Seventy years of diminishing biocomplexity of California Central Valley hatchery steelhead, Oncorhynchus mykiss

We present annual release data according to the California water year (i.e., 1 October to 30 September, hereafter “release year”) since the water year period is more relevant to the Central Valley O. mykiss life cycle than the calendar year (e.g., upstream migration in fall and winter, spawning and emergence in winter, rearing in spring and summer [Williams 2006]). When the release period spanned two water years (5.1% of the total number released, 90.5% of which occurred before the water year 1976), the water year possessing the largest share of the release period was assigned as the release year. In rare instances when the release period spanned two water years and was split equally between the first and second water years (0.2% of all releases), the second year was set as the release year. A 3-year centered moving average was applied to the annual numerical release data to highlight longer term trends.

Release group biomass was calculated as the product of the total number of fish released and the mean fish mass for that release group. The mean annual fish mass-at-release for a hatchery was used as the mean mass-at-release for any release group's missing weight, length, or life history stage-at-release information (3.8% of all releases). A 3-year centered moving average was also applied to the annual biomass release data to smooth the time series data.

We analyzed release timing on a monthly scale because the release day-of-month was missing for 63.5% of all releases. The release period usually occurred within the same calendar month (80.3% of all releases), but occasionally a range of calendar months were reported (16.7% of all releases). In limited cases, only one release year was reported (3.0% of all releases). Due to these inconsistencies, we restrict all release timing and growth rate analyses to cases where the release start and end months are the same.

Geographic coordinates of release locations and river km distances from the releasing hatchery to the release location were obtained from Sturrock et al. (2019) (92.0% of all releases), RMIS (1.4% of all releases), or the electronic database provided by CDFW (0.7% of all releases). An additional 5.7% of release location coordinates and hatchery distances were newly determined using the methods described by Sturrock et al. (2019). Coordinates and distances are unavailable for 0.2% of all released fish due to insufficient descriptions of release site locations.

Fish sizes were reported as mean mass (75.2% of all releases) or length (19.5% of total) for each release group. Length-at-release was converted to mass-at-release (and vice versa) according to a relationship for Sacramento River O. mykiss (Hallock et al. 1961) to facilitate dataset comparisons:

where mass is measured in grams and fork length in millimeters. Note that eq. 1 is modified from the original version to permit the use of metric units instead of English units. Also, note that this relationship was determined for fish with FLs ≥ 325 mm but predicts masses for smaller fish within 5% deviation from a similar length-weight transformation reported for California Central Coast O. mykiss (Huber 2018; 53–442 mm FL range; R² = 0.99) across nearly the entire O. mykiss size range encountered in this investigation (97.0% of all hatchery fish with reported lengths smaller than 325 mm FL were larger than 53 mm FL).

When size ranges were reported, the midpoint was assigned as the mean length or mass for the release group. Occasionally, missing size data could be gleaned from written descriptions of the release group's life history stage (e.g., “fry“, “fingerlings”, and “yearlings”). In these cases (1.5% of all releases), the midpoint of the life stage-at-release mass range (see “Life-stage-at-release” below) was used.

All age information was estimated based on an assumed 1 February spawn date (Satterthwaite et al. 2010) and, therefore, should be considered apparent ages. Age analyses were restricted to cases when the release group's beginning and end months of release were identical (80.3% of all releases). Apparent ages were estimated as the difference between the release month midpoint and 1 February of the brood year.

