Climate change impacts on population growth across a species’ range differ due to nonlinear responses of populations to climate and variation in rates of climate change

assumptions, but GAMs with only annual conditions satisfied all assumptions and showed similar results (S1 Appendix). Some responses of population growth rate to climate were roughly linear. For example, responses to average annual temperature at all populations besides the N population (Fig 7A) were roughly linear. However, many other relationships were highly nonlinear, showing unimodal responses (effect of coldest month temperature at the S and E2 populations, Fig 7C), decelerating responses (average annual temperature at the N population, Fig 7A), or more complex relationships (annual precipitation at the C popula- tion, Fig 7B). For most climate variables and populations, the LTRE approximation was inac- curate (Fig 7), resulting in a relatively weak correlation between the LTRE-predicted change in population growth rate and the GAM-predicted change in population growth rate (Pearson’s R = 0.62, n = 2500, S4 Fig in S1 Appendix). The direction and magnitude of bias of LTRE approximations differed dramatically across populations and across climate variables (Fig 7). Differences in climate conditions across populations resulted in variation in the response of population growth rate to particular climate conditions, and the resultant variation in climate responsiveness modulated the impact of each aspect of climate on population growth rate. Across populations’ climate conditions, the relationship between population growth rate and a particular climate variable differed in magnitude and in degree of nonlinearity. For example,

population growth rate increases linearly with changes in average annual temperature at most populations, but the N population’s response is decelerating (Fig 7A), resulting in only a Fig 5. Change in population growth rate due to climate change. Impact of climate change on population growth rate for all populations, expressed as the difference between future (2086–2100) and current (2008–2022) population growth rates. Points indicate the mean difference and error bars indicate 95% confidence intervals of differences (calculated across 500 bootstrapped samples from the distributions of model coefficients). As in other figures, populations are arranged by increasing current average annual temperature. See S2 Fig in S1 Appendix for population growth rates, and S3 Fig in S1 Appendix for GCM-specific results. https://doi.org/10.1371/journal.pone.0247290.g005 PLOS ONE Nonlinear responses of population growth rate to climate change PLOS ONE | https://doi.org/10.1371/journal.pone.0247290 March 3, 2021 11 / 20