US-IALE 2018

AUTHORS: Xin Chen, Laura Leites – Department of Ecosystem Science and Management, The Pennsylvania State University

ABSTRACT: Under rapid climate and land-use change at broad scales, forest management and conservation requires a better understanding on how tree species distributions respond to those changes across landscapes. In this context, species distribution models (SDMs) have been developed to identify the associations between species occurrence and abiotic factors. However, SDMs rarely consider effects of land-use legacies (LUL) on species distributions. Historical land use plays a key role in shaping tree species distributions by altering soil environment, ecosystem patterns, and ecological processes. SDMs that fail to account for LUL influence on species distributions are likely to yield biased projections of species distribution under future climate. Here, we address this shortcoming by incorporating LUL into modeling presence or absence of forest tree species in Pennsylvania. Using 3,336 USDA Forest Service’s Forest Inventory and Analysis (FIA) plots, we fitted the Random Forests models to predicting plot-level presence or absence for six forest tree species based on LUL related predictors in addition to climatic, topographic, and soil variables. The former included forest establishment year derived from FIA data, cropland area, and pasture percent area before establishment year at 5’ longitude/latitude grid resolution extracted from the History Database of the Global Environment. Results indicate that the forest establishment year and cropland area before establishment year had relative higher importance compared to other predictors while pasture percent area before establishment year was not significant in determining the performance of the SDMs. Additionally, accounting for LUL influence on tree species distributions lowered the misclassification error rates for absence of tree species, which suggests that, in addition to unsuitable abiotic conditions, the unoccupied areas of tree species can also result from unfavorable historical land use or the interaction between those two factors. The results highlight the importance in considering land-use legacies when predicting species distributions across landscapes.