Region-Specific Remote-Sensing Models for Predicting Burn Severity, Basal Area Change, and Canopy Cover Change following Fire in the Southwestern United States

Abstract

Estimates of burn severity and forest change following wildfire are used to determine changes in forest cover, fuels, carbon stocks, soils, wildlife habitat, and to evaluate fuel and fire management strategies and effectiveness. However, current remote-sensing models for assessing burn severity and forest change in the U.S. are generally based on data collected from California, USA, forests and may not be suitable in other forested ecoregions. To address this problem, we collected field data from 21 wildfires in the American Southwest and developed region-specific models for assessing post-wildfire burn severity and forest change from remotely sensed imagery. We created indices (delta normalized burn ratio (dNBR), relative delta normalized burn ratio (RdNBR), and the relative burn ratio (RBR)) from Landsat and Sentinel-2 satellite imagery using pre- and post-fire image pairs. Burn severity models built from southwest U.S. data had clear advantages compared to the current California-based models. Canopy cover and basal area change models built from southwest U.S. data performed better as continuous predictors but not as categorical predictors.