We explored coarse- and fine-scale trends in life-stages-at-release composition. We first classified O. mykiss as sub-yearling (y−) or yearling or older fish (y+). We followed hatchery program guidelines (CA HSRG 2012, Appendix VIII) and assumed O. mykiss became yearlings once they grew to 71.2 g (∼180 mm FL). We further classified life history stage-at-release diversity according to fish sizes and standardized nomenclature guidelines (Interagency Ecological Program Steelhead Project Work Team (IEP Steelhead PWT) 1998): “yolk-sac fry” were defined as fish with masses <0.3 g; “fry”: ≥0.3 to <1.4 g; “parr”: ≥1.4 g to <26.3 g; “silvery parr”: ≥26.3 g to <71.2 g; “small smolts”: ≥71.2 g to <131.6 g; “large smolts”: ≥131.6 g to <219.6 g; “subadults”: ≥219.6 g to <954.0 g; and “adults”: ≥954.0 g. For cases where size data were missing but life-stage-at-release was described, “fed fry” were assumed to be fry, “fingerlings” were assumed to be parr, “advanced fingerlings” were assumed to be silvery parr, and “smolts” were assumed to be small smolts.

We characterized life stage diversity by calculating the reciprocal Simpson's index (RSI; Simpson 1949) for each hatchery and all hatcheries combined per release year. The RSI measures the evenness of a community and ranges from 0 (all life stages were equally represented in every release group) to 1 (all fish were planted at the same life stage). The formula used to calculate RSI was:where n_i is the number of individuals in life stage i and n is the pooled number of individuals across all life stages.

To investigate interannual variation in release practices, we divided the 70-year time series (1948 to 2017) into seven 10-year intervals and calculated the decadal coefficient of variation (CV₁₀) for six metrics associated with hatchery releases summarized annually at each hatchery program and for all O. mykiss hatchery programs combined. The metrics examined included (1) total number released, (2) total biomass released, (3) mean release month, (4) mean release distance downstream of the hatchery, (5) mean mass-at-release, and (6) mean age-at-release:where is the 10-year or decadal mean of the release metric annual totals or means for each hatchery or all hatcheries combined, and SD₁₀ is the decadal standard deviation for each hatchery or all hatcheries combined:where p equals 1/10, x_i equals the annual release metric mean for release year i, and is the decadal mean of the annual release metric means for each hatchery or all hatcheries combined.

The four main O. mykiss Central Valley hatcheries (Fig. 1) released 137.5 million artificially propagated O. mykiss in Central Valley waters from 1948 to 2017 (Table A1 and Fig. 2). The number of fish increased rapidly from 1948 to the early 1970s as the experimental program at CNFH became permanent and the NFH, MRFH, and FRFH programs were established in 1957, 1965, and 1970, respectively (Table A1 and Fig. 2). According to an inspection of the plot, release numbers stabilized from 1974 to 1996 ( = 2.7×10⁶ fish·year⁻¹, CV = 22%) before a steep decline to a mean of 1.6×10⁶ fish·year⁻¹ (CV = 18%) from 1997 to 2017 (Fig. 2).

Approximately 8.8 million kg of O. mykiss were released from the four hatcheries propagating O. mykiss from 1948 to 2017 (Table A1 and Fig. 2A). Note that this value is underestimated because of unreported size-at-release information for 4.8% of all hatchery-year combinations. Total biomass released (Fig. 2B) resembles trends in the abundance data (Fig. 2A) during the early part of the time series (i.e., a rapid rise from 1948 to 1971); however, unlike the abundance data, total biomass released annually did not exhibit a steep decline during the mid-to-late 1990s but held steady and averaged ∼1.6×10⁵ kg·year⁻¹ from 1974 to 2017 (CV = 26%; Fig. 2).

Across all hatcheries and all years, the overall mean release date was 12 March. From an inspection of the plot (Fig. 3), there was a noticeable shift in 2003, so we have presented the results before and after that date. The mean annual release date for the period from 2003 to 2017 ( = 17 February) was over 1.5 months earlier than the mean date from 1953 to 1995 ( = 24 March) (Fig. 3). There was a linear decline in mean annual release date (quantified as months since 1 October) versus release water year for the combined hatchery stock complex in the full time series from 1970 (the first year all four O. mykiss release programs operated simultaneously) until 2017 (Fig. 3; R² = 0.68, slope = −0.07 months·year⁻¹, 95% CI [−0.08, −0.06]).

