to numerous individuals for field assistance. tion in Metolius River basin Bull Trout populations. U.S. Fish and Wildlife Mike Jakober and Rob Brassfield offered a tremendous amount Service, Final Report, Bend, Oregon. DeHaan, P. W., L. Godfrey, and W. R. Ardren. 2008b. Genetic assignment of of help and support. Angela Lodmell, Fred Allendorf, Steve Bull Trout collected at Libby Dam 2004 to 2007. Montana Fish, Wildlife, Amish, and Dan Bingham provided extensive assistance with and Parks, Final Report, Libby. a variety of laboratory work and genetic analyses. The maps DeHaan, P. W., L. Godfrey, D. Peterson, and D. Brewer. 2010. Bull Trout pop- were generated by Dawn Anderson (MFWP). We thank Tom ulation genetic structure and entrainment in Warm Springs Creek, Montana. McMahon, Mike Mitchell, and Patrick DeHaan for providing U.S. Fish and Wildlife Service, Final Report, Helena, Montana. Dunham, J. B., and B. E. Rieman. 1999. Metapopulation structure of Bull Trout: reviews of the manuscript and suggestions for improvement. influences of physical, biotic, and geometrical landscape characteristics. Eco- Funding was provided by MFWP and the Longwell Foundation. logical Applications 9:642–655. Excoffier, L., G. Laval, and S. Schneider. 2005. Arlequin ver. 3.0: an integrated software package for population genetics data analysis. Evolutionary REFERENCES Bioinformatics Online [online serial] 1:47–50. Allendorf, F. W., and G. Luikart. 2007. Conservation and the genetics of popu- Forbes, S. H., and F. W. Allendorf. 1991. Associations between mitochondrial lations. Blackwell Scientific Publications, Malden, Massachusetts. and nuclear genotypes in Cutthroat Trout hybrid swarms. Evolution 45:1332– Allendorf, F. W., and S. R. Phelps. 1981. Use of allelic frequencies to describe 1349. population structure. Canadian Journal of Fisheries and Aquatic Sciences Fraley, J. J., and B. B. Shepard. 1989. Life history, ecology and population 38:1507–1514. status of migratory Bull Trout (Salvelinus confluentus) in the Flathead Lake Angers, B., L. Bernatchez, A. Angers, and L. Desgroseillers. 1995. Specific and River systems, Montana. Northwest Science 63:133–143. microsatellite loci for Brook Charr reveal strong population subdivision on a Giles, B., and J. Goudet. 1997. A case study of genetic structure in a plant microgeographic scale. Journal of Fish Biology 47(Supplement A):177–185. metapopulation. Pages 429–454 in I. Hanski and M. Gilpin, editors. Metapop- Ardren, W. P., P. W. DeHaan, C. T. Smith, E. B. Taylor, R. Leary, C. Kozfkay, ulation biology: ecology, genetics, and evolution. Academic Press, San Diego, L. Godfrey, M. Diggs, W. Fredenberg, J. Chan, C. W. Kilpatrick, and California. D. K. Hawkins. 2011. Genetic structure, evolutionary history, and conser- Goudet, J. 2001. FSTAT, a program to estimate and test gene diversi- vation units of Bull Trout in the coterminous United States. Transactions of ties and fixation indices, version 2.9.3. Available: http://www.unil.ch/ the American Fisheries Society 140:506–525. izea/softwares/fstat.html. (February 2013). Bahr, M. A., and J. M. Shrimpton. 2004. Spatial and quantitative patterns of Hallerman, E. M., editor. 2003. Population genetics: principles and applications movement in large Bull Trout (Salvelinus confluentus) from a watershed in for fisheries scientists. American Fisheries Society, Bethesda, Maryland. north-western British Columbia, Canada, are due to habitat selection and not Hansen, M. M., and K. L Mensberg. 1998. Genetic differentiation and relation- differences in life history. Ecology of Freshwater Fish 13:294–304. ship between genetic and geographical distance in Danish Sea Trout (Salmo Burton, T. A. 2005. Fish and stream habitat risk from uncharacteristic wildfire: trutta L.) populations. Heredity 81:493–504. observations from 17 years of fire-related disturbances on the Boise National Hansen, M. M., E. E. Nielsen, and K. L. D. Mensberg. 1997. The problem of Forest, Idaho. Forest Ecology and Management 211:140–149. sampling families rather than populations: relatedness among individuals in Cegelski, C. C., M. R. Campbell, K. A. Meyer, and M. S. Powell. 2006. Mul- samples of juvenile Brown Trout Salmo trutta L. Molecular Ecology 6:469– tiscale genetic structure of Yellowstone Cutthroat Trout in the upper Snake 474. River basin. Transactions of the American Fisheries Society 135:711–726. Hanski, I. 1998. Metapopulation dynamics. Nature 396:41–49. Cornuet, J. M., and G. Luikart. 1996. Description and power analysis of two Hanski, I., and D. Simberloff. 1997. The metapopulation approach, its his- tests for detecting recent population bottlenecks from allele frequency data. tory, conceptual domain, and application to conservation. Pages 5–26 in I. Genetics 144:2001–2014. Hanski and M. Gilpin, editors. Metapopulation biology: ecology, genetics, Costello, A. B., T. E. Downs, S. M. Pollard, C. J. Pacas, and E. B. Taylor. and evolution. Academic Press, San Diego, California. 2003. The influence of history and contemporary stream hydrology on the Harrison, S., and A. Hastings. 1996. Genetic and evolutionary consequences of evolution of genetic diversity within species: an examination of microsatellite metapopulation structure. Trends in Ecology and Evolution 11:180–183. DNA variation in Bull Trout, Salvelinus confluentus (Pisces: Salmonidae). Homel, K., P. Budy, M. E. Pfrender, T. A. Whitesel, and K. Mock. 2008. Evolution 57:328–344. Evaluating genetic structure among resident and migratory forms of Bull Downloaded by [Department Of Fisheries] at 19:52 28 May 2013 Crane, P. A., C. J. Lewis, E. J. Dretschmer, and S. J. Miller. 2004. Charac- Trout (Salvelinus confluentus) in northeast Oregon. Ecology of Freshwater terization and inheritance of seven microsatellite loci from Dolly Varden, Fish 18:465–474. Salvelinus malma, and cross-species amplification in Arctic Char, S. alpinus. Jakober, M. J., T. E. McMahon, R. F. Thurow, and C. G. Clancy. 1998. Role Conservation Genetics 5:737–741. of stream ice on fall and winter movement and habitat use by Bull Trout and DeHaan, P. W., B. Adams, W. Fredenberg, and B. Gardner. 2011a. Identification Cutthroat Trout in Montana headwater streams. Transactions of the American of population origin for Bull Trout captured in Swan Lake, Montana. U.S. Fisheries Society 127:223–235. Fish and Wildlife Service, Final Report, Longview, Washington. Jonsson, B., and N. Jonsson. 1993. Partial migration: niche shift versus sexual DeHaan, P. W., and W. R. Ardren. 2005. Characterization of 20 highly vari- maturation in fishes. Reviews in Fish Biology and Fisheries 3:348–365. able tetranucleotide microsatellite loci for Bull Trout (Salvelinus confluen- Kalinowski, S. T. 2005. HP-Rare 1.0: a computer program for performing rar- tus) and cross-amplification in other Salvelinus species. Molecular Ecology 5: efaction on measures of allelic richness. Molecular Ecology Notes 5:187–189. 582–585. Kalinowski, S. T. 2008. ONCOR. Montana State University, Bozeman. Avail- DeHaan, P. W., C. A. Barfoot, and W. R. Ardren. 2007. Genetic analysis of able: http://www.montant.edu/kalinowski/software.htm. (March 2013). Bull Trout populations on the Flathead Indian Reservation, Montana. Pages Kanda, N., and F. W. Allendorf. 2001. Genetic population structure of Bull Trout 145–153 in R. F. Carline and C. LoSapio, editors. Wild trout IX sustaining from the Flathead River basin as shown by microsatellites and mitochon- wild trout in a changing world. Bozeman, Montana. drial DNA markers. Transactions of the American Fisheries Society 130:92– DeHaan, P. W., S. J. Brenkman, B. Adams, and P. Crain. 2011b. Genetic popula- 106. tion structure of Olympic Peninsula Bull Trout populations and implications Knutsen, H., J. A. Knutsen, and P. E. Jorde. 2001. Genetic evidence for mixed for Elwha Dam removal. Northwest Science 85:463–475. origin of recolonized Sea Trout populations. Heredity 87:207–214. BULL TROUT GENETIC POPULATION STRUCTURE 445

