Threatened salmon rely on a rare life history strategy in a modi � ed and warming landscape

Flora Cordoleani (  ora.cordoleani@noaa.gov ) University of California Santa Cruz / NOAA Corey Phllis The Metropolitan Water District of Southern California Anna Sturrock School of Life Sciences, University of Essex Alyssa FitzGerald University of California Santa Cruz / NOAA George Whitman UC Davis Anthony Malkassian Mediterranean Institute of Oceanography, Aix Marseille Université Peter Weber Lawrence Libermore National laboratory https://orcid.org/0000-0001-6022-6050 Rachel Johnson NOAA / UC Davis


Introduction
Climate change is arguably the greatest emerging threat to global biodiversity and ecosystems functioning in this period of unprecedented change 1,2 . To track changing climate regimes, many species have shifted their phenology 3 , distribution 4,5 , and abundances 6 . While there has been considerable attention given to predicting species and community-level phenological responses to climate change 7,8 , far less attention has been given to understanding how the loss of within population variation and rare phenotypes might modulate population resilience to future climate forcings 9,10 .
Phenotypic diversity is one way for populations to buffer themselves against natural or anthropogenic perturbations [11][12][13][14] . Plasticity in migratory timing may be particularly important for riverine species, as their ability to respond to adverse ambient conditions is constrained by the stream network, leaving fewer options for lateral movements compared with terrestrial or marine species [15][16][17][18] . Furthermore, in many cases, anthropogenic land use changes have restricted access to high elevation habitats that would have otherwise provided thermal refugia 19 .
Salmonids exhibit extensive phenotypic plasticity, which could enhance population stability against disturbances by spreading risk across time and space (portfolio effect concept 11,13,20,21 ). However, multiple concurrent environmental forcings could weaken this portfolio effect and challenge species resilience to future climate change 22 . In particular, the combination of warming and habitat contraction, caused by dam construction and other water projects, has resulted in large population declines and erosion of salmon life history diversity, particularly for runs that rely on cooler high elevation habitats [23][24][25][26] . To understand how life history diversity may in uence salmon resilience to climate change, we tracked the relative success of different juvenile migratory strategies in California's Chinook salmon (Oncorhynchus tshawytscha) populations that spawn at the southern edge of the native species' range 27 , in a heavily modi ed environment. These salmon serve as a model system for early indication of the challenges faced by cold-water shes when access to thermal refugia has been drastically reduced 28 .
While juvenile salmonids at higher latitudes often spend multiple years in freshwater before emigrating to the ocean 27,29 , today, most salmon in the California Central Valley emigrate in their rst winter and spring before river temperatures become intolerable. An exception is found among spring-run Chinook salmon that used to dominate the region before the construction of impassable dams 23 . Two populations still have access to high elevation reaches, and exhibit a rare phenotype where juveniles remain in the river over-summer before emigrating the following fall 30 . This late-migrating phenotype relies on access to cool water for the entire rearing period and is therefore most likely to be negatively impacted by warming temperatures and impaired access to high elevation reaches. Here, we used otolith strontium isotope ratios and daily growth increments to reconstruct the juvenile emigration patterns and growth rates of returning (i.e., successful) adult spring-run Chinook salmon, and to estimate the contribution of different migration strategies to the reproductive population across environmental extremes. Speci cally, we  Table S2), allowing us to explore potential mechanisms underpinning the expression and success of alternative life history strategies. We also investigated how predicted future river temperatures will affect the availability of suitable over-summering habitat and the long-term viability of the late-migrating phenotype. In summary, we show how climate change may truncate salmon life history diversity, and how the loss of the late-migrant phenotype could negatively affect the long-term resilience of threatened spring-run Chinook salmon populations.

Results
The importance of rare phenotypes and life history diversity Otolith isotope pro les revealed three distinct juvenile life-history types (hereon referred to as "early", "intermediate" and "late" migrants; Figure 1  Late migrants may thus experience very different freshwater, and estuarine and nearshore marine conditions, potentially resulting in differential feeding, growth and survival opportunities 31,32 . While late migrants were the least commonly observed phenotype in juvenile monitoring traps 33 (Supplementary Figure S2a), they represented the majority of the returning adults (mean across years = 60%; Figure 2a). Conversely, on average, 19% of surviving adults were represented by intermediate migrants (mean natal rearing period = 84 days ± 27 days SD) and 21% by early migrants (mean natal rearing period = 15 days ± 14 days SD). Importantly, the contribution rate of each life history type varied considerably among years ( Figure 3). Half of the return years (2007,2008,2013)  Thermally suitable habitat in a warming climate Temperature strongly in uences salmonid physiology, growth and survival 36 . Thus populations with access to diverse water temperatures during incubation and natal rearing are predicted to exhibit increased phenotypic and phenological diversity 37 . To support late migrants, stream temperatures need to remain suitably cool over the summer to accommodate the extended rearing period. Mill and Deer Creek watersheds, along with upstream reaches of the Battle and Clear Creeks and the Yuba River, are among the few accessible and populated spring-run streams in the system that still provide suitable rearing temperatures to support all three phenotypes ( Figure 5 top panels). In accessible stream reaches where spring run were historically present but are now extirpated, only the Stanislaus River has temperatures that could support the late migrating phenotype. Importantly, increases in spring and summer stream temperatures by 2040 ( Figure 5  Yuba, American and Tuolumne Rivers ( Figure 5). Here we used a xed temperature threshold of 15°C after Richter and Kolmes 38 , yet we acknowledge that there is likely some variation in this threshold according to local water quality and food availability 39 .

