Harvest, Return, and Stray Rates of Recycled Summer Steelhead in The

RIVER RESEARCH AND APPLICATIONS River Res. Applic. 32: 1790–1799 (2016) Published online 23 March 2016 in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/rra.3023

ANGLER HARVEST, HATCHERY RETURN, AND TRIBUTARY STRAY RATES OF RECYCLED ADULT SUMMER STEELHEAD ONCORHYNCHUS MYKISS IN THE COWLITZ RIVER, WASHINGTON

T. J. KOCKa*, R. W. PERRYa, C. GLEIZESb, W. DAMMERSc AND T. L. LIEDTKEa a US Geological Survey, Western Fisheries Research Center, Columbia River Research Laboratory, Cook, Washington USA b Washington Department of Fish and Wildlife, Vancouver, Washington USA c Washington Department of Fish and Wildlife, Salkum, Washington USA

ABSTRACT Hatchery ‘recycling’ programs have been used to increase angling opportunities by re-releasing fish into a river after they returned to a hatchery or fish trap. Recycling is intended to increase opportunities for fishermen, but this strategy could affect wild fish populations if some recycled fish remain in the river and interact with wild fish populations. To quantify hatchery return and angler harvest rates of recycled steelhead, we conducted a 2-year study on the Cowlitz River, Washington. A total of 1051 steelhead were recycled, including 218 fish that were radio-tagged. Fates of recycled steelhead were similar between years: 48.4% returned to the hatchery, 19.2% were reported captured by anglers, and 32.4% remained in the river. A multistate model quantified the effects of covariates on hatchery return and angler harvest rates, which were positively affected by river discharge and negatively affected by time since release. However, hatchery return rates increased and angler harvest rates decreased during periods of increasing discharge. A total of 21.1% (46 fish) of the radio-tagged steelhead failed to return to the hatchery or be reported by anglers, but nearly half of those fish (20 fish) appeared to be harvested and not reported. The remaining tagged fish (11.9% of the radio-tagged population) were monitored into the spawning period, but only five fish (2.3% of the radio- tagged population) entered tributaries where wild steelhead spawning occurs. Future research focused on straying behaviour, and spawning success of recycled steelhead may further advance the understanding of the effects of recycling as a management strategy. Copyright © 2016 John Wiley & Sons, Ltd. key words: Oncorhynchus mykiss; steelhead; telemetry; recycling; multistate model

Received 23 October 2015; Revised 26 February 2016; Accepted 03 March 2016

INTRODUCTION in the Cowlitz River basin of Washington State. The Cowlitz River historically supported large populations of wild fall Fishery management issues can be complex in riverine Chinook salmon Oncorhynchus tshawytscha, coho salmon systems where wild and hatchery Pacific salmon Oncorhyn- Oncorhynchus kisutch, chum salmon Oncorhynchus keta, chus spp. populations coexist. Many of the wild salmon and and winter steelhead (LCFRB, 2004), but these populations steelhead O. mykiss stocks in the Pacific Northwest of the are currently listed as threatened (NMFS, 2005, 2008). Voli- USA are imperiled, and several have been listed as threat- tional upstream passage for anadromous fish species was ened or endangered under the US Endangered Species Act eliminated for a large section of the Cowlitz River in the (NMFS, 2005, 2008). Because of the low abundance of wild 1960s when Mayfield Dam and Mossyrock Dam were con- stocks, hatchery-produced salmon and steelhead support structed. Two hatcheries, the Cowlitz Salmon Hatchery and most of the commercial and recreational fisheries that occur the Cowlitz Trout Hatchery, are operated to mitigate for the in marine and freshwater environments. Hatchery popula- construction of Mayfield and Mossyrock dams and to main- tions must be carefully managed, however, because hatchery tain sport fishing opportunities in the river. These hatcheries salmon have been shown to negatively affect wild salmon produce more than 13 million juvenile salmon and steelhead populations when interbreeding occurs (Waples, 1991; each year including spring Chinook salmon, fall Chinook Chilcote, 2003; Kostow et al., 2003; Chilcote et al., 2011). salmon, coho salmon, summer steelhead, winter steelhead, The management of wild and hatchery salmon and and coastal cutthroat trout Oncorhynchus clarkii (Tacoma steelhead populations is a central focus for fishery managers Power, 2011). The potential effects of the hatchery-produced fish have to be carefully monitored downstream of Mayfield Dam where wild salmon and steelhead populations exist. *Correspondense to: T. J. Kock, US Geological Survey, Western Fisheries Re- search Center, Columbia River Research Laboratory, Cook, Washington, USA. Adult returns to hatcheries often exceed broodstock require- E-mail: [email protected] ments, and one use of the excess fish has been to ‘recycle’ them