Leary, R. F., F. W. Allendorf, and S. H. Forbes. 1993. Conservation genetics Rieman, B. E., and J. B. Dunham. 2000. Metapopulations and salmonids: a of Bull Trout in the Columbia and Klamath river drainages. Conservation synthesis of life history patterns and empirical observations. Ecology of Biology 7:856–865. Freshwater Fish 9:51–64. Lockard, L. 2006. Bull Trout flip over Cabinet Gorge Dam. U.S. Fish and Rieman, B. E., D. Lee, G. Chandler, and D. Myers. 1997. Does wildfire threaten Wildlife Service Endangered Species Bulletin (March 2006):22–23. extinction for salmonids? Responses of redband trout and Bull Trout follow- Lowe, W. H., and F. W. Allendorf. 2010. What can genetics tell us about ing recent large fires on the Boise National Forest. Pages 47–57 in J. Greenlee, population connectivity? Molecular Ecology 19:3038–3051. editor. Proceedings of the symposium on fire effects on rare and endangered McCart, P. 1997. Bull Trout in Alberta: a review. Pages 191–207 in W. C. species and habitats. International Association of Wildland Fire, Fairfield, Mackay, M. K. Brewin, and M. Monita, editors. Friends of the Bull Trout Washington. conference proceedings. Trout Unlimited Canada, Calgary, Alberta. Rieman, B. E., and J. D. McIntyre. 1993. Demographic and habitat requirements McPhail, J. D., and J. S. Baxter. 1996. A review of Bull Trout (Salvelinus for conservation of Bull Trout. U.S. Forest Service General Technical Report confluentus). Life-history and habitat use in relation to compensation and im- INT-302 provement opportunities. British Columbia Ministry of Environment, Lands, Rousset, F. 2008. GENEPOP’007: a complete reimplementation of the and Parks, Fisheries Management Report 104, Victoria. GENEPOP software for Windows and Linux. Molecular Ecology Resources Meeuwig, M. H., C. S. Guy, S. T. Kalinowski, and W. A. Fredenberg. 2010. 8:103–106. Landscape influences on genetic differentiation among Bull Trout populations Schindler, D. E., R. Hilborn, B. Chasco, C. P. Boatright, T. P. Quinn, L. A. in a stream-lake network. Molecular Ecology 19:3620–3633. Rogers, M. S. Webster. 2010. Population diversity and the portfolio effect in Minitab. 2007. Minitab 15 Statistical Software. Minitab, Inc., State College, an exploited species. Nature 465:609–612. Pennsylvania. Sestrich, C. M., T. E. McMahon, and M. K. Young. 2011. Influence of fire on Montana Bull Trout Restoration Team. 2000. Restoration plan for Bull Trout native and nonnative salmonid populations and habitat in a western Montana in the Clark Fork River basin and Kootenai River basin, Montana. Montana basin. Transactions of the American Fisheries Society 140:136–146. Department of Fish, Wildlife, and Parks, Helena. Slatkin, M. 1987. Gene flow and the geographic structure of natural populations. Montana Bull Trout Scientific Group. 1995. Bitterroot River drainage Bull Trout Science 236:787–792. status review. Montana Fish, Wildlife, and Parks, Helena. Spruell, P., A. R. Hemmingsen, P. J. Howell, N. Kanda, and F. W. Allendorf. Morita, K., S. H. Morita, and S. Yamamoto. 2009. Effects of habitat fragmen- 2003. Conservation genetics of Bull Trout: geographic distribution of varia- tation by damming on salmonid fishes: lessons from White-spotted Charr in tion at microsatellite loci. Conservation Genetics 4:17–29. Japan. Ecological Resources 24:711–722. Spruell, P., B. E. Rieman, K. L. Knudsen, F. M. Utter, and F. W. Allendorf. Morita, K., and S. Yamamoto. 2002. Effects of habitat fragmentation by 1999. Genetic population structure within streams: microsatellite analysis of damming on the persistence of stream-dwelling charr populations. Conser- Bull Trout populations. Ecology of Freshwater Fish 8:114–121. vation Biology 16:1318–1323. Swanberg, T. R. 1997. Movements of and habitat use by fluvial Bull Trout in the Nelson, M. L., T. E. McMahon, and R. F. Thurow. 2002. Decline of the migratory Blackfoot River, Montana. Transactions of the American Fisheries Society form in Bull Charr, Salvelinus confluentus, and implications for conservation. 126:735–746. Environmental Biology of Fishes 64:321–332. Thorpe, J. E. 1994. Thresholds and life-history flexibility in salmonids. Conser- Neraas, L. P., and P. Spruell. 2001. Fragmentation of riverine systems: the vation Biology 8:877–879. genetic effects of dams on Bull Trout (Salvelinus confluentus) in the Clark USFWS (U.S. Fish and Wildlife Service). 2008. Bull Trout recovery: monitoring Fork River system. Molecular Ecology 10:1153–1164. and evaluation guidance. Report prepared for the U.S. Fish and Wildlife Neville, H. M., J. B. Dunham, and M. M. Peacock. 2006. Landscape attributes Service by the Bull Trout Recovery and Monitoring Technical Group, Version and life history variability shape genetic structure of trout populations in a 1, Portland, Oregon. stream network. Landscape Ecology 21:901–916. USOFR (U.S. Office of the Federal Register). 1998. Endangered and threatened Northcote, T. G. 1997. Potamodromy in Salmonidae-living and moving in the wildlife and plants; determination of threatened status for the Klamath River fast lane. North American Journal of Fisheries Management 17:1029–1045. and Columbia River distinct population segments of Bull Trout. Federal Nyce, L. G. 2011. Genetic population structure and conservation of Bull Trout in Register 63:111(10 June 1998):31647–31674. the East Fork Bitterroot River drainage, Montana. Master’s thesis. University Waples, R. S., and P. R. England. 2011. Estimating contemporary effective of Montana, Missoula. population size on the basis of linkage disequilibrium in the face of migration. Peakall, R., and P. E. Smouse. 2006. GENALEX 6: genetic analysis in Excel. Genetics 189:633–644. Population genetic software for teaching and research. Molecular Ecology Warnock, W. G., J. B. Rasmussen, and E. B. Taylor. 2010. Genetic cluster- Downloaded by [Department Of Fisheries] at 19:52 28 May 2013 Notes 6:288–295. ing methods reveal Bull Trout (Salvelinus confluentus) fine-scale population Piry, S., A. Alapetite, J. M. Cornuet, D. Paetkau, L. Baudouin, and A. Estoup. structure as a spatially nested hierarchy. Conservation Genetics 11:1421– 2004. GeneClass2: a software for genetic assignment and first-generation 1433. migrant detection. Journal of Heredity 95:536–539. Weir, B. S., and C. C. Cockerham. 1984. Estimating F-statistics for the analysis Rannala, B., and J. L. Mountain. 1997. Detecting immigration by using mul- of population structure. Evolution 38:1358–1370. tilocus genotypes. Proceedings of the National Academy of Sciences of the Whiteley, A. R., P.Spruell, and F. W. Allendorf. 2004. Ecological and life history USA 94:9197–9201. characteristics predict population genetic divergence of two salmonids in the Rexroad, C. E. III, R. L. Coleman, A. M. Martin, W. K. Hershberger, and same landscape. Molecular Ecology 13:3675–3688. J. Killefer. 2001. Thirty-five polymorphic microsatellite markers for Rainbow Whiteley, A. R., P. Spruell, B. E. Rieman, and F. W. Allendorf. 2006. Fine- Trout (Oncorhynchus mykiss). Animal Genetics 32:316–331. scale genetic structure of Bull Trout at the southern limit of their distribution. Rice, W. R. 1989. Analyzing tables of statistical tests. Evolution 43:223– Transactions of the American Fisheries Society 135:1238–1253. 225. Yamamoto, S., K. Morita, I. Koizumi, and K. Maekawa. 2004. Genetic differ- Rieman, B. E., and F. W. Allendorf. 2001. Effective population size and genetic entiation of White-spotted Charr (Salvelinus leucomaenis) populations after conservation criteria for Bull Trout. North American Journal of Fisheries habitat fragmentation: spatial-temporal changes in gene frequencies. Conser- Management 21:756–764. vation Genetics 5:529–538. This article was downloaded by: [Department Of Fisheries] On: 28 May 2013, At: 19:53 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 Comment: Detection and Population Estimation for Small-Bodied Fishes in a Sand-Bed River Thomas P. Archdeacon a & Stephen R. Davenport a a U.S. Fish and Wildlife Service, New Mexico Fish and Wildlife Conservation Office, 3800 Commons Avenue North East, Albuquerque, New Mexico, 87109, USA Published online: 02 Apr 2013.

