Species distribution models and pollen models predict different rates of climate tracking across the last 21,000 years

Background/Question/Methods

Historic shifts in species’ distributions reveal their tolerance, potential for survival, and patterns of adaptation in response to climate changes. Yet, researchers use a variety of models to represent and interpret these shifts. Hindcasted species distribution models (SDMs) determine habitat suitability for a species in response to climate variables. Pollen density models (PDMs) use fossil pollen data from lake cores to estimate spatiotemporal patterns of species abundance around lakes. In this study, we compared the SDMs and PDMs of multiple tree genera in their ability to predict past glacial refugia and rates of migration in North America since the Last Glacial Maximum, 21 Kybp. Specifically, we constructed individual SDMs for each species based on three different paleoclimate models and the Maxent modeling algorithm. Using pollen records from the Neotoma database, we built PDMs and used them to estimate relative abundances at the genus level. We aggregated SDMs to the genus level and compared outputs to the pollen model at ~1000-yr intervals, investigating differences in the predicted rate of range migration, locations of past glacial refugia, and spatial similarity across time periods.

Results/Conclusions

The SDMs for individual species were different from pollen model output at the genus level. For example, SDMs for Fraxinus predicted a large southern refuge for green ash oriented in an east-west direction in the Last Glacial Maximum. Contrarily, our PDM predicted glacial refugia distribution along a north-south axis, aligning with the Mississippi River Valley up to the Great Lakes, with a disjunct portion in the Florida peninsula. Additionally, Fraxinus’ migration rate predicted by the pollen model was nearly 4 times greater, up to 450 m/yr, than the rate of migration of some individual species modeled with SDMs, only up to 175 m/yr. The estimated periods of fastest migration of Fraxinus also differed, with SDM predictions fastest around 13-15 Kybp and around 8-7 Kybp for the PDM. The predictions from the SDMs align with the Bølling-Allerød warming event (~13-15 Kybp), while the later predictions from the PDM may suggest a lag in tree migration. Overall, there is more variability in the migration rate estimates for pollen. While both SDMs and PDMs are used independently to reconstruct species’ biogeographic histories, each data type has its own strengths and weaknesses.