Copyright © 2016 John Wiley & Sons, Ltd. RETURN RATES OF RECYCLED STEELHEAD 1791 back into the river to increase angling opportunities. Recycling METHODS refers to the process of transporting fish from a hatchery or fish Study area trap and releasing them back into a river where they can be harvested by anglers. Although recycling is popular with Our study was conducted in the unimpounded 82-km fishermen, studies have shown that harvest rates of recycled reach of the lower Cowlitz River in southwest fish are often low. For example, Evenson and Cramer (1984) Washington State (Figure 1). The upper bound of the estimated that anglers harvested 10% of the spring Chinook study area was the Barrier Dam/Salmon Hatchery salmon that were recycled in the Rogue River, Oregon. complex, and the lower bound was the confluence of the Similarly, Schemmel (2009) estimated angler harvest of Cowlitz and Columbia rivers. Several tributaries in the recycled summer steelhead in the Clackamas River, Oregon, lower Cowlitz River are known spawning areas for wild to be 10%, and Jepson et al. (2013) found that anglers winter steelhead (LCFRB, 2004). These include (from up- harvested 4 and 17% of the summer steelhead that were stream to downstream) Brights Creek, Salmon Creek, recycled at two locations in the Willamette River. Hatchery Lacamas Creek, Olequa Creek, the Toutle River, return rates for recycled fish in these studies were low as well, Arkansas Creek, Delameter Creek, Ostrander Creek, and ranging from 10 to 39% (Evenson and Cramer, 1984; the Coweeman River (Figure 1). Schemmel, 2009; Jepson et al., 2013). Less than half of the Wild winter steelhead return to the Cowlitz River recycled fish from these studies were known to have been between December and April and spawn from March to removed from the river, either through hatchery return or May each year. Summer steelheads are not native to the angler harvest, which raised concerns that a large percentage Cowlitz River and are maintained through hatchery of hatchery fish that would have otherwise been removed from production to support a recreational sport fishery. The the river were available to spawn with wild fish. Owing to summer steelhead population enters the river from May these findings, recycling programs have become less common through October and is spawned in the hatchery during in recent years (Evenson and Cramer, 1984; Schemmel, 2009; December and January. Schemmel et al., 2012; Jepson et al., 2013). In this 2-year study, our goal was to evaluate steelhead recycling in the Cowlitz River to quantify hatchery return, Fish tagging and release angler harvest, and tributary stray rates. Information from previous studies failed to completely address the potential For our study, hatchery summer steelhead were recycled effects of steelhead recycling for several reasons: tag loss from mid-June to late-August in 2012 and 2013 (Table I issues likely resulted in the underestimation of angler ). Collection and tagging were conducted at the Cowlitz harvest and hatchery return rates (Tipping, 1998); hatchery Salmon Hatchery when fish were being processed during trap closures restricted fish from re-entering the hatchery the daily collection routine. During this process, groups which likely resulted in underestimating hatchery return of fish (≈20 fish/group) were crowded into a large basket rates and overestimating tributary stray rates (Schemmel and anesthetized using an electroanesthesia system et al., 2012), and important spawning tributaries for wild (model EA-100; Smith-Root, Vancouver, Washington, steelhead were not adequately monitored to determine if USA). Anesthetized fish were then sorted, and hatchery recycled steelhead moved into those locations (Schemmel summer steelhead were selected for tagging if they had et al., 2012; Jepson et al., 2013). Consequently, more rigor- not previously been tagged. Each fish was measured for ous studies are needed to better understand the effects of fork length (to the nearest centimeter), visually catego- recycling programs on the prevalence of hatchery/wild rized as male or female, and then Floy-tagged (model interactions. Because of uncertainty in stray rates and FD-68BC, Floy Tag, Inc., Seattle, Washington, USA) potential hatchery/wild fish interactions, fishery managers and opercle-punched. Individuals were tagged with a in the Cowlitz River basin abandoned steelhead recycling single Floy tag in 2012 and with two Floy tags in 2013. until its effects could be thoroughly evaluated. Thus, our A subset of the recycled population was also radio-tagged study was designed to estimate harvest and hatchery return (Table I). Radio transmitters (model Pices, Sigma Eight, rates while also examining how factors such as river Inc., Newmarket, Ontario, Canada; model MCFT2-3EM, discharge and time of year affected these rates. We also Lotek Wireless, Inc., Newmarket, Ontario, Canada) were focused on describing behaviour patterns of recycled fish gastrically implanted using procedures described in that remained in the river to determine if these fish were Keefer et al. (2004). Following tagging, steelhead were using tributaries where wild winter steelhead spawn. These transferred to a concrete holding tank and held overnight. findings could then be used to guide future decisions about The following day, tagged fish were loaded onto a fish- whether to continue or terminate steelhead recycling in the hauling truck and transported downstream and released Cowlitz River basin. at rkm 48 (Figure 1).