To cite this article: Thomas P. Archdeacon & Stephen R. Davenport (2013): Comment: Detection and Population Estimation for Small-Bodied Fishes in a Sand-Bed River, North American Journal of Fisheries Management, 33:2, 446-452 To link to this article: http://dx.doi.org/10.1080/02755947.2012.681011

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 33:446–452, 2013 American Fisheries Society 2013 ISSN: 0275-5947 print / 1548-8675 online DOI: 10.1080/02755947.2012.681011

COMMENT

Comment: Detection and Population Estimation for Small-Bodied Fishes in a Sand-Bed River

Fisheries researchers often rely on single-pass sampling to we show that Widmer et al. (2010) misinterpret their data and monitor changes in fish distributions, communities, and popu- present our own data to show that the single-pass data collected lations. Single-pass catch-per-effort (S-CPE) density estimates by Widmer et al. (2010) do not represent the type of data that are frequently used when budgets cannot support absolute abun- can be collected with single-pass seining, thereby invalidating dance estimates or it is determined that an unbiased, precise their comparisons of S-CPE data with estimates of abundance estimate of absolute abundance is not required for manage- generated by multipass removal sampling schemes. ment objectives (Simonson and Lyons 1995; Hubert 1996). In Specifically, we compare the fish communities collected by many cases, a single-pass approach provides adequate informa- the two sampling designs to show that further comparisons be- tion about population trends, species distributions, and commu- tween the data sets are valid. Next, we compare the effort re- nity composition. In other cases, however, the effort required quired by our single-pass methods with that required by S-CPE, to generate estimates of absolute abundance is often in excess comprehensive catch-per-effort (C-CPE), and the removal sam- of monitoring budgets and can be prone to violations of model pling reported by Widmer et al. (2010). We then use the methods assumptions, resulting in biases. Many monitoring programs of Widmer et al. (2010) to generate a species accumulation curve fail because there is no underlying question for monitoring to and extrapolate the required number of sites to generate a full address, managers cannot agree on what to monitor, or insuffi- species list for our single-pass data. Finally, we use linear re- cient data are collected to monitor changes in communities and gression to relate the species occupancy rates and densities (fish populations (Lindenmayer and Likens 2009). Addressing these per 100 m2) from our single-pass methods to the estimates gen- questions will help guide researchers in the choice of methods erated by the removal design employed by Widmer et al. (2010), for data collection in fisheries monitoring programs. comparing the results of the least labor-intensive method with Widmer et al. (2010) do not recommend monitoring fish those of the most labor-intensive method. communities, distributions, or populations in the middle Pecos River, New Mexico, by single-pass seining because of its poor species detection probability and poor correlation with estimates COMPARISON OF FISH COMMUNITIES AND EFFORT of abundance and because it has the lowest efficiency in gener- We collected fish community data at 12 nonrandom sites ating a full species list among the three methods tested. Their (Figure 1) between 30 October and 8 November 2007 (sam- low estimates of detection probability contradict one previous pling did not occur daily). We used a 3.0-m seine (mesh stretch

Downloaded by [Department Of Fisheries] at 19:53 28 May 2013 study in North American Great Plains streams in which all but measure, 3.2 mm) and performed 10–13 seine hauls per site the rarest fishes (those for which few individuals are collected) (mean = 11.8, SD = 1.0). Each seine haul was limited to a were highly detectable and the probability of missing common discrete, visually identified mesohabitat, following Vadas and species on a single pass was low (Scheurer et al. 2003). The Orth (1998) and Widmer et al. (2010). We calculated the total single-pass methods described by Widmer et al. (2010) are the catch, percent abundance, density, and observed occupancy for same methods that have been used by the U.S. Fish and Wildlife each species (Table 1). For each site, we calculated the species Service to monitor changes in fish communities and population richness and area sampled (Table 2). trends for Pecos Bluntnose Shiners Notropis simus pecosensis We compared the fish community from our sample with that in the same segment of the Pecos River since 1992. If the results of Widmer et al. (2010) using a modified Jaccard index of sim- and conclusions of Widmer et al. (2010) are correct, the mon- ilarity, hereafter JˆChao (Chao et al. 2005). A value of 1.0 for itoring program would require significant redesign to ensure this index indicates complete overlap of fish communities. We collection of adequate data for monitoring the fish community also calculated the more familiar Morisita index of similarity and Pecos Bluntnose Shiner populations. (C), which compares the relative abundance of species in two We collected fisheries community data in the same segment assemblages (Kwak and Peterson 2007). As with the JˆChao es- of river and time period as Widmer et al. (2010). In this comment, timator, a value of 1.0 indicates identical relative abundance in

446 COMMENT 447 Downloaded by [Department Of Fisheries] at 19:53 28 May 2013

FIGURE 1. Map of the study area showing the sampling locations for Pecos River ﬁsh community monitoring. 448 ARCHDEACON AND DAVENPORT

2 TABLE 1. List of species sampled, number captured (N), relative abundance (AR), observed occupancy (ψ), and observed density (D; ﬁsh per 100 m ), in the Pecos River, October–November 2007; NA = not applicable.

Species NAR (%) ψ (%) D Common Carp Cyprinus carpio 48 1.4 33.3 1.7 Red Shiner Cyprinella lutrensis 1,352 39.3 100 46.7 Plains Minnow Hybognathus placitus 247 7.2 83.3 8.5 Speckled Chub Macrhybopsis aestivalis 177 5.1 100 6.1 Arkansas River Shiner Notropis girardi 115 3.3 100 4.0 Rio Grande Shiner N. jemezanus 466 13.5 100 16.1 Pecos Bluntnose Shiner N. simus pecosensis 589 17.1 100 30.3 Sand Shiner N. stramineus 42 1.2 50.0 1.5 Fathead Minnow Pimephales promelas 47 1.4 75.0 1.6 River Carpsucker Carpiodes carpio 38 1.1 58.3 1.3 Mexican Tetra Astyanax mexicanus 3 <0.1 25.0 0.1 Catfishes Ictalurus spp. 11 0.3 58.3 0.4 Pecos Pupfish Cyprinodon pecosensis 12 0.3 0.8 0.4 Plains Killifish Fundulus zebrinus 155 4.5 50.0 5.4 Rainwater Killifish Lucania parva 2 <0.1 0.8 0.1 Western Mosquitofish Gambusia affinis 126 3.7 66.7 4.4 Inland Silverside Menidia beryllina 8 0.2 0.8 0.3 White Bass Morone chrysops 1 <0.1 0.8 <0.1 Bluegill Lepomis macrochirus 3 <0.1 0.8 0.1 Spotted Bass Micropterus punctatus 1 <0.1 0.8 <0.1 Largemouth Bass Micropterus salmoides 1 <0.1 0.8 <0.1 Total 3,444 NA NA NA

species composition. We used vegan version 2.0-2 (Oksanen The total area sampled by Widmer et al. (2010) at all sites et al. 2011) in Program R version 2.14.0 (R Development Core combined was 24,000 m2 for removal sampling (not including Team 2011) to calculate the indices. Both indices showed highly multiple passes) and C-CPE and 7,038 m2 for S-CPE, compared 2 similar communities, with JˆChao = 0.99 and C = 0.93, in spite with 2,769 m for S-CPE during our study, which amounts to of large differences in effort, as described below. 12.0% and 41.1%, respectively, of the area that they sampled. Per site, we averaged 231 m2 (SD = 30 m2) of area sampled (Table 2), which amounts to 56% of the area sampled per site TABLE 2. Number of species collected, area sampled, and number of seine hauls performed at each sampling site during fish community monitoring in the for S-CPE reported by Widmer et al. (2010). Pecos River, 30 October–8 November 2007. We collected 21 species in the Pecos River (78% of total species in October 2007), 2 of which were not collected by 2 Downloaded by [Department Of Fisheries] at 19:53 28 May 2013 Site Species Area (m ) Seine hauls Widmer et al. (2010), bringing the species richness to 27 in both Willow Creek 13 273 13 studies combined. Widmer et al. (2010) collected only 17 species Six Mile Draw 13 247 11 in 17 samples with single-pass methods, while covering nearly Bosque Draw 11 246 12 2.5 times the total area. Per site, our average species richness = Crockett Draw 11 228 12 was 10.6 (SD 2.2), 3.6 more species (51%) per single-pass Gasline 12 207 11 sample than in Widmer et al. (2010) in slightly over half the Highway 70 10 206 12 average area sampled by Widmer et al. (2010). Scout Camp 11 267 13 The C-CPE method of Widmer et al. (2010) requires the same Highway 380 7 216 13 amount of time as the S-CPE method but a crew of 10 instead of Dexter 8 265 11 3, equaling 3.3 times the effort. The removal method required Lake Arthur 7 239 13 a crew 3.3 times larger and took five times as long, enabling Highway 82 12 188 11 them to sample only one site per day as opposed to the five sites Brantley Inflow 12 187 10 per day that we were able to sample. We found that a 3-person Mean ± SD 10.6 ± 2.2 231 ± 30 11.8 ± 1.0 crew could reasonably be expected to sample five sites per day, agreeing with Widmer et al. (2010). Therefore, C-CPE requires COMMENT 449

3.3 times the effort of single-pass sampling, and removal sam- person crew (22.8 personnel-days), while the estimate with 167 pling requires 16.5 times the effort. In percentages, single-pass S-CPE samples by Widmer et al. (2010) requires 33.4 d with sampling requires only 33% of the effort required by C-CPE a 3-person crew (100.2 personnel-days). Widmer et al. (2010) sampling and 6% of the effort required by removal sampling. estimate that only 13 samples are needed if the removal method Requiring only 6% of the human effort of removal sampling is used, equaling 13 d with a 10-person crew (130 personnel- and covering 12% of the area sampled, our single-pass sampling days). Both our data and Widmer et al. (2010) clearly show produced a fish community nearly identical to that reported by that the most efficient method for generating a full species list Widmer et al. (2010). In this segment of the Pecos River, single- is single-pass sampling with more sites, as opposed to more pass sampling is clearly the most efficient method for monitoring intense sampling at fewer sites. fish communities among the three methods under discussion.