The overall mean of the mean annual release location distances downstream from all hatcheries combined was 40.2 river kilometers (rkm) (CV = 63%; Fig. 4). Based on the plots, average stocking distances from the releasing hatchery (Fig. 1) were above the long-term average (i.e., farther from the hatchery) between 1976 and 2017 ( = 50.5 rkm, CV = 42%) but below the long-term average earlier in the time series (1948–1975, = 23.6 rkm, CV = 96%), respectively (Fig. 4). The overall mean of the mean annual release location distances upstream of the Pacific Ocean at the Golden Gate was 337.4 rkm (CV = 24%). Below and above average stocking distances from the Pacific Ocean occurred from 1976 to 2017 ( = 308.6 rkm, CV = 16%) and 1948 to 1975 ( = 382.4 rkm, CV = 25%), respectively.

The number of unique release sites during the period when all four release programs were operational (i.e., 1970–2017) ranged from six (1978 and 2017) to 21 sites (1974). From 1948 to 1974, as new hatcheries were constructed and early programmatic adjustments were made, the total number of release sites (R² = 0.86, slope = 0.73 sites·year⁻¹, 95% CI [0.61, 0.86]) and average number of fish released per site (R² = 0.44, slope = 8627 fish·site⁻¹, 95% CI [5233, 12052]) increased linearly. The opposite was observed from release year 1975–2017; the total number of release sites (R² = 0.21, slope = −0.09 sites·year⁻¹, 95% CI [−0.15, −0.04]) and average number of fish released per site (R² = 0.15, slope = −1774 fish·site⁻¹, 95% CI [−2909, −570]) declined linearly.

The mean fish mass-at-release across all 70 years was 64.3 g or 174 mm fork length (eq. 1; Table A1 and Fig. 5A). A positive linear relationship for mean annual fish mass-at-release versus release year is observed throughout the 70-year time series (Fig. 5A; R² = 0.80, slope = 1.44 g·year⁻¹, 95% CI [1.28, 1.59]).

A bimodal pattern is evident for age-at-release with peaks at ∼120–130 days old (1970s to 1990s) and ∼370–380 days old (all years) (Fig. 5B). Overall, the average annual age-at-release was 322 days (CV = 17%). Generally, fish were released at older and more consistent ages later in the time series (Fig. 5B). From an inspection of the plot, an apparent shift occurred in 1998 (Fig. 5B). In particular, the mean age-at-release from 1998 to 2017 was 359 days (CV = 8%) compared with 305 days for the pre-1998 value (Fig. 5B).

Small smolts, silvery parr, and parr were the most common life stages released and comprised 78.8% of all releases (Table A1 and Fig. 6). As fish at a given time of year were released at larger sizes (Supplement 2, Fig. S2.1), the corresponding life stage composition shifted to smaller smolts and less parr and silvery parr were released (Fig. 6). For example, after 1997, small smolts comprised more than 80% of all O. mykiss released (Table A1 and Fig. 6).

Over time, increased growth performance in the hatchery environment produced an apparent “substitution of growth for time” effect (Supplement 2, Fig. S2.1 [left panel]). These factors facilitated a rapid increase in the release of “advanced yearlings” (i.e., O. mykiss released at smolt sizes [71.2 g] or larger; Fig. 7). Logistic regression indicates that advanced yearlings comprised 25%, 50%, and 75% of all releases in 1978, 1988, and 1998, respectively (Fig. 7).

The decadal CVs (eqs. 3 and 4) for all six release metrics (total number released, total biomass released, mean release month, mean release distance downstream of the hatchery, mean mass-at-release, and mean age-at-release) decreased through time for the Central Valley O. mykiss population complex, as indicated by negative correlations with decade for each release metric (Fig. 8). The linear temporal declines were significant for total number released (R² = 0.71, slope = −8.3%·decade ⁻¹, 95% CI [−15.2, −2.7]), total biomass released (R² = 0.84, slope = −13.5%·decade ⁻¹, 95% CI [−21.4, −7.7]), mean release distance downstream of hatchery (R² = 0.75, slope = −18.0%·decade ⁻¹, 95% CI [−31.2, −10.2]), and mean mass-at-release (R² = 0.83, slope = −8.9%·decade ⁻¹, 95% CI [−13.6, −5.3]) (Fig. 8).

