Genetics-based species distribution models: Building better predictions of global change
Most species distribution models (SDM) operate under the simplifying assumptions that 1) there are no barriers to gene flow (e.g., species are not genetically differentiated throughout their ranges), and 2) all populations share the same climate niche (e.g., no local adaptation). However, we know that most species are locally adapted in response to climate-related selection pressure, and barriers to migration and gene flow are common. We propose an evolutionary modeling strategy to improve predictions of future climate change impacts on species distributions, using the broadly distributed foundation riparian tree Populus angustifolia as a model. We collected leaf samples from 696 trees at 34 sites spanning the full range of P. angustifolia from Arizona to Alberta. Geographic patterns of genetic diversity and structure were assessed using 12 microsatellite loci. How genetic variation is structured across the landscape is valuable for delineating evolutionarily significant units (ESUs) for land management policy and future restoration efforts. We then generated both species level SDM and regional models for each ESU to illustrate 1) differences in predicted suitable habitat, and 2) implications for contact zones among ESUs with and without the inclusion of genetic data.
Genetic analyses revealed that P. angustifolia is differentiated into six genetic clusters across its range. We then tested whether the six genetically distinct P. angustifolia lineages also occupy different climate space and found strong support for unique niche separation by climate (perMANOVA: R2 = 0.78, p = 0.001). Next, we explored how including genetic information in SDM can impact projections. Comparing the full species model (FM) with six regional SDM based on ESUs (RM), we found the traditional full model predicted 40% less current suitable habitat than the six regional models combined. Projecting forward to climate scenarios in 2099, FM predicted 75% gain, whereas RM predicted 20% loss in suitable habitat. Contact zones between different lineages often represent evolutionary hotspots where novel adaptations can arise. These regions are particularly important in light of climate change, where cross-region gene exchange may assist populations in adapting to novel climates. Using the six RM, models exhibit the general trend of progressively decreasing overlap among contact zones from present - 2099. Our findings illustrate how FM and RM can produce drastically different predictions. We suggest evolutionary-based SDM can provide more accurate predictions to help land managers prepare for the impacts of climate change.