COMPARISON OF OCCUPANCY AND DENSITY SPECIES ACCUMULATION Widmer et al. (2010) report poor detection probability for To determine the rate of species accumulation, we random- S-CPE methods, with only a single species being observed ized the site order 10,000 times and calculated the cumula- at all 17 sites by single-pass seining (Red Shiner), while six tive mean species richness for each cumulative sample in our species were observed at all 17 sites by removal sampling. We data set. We used log-linear regression to describe the relation- found five species—Red Shiner, Speckled Chub, Arkansas River ship between cumulative samples and mean species richness Shiner, Rio Grande Shiner, and Pecos Bluntnose Shiner—at all (Colwell and Coddington 1994; Kwak and Peterson 2007; 12 sites in our study (Table 1), a highly improbable result if the Widmer et al. 2010). We used vegan 2.0-2 (Oksanen et al. 2011) single-pass seining detection probabilities presented by Widmer in Program R 2.14.0 (R Development Core Team 2011) to ran- et al. (2010) are correct. Specifically, if their estimated detec- domize the sample order and calculate mean cumulative species tion probabilities are correct, the probability of finding Speckled richness. Our data show that only 38 single-pass samples are Chub at all 12 sites with single-pass sampling is 4%, while the needed to collect 25 species (Figure 2), not 167 as predicted by probabilities of finding the Arkansas River Shiner, Rio Grande Widmer et al. (2010). In fact, Widmer et al. (2010) estimate that Shiner, and Pecos Bluntnose Shiner are 9.8, 22.2, and 1.1%, 50 C-CPE samples are required to capture 25 species, a much respectively. The joint probability of finding all four of those greater amount of effort than 38 S-CPE samples when consid- species at all 12 sites if the detection probabilities presented ering crew size. Widmer et al. (2010) conclude that the removal by Widmer et al. (2010) are correct is <0.001%. Previous re- method is the most efficient method to generate a full species search in Great Plains streams of North America also contradicts list; however, our results and their own data contradict this. Our these low detection probabilities, where common fish are highly estimate with 38 single-pass samples requires 7.6 d with a 3- detectable with a single pass (Scheurer et al. 2003). Clearly, the S-CPE detection probabilities determined by Widmer et al. (2010) are not representative of S-CPE detection probabilities in general. For the following calculations, we assumed that the occupancy and density reported by Widmer et al. (2010) from removal sampling represent the true (or as close as possible) occupancy rates and densities of the fishes in the Pecos River. If the

Downloaded by [Department Of Fisheries] at 19:53 28 May 2013 conclusions of Widmer et al. (2010) are correct, our observed densities and occupancy rates will correlate poorly with theirs because of poor detection probability and catchability. We used a log–log linear model (both axes loge + 1 transformed) to test how closely our single-pass observed fish density would predict the occupancy rates obtained by removal sampling (Widmer et al. 2010). We calculated a jackknife R2 to evaluate the fit of the regression model. The jackknife R2 is a leave-one-out cross validation (Efron and Gong 1983). We also calculated the 95% confidence intervals for the slopes to determine whether they differed significantly from 1.0. Fish density as determined by removal sampling was positively correlated with the fish densities produced by our single- FIGURE 2. Mean species richness from 10,000 randomizations of the sample 2 order of 12 single-pass seining samples in October and November 2007. The pass seining (Figure 3). A relatively high jackknife R indicated points represent the actual cumulative species richness values observed for the that most of the variation in our observed single-pass densities 12 sites. could be predicted from the true fish densities, suggesting that 450 ARCHDEACON AND DAVENPORT

can be accurately predicted by single-pass seining. The 95% confidence interval for the slope was 0.72–1.08, suggesting that true occupancy can be estimated at a 1:1 ratio from single-pass data. We emphasize that while not all fish were collected with a single pass and not all densities or occupancies were accurately predicted for a given species, most species were predicted accurately by a single pass, with a fraction of the effort required by removal sampling. We find it remarkable that there is such a strong relationship between our single-pass densities and the removal densities recorded by Widmer et al. (2010), given that we did not sample the same locations. Widmer et al. (2010) show poor correlation between their S-CPE densities and population estimates, but use inappropriate analyses in this regard. Widmer et al. (2010) use species-, age-, and site-specific combinations of population estimates as a response to species-, age-, and site-specific combinations of S-CPE density in a simple linear regression but do not account for the known sources of variation in the model and violate the assumption of independence. The effects of species, age-group, FIGURE 3. Relationship between single-pass catch per effort and the assumed true density of fishes (Widmer et al. 2010) in October 2007 in the Pecos River. and site might mask any relationship between S-CPE and popu- The dotted lines represent the 95% prediction band. lation estimates. Instead, the authors assume the slope of the line showing the relationship between S-CPE density and the population estimate is constant for all species, age-groups, and sites S-CPE can be used to detect trends in abundance. The 95% but offer no evidence that this assumption is correct and later confidence interval for the slope was 0.62–0.92, suggesting that present data showing that the slopes are different among species there is no direct 1:1 relationship between true fish density and and age-groups. A mixed-effects model (Pinheiro and Bates S-CPE fish density. We found a significant relationship between 2004) is more appropriate than simple linear regression for deter- our single-pass-seining occupancies and the occupancies determining this relationship, but even analysis of covariance would mined by removal sampling (Figure 4). As with fish density, account for the differences in slope among combinations. The single-pass occupancy rates were predicted from removal oc- tight relationship between C-CPE and population estimates is cupancy rates, suggesting that reachwide the true occupancy also more appropriately analyzed by mixed-effects models than by simple linear regression, but the increased effort—essentially sampling the entire closed river section—increases capture efficiency, reducing the variation between species, age-groups, and sites. The tight relationship between C-CPE and the population estimate is expected, as otherwise population estimates generated by the removal method would fail. Downloaded by [Department Of Fisheries] at 19:53 28 May 2013 DISCUSSION Single-pass seining might not be adequate for obtaining precise estimates of population size; however, such estimates might not be a worthwhile goal for short-lived species with densities that can fluctuate by an order of magnitude among months and years, such as the Pecos Bluntnose Shiner and other small- bodied fishes (Hatch et al. 1985; Hoagstrom et al. 2008). Further, multipass removal sampling estimates can be biased compared with mark–recapture estimates (Peterson and Cederholm 1984) or when capture probabilities decrease with each pass (Riley and Fausch 1992). When dealing with large numbers of fish, there is often so much uncertainty in the estimate, as in Widmer FIGURE 4. Relationship between the occupancy of fish species estimated from single-pass seining and that estimated by multipass removal sampling et al. (2010), that the estimate is not informative. Much of this (Widmer et al. 2010) in October 2007 in the Pecos River. The dotted lines uncertainty is likely because the proportion of river sampled represent the 95% prediction band. with removal sampling is very small; studies that successfully COMMENT 451

employ removal estimators are typically done on smaller searchers must choose methods that balance cost with the collec- streams. As stream size increases, the amount of effort required tion of sufficient data to meet research objectives. Many studies to sample a sufficient proportion of stream also increases and of fisheries methods were designed to help researchers choose the ability of block nets to maintain closure decreases as water methods by comparing the costs and results of less costly meth- scours beneath lead lines and around anchors. Typically, elec- ods with those of more labor-intensive—and hopefully more trofishing studies have shown that unless precise estimates of accurate—sampling designs. The comparison of single-pass abundance are needed, single-pass sampling is sufficient for de- and removal sampling population estimates in the Pecos River tecting population trends and determining community structure would have been extremely beneficial for managers working and that sampling more sites with a single pass is more efficient with threatened and endangered species in Great Plains streams. than making greater effort at fewer sites when collecting com- Unfortunately, Widmer et al. (2010) misinterpret and fail to cor- munity data (Paller 1995; Simonson and Lyon 1995; Kruse et al. rectly analyze their own data and provide a single-pass data 1998; Bateman et al. 2005; Bertrand et al. 2006; Reid et al. 2008, set that is not representative of the quality of data that can be 2009), mainly because the precision of the estimates increases collected with single-pass methods. Many of the conclusions as the fraction of habitat sampled is increased (Hankin and reached by Widmer et al. (2010) about methods for monitor- Reeves 1988; Bateman et al. 2005). The U.S. Fish and Wildlife ing fish communities are incorrect, and their comparisons of Service’s monitoring program for Pecos Bluntnose Shiners un- S-CPE with population estimates are not valid. Our S-CPE data derwent peer review in 2006 (Fagan 2006), showing that the set is more similar to the C-CPE data set presented by Widmer monitoring program would benefit from standardized sampling et al. (2010), which was strongly correlated with population locations (implemented in 2004) and that CPE was a useful and estimates, suggesting that replacement of their S-CPE data set defensible metric for monitoring Pecos Bluntnose Shiner pop- with a more robust single-pass data set would lead to different ulations. Single-pass CPE data were sufficient to show Pecos results and recommendations as to sampling methods. Bluntnose Shiners in steep decline over the 2002–2004 period (Fagan 2006). We find the differences in species richness, accumulation, ACKNOWLEDGMENT and detection probability between our single-pass samples and We thank the U.S. Bureau of Reclamation, Albuquerque Area the single-pass samples reported by Widmer et al. (2010) to be Office, for funding (contract 04-AA402212). C. Hoagstrom, K. difficult to explain. We can only attribute these differences to Gido, and four anonymous reviewers provided many helpful crew experience, that is, knowing how, where, and how much to comments on a previous version of the manuscript. We thank seine. Widmer et al. (2010) collected only 2 species by single- T. Knecht and N. Zymonas for field assistance. The views ex- pass sampling at a site at which 13 had been collected dur- pressed here are the authors’ and do not necessarily reflect those ing removal sampling, and 3 at another site at which 12 were of the U.S. Fish and Wildlife Service. collected during removal sampling. Seining based on discrete mesohabitats requires some judgment as to where to seine and THOMAS P. A RCHDEACON* how many seine hauls to perform. We do not find it logical to AND STEPHEN R. DAV E N P O RT cease sampling if only two or three species are found and would not recommend this, nor would we consider a site to be com- U.S. Fish and Wildlife Service, pletely sampled after only four seine hauls, as occurred at at New Mexico Fish and Wildlife Conservation Office, least one site in the Widmer et al. (2010) study. Another source 3800 Commons Avenue North East,

