Do genetically informed distribution models improve range predictions in past climates? A case study with balsam poplar

Species distribution models (SDMs) are one of the most widely used approaches to predict changes in habitat suitability in response to climate change. However, as typically implemented, SDMs treat species as genetically uniform throughout their ranges and thereby ignore potentially important genetic differences between populations. While numerous studies have used SDMs to model genetically based subgroupings within species, the ability of such models to be transferred to new times has rarely been evaluated. Here, we used standard and genetically informed distribution models (gSDMs) to predict the future and past range of balsam poplar ( Populus balsamifera L.). We then assessed model transferability of standard SDMs and gSDMs using balsam poplar fossil pollen and macrofossil occurrences. In general, standard and gSDMs performed similarly through time, with both predicting a northward expanding range from refugia as glaciers receded over the past 22 ky BP and declining suitable area in future climates. Both standard and gSDMs showed moderate abilities to distinguish balsam poplar fossils from pseudo-absences but tended to predict lower suitability at fossil sites during the Pleistocene-Holocene transition. Although gSDMs applied to balsam poplar did not prove more transferable than standard SDMs, they provided numerous unique insights, such as the change in suitable area of genetic clusters through time and potential refugial locations. We argue more research is needed to determine which species may benefit most from the gSDM approach and to test gSDMs with temporally or spatially independent occurrences, as is often recommended for standard SDMs.