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

Variation in the impacts of climate change on population growth rate is driven primarily by differences in populations’ climate responsiveness, suggesting that future studies should recog- nize that climate responsiveness, and thus the degree to which current response to climate pre- dicts future response, could vary across populations. Differences in climate responsiveness across populations are not surprising; we know that interspecific variation in life history char- acteristics can modulate the impact of a given amount of climate change. For example, species that are shorter lived [34] or have shorter generation times [35] tend to be more impacted by changes in environmental conditions, such as climate change. We also know that at the intraspecific level, life history characteristics can vary substantially across populations [12], meaning that populations could vary in their responsiveness to changes in environmental con- ditions. For example, perhaps populations with lower survival (and thus shorter generation times) have higher climate responsiveness. Among-population variation in climate itself can generate variation in life history that then modulates the impact of climate. For example, rela- tive to other populations, the N population is less sensitive to the positive effects of average annual temperature on population growth rate (Fig 8), likely because changes in other climate variables will modulate its sensitivity to mean annual temperature. Namely, large increases in coldest month temperature at this population (Fig 4B) will lead to reduced probability of

reproduction of those larger plants (Table 1), tempering the positive effects of higher average annual temperature on reproduction (mediated through larger sizes due to higher mean growth rates; Table 1). Nonlinear responses of population growth rate to climate are very common, and the degree of nonlinearity, as well as the shape of the nonlinear relationship, varies dramatically across populations. The sensitivities to climate variables at current climate conditions are very different, in both magnitude and sign, than the climate responsiveness, which incorpo- rates the entire range of temperature and precipitation conditions. For example, at the E1 population, the (linear) sensitivity to average annual temperature at current climate condi- tions is negative (Fig 6), which suggests that increases in average annual temperature should decrease population growth rate. However, when we incorporate nonlinearities in the response to changes in temperature, we see a net positive effect of increases in average annual temperature on population growth rate (Fig 8). The direction and magnitude of bias in LTRE predictions varies across populations (due to variation in their climate conditions; Fig 7A), with similar effects for other climate variables (Fig 7B–7D). While the possibility of non- linearities in responses of population growth rate to climate and other drivers are often acknowledged in population-level analyses, we know little about their prevalence [36, 37], let alone the degree to which the presence or degree of nonlinearity in responses differs

across populations. Particularly concerning is the growing body of evidence that nonlinear population responses to climate are common and widespread [26, 38], and that it is difficult to quantify how large increases in temperature will affect population dynamics due to the dearth of data on population responses at historically unprecedented temperatures [39]. While past work has shown that a linear approximation is generally reasonable when calcu- lating elasticities to vital rates [40], these analyses ignore covariances among vital rates, which are present in D. alaskana, as well as any other species in which one climate variable affects multiple vital rates (as likely occurs for most species). Thus, our results suggest that population-specific nonlinearities (effected in our study by variation in climate conditions) may be critical in mediating overall population response to changes in climate or other driv- ers for a variety of species. Our metrics of climate responsiveness and of sensitivity to climate variables assume that variation in climate response arises only from climate differences across populations, but we know that natural selection could also lead to variation in response across populations. Only a PLOS ONE Nonlinear responses of population growth rate to climate change PLOS ONE | https://doi.org/10.1371/journal.pone.0247290 March 3, 2021 16 / 20