2.2 Environmental datasets
We extracted environmental data for each sampling location from publicly available geospatial raster layers (Table S1). These 26 environmental data layers included 19 bioclimatic variables summarizing the mean, maximum, minimum and range values of temperature and precipitation across the study region (Fick & Hijmans, 2017), four vegetation layers (Huete et al., 2002), elevation (Rodriguez et al., 2006), net primary productivity (NPP; Running et al., 2015), and potential evapotranspiration (PET; Mu et al., 2011). We then tested for collinearity in these raster layers (Table S2) using the ‘removeCollinearity’ function in the R package virtualSpecies to select a subset of variables where no two variables had a Pearson’s correlation coefficient > 0.7, resulting in the following twelve uncorrelated variables: isothermality (i.e. temperature evenness; BIO3), minimum temperature of coldest month (BIO6), mean temperature of driest quarter (BIO9), precipitation seasonality (BIO15), precipitation of wettest quarter (BIO16), precipitation of warmest quarter (BIO18), precipitation of coldest quarter (BIO19), elevation, percent tree cover, leaf area index (LAI), NPP, and PET. Future projections of bioclimatic variables were taken from aggregated global climate models (Sesink Clee, 2017) for two representative concentration pathways (RCPs) 2.6 and 8.5, projected for year 2080 based on the Intergovernmental Panel on Climate Change (IPCC) 5th assessment report. RCP 2.6 represents a “best case” scenario in that global mean temperature is projected to rise by 0.3 to 1.7 °C by the late-21st century, whereas RCP 8.5 represents a “worst case” scenario and projects global mean temperatures to rise by 2.6 to 4.8°C. All climate layers have a spatial resolution of 30 arc seconds (approximately 1km2).