Impacts of spatial scale and resolution on species distribution models of American chestnut (Castanea dentata) in Pennsylvania, USA

Abstract

American chestnut (Castanea dentata) was once a foundational tree species of eastern North American hardwood forests before its functional extinction due to an invasive chestnut blight first described at the beginning of the 20th century. Ongoing research to develop a blight-resistant American chestnut through traditional breeding or transgenic means is promising. However, restoration of American chestnut will require a thorough understanding of its spatial distribution and habitat suitability to both find surviving mature trees to enhance the genetic diversity of experimental blight-resistant trees and to identify optimal restoration sites. While many studies have modeled the spatial distribution of American chestnut trees, few have addressed how adjusting spatial scale and resolution impact the predictive ability of species distribution models (SDMs) for the tree species. In this study, we used Maxent to model American chestnut distribution at coarse resolution (238 m) across the state of Pennsylvania and fine resolution (3 m) within French Creek State Park in the southeastern portion of the state. Pennsylvania was chosen as our study area due to its central location in American chestnut’s pre-blight range. Results indicate that the spatial extent of the study area and ranges of environmental variables included in the models influence which variables are most predictive of chestnut distribution. This study suggests that environmental variables such as soil content, soil pH, and elevation are all strong predictors of American chestnut habitat, but their importance varies by resolution. This study also investigates how spatial scale and resolution influence the predictive ability of American chestnut SDMs and offers a workflow for locating restoration sites by first modeling species distribution at a coarse scale and broad extent to identify patches of suitable habitat, then focusing resources on fine scale modeling of species distribution local to those suitable patches.