A major challenge in face of climate change is to maintain biodiversity and nature’s contributions to people (NCP). In order to inform policy formulation, it is crucial to estimate rates of vegetation change and their causes. In this study, we integrate a 20-year record of forest loss with meteorological, fire, and socio-economical data to assess the susceptibility of Environmental Protection Agency level 3 ecoregions to forest loss. Specifically, forest cover data derived from change detection analysis of Landsat imagery (30-m resolution) were modeled using a multivariate hierarchical Bayesian approach that provides simultaneous inference for temporal and spatial dependence among variables. Models were compared based on test-set log likelihood and posterior predictive checks.
Results/Conclusions
In most ecoregions, forest loss has led to fragmentation that is close to the critical point of percolation, meaning that connectivity between forest patches is likely to be lost if reforestation policies are not implemented. Additionally, our models suggest that ecotones are more responsive to climate variability and more prone to fire than patches that are distant from the edges of the forest. These results highlight important challenges for conservation and sustainability of forests across the US.