Figure 1. Map of the lower Cowlitz River and its major tributaries. Black circles are locations where ﬁxed radiotelemetry sites were located during both years of the study, and grey circles are locations where ﬁxed radiotelemetry sites were located during one of the study years. The grey shaded box is where recycled steelhead were released

Monitoring methods opercle punch, radio transmitter) used during the study. When recycled steelhead were encountered, the Floy tag Data were collected to quantify fish that returned to the and radio transmitter (when present) were removed and the hatchery and that were reported as harvested by anglers identification numbers and recapture date were recorded. If and to describe movement patterns of radio-tagged fish. a recycled steelhead returned without a Floy tag or radio The fish trap at the Cowlitz Salmon Hatchery is operated transmitter (identified by the opercle punch only), then only 5 days per week (Monday–Friday). Hatchery personnel the recapture date was recorded. Recycled steelhead that examined summer steelhead that were captured in the trap were recaptured at the hatchery were not recycled a second during June–December and counted recycled fish when they time. Angler harvest data were collected using voluntary were encountered. Recycled steelhead were identified by the angler reports and a creel survey. Contact information was presence of at least one of the three tags or marks (Floy tag, available on Floy tags and radio transmitters, and signs were posted throughout the study area to encourage anglers to report when recycled steelhead were harvested. The creel Table I. Summary of release dates and number of steelhead survey was conducted during June–January each year. Creel recycled during 2012 and 2013 survey technicians visited popular fishing locations in the system, interviewed anglers, and examined harvested fish Month and week 2012 releases 2013 releases Total to identify recycled steelhead. June, week 3 37 (12) 42 (7) 79 (19) Five fixed site telemetry stations were operated on the June, week 4 31 (1) 47 (7) 78 (8) mainstem Cowlitz River, and eight fixed sites were operated July, week 1 56 (16) 62 (17) 118 (33) on lower Cowlitz River tributaries to collect data on move- July, week 2 55 (5) 69 (14) 124 (19) ment patterns of radio-tagged recycled steelhead. Mainstem July, week 3 80 (15) 76 (11) 156 (26) fi July, week 4 77 (7) 61 (16) 138 (23) xed sites were located near the Cowlitz Salmon Hatchery, July, week 5 77 (17) 56 (11) 133 (28) the Cowlitz Trout Hatchery, the Interstate 5 boat ramp, the August, week 1 53 (3) 59 (14) 112 (17) mouth of the Toutle River, and near the mouth of the August, week 2 55 (25) 30 (12) 85 (37) Cowlitz River (Figure 1). The Interstate 5 and Toutle River August, week 3 28 (8) 28 (8) mouth sites were only operated during 2013. Tributary fixed Total 549 (109) 502 (109) 1051 (218) sites were located on Brights Creek, Salmon Creek, Cell numbers include the number of fish released with Floy tags and the Lacamas Creek, Olequa Creek, the Toutle River, at the con- number of recycled fish that were radio-tagged (numbers in parentheses). fluence of Arkansas and Delameter creeks, Ostrander Creek,