Discussion
Here, using archived otolith tissues, we revealed how a diversity of growth rates and behaviors expressed during early life stages can shape population dynamics and resilience, via within-population portfolio effects, and why it is essential that conservation strategies developed for the recovery of vulnerable species support both common as well as rare phenotypes. The phenotypic diversity expressed by California Central Valley spring-run Chinook salmon has thus far enabled these populations to persist, despite habitat loss and degradation along their migratory corridor, warming temperatures, and an increasingly volatile Mediterranean climate 23,40 . We show for the rst time that the late-migrating strategy is the life-support for spring-run populations during current periods of warming. Therefore, conservation priorities should be placed on supporting this rare and climate-adapted behavior to promote the long-term persistence of spring-run populations predicted to confront an increase in future climatic extremes, such as extended droughts and marine heatwaves future [41][42][43] . Late migrants likely experience very different selective pressures to the other phenotypes; for example, entering the ocean in different seasons at larger sizes potentially reduces interspeci c competition and risk of mismatch with peak salmonid prey production during early ocean residence, a critical period for cohort success 44 .
For salmon and other species impacted by habitat contraction, restoring and maintaining a diverse mosaic of the habitats they require to support life history diversity will also be critical for their persistence 21 . For spring-run Chinook salmon, predicted stream temperatures under our climate change scenarios demonstrates the necessity for maintaining and expanding thermally suitable rearing habitat in order to support diverse growth rates and a broader spread of emigration timings. Spring-run Chinook salmon currently face high mortality during migration to the sea in the spring as early and intermediate migrants 45 , which is especially pronounced during drought conditions as evidenced by their poor representation in the adult returns in 2012, 2014 and 2018. Late migrants have evolved a drought-resilient strategy of leaving later in the fall when conditions are cooler, but they must be able to survive the heat of the spring and summer in headwater habitats. Improving access to cold water refugia, through habitat restoration and/or reintroductions to high elevation habitats above impassable dams, might be vital for preserving the late migrant life-history type now and in the climate future 46,47 . Substantially improving conditions along degraded migratory corridors could also be instrumental to bolster salmon resilience, particularly in wetter years when earlier migrants play an important role in population success.

Otoliths microchemistry
Otoliths were prepared at UC Davis per established techniques 48 . The otoliths' sagittal plane was sectioned on both sides using 600 and 1500 grit wet/dry sandpaper to expose the primordia and surrounding microstructure. The surface achieved a further ne polish using 3µm and 1 µm Al 2 O 3 lapping lms. Finished samples were mounted to a 1cm square glass pedestal using Gorilla Glue TM . The otoliths' dorsal side was photographed in 20x magni cation using a Qimaging digital camera (MicroPublisher 5.0 RTV) mounted to a Olympus BX60 microscope. Following imaging otoliths were analyzed for strontium isotopes at the UC Davis Interdisciplinary Center for Inductively-Coupled Plasma Mass Spectrometry by Laser Ablation on their Multi Collector Inductively Coupled Mass Spectrometer. We used the otolith strontium isotope methods described in Barnett-Johnson et al. 49,50 to reconstruct juvenile freshwater habitat-use and migration histories. In brief, the strontium isotope ratio ( 87 Sr/ 86 Sr) of freshwater habitats (the "isoscape") varies as a function of rock geology and weathering patterns 51 , and because there is no biological fractionation of strontium isotopes, the otoliths faithfully record the signature of the surrounding water and dietary sources. Strontium isotopes are a particularly powerful tool in the California Central Valley, because the spatial heterogeneity in rock types results in signi cant differences in isotope signatures among most of the salmon-bearing watersheds. Consequently, variations in 87 Sr/ 86 Sr and strontium concentration across Central Valley watersheds has proven useful for determining population of origin 52,50 and reconstructing juvenile rearing and migration behavior 53,54 .

