Integrating ensemble species distribution modeling and statistical phylogeography to inform projections of climate change impacts on species distributions

Species distribution models (SDMs) are commonly used to forecast climate change impacts on species and ecosystems. These models, however, are subject to important assumptions and limitations. By integrating two independent but complementary methods, ensemble SDMs and statistical phylogeography, I was able to address key assumptions and create robust assessments of climate change impacts on species\u27 distributions while improving the conservation value of these projections. This approach was demonstrated using Rhodiola integrifolia, an alpine-arctic plant distributed at high elevations and latitudes throughout the North American cordillera. SDMs for R. integrifolia were fit to current and past climates using eight model algorithms, two threshold methods, and between one and three climate data sets (downscaled from general circulation models, GCMs). This ensemble of projections was combined using consensus methods to create a map of stable climate (refugial habitat) since the Last Interglacial (124,000 years before present). Four biogeographic hypotheses were developed based on the configuration of refugial habitat and were tested using a statistical phylogeographic approach. Statistical phylogeography evaluates the probability of alternative models of population history given uncertainty about past population parameters, such as effective population sizes and the timing of divergence events. The multiple-refugia hypothesis was supported by both methods, validating the assumption of niche conservatism in R. integrifolia, and justifying the projection of SDMs onto future climates. SDMs were projected onto two greenhouse gas scenarios (A1B and A2) for 2085 using climate data downscaled from five GCMs. Ensemble and consensus methods were used to illustrate variability across these GCMs. Projections at 2085 showed substantial losses of climatically suitable habitat for R. integrifolia across its range. Southern populations had the greatest losses, though the biogeographic scale of modeling may overpredict extinction risks in areas of topographic complexity. Finally, past and future SDM projections were assessed for novel values of climate variables; projections in areas of novel climate were flagged as having higher uncertainty. Integrating molecular approaches with spatial analyses of species distributions under global change has great potential to improve conservation decision-making. Molecular tools can support and improve current methods for understanding species vulnerability to climate change, and provide additional data upon which to base conservation decisions, such as prioritizing the conservation of areas of high genetic diversity in order to build evolutionary resiliency within populations