and the Coweeman River (Figure 1). The Brights Creek site transition rate matrix Q(t) with elements qrs(t) that represent was only operated during 2012. Fixed sites continuously the instantaneous risk of transitioning from state r to state s. monitored for the presence of tagged fish from June 2012 Under this model, the times between state transitions are to January 2013 and from June 2013 to February 2014. assumed to be distributed as exponential random variables Mobile tracking surveys were conducted from a boat on with rate qrs and mean time to transition 1/qrs. Transition the mainstem Cowlitz River on 9 October 2012, 19 intensities may be allowed to vary over time or as a function November 2012, 9 January 2013, 25 October 2013, 21 of individual or time-specific covariates. November 2013, 26 November 2013, 4 December 2013, In our study, recycled fish could have occupied three pos- and 22 January 2014. Boat mobile tracking surveys covered sible states: at large in the river (R), returned to the hatchery the reach between the Barrier Dam and the town of (H), or reported as harvested by an angler (A). Only transi- Longview, Washington (rkm 3). The 19 November 2012 tions from R to A and R to H were possible because mobile tracking event was terminated prematurely because recycled fish were not returned to the river after entering of weather conditions and included the reach between the the hatchery or being harvested by an angler (Figure 2). Barrier Dam and the mouth of the Toutle River. Mobile Data for the analysis represented a daily time series of occu- tracking was conducted from an automobile on 18 pied states, S(t = 0), S(t = 1), …, S(t = T) for days t =0,1,…, December 2013 and focused on tributaries in the lower T days after release, where T is the either the number of days Cowlitz River. Daily discharge data was obtained from the from release to return to the hatchery (H) or harvest by an US Geological Survey gage located downstream of angler (A) or the number of days between release and Mayfield Dam (#14238000). December 1 for fish that remained in the river (R) until the onset of the spawning period. For example, the data for a recycled fish that was reported as harvested by an angler Data analysis on the third day after release was represented as the state Hatchery return and angler harvest rates were quantified vector S(t)=(R, R, R, A) at times t = (0, 1, 2, 3). This struc- using data collected at the hatchery, from angler reports ture allowed us to express transition rates as a function of and from creel surveys. Recycled steelhead unaccounted time-varying daily covariates. for by harvest or hatchery return were assigned to a group To assess factors affecting the transition rates, we com- that was not known to have been removed from the river. pared an a priori set of alternative models that expressed This group consisted of fish that may have experienced transition rates as a log-linear function of covariates: fi one of several possible fates. These sh could have remained in the river and spawned, died prior to spawning, ðÞ¼ðÞ β βT ðÞ qrs zi t exp rs;0 exp rszi t been harvested by an angler and not reported, or migrated downstream and left the river. The primary purpose of the where zi(t) is a vector of covariate values for individual i at telemetry data was to distinguish between these possible time t, exp(βrs,0) is the baseline hazard rate, βrs is a vector of fates. Detection records from radio-tagged fish were summa- slope coefficients, and exp(βrs,k) is the hazard ratio for the rized to describe movement patterns and to assess potential fates of fish that were not known to have been removed from the river. The number of tagged fish that moved upstream to the Cowlitz Trout Hatchery and Cowlitz Salmon Hatchery (from the release site) was quantified, and travel times to these locations were determined. The elapsed time from release to arrival at these sites was calculated by subtracting the first date/time of detection at the site from the date/time when fish were released at rkm 48. Observations from fixed sites and mobile tracking were used to determine if recycled steelhead entered lower Cowlitz River tributaries and to identify fish that may have been harvested but not reported. A multistate model was used to estimate hatchery return and angler harvest rates and to quantify how specific covariates affected these rates. We used continuous-time Markov models known as multistate models (Jackson, 2014) in which individuals occupy a discrete set of states Figure 2. Schematic of the multistate model and the associated state and are allowed to transition from state r(t) at time t to state transition matrix used to estimate rates of transition (qrs)of s(t +dt) at time t +dt. Interest then centers on estimating the recycled steelhead from the river to the hatchery and to anglers

Copyright © 2016 John Wiley & Sons, Ltd. River Res. Applic. 32: 1790–1799 (2016) DOI: 10.1002/rra 1794 T. J. KOCK ET AL. kth covariate. We fit alternative models to the data using the RESULTS msm package (Jackson, 2014) in R (R Core Team, 2013), A total of 1051 steelhead were recycled during the 2-year study, which uses maximum likelihood to estimate model parame- including 549 fish in 2012 and 502 fish in 2013 (Table I). ters and standard errors. Recycling occurred weekly from mid-June through mid- We fit eight alternative models that expressed transition August and release groups ranged from 28 to 80 fish per week. rates as a function of covariates. The simplest model was the Radio-tagged fish were included in each of the weekly release null model that assumed that transition rates were constant. groups (Table I). Weekly release numbers peaked during We hypothesized that the magnitude of daily river discharge mid-July to mimic the run timing of hatchery summer steelhead (Q) and the daily change in discharge (ΔQ) might affect tran- in the Cowlitz River. sition rates and, in turn, the time between release and return to We found that 48.4% of the recycled steelhead returned to the hatchery or harvest by an angler. Therefore, we fitthree the hatchery, 19.2% were harvested by anglers, and the models: one with Q only, one with ΔQ, and one with both Q remaining 32.4% were not known to have been removed and ΔQ. Next, we added time to the model with Q and ΔQ from the river (Table II). There was no difference in the to assess whether transition rates varied as a function of time proportion of recycled steelhead that were assigned to the after accounting for flowcovariates.Tothismodel,weformed three fates during 2012 and 2013 (chi-squared test χ2 = 2.3; two additional models by adding either the month of release as p = 0.322). The median time from release to hatchery return a factor or the presence of a radio transmitter as an indicator was 10 days, and 87.2% of these fish returned to the variable (1 = radio transmitter, 0 = no radio transmitter). We hatchery within 30 days of release. Similarly, median time included release month to determine whether transition rates from release to angler harvest was 10 days, and 83.0% of differed among fish released in different months, after the angler harvest occurred within 30 days of release. accounting for other effects in the model. The presence of the radio transmitter was included to assess whether the radio Multistate model transmitter had an effect on transition rates. Finally, we fita model using only time since release to determine the isolated Three models had similar AIC values, and we selected the effects of that covariate on transition rates. For all models, we model with the fewest parameters and lowest AIC as the assumed the same model structure for both qRA and qRH.We best model for inference. The model with daily river ranked models using Akaike’s information criterion (AIC) discharge, daily change in river discharge, and time since and then selected the model with the lowest AIC for inference release had the lowest AIC value, but two other models (Burnham and Anderson, 2002). had AIC values that were within 2 AIC points of this model Given the best fit model, we examined effects of covariates (Table III). The model with daily river discharge, daily on transition rates by (1) using the sign and magnitude of the change in river discharge, time since release, and month of slope coefficients, (2) plotting the effect of each covariate on release had an AIC value that was 0.6 points greater, but a the daily probability of transition when the remaining deviance that was 7.4 points lower than the lowest AIC covariates were held at constant values, and (3) plotting the model. These differences indicated slightly better goodness cumulative probability of transitioning from state r to state s of fit but at a cost of four more parameters for the second- for a given release date and time series of covariates. Condi- ranked model, which resulted in a slightly higher AIC than tional on a transition not having occurred, the probability of the lowest AIC model. The model with daily river discharge, transitioning from state r to state s was calculated as daily change in river discharge, time since release, and an indicator for radio-tagged fish also had an AIC value that Ptð þ 1jztðÞÞ¼ExpðÞQztðÞðÞ was similar to the first model (ΔAIC = 1.7). This model fit the data only slightly better than the first model (<2-point where P(t +1|z(t)) is the 1-day transition probability matrix conditional on covariates z(t)ondayt, ‘Exp’ denotes the matrix exponential, and Q(z(t)) is the transition rate matrix that Table II. Estimated fate of recycled steelhead during 2012 and depends on covariate values z(t). The elements of the 2013 transition probability matrix, prs(t +1|z(t)), yield the daily probability of transitioning from state r to state s, conditional Fate on covariates on day t. Given a time series of time-varying Year Angler Hatchery River covariates, the cumulative probability of transitioning from state r to state s by day T was calculated as 2012 102 (18.6%) 258 (47.0%) 189 (34.4%) T 2013 100 (19.9%) 251 (50.0%) 151 (30.1%) ∏ ð þ j ðÞÞ prs t 1 zt . Total 202 (19.2%) 509 (48.4%) 340 (32.4%) t¼0