Movement reconstruction
Otolith radius was used as a proxy for sh size at natal and freshwater exit. The otolith radius for each 87 Sr/ 86 Sr measurement was estimated by measuring the distance from the otolith core to the center of each laser pit along a standardized 90˚ axis 48 . Strontium isotope pro les representing changes in 87 Sr/ 86 Sr values as a function of otolith distance from the core were created for each otolith. Speci c location 87 Sr/ 86 Sr threshold values were used to identify the movement of Central Valley spring-run Chinook juveniles from one rearing region to the other. These values come from a Central Valley isoscape database 50,53,54 . We considered four distinct regions in this study: Natal tributary (i.e., Mill and Deer Creeks), Sacramento River, Sacramento-San Joaquin Delta (hereon "Delta"), and San Francisco-San Pablo Bay (hereon "Bay") & Ocean. We used changes in 87 Sr/ 86 Sr along the otolith transect to identify two key habitat shifts to reconstruct the size at which individuals exited (1) the natal tributary, and (2) freshwater (exit location is Chipps Island, river kilometer 73). Otolith radius at natal exit was calculated by linearly interpolating between otolith distances at the 87 Sr/ 86 Sr measurements on either side of the upper Sacramento River (point of Mill or Deer Creek exit and Sacramento River entry) strontium threshold value.
We used the lowest 87 Sr/ 86 Sr value found for the upper Sacramento River region in the Central Valley isoscape database. If for a given sh this threshold was never crossed, we determined it by visually identifying the closest point to the Mill/Deer Creek habitat 87 Sr/ 86 Sr threshold value in the strontium pro les. Similarly, otolith radius for freshwater exit was calculated by linearly interpolating between otolith distances at the 87 Sr/ 86 Sr measurements on either side of the Chipps Island (point of Delta exit and Bay entry) strontium threshold value. Finally, the Sacramento River at Freeport 87 Sr/ 86 Sr value threshold was used to identify the migration of spring-run juveniles from the mainstem Sacramento River to the Delta.

Clustering analysis
We conducted a clustering analysis 55 on the strontium pro les obtained from the otolith microchemistry analysis to investigate whether we could statistically identify groups of sh exhibiting similar juvenile rearing strategies among Mill and Deer Creek populations. Strontium pro les were considered as smooth curves or functions sampled at a nite subset of some interval (here the distance from the otolith core); the statistical methods for analyzing such data are described as "Functional Data Analysis" (FDA; see Ramsay and Silverman 56 for an overview of FDA). With FDA methods each pro le is modeled in an in nite functional space rather than considered as a discrete vector in a multidimensional space (as modeled in multivariate data analysis). The clustering analysis performed in this paper included the following steps: 1. A smoothing spline was tted to each pro le to predict continuous 87Sr/86Sr values for otolith radius distances between 0 and 1000 µm (using thespline function in R 57 ). This allows the direct comparison of all Mill and Deer Creek strontium pro les of different lengths.
2. Each smoothed pro le was then transformed into a functional data object, using a B-splines basis (using the fda package in R 56 ).

A principal component analysis (PCA) was performed on those functional objects (using fda package in R)
. This allowed us to identify the principal modes of variation of each functional data object, and reduce data dimensions which has been shown to help for clustering pattern recognition and processing time 55 . 4. We used a model-based clustering method, where the data were represented by a series of Gaussian Mixture Models (GMM) for which each point was associated with a probability of belonging to each potential cluster 58 . The mixture model parameters were estimated using the Expectation-Maximization (EM) algorithm.
5. The Bayesian Information Criterion (BIC) was used to select the best model with the optimum cluster number (using mclust package in R 59 ).

Early-life growth rate estimation
To estimate habitat-speci c juvenile growth rates we measured the otolith increment widths using Image Pro Premier 9.0 (Media Cybernetics) in each isotopically distinct habitat region 48 . Each otolith reading was assigned a score of "certainty" on a scale of 1-5, 5 being the highest certainty. This index is a combination of the reader's con dence in the accuracy of the increment placement and the quality or readability of the image (i.e., how likely it is that another reader would get the exact same increment width measurements). Otoliths with poor readability were eliminated from the analysis. A total of 86 otoliths were used for growth rate estimations.      Central Valley habitat suitability mapping under current and future climate conditions. Rearing temperature suitability (temperature < 15°C38) in accessible (orange lines) and inaccessible (i.e., blocked by impassable dam; blue lines) river reaches in the California Central Valley, focusing on months when temperature stress is most likely to impact rearing success for early and intermediate migrants (May; left panels) and late migrants (August; right panels). We examined suitability during our study period (top