Rare phenotypes and behaviours within a population are often overlooked, yet they may serve a heightened role for species imperilled by rapid warming. In threatened spring-run Chinook salmon spawning at the southern edge of the species range, we show late-migrating juveniles are critical to cohort success in years characterized by droughts and ocean heatwaves. Late migrants rely on cool river temperatures over summer, increasingly rare due to the combined effects of warming and impassable dams. Despite the dominance of late migrants, other strategies played an important role in many years. Our results suggest that further loss of phenotypic diversity will have critical impacts on population persistence in a warming climate. Predicted thermally suitable river conditions for late migrants will shrink rapidly in the future and will be largely relegated above impassable dams. Reconnecting diverse habitat mosaics to support phenotypic diversity will be integral to the long-term persistence of this species.

These datasets comprise the part of the manuscript that defines thermally suitable habitat in a warming climate. Temperature strongly influences salmonid physiology, growth and survival. Thus populations with access to diverse water temperatures during incubation and natal rearing are predicted to exhibit increased phenotypic and phenological diversity. 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 support all three phenotypes.

To explore why Deer and Mill Creeks may exhibit multiple juvenile life histories and how life history expression may change with climate change, we compared current and future thermal conditions along every current and historical spring-run stream. Temperature was obtained from a mean monthly stream temperature model. In brief, FitzGerald et al. (2021) employed a spatial stream temperature model to predict mean monthly stream temperature for nearly every river km in the western U.S. In the Central Valley, the test sample r2 was 0.813 and the mean absolute prediction error (MAPE) was 1.024°C. We first clipped this stream temperature dataset with the current and historical Central Valley spring-run distributions, and defined accessible reaches as those downstream of impassable barriers, such as dams. In general, the distribution and stream networks matched, but a few reaches with spring-run did not have stream temperature. We averaged the monthly temperature at each stream segment from 2005-2015, representing our study period. In the Central Valley, stream temperatures are predicted to increase by 0.6°C by 2040 and 1.0°C by 2080 (Isaak et al. 2017), so we applied these deltas to the temperature dataset.

We then examined stream temperature suitability (i.e, < 15°C) for juveniles rearing in May and August of 2005-2015, 2040, and 2080. Here we used a fixed temperature threshold of 15°C because temperatures greater than ~15°C result in decreased growth rates and increased mortality rates, yet we acknowledge that there is likely some variation in this threshold according to local water quality, food availability, and the life stage considered.

Attached are two shapefiles for data reproduction. The shapefiles are Central Valley spring-run Chinook salmon spatial distributions (1: current and historical distributions; 2: accessible distribution) merged with monthly stream temperature. We include the months (May and August) and years of interest from our paper (2005-2015 averaged, 2040, 2080). We also include other months for the above year categories (June, July, September, October, November) and each individual year from 1993-2015 for other research. For our paper, we applied a threshold of 15°C, but we include all temperatures in these files for other research. See associated README_CVSCStemps file.