Table III. Model selection table for evaluating effects of covariates harvest and hatchery return across the range of daily on hatchery and angler transition rates of recycled steelhead in the changes in discharge. Slope coefficients for time since Cowlitz River release were negative and similar in magnitude for both Model K Deviance AIC ΔAIC angler harvest and hatchery return rates. This suggests that recycled steelhead were less likely to return to the hatchery Q + ΔQ + t 8 7237.3 7253.3 0.0 or be harvested by anglers as time since release increased. Δ Q + Q + t + m 12 7229.9 7253.9 0.6 To illustrate the simultaneous effect of covariates, we Q + ΔQ + t + r 10 7235.0 7255.0 1.7 t 4 7276.1 7284.1 30.8 used the multistate model to predict daily and cumulative Q + ΔQ 6 8011.1 8023.1 769.8 transition probabilities for steelhead recycled on 1 June Q 4 8035.1 8043.1 789.8 2012. Daily discharge data from 2012 were used for these ΔQ 4 8040.6 8048.6 795.3 predictions. For reference, discharge patterns in 2013 were No covariates 2 8057.6 8061.6 808.2 generally similar to those shown for 2012 (Figure 4). How- (constant rate) ever, 2013 flows were lower than 2012, particularly during 3 1 Q, discharge; ΔQ, daily change in discharge; t, time since release; m, month June and July when discharge was 100 to 150 m s lower of release as a categorical factor; r, indicator variable for radio-tagged fish; on most days. Daily probabilities of return to the hatchery K, number of parameters; AIC, Akaike’s information criterion; ΔAIC, dif- fi ference from the best model. varied considerably with discharge within the rst month after release but varied less thereafter owing to the effect of time since release on hatchery return rates (Figure 4). In difference in deviance values between the two models) and addition, the highest probabilities of return to the hatchery required two more parameters, which again favored the first occurred during the higher discharge periods. Daily model in terms of the best model for analysis (Table III). probabilities of angler harvest contrasted with daily The best-fit model revealed a positive effect of river transition probabilities of hatchery return; angler harvest discharge on angler and hatchery return rates (Table IV). probabilities increased during periods when river discharge However, the slope coefficient for angler harvest rate was decreasing (Figure 4). The cumulative probabilities (0.0037) was nearly three times greater than the slope revealed that most (65%) recycled steelhead would be coefficient for hatchery return (0.0015) which indicated that removed from the river within the first month after angler harvest rates were more influenced by river discharge release and that about 70% would be removed after than were hatchery return rates (Table IV). Over the 3 months (Figure 4). observed range in discharge, these rates translated to an increase from about a 0.01 to 0.03 daily probability of Fates of radio-tagged fish harvest and from a 0.027 to 0.045 daily probability of return to the hatchery (Figure 3). Daily change in river discharge Many radio-tagged fish moved quickly upstream following had opposite effects on angler harvest and hatchery return. release. We found that 82.1% of the radio-tagged fish The slope coefficient for angler harvest was negative that moved upstream and were detected at the Cowlitz Trout indicated that angler harvest rates decreased as daily change Hatchery fixed site following release. The median elapsed in river discharge increased (i.e. on the ascending limb of time from release to first detection at the site was 4.6 days the hydrograph). Alternately, hatchery return rates increased (range = 0.5–99.1 days). Similarly, 72.3% of the radio- as daily change in river discharge increased (Table IV). tagged fish moved upstream from the release site to the Figure 3 illustrates the differences in probability of angler Cowlitz Salmon Hatchery and the median elapsed time to