Downloaded by [Department Of Fisheries] at 19:53 28 May 2013 of the differences could be the misidentification of fish species. Albuquerque, New Mexico 87109, USA For example, we collected 589 Pecos Bluntnose Shiners at 12 sites, whereas Widmer et al. (2010) found only 496 in an area *Corresponding author: thomas [email protected] 8.6 times as large and sampled by closed system, multipass seining. Widmer et al. (2010) report collecting far more Arkansas REFERENCES River Shiners, a morphologically similar and usually less abun- Bateman, D. S., R. E. Gresswell, and C. E. Torgersen. 2005. Evaluating single- dant minnow than Pecos Bluntnose Shiners (Davenport 2011), pass catch as a tool for identifying spatial pattern in fish distribution. Journal as was the case in our single-pass data set. Confusion of these of Freshwater Ecology 20:335–345. Bertrand, K. N., K. B. Gido, and C. S. Guy. 2006. An evaluation of single-pass two species might account for some of the differences in ob- versus multiple-pass backpack electrofishing to estimate trends in species served total catch between data sets. A subset of fishes collected abundance and richness in prairie streams. Transactions of the Kansas by the U.S. Fish and Wildlife Service are retained as vouchers Academy of Science 109:131–138. and to identify difficult specimens, whereas none are available Chao, A., R. L. Chazdon, R. K. Colwell, and T. J. Shen. 2005. A new statistical for the Widmer et al. (2010) study. approach for assessing similarity of species composition with incidence and abundance data. Ecology Letters 8:148–159. Fisheries managers must make choices about how to moni- Colwell, R. K., and J. A. Coddington. 1994. Estimating terrestrial biodiversity tor fish communities and populations. Monitoring methods are through extrapolation. Philosophical Transactions of the Royal Society of often chosen based on the amount of funding available. Re- London B 345:101–118. 452 ARCHDEACON AND DAVENPORT

Davenport, S. R. 2011. Status and trends of Pecos Bluntnose Shiner Notropis tical Computing, Vienna. Available: CRAN.R-project.org/package=vegan. simus pecosensis, Pecos River, New Mexico. Report to the U.S. Bureau of (March 2012). Reclamation, U.S. Fish and Wildlife Service, New Mexico Fish and Wildlife Paller, M. H. 1995. Relationships among number of fish species sampled, reach Conservation Office, Albuquerque. length surveyed, and sampling effort in South Carolina coastal plain streams. Efron, B., and G. Gong. 1983. A leisurely look at the bootstrap, the jackknife, North American Journal of Fisheries Management 15:110–120. and cross-validation. American Statistician 37:36–48. Peterson, N. P., and C. J. Cederholm. 1984. A comparison of the removal and Fagan, W. F. 2006. Peer review of Pecos Bluntnose Shiner database and sampling mark–recapture methods of population estimation for juvenile Coho Salmon protocol. Final Report to the U.S. Fish and Wildlife Service, Albuquerque, in a small stream. North American Journal of Fisheries Management 4:99– New Mexico. 102. Hankin, D. G., and G. H. Reeves. 1988. Estimating total fish abun- Pinheiro, J. C., and D. M. Bates. 2004. Mixed-effects models in S and S-PLUS. dance and total habitat area in small streams based on visual estimation Springer-Verlag, New York. methods. Canadian Journal of Fisheries and Aquatic Sciences 45:834– R Development Core Team. 2011. R: a language and environment for statisti- 844. cal computing. R Foundation for Statistical Computing, Vienna. Available: Hatch, M. D., W. H. Baltosser, and C. G. Schmitt. 1985. Life history and ecology www.R-project.org. (March 2012). of the Bluntnose Shiner (Notropis simus pecosensis) in the Pecos River of Reid, S. M., N. E. Jones, and G. Yunker. 2008. Evaluation of single-pass elec- New Mexico. Southwestern Naturalist 30:555–562. trofishing and rapid habitat assessment for monitoring Redside Dace. North Hoagstrom, C. W., J. E. Brooks, and S. R. Davenport. 2008. Spatiotemporal American Journal of Fisheries Management 28:50–56. population trends of Notropis simus pecosensis in relation to habitat con- Reid, S. M., G. Yunker, and N. E. Jones. 2009. Evaluation of single-pass back- ditions and the annual flow regime of the Pecos River, 1992–2005. Copeia pack electric fishing for stream fish community monitoring. Fisheries Man- 2008:5–15. agement and Ecology 16:1–9. Hubert, W. A. 1996. Passive capture techniques. Pages 157–192 in B. R. Mur- Riley, S. C., and K. D. Fausch. 1992. Underestimation of trout population size phy and D. W. Willis, editors. Fisheries techniques, 2nd edition. American by maximum-likelihood removal estimates in small streams. North American Fisheries Society, Bethesda, Maryland. Journal of Fisheries Management 12:768–776. Kruse, C. G., W. A. Hubert, and F. J. Rahel. 1998. Single-pass electrofishing Scheurer, J. A., K. D. Fausch, and K. R. Bestgen. 2003. Multiscale processes predicts trout abundance in mountain streams with sparse habitat. North regulate Brassy Minnow persistence in a Great Plains river. Transactions of American Journal of Fisheries Management 18:940–946. the American Fisheries Society 132:840–855. Kwak, T. J., and J. T. Peterson. 2007. Community indices, parameters, and Simonson, T. D., and J. Lyons. 1995. Comparison of catch per effort and removal comparisons. Pages 677–763 in C. S. Guy and M. L. Brown, editors. Analysis procedures for sampling stream fish assemblages. North American Journal of and interpretation of freshwater fisheries data. American Fisheries Society, Fisheries Management 15:419–427. Bethesda, Maryland. Vadas, R. L., Jr., and D. J. Orth. 1998. Use of physical variables to discriminate Lindenmayer, D. B., and G. E. Likens. 2009. Adaptive monitoring: a new visually determined meso-habitat types in North American streams. Rivers paradigm for long-term research and monitoring. Trends in Ecology and 6:143–159. Evolution 24:482–486. Widmer, A. M., L. L. Burckhardt, J. W. Kehmeier, E. J. Gonzales, C. N. Medley, Oksanen, J., F. G. Blanchet, R. Kindt, P. Legendre, P. R. Minchin, R. B. and R. A. Valdez. 2010. Detection and population estimation for small-bodied O’Hara, G. L. Simpson, P. Solymos, M. H. H. Stevens, and H. Wag- fishes in a sand-bed river. North American Journal of Fisheries Management ner. 2011. VEGAN: community ecology package. R Foundation for Statis- 30:1553–1570. Downloaded by [Department Of Fisheries] at 19:53 28 May 2013 This article was downloaded by: [Department Of Fisheries] On: 28 May 2013, At: 19:54 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 Detection and Population Estimation for Small-Bodied Fishes in a Sand-Bed River: Response to Comment Ann M. Widmer a , Laura L. Burckhardt a , Jon W. Kehmeier a , Eric J. Gonzales b , C. Nicolas Medley c e & Richard A. Valdez d a SWCA Environmental Consultants, 295 Interlocken Boulevard, Suite 300, Broomfield, Colorado, 80021, USA b SWCA Environmental Consultants, 5647 Jefferson Street North East, Albuquerque, New Mexico, 87109, USA c New Mexico Interstate Stream Commission, Bataan Memorial Building, Suite 101, Don Gaspar Avenue, Santa Fe, New Mexico, 87504, USA d SWCA Environmental Consultants, 172 West 1275 South, Logan, Utah, 84321, USA e National Park Service, 1201 Oakridge Drive, Fort Collins, Colorado, 80525, USA Published online: 04 Apr 2013.

