Increasing the realism of projected tree species ranges by incorporating migration potential: an eastern US case study

Abstract:

Models of vegetation distributions are used for a wide variety of purposes, from global assessments of biome shifts and biogeochemical feedbacks to local management planning. Dynamic vegetation models, mostly mechanistic in origin, are valuable for regional to global studies but remain limited for more local-scale applications, especially those that require species-specific responses to climate change. Species distribution models (SDMs) are broadly used for such applications, but these too have several outstanding limitations, one of the most prominent being a lack of dispersal and migration. Several hybrid models have recently been developed, but these generally require detailed parameterization of species-level attributes that may not be known.

Here we present an approach to couple migration potential with SDM output for a large number of species in order to more realistically project future range shifts. We focus on 40 tree species in the eastern US of potential management concern, either because of their canopy dominance, ecosystem functions, or potential for utilizing future climates. Future climates were taken from a CMIP5 model ensemble average using RCP 4.5 and 8.5 scenarios. We used Random Forests to characterize current and future environmental suitability, and modeled migration as a negative exponential kernel that is affected by forest fragmentation and the density of current seed sources. We present results in a vulnerability framework relevant for a number of ongoing management activities in the region. We find an overarching pattern of northward and eastward range shifts, with high-elevation and northern species being the most adversely impacted. Because of limitations to migration, many newly suitable areas could not be utilized without active intervention. Only a few areas exhibited consistently favorable conditions that could be utilized by the relevant species, including the central Appalachian foothills and the Florida panhandle. We suggest that a continued effort to include migration potential into vegetation models can lead to more realistic results and management-relevant products.