Increased phenotypic diversity of released fish can be readily achieved by considering changes to broodstock collection schedules, release sites, and rearing conditions. Run timing in salmonids is heritable, and selecting broodstock based on the earliest returning individuals can shift population run timing (Tillotson et al. 2019); therefore, it is important to collect hatchery broodstock over the entire return distribution and hatch fry over periods that track natural temporal patterns, including interannual shifts. Salmonids have evolved to use a mosaic of habitats that vary in temperature, energetic demands, and food resources, resulting in a diversity of growth rates and outmigration timing (Brennan et al. 2015; Lusardi et al. 2020; Cordoleani et al. 2021; Rossi et al. 2022). A diversity of juvenile rearing habitats encourages phenotypic variability in wild salmonids, both in terms of size and migration timing.

Provided continuing hatchery releases, hatchery managers could re-incorporate releasing hatchery juveniles at multiple life stages, including on-site releases of fry and parr that are more likely to use non-natal rearing habitats (e.g., tributaries, side/secondary channels, floodplains, and estuaries) and encourage phenotypic diversity. While smaller (younger) fish often have lower survival and return rates, they can contribute significantly to adult returns when they encounter favorable conditions in non-natal rearing habitats (Phillis et al. 2018; Sturrock et al. 2020). Additionally, with fixed rearing space in hatchery environments, producing a larger proportion of smaller (younger) fish is possible. This practice can also reduce the use of antibiotics or chemical treatments to treat pathogens, buffer against potential losses caused by low fry-to-smolt survival, and reduce the time that fish are subjected to artificial selection.

Due to their conservation status, all hatchery-origin steelhead released in the Central Valley must be marked with an adipose fin clip. Individuals smaller than the minimum size for adipose fin clipping at maximum rates (∼50 mm FL; Skalski et al. 2009) could be identified by parentage-based tagging (Pepping et al. 2020). Unlike mass marking, parentage-based tagging can uniquely identify the offspring of hatchery broodstock. Hatcheries can treat the releases of parentage-based tagged fry and larger (older) fish as “experimental”, creating an opportunity to attain valuable information (e.g., outmigration survival, return rates, straying rates, adult return age/size structure, etc.) critical to assess the outcomes of interventions in release practices.

Hatchery infrastructure such as cement raceways with uniform flow and thermal environments and predictable feeding schedules and food quantities can homogenize growth among fish. Conversely, implementing “natural growth regimes” (sensu Berejikian et al. 2012) that mimic natural variations in water temperatures, feeding rates, and rearing densities can promote proper smoltification and diversify age structure and size-at-age before release into the environment. The MRFH's “Natural Rearing Enhancement System” (NATUREs) pilot program utilizes naturalized rearing regimes, enriched environments, and predator training to increase post-release O. mykiss survival rates (Williams 2006; CA HSRG 2012, Appendix VIII) and serves as a model that could be adopted and expanded elsewhere.

Today, the Central Valley O. mykiss population complex is dominated by hatchery-origin fish (NMFS 2003; Lindley et al. 2007). Best available information indicates that natural-origin O. mykiss comprise only ∼6% to 16% of the O. mykiss in the northern Sacramento-San Joaquin Delta and San Francisco Estuary. While diversifying the portfolio of release practices is expected to benefit stock complex resiliency and stability, it is also important to consider the potential genetic and ecological impacts of releasing hatchery-origin pre-smolts on natural-origin O. mykiss and other life stages of hatchery-origin O. mykiss. Genetic risks include inbreeding depression, outbreeding depression, and domestication selection, and ecological risks include competition, predation, and disease (reviewed by Claussen and Philipp 2022). The degree to which these potential unintended consequences outweigh the possible benefits of a diversified release portfolio for Central Valley O. mykiss (and native salmonids in hatchery-supplemented systems, more generally) requires further study.