To cite this article: Ann M. Widmer , Laura L. Burckhardt , Jon W. Kehmeier , Eric J. Gonzales , C. Nicolas Medley & Richard A. Valdez (2013): Detection and Population Estimation for Small-Bodied Fishes in a Sand-Bed River: Response to Comment, North American Journal of Fisheries Management, 33:2, 453-458 To link to this article: http://dx.doi.org/10.1080/02755947.2013.768566

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 33:453–458, 2013 C American Fisheries Society 2013 ISSN: 0275-5947 print / 1548-8675 online DOI: 10.1080/02755947.2013.768566

COMMENT

Detection and Population Estimation for Small-Bodied Fishes in a Sand-Bed River: Response to Comment

The U.S. Fish and Wildlife Service (USFWS) currently uses estimates derived by Widmer et al. (2010). This validation single-pass seining without block nets (i.e., open catch per effort was not scientifically legitimate and does not support effort [S-CPE]) to monitor the status of the Pecos Bluntnose their conclusion that S-CPE is a reliable index of population Shiner Notropis simus pecosensis in the Pecos River. Widmer size. et al. (2010)1 estimated low species detection for S-CPE sam- 3. Archdeacon and Davenport (2013) suggested changes to a ples compared with closed multiple-pass removal samples in regression analysis published in Widmer et al. (2010), which the Pecos River. In their comment, Archdeacon and Davenport we implemented and which improved the fit of S-CPE data (2013, this issue) compare the S-CPE data sets collected from to removal population estimates. different sampling reaches of the Pecos River by the USFWS and Widmer et al. (2010) during October and November 2007. These two data sets were collected for different purposes, which COMPARISON OF FISH COMMUNITY RESULTS are reflected in their methods. Given the differences in method- AND EFFORT ology, Archdeacon and Davenport (2013) should not have found Comparison of Single-Pass Seining Results the differences in species richness, accumulation, and detection prob- Archdeacon and Davenport (2013) reported that the USFWS ability between our single-pass samples and the single-pass samples collected more species with less effort (measured as crew time reported by Widmer et al. (2010) to be difficult to explain. or area sampled) than Widmer et al. (2010) did using the S-CPE These differences are a result of the methods used and the habi- technique in the Pecos River in the fall of 2007. Key differences tats sampled, not the relative quality of the data sets or the in the methodologies used account for this greater efficiency. experience of the field crews, as Archdeacon and Davenport It is misleading for Archdeacon and Davenport (2013) to (2013) suggested. state that Archdeacon and Davenport (2013) expressed criticisms of [t]he single-pass methods described by Widmer et al. (2010) are the the study published by Widmer et al. (2010), none of which same methods that have been used by the U.S. Fish and Wildlife Ser- invalidate or diminish the value of that study. We have chosen vice to monitor changes in fish communities and population trends not to respond to each one but to focus our response on those in Pecos Bluntnose Shiners Notropis simus pecosensis in the same segment of the Pecos River since 1992. with scientific merit. The response is organized by the headings

Downloaded by [Department Of Fisheries] at 19:54 28 May 2013 used by Archdeacon and Davenport (2013). Although both studies collected S-CPE samples at discrete, visu- The major points made are as follows: ally identified mesohabitats (following Vadas and Orth 1998), there are major differences in the study objectives, sampling 1. Archdeacon and Davenport (2013) made valid comparisons reach selection, and sample placement that affect the sampling between fish community results and effort that demon- results. strate that their S-CPE method can be used to compile Widmer et al. (2010) selected sampling reaches using a gener- species lists at permanent sampling reaches using nonran- alized random tessellation stratified (GRTS) study design. The dom methods with less effort than expended by Widmer et al. GRTS is a type of stratified survey design that is probability (2010). based and provides spatial balance (Stevens and Olsen 2004). 2. Archdeacon and Davenport (2013) attempted to validate the At each sampling reach, Widmer et al. (2010) set up one to two USFWS S-CPE sampling data by comparison with an un- enclosures around sites (43–145 m in length) that had a wet- biased estimate of population size, the removal population ted width, water depth, and mesohabitat composition that was

1See also the erratum in North American Journal of Fisheries Management 32:1032 (2012). 453 454 WIDMER ET AL.

representative of the larger reach (i.e., average habitat quality). several passes) over larger areas. This, too, is an artifact of the All mesohabitat types were sampled in proportion to their avail- different sampling objectives of the two studies. The USFWS ability within the enclosure using S-CPE samples. Seine hauls monitors species presence/absence and samples in a manner to were spaced sufficiently far apart to minimize the likelihood of maximize species detection. The primary intent of the Widmer recapturing fish and disturbing other sampling locations within et al. (2010) study was to develop a method for estimating the the enclosure. A representative selection of mesohabitats was population sizes of Pecos River fishes, which involved sampling sampled so that the data collected would be comparable with the representative reaches of river (i.e., average habitats) and statis- removal estimates calculated for the same enclosure and so that tically expanding the results to a larger area. If Widmer et al. the results would be appropriate for making inferences about a (2010) had designed its sampling to maximize the number of larger area. species detected or to collect the greatest number of fish at a By contrast, the USFWS selected sampling reaches nonran- sampling reach, as the USFWS did, the resulting population domly, establishing index reaches that were concentrated in the estimates would have been biased upward. Removal sampling highest-quality habitat. The USFWS crew members have sam- collected all or nearly all of the species present in the blocked pled those reaches many times before (some since 1992) and use sampling sites, but not necessarily all of the species present at that experience to maximize fish collection at those specific loca- the corresponding sampling reaches. tions. The USFWS did not sample a representative mix of meso- Although the data are not provided for comparison, we hy- habitats but focused their effort on the high-quality mesohabitats pothesize that the linear length of river that the USFWS sampled where they knew they would collect fish (especially Pecos Blunt- (i.e., the length of river from which mesohabitats were selected nose Shiners) based on prior experience. This strategy is consis- for sampling) at a sampling reach is actually greater than the tent with their objective of detecting Pecos Bluntnose Shiners sum of the blocked site lengths sampled at a sampling reach by at every sampling reach in which they are present. Hoagstrom Widmer et al. (2010). Given the large differences in methodol- et al. (2008) described the USFWS’s Pecos Bluntnose Shiner ogy and the patchy distribution of high-quality habitats, stan- S-CPE monitoring as being based on “exploratory” sampling dardizing by river length may provide a more relevant compari- methods, appropriately acknowledging the potential bias in the son of the effort required per species detected than area seined. data. The crews’ relative experience sampling these specific sampling reaches represents another difference between the two SPECIES ACCUMULATION studies. We agree that these differences in fish collection and For reasons discussed previously, the USFWS was more ef- species detection rates could be influenced by “crew experi- ficient at sampling species richness than Widmer et al. (2010) ence, that is, knowing how, where, and how much to seine” when effort is standardized by area seined or crew time ex- (Archdeacon and Davenport 2013). However, the allegations pended. Thus, a full species list can be generated from fewer made by Archdeacon and Davenport (2013) that the Widmer of the USFWS S-CPE samples than the Widmer et al. (2010) et al. (2010) crew was inexperienced in sampling and in identi- S-CPE samples (Archdeacon and Davenport 2013). The Wid- fying Pecos River fishes are false and unsupported. The Widmer mer et al. (2010) removal sampling still detected the full species et al. (2010) field crew collectively had over 150 years of ex- list in fewer samples than the USFWS S-CPE samples but at perience conducting fisheries work in the southwestern United a much greater cost per sample due to larger field crew sizes, States and included an author of The Fishes of New Mexico larger sampling areas, and multiple passes through each area. (Sublette et al. 1990). The USFWS crew knew where and how The Widmer et al. (2010) estimates of sampling effort re-

Downloaded by [Department Of Fisheries] at 19:54 28 May 2013 to seine to maximize fish collection because they had sampled quired to detect all species present are representative of sam- those sampling reaches tens or hundreds of times previously. pling a representative mix of habitats at GRTS or randomly The Widmer et al. (2010) crew, although experts at sampling selected sampling reaches. The USFWS estimates are represen- with seines, sampled reaches they had never seen before. We tative of sampling to maximize species detection at permanent observed a similar increase in S-CPE sampling efficiency when sampling locations. The differences in methodology and objec- we repeatedly sampled a set of permanent sampling reaches for tives should be considered when interpreting and applying the an unrelated Pecos River fisheries study (Kehmeier et al. 2007). respective estimates of effort required. Based on the differences observed between the two data sets, it Consistent with a large body of fisheries literature (e.g., seems that years of experience sampling a reach may be a useful Angermeier and Smogor 1995; Paller 1995; Quinn and Deriso covariate when evaluating trends in catch-per-effort data. 1999; Patton et al. 2000; Dauwalter and Pert 2003; Hubert and Fabrizio 2007; Fischer and Paukert 2009), Widmer et al. (2010) Comparison of Removal Method and Single-Pass found that additional sampling effort yielded a greater catch Techniques and larger species list. While the USFWS’s sampling technique The USFWS collected nearly as many species using S-CPE appears to be efficient at generating a species list at perma- seining of much smaller areas as Widmer et al. (2010) did us- nent sampling reaches, they, too, would find more species with ing the removal method (depletion of a closed population over additional sampling effort. Periodic validation of their current COMMENT 455

species sampling using a more intensive sampling technique, [a] relatively high jackknife R2 indicated that most of the variation in such as the removal method, would add value to their monitor- our observed single-pass densities could be predicted from the true ing program (e.g., more reliable detection of rare species) even ﬁsh densities, suggesting that S-CPE can be used to detect trends in abundance if it is not cost-effective to employ it regularly. is not tested or supported by their analysis.