ﬁ βb ðβb Þ ﬁ Table IV. Parameter estimates for slope coef cients ( rs) and hazard ratios (exp rs ) from the best- t multistate model evaluating effects of covariates on angler (qRA) and hatchery (qRH) transition rates of recycled steelhead βb βb Transition rate Covariate rs (95% CI) exp rs (95% CI) qRA Baseline 6.5274 (6.7336, 6.3212) NA Q 0.0037 (0.0022, 0.0053) 1.0037 (1.0015, 1.0060) ΔQ 0.0065 (0.0111, 0.0019) 0.9935 (0.9871, 1.0000) t 0.0379 (0.0426, 0.0333) 0.9628 (0.9565, 0.9691) qRH Baseline 5.7213 (5.8631, 5.5795) NA Q 0.0015 (0.0004, 0.0027) 1.0016 (1.0000, 1.0031) ΔQ 0.0092 (0.0061, 0.0124) 1.0092 (1.0048, 1.0137) t 0.0401 (0.0432, 0.0369) 0.9607 (0.9565, 0.9650)

Baseline, transition rate when covariate values are set to zero; Q, discharge; ΔQ, daily change in discharge; t, time since release.

fish were harvested by anglers (25.2%; 55 fish) or were not known to have been removed from the river (21.1%; 46 fish). Analysis of detection records from the 46 radio-tagged fish that were not known to have been removed from the river suggests that few recycled steelhead entered tributaries where wild steelhead spawning occurs. Twenty of the radio- tagged steelhead (9.2% of all radio-tagged fish) had detection histories which indicated that they may have been harvested but not reported, based on repeated detections by mobile trackers near boat ramps in the lower Cowlitz River. The remaining 26 fish (11.9% of all radio-tagged fish) may have had the opportunity to interact with wild steelhead during the spawning period. However, only five recycled steelhead (2.3% of all radio-tagged fish) were ever detected in lower Cowlitz River tributaries. In 2012, one radio-tagged steelhead spent 7 days in the Toutle River during late August and early September, several months before spawning occurs, and then returned to the lower Cowlitz River. In 2013, four radio-tagged steelhead entered lower Cowlitz River tributaries, where they remained. Two of the fish were detected in Salmon Creek, and the other two fish were detected in the Toutle River. Of the 22 radio-tagged steelhead that were last known to have been detected in the lower Cowlitz River, 21 were located in the reach between the Cowlitz Trout Hatchery and the mouth of Ostrander Creek. The one remaining steelhead left the Cowlitz River in October 2013 and was later found in the Kalama River, Washington.

DISCUSSION About two thirds of the steelhead that were recycled for this study were removed from the lower Cowlitz River by either returning to the hatchery or being harvested by anglers. The proportion of recycled steelhead assigned to each of these groups was nearly identical in each year and closely matched results from a previous study. Tipping (1998) evaluated steelhead recycling in the lower Cowlitz River and found that 55% of the fish in his study returned to the hatchery while 14% were harvested by anglers. Results from the 3 years of research are consistent and showed that recycled steelhead were more than twice as likely to return to the hatchery, than to be harvested by anglers. The Figure 3. Effect of covariates on daily probability of being caught multistate model revealed how covariates affected hatchery by an angler or returning to the hatchery. For the effect of each return and angler catch rates and provided a better covariate, remaining covariates were held at constant values of t = 0 (days since release), ΔQ =0m3 s 1 (change in discharge), or understanding of the mechanisms responsible for the overall the mean discharge of Q = 152.4 m3 s 1 harvest and hatchery return fractions. Hatchery return and angler harvest rates were affected by first detection at that site was 6.3 days (range = 0.9– river discharge, and steelhead were most likely to be 80.8 days). About half (53.7%; 117 fish) of the radio-tagged removed from the river within 30 days of release. Our results steelhead were recaptured at the hatchery, and the remaining showed that hatchery return and angler harvest rates were

