Incorporating population-level variation in thermal performance into predictions of geographic range shifts

Determining how species' geographic ranges are governed by current climates and how they will respond to rapid climatic change poses a major biological challenge. Geographic ranges are often spatially fragmented and composed of genetically differentiated populations that are locally adapted to different thermal regimes. Tradeoffs between different aspects of thermal performance, such as between tolerance to high temperature and tolerance to low temperature or between maximal performance and breadth of performance, suggest that the performance of a given population will be a subset of that of the species. Therefore, species-level projections of distribution might overestimate the species' ability to persist at any given location. However, current approaches to modeling distributions often do not consider variation among populations. Here, we estimated genetically-based differences in thermal performance curves for growth among 12 populations of the scarlet monkeyflower, Mimulus cardinalis, a perennial herb of western North America. We inferred the maximum relative growth rate (RGR(max)), temperature optimum (T(opt)), and temperature breadth (T(breadth)) for each population. We used these data to test for tradeoffs in thermal performance, generate mechanistic population-level projections of distribution under current and future climates, and examine how variation in aspects of thermal performance influences forecasts of range shifts. Populations differed significantly in RGR(max) and had variable, but overlapping, estimates of T(opt) and T(breadth). T(opt) declined with latitude and increased with temperature of origin, consistent with tradeoffs between performances at low temperatures versus those at high temperatures. Further, T(breadth) was negatively related to RGR(max), as expected for a specialist-generalist tradeoff. Parameters of the thermal performance curve influenced properties of projected distributions. For both current and future climates, T(opt) was negatively related to latitudinal position, while T(breadth) was positively related to projected range size. The magnitude and direction of range shifts also varied with T(opt) and T(breadth), but sometimes in unexpected ways. For example, the fraction of habitat remaining suitable increased with T(opt) but decreased with T(breadth). Northern limits of all populations were projected to shift north, but the magnitude of shift decreased with T(opt) and increased with T(breadth). Median latitude was projected to shift north for populations with high T(breadth) and low T(opt), but south for populations with low T(breadth) and high T(opt). Distributions inferred by integrating population-level projections did not differ from a species-level projection that ignored variation among populations. However, the species-level approach masked the potential array of divergent responses by populations that might lead to genotypic sorting within the species' range. Thermal performance tradeoffs among populations within the species' range had important, but sometimes counterintuitive, effects on projected responses to climatic change.