Before its decline in the wake of agricultural conversion, timber harvesting, thenaval stores production industry, and fire suppression, longleaf pine (Pinus palustris) stands were a prominent feature of the natural ecosystem of the southeastern U.S. In lightof this, longleaf pine has become a strong candidate species for restoration projects.However, previous studies have mainly focused on restoration within the historic rangeof the longleaf pine ecosystem. To evaluate persistence of longleaf pine into the future,this study focused on the potential impacts of climate change on longleaf pinedistribution. In order to help longleaf pine restoration projects to be successful, it iscritical to estimate the potential change in suitable sites for longleaf pine over the courseof the next century. In order to model potential suitable site for longleaf pine, twoapproaches were compared: the Maximum Entropy approach (MaxEnt) and the RandomForests (RF) approach. Changes in suitable habitat were evaluated for three time points(2020, 2050, and 2080) using two general circulation models (Canadian Center forClimate Modeling and Analysis Coupled Model Version 3 (CGCM3) and the HadleyCenter Coupled Model Version 3 (HadCM3)) with two emission scenarios (medium tohigh emissions and low to medium emissions) and two thresholds for determining therange considered "suitable" given model output (relatively liberal and strict). Thus, eachapproach has eight scenarios, for a total of 16 scenarios in this study.Most of the scenarios estimated that habitats suitable to longleaf pine willdecrease by 2080, and that the maximum latitude of the area designated as suitablehabitat would move farther north than that of current longleaf pine distribution. Eachmodel produced different predictions and, while the degree of difference was significantin terms of the size of suitable habitat, the difference was slight for northward shift.General circulation models and emission scenarios did not result in large differencesbetween the 16 model predictions regarding the size of the future suitable habitat and thelocation of suitable habitat. However, the approach (MaxEnt or Random Forests) andthresholds (the value for determining suitable and unsuitable habitat) produced significantdifferences among the various model predictions.