Figure 4. Predicted effects of discharge, change in discharge, and time since release on daily and cumulative probability of being caught by an angler or returning to the hatchery for a recycled fish released on 1 June 2012 highest when discharge was high, but the daily change in spawning occurs. First, we suspect that our estimate of discharge was the factor that was found to be the best angler harvest was negatively biased because we relied determinant of whether fish returned to the hatchery or were heavily on voluntary angler reporting as a means of harvested by anglers. During periods of increasing flow, determining when recycled steelhead were removed from steelhead were most likely to return to the hatchery, whereas the river by anglers. Several studies have found that angler harvest was highest during periods of decreasing voluntary angler reporting results in underestimates of true flow. These factors were particularly important shortly after harvest rates (Green et al., 1983; Bray and Schramm, release because >80% of the steelhead that were harvested 2001; McCormick et al., 2013). Data collected during by anglers or that returned to the hatchery were removed mobile tracking supports our hypothesis that some fish were from the river in less than 30 days. It is possible that angler harvested but not reported by anglers. Twenty radio effort varied with discharge. For example, anglers could transmitters (9.2% of the fish released) were repeatedly have been less likely to fish during periods when discharge detected near boat ramps in the lower Cowlitz River by was increasing and more likely to fish as discharge mobile tracking crews. It is possible that some transmitters decreased. We were not able to address this factor in our were regurgitated or that tagged fish died at these locations, analysis, but it should be considered when reviewing our but we believe that angler harvest was the most likely results. explanation. This belief is supported by the observation that Approximately one-third (32.4%) of the recycled these transmitters were not randomly distributed within the steelheads were not known to have been removed from the study area but were clustered near boat ramps commonly river, but there are several lines of evidence to suggest that used by fishermen. Second, we found that radio-tagged these fish did not enter tributaries where wild steelhead hatchery summer steelhead rarely entered lower Cowlitz