COMPARISON OF OCCUPANCY AND DENSITY Species Occupancy Archdeacon and Davenport (2013) repeated this error when Species Detection comparing species occupancy between methods using regres- Archdeacon and Davenport (2013) expanded the S-CPE de- sion. The results of their analysis (their Figure 4) simply demon- tection probabilities calculated by Widmer et al. (2010) for four strate that both methods detected common species at a greater species to demonstrate that the USFWS detected those four number of sampling reaches than they did less common species. species at a much higher rate than predicted by Widmer et al. Their interpretation that (2010). This finding appears to be consistent with their other results. However, it is worth noting that the USFWS did not have [a]s with fish density, single-pass occupancy rates were predicted the data to validate their own samples and provide comparable from removal occupancy rates, suggesting that reach-wide the true estimates of species detection probabilities. occupancy can be accurately predicted by single-pass seining Archdeacon and Davenport (2013) cite Scheurer et al. (2003) is again not tested or supported by their analysis. as a study that shows that common North American Plains Widmer et al. (2010) compared S-CPE samples and re- species are highly detectable with a single seine pass, which moval population estimates of density across multiple sam- is contrary to the Widmer et al. (2010) results using the S-CPE pling reaches (i.e., in a replicated study) for each of the species method. Archdeacon and Davenport (2013) omitted the fact that included in the analysis. To correctly validate their CPE and Scheurer et al. (2003) were seining into a block net, a technique species occupancy data, the USFWS will need samples that that was not used by either the USFWS or Widmer et al. (2010). are similarly paired by sampling reach and date so that species The technique used by Scheurer et al. (2003) is more comparable samples are replicated. to the C-CPE method of Widmer et al. (2010) (i.e., a single pass through an area blocked with nets) than to the S-CPE method. Comparison of S-CPE and Removal Population Estimates Widmer et al. (2010) documented detection rates for common Archdeacon and Davenport (2013) assert that Widmer et al. species with the C-CPE method that were similar to those ob- (2010) used inappropriate analyses to compare S-CPE and retained with the removal method, which is consistent with the moval samples because the data sets were not independent and high species detection rates reported by Scheurer et al. (2003). the simple linear regression used (Figure 3 of Widmer et al. 2010) did not account for known sources of variation (species, Species Density age, and sampling reach). Regarding data independence, there Archdeacon and Davenport (2013) misinterpreted the mean- is a violation of independence if the Y value at a particular Xi is ing of the regression model they used to compare fish density influenced by other Xi (Quinn and Keough 2002). The S-CPE derived from the USFWS’s S-CPE samples with the fish den- sample (Y) associated with the population estimate at a sam- sity derived from the Widmer et al. (2010) removal population pling site (Xi) is not influenced by the population estimates at estimates (fish per area). This misinterpretation led them to the other sampling sites, so there is no violation of independence

Downloaded by [Department Of Fisheries] at 19:54 28 May 2013 unsupported conclusion that there was a positive correlation in even though the S-CPE and population estimate at a sampling fish density between the two methods. To calculate fish density site are nested. for each species, they divided the sum of fish collected at all Regarding the use of covariates, we acknowledge that some sampling reaches by the sum of the area sampled at all sampling of the variation in the regression (Figure 3 of Widmer et al. 2010) reaches for both the USFWS data and the Widmer et al. (2010) could have been explained by the use of covariates. The species- removal data. They then regressed the results of one method and age-specific linear regressions (Table 8 in Widmer et al. against those of the other, so that each point in their Figure 2010) adequately answered the primary research questions that 3 represents one species (i.e., no replication). By comparing were posed in that study, that is, to determine the differences in a single estimate of density for each species between the two the effectiveness of the different methods (S-CPE, C-CPE, and methods, Archdeacon and Davenport (2013) are simply com- removal) by species and age-group. However, we conducted the paring the species composition of the catch between the two mixed-models analysis suggested by Archdeacon and Daven- data sets. Their Figure 3 illustrates that both surveys collected port (2013) to improve our estimates of the relationship between similar fish community data and that the most common species S-CPE data and removal population estimates when considering occurred at higher densities for both methods for all sampling all data combined. reaches combined. Archdeacon and Davenport’s (2013) inter- We transformed the paired S-CPE and removal population es- pretation that timates (loge[x + 1]) and compared them using a least-squares 456 WIDMER ET AL.

TABLE 1. Post hoc comparison of the least-squares means for the sampling TABLE 2. Post hoc comparison of the least-squares means for the species site covariate in the comparison of the S-CPE and removal population esti- (age-group) covariate in the comparison of the S-CPE and removal population mates. The first two characters of the site codes denote the reaches (upstream estimates. Species (age-groups) with different letters are significantly (α < 0.05) to downstream) in the upper (U), middle (M), and lower (L) strata, respectively different. (see Figure 1 in Widmer et al. 2010); the two digits that follow denote the sites sampled. Sites with different letters are significantly (α < 0.05) different. Species (age [years]) Least-squares mean

Sampling site Least-squares mean Red Shiner Cyprinella lutrensis (0) 0.2083 z Speckled Chub Macrhybopsis 0.0579 y L301 0.0723 z aestivalis (0) U502 0.0557 zy Arkansas River Shiner Notropis 0.0551 yx U302 0.0552 zy girardi (0) M201 0.0494 zyx Plains Minnow Hybognathus 0.0510 yxw U601 0.0491 zyx placitus (0) M302 0.0490 zyx Red Shiner (1+) 0.0427 yxwv M202 0.0442 yxw Plains Killifish Fundulus zebrinus 0.0388 yxwv M301 0.0435 yxw (1 + ) L201 0.0416 yxw Rio Grande Shiner Notropis 0.0379 yxwv M401 0.0372 yxw jemezanus (0) U401 0.0371 yxw Rio Grande Shiner (1+) 0.0298 xwv U201 0.0362 yxw Sand Shiner Notropis stramineus (0) 0.0274 xwv M101 0.0335 yxw Plains Minnow (1+) 0.0216 wv L101 0.0325 yxw Plains Killifish (0) 0.0188 v U501 0.0297 xw Pecos Bluntnose Shiner Notropis 0.0178 v U101 0.0248 xw simus pecosensis (0) U301 0.0229 w Sand Shiner (1 + ) 0.0171 v Pecos Bluntnose Shiner (1 + ) 0.0168 v Speckled Chub (1 + ) 0.0158 v + analysis with restricted maximum likelihood variance estimates Arkansas River Shiner (1 ) 0.0155 v in JMP 9. The covariates in the analysis were sampling site and nested species (age-group). Both sampling site and species (age-group) were treated as random variables because it was the first-pass catch and the removal population estimate had to be consistent with the study design and enabled us to maintain highly correlated or the population estimation procedure would the power to extrapolate the results to the whole study area have failed. The results from the mixed-model analysis suggest (Zuur et al. 2009). The inclusion of these covariates increased that age-group does not need to be included for analysis of C- 2 the adjusted R of the regression from 0.16 (Widmer et al. CPE data but should be included for analysis of S-CPE data for 2010) to 0.67. The sampling site covariate explained 5.55% of monitoring trends in the abundance of individual Pecos River the variation, and the species (age-group) covariate explained fish species. 53.52%. A post hoc comparison of the least-squares means us-

