New Mexico Geological Society Annual Spring Meeting — Abstracts

In recent years, fires in western US have increased in size and severity. The impact of wildfires on vegetation recovery, ecosystem services, erosion, and debris flow generation can depend on the severity of the burn. Post-fire debris flows threaten lives, infrastructure, and property at the Wildland Urban Interface, and their likelihood depends critically on the severity of the burn. Current hazard models rely on the differenced normalized burn ratio (dNBR) to characterize burn severity, yet pre-fire estimates of dNBR remain limited in their ability to capture landscape-scale uncertainty. We present a Bayesian statistical model that estimates the Weibull probability distribution of dNBR across 126 vegetation types that have burned throughout the Intermountain West over the last 13 years. We analyze forested and non-forested vegetation regimes that frequently experience fire to reveal patterns across physiognomy. Unlike deterministic approaches, our model explicitly incorporates uncertainty in burn severity predictions. Distribution parameters for each vegetation type are conditioned on vegetation density, annual climatic aridity, and topographic predictors selected through systematic model comparison. We find that the topographic controls on dNBR are vegetation-type dependent: aspect-derived solar radiation indices are the dominant topographic predictor for forested and grassland vegetation types, while the standard deviation of landscape elevation (a proxy for slope and ruggedness) is a powerful predictor over numerous vegetation types. Our model successfully reproduces the observed statistical distributions of dNBR across fire-prone landscapes of the Intermountain West, offering a probabilistic framework for pre-fire burn severity prediction that can be directly integrated into hazards mitigation planning.