River tributaries where wild winter steelhead spawning Funding for this study was provided by the Washington occurs. We found that five radio-tagged steelhead (2.3% of Department of Fish and Wildlife with funds provided the tagged fish that were released) were detected in lower through the Columbia River Salmon and Steelhead Cowlitz River tributaries and one of those fish only spent Endorsement Program. Any use of trade, firm, or product about a week in the Toutle River during late August and names is for descriptive purposes only and does not imply early September. The Washington Department of Fish and endorsement by the US government. Wildlife operated resistance board weirs on Ostrander Creek, Delameter Creek, Olequa Creek, and Lacamas Creek during our study and never collected a summer steelhead REFERENCES (recycled or otherwise) at these locations (Gleizes et al., Bray GS, Schramm HL. 2001. Evaluation of statewide volunteer diary 2014). Finally, there is a strong likelihood that spawn timing program for use as a fishery assessment tool. North American Journal differs substantially between hatchery summer steelhead of Fisheries Management 21: 606–615. and wild winter steelhead in the lower Cowlitz River. Burnham KP, Anderson DR. 2002. Model selection and multimodel Summer steelhead spawning occurs primarily during inference: a practical information-theoretic approach. Springer Science December at the Cowlitz Trout Hatchery, and winter & Business Media: New York. – Chilcote MW. 2003. Relationship between natural productivity and the steelhead spawning occurs during March May in lower frequency of wild fish in mixed spawning populations of wild and Cowlitz River tributaries (Gleizes et al., 2014). Thus, it hatchery steelhead (Oncorhynchus mykiss). Canadian Journal of appears that there is a low likelihood of interbreeding Fisheries and Aquatic Sciences 60: 1057–1067. occurring between recycled summer steelhead and wild Chilcote MW, Goodson KW, Falcy MR. 2011. Reduced recruitment winter steelhead, but additional research may be required performance in natural populations of anadromous salmonids associated with hatchery-reared fish. Canadian Journal of Fisheries and Aquatic to assess straying rates and spawning success of recycled Sciences 68: 511–522. steelhead. Evenson MD, Cramer SP. 1984. An evaluation of recycling hatchery spring It was possible for recycled steelhead to remain in the Chinook salmon through the sport fishery in the upper Rogue River. Or- mainstem Cowlitz River or stray into other nearby rivers egon Department of Fish and Wildlife Final Information Report 84–10, and spawn, so there is still uncertainty regarding the 19 p. Available: https://nrimp.dfw.state.or.us/CRL/Reports/Info/84-10. pdf (April 2015). cumulative effects of recycling. Keefer and Caudill (2014) Gleizes, C, Serl J, Zimmerman M, Glaser B, Dammers, W. 2014. Lower reviewed the topic of straying by anadromous salmonids cowlitz river monitoring and evaluation, 2013. Report by the Washington and showed that hatchery steelhead strays have comprised Department of Fish and Wildlife, Vancouver, Washington, USA. a large proportion (50–70%) of certain steelhead runs that Green AW, Matlock GC, Weaver JE. 1983. A method for directly have returned to rivers in the Pacific Northwest during some estimating the tag-reporting rate of anglers. Transactions of the American Fisheries Society 112: 412–415. years. We did not design our study to evaluate straying rates Jackson CH. 2014. Multi-state models for panel data: the msm package for of steelhead that were recycled, but we did receive reports of R. Journal of Statistical Software 38:1–28. one recycled fish that was recovered in the Kalama River. Jepson MA, Keefer ML, Clabough TS, Caudill CC, Sharpe CS. 2013. Migra- Tipping (1998) also found that some recycled steelhead tory behavior, run timing, and distribution of radio-tagged adult winter were recovered in rivers located near the Cowlitz River in steelhead, summer steelhead, and spring Chinook salmon in the Willamette River, 2012. University of Idaho Technical Report 2013-1, 96 p. Available: his study. Additional factors that could be considered http://www.nwp.usace.army.mil/References.aspx (April 2015). include determining spawning success by recycled steelhead Keefer ML, Peery CA, Ringe RR, Bjornn TC. 2004. Regurgitation rates of and quantifying juvenile production from these fish, because intragastric radio transmitters by adult Chinook salmon and steelhead of the potential affects these progeny could have on during upstream migration on the Columbia and Snake Rivers. 24 – juveniles produced by wild steelhead (Tatara and Transactions of the American Fisheries Society :47 54. Keefer ML, Caudill CC. 2014. Homing and straying by anadromous Berejikian, 2012). These factors may be important to salmonids: a review of mechanisms and rates. Reviews in Fish Biology evaluate in future studies to further determine the effects and Fisheries 24: 333–368. of recycling adult steelhead. Kostow KE, Marshall AR, Phelps SR. 2003. Naturally spawning hatchery steelhead contribute to smolt production but experience low reproductive success. Transactions of the American Fisheries Society 132: 780–790. ACKNOWLEDGEMENTS [LCFRB] Lower Columbia Fish Recovery Board. 2004. Lower Columbia Salmon Recovery and Fish and Wildlife Subbasin Plan. Available: We would like to thank the following people for their assis- http://www.lcfrb.gen.wa.us (May 2015). tance with this project: Wade Heimbigner with the Pacific McCormick JL, Quist MC, Schill DJ. 2013. Self-reporting bias in Chinook States Marine Fisheries Commission; Missy Baier, Scott salmon sport fisheries in Idaho-implications for roving creel surveys. 33 – Gibson, Jamie Murphy, and Mark LaRiviere with Tacoma North American Journal of Fisheries Management : 723 731. [NMFS] National Marine Fisheries Service. 2005. Endangered and Power; Mike Blankenship, Teresa Fryer, and John Serl with threatened species: final listing determinations for 16 ESUs of West the Washington Department of Fish and Wildlife; and our Coast salmon, and final 4(d) protective regulations for threatened colleagues the Columbia River Research Laboratory. salmonid ESUs. Federal Register 70:123(28 June 2005):37160-37204.

[NMFS] National Marine Fisheries Service. 2008. Endangered species act Tacoma Power. 2011. Cowlitz River Fisheries and Hatchery Manage- section 7(a)(2) consultation biological opinion and Manguson-Stevens ment Plan. Available: http://www.mytpu.org/file_viewer.aspx? fishery conservation and management act essential fish habitat id=2979 (May 2015). consultation: NOAA, log number F/NWR/2005/05883. Tatara CP, Berejikian BA. 2012. Mechanisms influencing competition R Core Team. 2013. R: A language and environment for statistical between hatchery and wild juvenile anadromous Pacific salmonids in computing. R Foundation for Statistical Computing, Vienna, Austria. fresh water and their competitive abilities. Environmental Biology of http://www.R-project.org/. Fishes 94:7–19. Schemmel EM. 2009. Managing adult hatchery summer steelhead for a Tipping JM. 1998. Summer steelhead recycling experiment. recreational fishery with reduced hatchery and wild interactions. Washington State Department of Fish and Wildlife Report Number (Master’s thesis, Oregon State University, Corvallis, OR). 99-01. Schemmel E, Clements S, Schreck C, Noakes D. 2012. Using radio Waples RS. 1991. Genetic interactions between hatchery and wild telemetry to evaluate the success of a hatchery steelhead recycling salmonids: lessons from the Pacific Northwest. Canadian Journal of program. American Fisheries Society Symposium 76: 371–378. Fisheries and Aquatic Sciences 48(Supplement 1): 124–133.