Downloaded by [Department Of Fisheries] at 19:54 28 May 2013 ing Student’s t-test revealed a few significant differences (α < 0.05) among sampling sites (Table 1) and species (age-group) DISCUSSION (Table 2). This result is consistent with the differences observed The primary intent of the Widmer et al. (2010) paper was to among the species- and age-specific regressions presented by develop a method for estimating the population sizes of Pecos Widmer et al. (2010; their Table 8). River fishes that would not be prone to the errors and uncertainties associated with catch-per-effort (CPE) data. The paper Comparison of C-CPE and Removal Population Estimates provides the first population estimates for these species in the We performed the same regression analysis with the paired Pecos River and the only estimates available for many of these C-CPE and removal population estimates. The inclusion of the species in any river. This study also provided us the opportunity sampling site and species (age-group) covariates increased the to assess the efficacy of other methods that may be or are be- R2 of the regression from 0.90 (Widmer et al. 2010) to 0.93. ing used to derive indicators of population status and trends in The sampling site covariate explained 8.09% of the variance, small-bodied fishes. and the species (age-group) covariate explained 5.83%. Clearly, The differences in the S-CPE data sets from the USFWS sampling reach, species, and age covariates are far less important and Widmer et al. (2010) are easily explained by differences to consider when using C-CPE data than when using S-CPE data. in sampling and site selection methods and do not invalidate We acknowledge Archdeacon and Davenport’s (2013) point that the results and conclusions of Widmer et al. (2010). Methods COMMENT 457

that attempt to provide unbiased population estimates and as- pling errors that continue to confound the interpretation of these sociated estimates of capture probability are the only reliable data. basis for determining population status (Hubert and Fabrizio Archdeacon and Davenport (2013) have implied that the US- 2007). Archdeacon and Davenport’s (2013) defense of the S- FWS does not have unbiased population estimates (from which CPE statistic for monitoring the Pecos Bluntnose Shiner was occupancy, species density, species richness, and relative abun- not supported by their analyses and did not resolve the funda- dance can be derived) with which to validate their S-CPE data. mental limitations of CPE data. Without any estimates of chang- Any comparisons with the Widmer et al. (2010) data set may ing capture probability (e.g., among species, ages, sampling be limited by differences in the sampling methods, locations, reaches, and occasions), the results of CPE sampling are po- and timing even if appropriate statistical techniques are used. If tentially confounded by a suite of untested assumptions (Fagan the USFWS intends to use their S-CPE samples as an index of 2006). abundance—not just to detect species presence/absence—they Archdeacon and Davenport (2013) stated that a peer review will need unbiased population estimates that are paired with of the USFWS monitoring program for Pecos Bluntnose Shiners their S-CPE samples to validate their index. This validation is a (Fagan 2006) indicated that “CPE was a useful and defensible necessary part of monitoring using an index measurement and metric for monitoring Pecos Bluntnose Shiner populations.” It is could be could be accomplished using techniques described by true that the review by Fagan (2006) favored S-CPE for monitor- Widmer et al. (2010) at the sampling reach level. ing the status of Pecos Bluntnose Shiners over an uninformative species percent composition metric. However, the review also stated that management goals based on the S-CPE metric needed ACKNOWLEDGMENTS to reflect the condition of the Pecos Bluntnose Shiner popula- The original project was funded by the New Mexico Interstate tion (Fagan 2006). In order to develop meaningful management Stream Commission and this response by SWCA Environmental goals based on an S-CPE metric, the Pecos River Bluntnose Consultants. We thank the North American Journal of Fisheries Shiner population must be characterized with population esti- Management for the opportunity to respond to this comment and mates (Widmer et al. 2010) and S-CPE must be proven a reliable two anonymous reviewers for their constructive feedback. This index of population size (Hubert and Fabrizio 2007). paper represents the views of the authors and not necessarily We do not argue with Archdeacon and Davenport’s (2013) those of any aforementioned agency or group. conclusion that seining less intensively at a greater number of sampling reaches may result in a nearly full species list in fewer ANN M. WIDMER,* LAURA L. BURCKHARDT, field crew person-hours than it takes to collect removal sam- AND JON W. KEHMEIER ples so long as the samples are collected from the highest- quality habitats. This capture efficiency may be further increased SWCA Environmental Consultants, with experience as a crew repeatedly visits permanent sampling 295 Interlocken Boulevard, Suite 300, reaches. However, the fish densities estimated from these sam- Broomfield, Colorado 80021, USA ples will be biased upward and this should be acknowledged; they may also be inappropriate for monitoring abundance (Cowx ERIC J. GONZALES 1991; Gryska et al. 1997). Unlike the removal samples collected SWCA Environmental Consultants, by Widmer et al. (2010), they cannot be expanded to produce 5647 Jefferson Street North East, statistically valid riverwide population estimates. Albuquerque, New Mexico 87109, USA

Downloaded by [Department Of Fisheries] at 19:54 28 May 2013 As stated in Widmer et al. (2010), we recommend the use of more robust methods to assess population status. Our re- C. NICOLAS MEDLEY1 sults suggest that CPE metrics (e.g., C-CPE) can be useful in assessing the population trends of individual fish species, but New Mexico Interstate Stream Commission, only if they are implemented as part of a highly standardized Bataan Memorial Building, Suite 101, study in which effort is made to minimize the variability in Don Gaspar Avenue, factors that affect fish capture probabilities (e.g., discharge, sea- Santa Fe, New Mexico 87504, USA son, temperature, sampling reach characteristics, time of day, water clarity, etc.). Currently, there is considerable unexplained RICHARD A. VALDEZ variability in the spatial and temporal distributions in the US- SWCA Environmental Consultants, FWS’s Pecos Bluntnose Shiner data set (Hoagstrom et al. 2008). 172 West 1275 South, While the recommendation made by Fagan (2006) to standard- Logan, Utah 84321, USA ize sampling reaches has been implemented, there is still no standard sampling protocol at each reach or documentation of *Corresponding author: [email protected] the environmental conditions under which the data were col- 1Present address: National Park Service, 1201 Oakridge Drive, lected, creating the opportunity for significant unexplained sam- Fort Collins, Colorado 80525, USA. 458 WIDMER ET AL.

REFERENCES and interpretation of freshwater fisheries data. American Fisheries Society, Angermeier, P. L., and R. A. Smogor. 1995. Estimating number of species and Bethesda, Maryland. relative abundances in stream-fish communities: effects of sampling effort Kehmeier, J. W., R. A. Valdez, C. N. Medley, and O. B. Myers. 2007. Rela- and discontinuous spatial distributions. Canadian Journal of Fisheries and tionship of fish mesohabitat to flow in a sand-bed southwestern river. North Aquatic Sciences 52:936–949. American Journal of Fisheries Management 27:750–764. Archdeacon, T. P., and S. R. Davenport. 2013. Comment: detection and popula- Paller, M. H. 1995. Relationships among number of fish species sampled, reach tion estimation for small-bodied fishes in a sand-bed river. North American length surveyed, and sampling effort in South Carolina coastal plain streams. Journal of Fisheries Management 33:446–452. North American Journal of Fisheries Management 15:110–120. Cowx, I. G., editor. 1991. Catch effort sampling strategies: their applica- Patton, T. M., W. A. Hubert, F. J. Rahel, and K. G. Gerow. 2000. Effort needed tion in freshwater fisheries management. Fishing News Books, Oxford, to estimate species richness in small streams on the Great Plains in Wyoming. UK. North American Journal of Fisheries Management 20:394–398. Dauwalter, D. C., and E. J. Pert. 2003. Electrofishing effort and fish species Quinn, G. P., and M. J. Keough. 2002. Experimental design and data analysis richness and relative abundance in Ozark Highland streams of Arkansas. for biologists. Cambridge University Press, Cambridge, UK. North American Journal of Fisheries Management 23:1152–1166. Quinn, T. J., II, and R. B. Deriso. 1999. Quantitative fish dynamics. Oxford Fagan, W. F. 2006. Peer review of Pecos Bluntnose Shiner database and sampling University Press, New York. protocol. Final Report to the U.S. Fish and Wildlife Service, Albuquerque, Scheurer, J. A., K. D. Fausch, and K. R. Bestgen. 2003. Multiscale processes New Mexico. regulate Brassy Minnow persistence in a Great Plains river. Transactions of Fischer, J. R., and C. P. Paukert. 2009. Effects of sampling effort, assemblage the American Fisheries Society 132:840–855. similarity, and habitat heterogeneity on estimates of species richness and rel- Stevens, D. L., Jr., and A. R. Olsen. 2004. Spatially balanced sampling of natural ative abundance of stream fishes. Canadian Journal of Fisheries and Aquatic resources. Journal of the American Statistical Association 99:262–278. Sciences 66:277–290. Sublette, J. E., M. D. Hatch, and M. Sublette. 1990. The fishes of New Mexico. Gryska, A. D., W. A. Hubert, and K. G. Gerow. 1997. Use of power analy- University of New Mexico Press, Albuquerque. sis in developing monitoring protocols for the endangered Kendall Warm Vadas, R. L., Jr., and D. J. Orth. 1998. Use of physical variables to discriminate Springs dace. North American Journal of Fisheries Management 17:1005– visually determined meso-habitat types in North American streams. Rivers 1009. 6:143–159. Hoagstrom, C. W., J. E. Brooks, and S. R. Davenport. 2008. Spatiotemporal Widmer, A. M., L. L. Burckhardt, J. W. Kehmeier, E. J. Gonzales, C. N. Medley, population trends of Notropis simus pecosensis in relation to habitat con- and R. A. Valdez. 2010. Detection and population estimation for small-bodied ditions and the annual flow regime of the Pecos River, 1992–2005. Copeia fishes in a sand-bed river. North American Journal of Fisheries Management 2008:5–15. 30:1553–1570. Hubert, W. A., and M. C. Fabrizio. 2007. Relative abundance and catch per Zuur, A. F., E. N. Ieno, N. J. Walker, A. A. Saveliev, and G. M. Smith. 2009. unit effort. Pages 279–325 in C. S. Guy and M. L. Brown, editors. Analysis Mixed effects models and extensions in ecology with R. Springer, New York. Downloaded by [Department Of Fisheries] at 19:54 28 May 2013