Survival to adulthood appears to be extremely low for Central Valley steelhead (CA HSRG 2012, Appendix VIII). Releasing more O. mykiss at or near hatcheries (i.e., on-site releases) may improve homing and survival by allowing more individuals to access diverse habitats and environmental conditions during outmigration (CA HSRG 2012). Such habitats likely offer variable growth opportunities and contribute to the expression of diverse life histories (e.g., Bourret et al. 2016). Furthermore, juveniles migrating over a variety of time periods likely spread mortality risk by reducing the chance that a large fraction of fish encounter, for example, stressful thermal conditions or predatory hot spots. Moreover, increased variability in release timing, either within or among hatcheries, could help ensure that some smolts enter nearshore habitats during or after the spring transition and the onset of coastal upwelling (Holtby et al. 1990; Lindley et al. 2009). Normally, productive marine food webs establish in the California Current by mid-March, with occasional delays occurring into late May or June (Barth et al. 2007; Lindley et al. 2009; Satterthwaite et al. 2014). Volitional releases at hatcheries or release of multiple age classes (e.g., age-1 and age-2 smolts; sensu Berejikian et al. 2012) would likely diversify maturation schedules and population age structure to better match historical conditions (Hallock et al. 1961; CA HSRG 2012, Appendix VIII) and narrow Central Valley population complex mortality by spreading risk through both space and time. Release programs that mimic size-at-age and outmigration timing variations within natural ranges of variability due to differences in spawning and emergence timing and growth rates are expected to buffer population dynamics and hedge against unpredictable environmental conditions.

One issue that exacerbates hatchery dominance is that program goals often focus on achieving a total number of fish released rather than maximizing biocomplexity and resiliency. Focusing on improving post-release survival and interannual stability in returns may allow fewer fish to be produced and released while sustaining the popular recreational inland fishery and allowing hatchery managers to meet broodstock needs consistently. For instance, reducing smolt release group sizes and diversifying release timing, locations, and sizes could minimize size-based interference competition, density-dependent effects, and (or) predation effects on hatchery- and natural-origin O. mykiss. During dry years when minimal outflow occurs and transporting fish is deemed necessary, post-release survival may be improved by continuing and expanding practices such as allowing trucked fish to acclimate to environmental conditions before release, releasing fish closer to in-stream and riparian cover, and matching smolt releases with environmental factors that enhance survival like rain events and turbidity spikes, low light conditions, and (or) reservoir releases.

Historical land use changes have profoundly altered California's Central Valley and reduced environmental variation and biocomplexity for Central Valley salmonids, including steelhead (Lindley et al. 2006; Carlson and Satterthwaite 2011). Restoring habitat complexity and connectivity is essential to promoting life history diversity, particularly in degraded environments. The concept of shifting habitat mosaics (Stanford et al. 2005; Brennan et al. 2019) considers how habitat diversity contributes to population resiliency by providing options that contribute to the expression of diverse life histories. For example, the relative productivity of Sockeye salmon (O. nerka) in a large Alaskan watershed varies temporally across locations but is ultimately stable at the basin scale, providing evidence of the importance of the shifting habitat mosaic for stabilizing salmon production across years with different conditions (Brennan et al. 2019). Similarly, juvenile Central Valley fall-run Chinook Salmon exhibited multiple rearing and migration timing strategies, enabling them to utilize spatiotemporal differences in growth opportunities between the American River mainstem and Sacramento-San Joaquin Delta during low and normal/high flow years (Coleman et al. 2022). Ultimately, integrating diversified hatchery release strategies with an increased pace and scale of environmental restoration activities is necessary to reconcile the water demand of humans and aquatic ecosystems in California's Central Valley.