Modeling of fire spread in sagebrush steppe using FARSITE: an approach to improving input data and simulation accuracy

Abstract Background Model simulations of wildfire spread and assessments their accuracy are needed for understanding managing altered fire regimes in semiarid regions. The can be evaluated from post hoc comparisons simulated actual perimeters, but this requires information on pre-fire vegetation fuels that is typically not available. We assessed the Fire-Area Simulator (FARSITE) model parameterized with maps behavior fuel models (FBFMs) obtained widely used LANDFIRE, as well alternative means which utilized classification Rangeland Analysis Platform (RAP) satellite-derived cover to create FBFM maps. focused 2015 Soda wildfire, burned 113,000 ha sagebrush steppe western USA, then transferability our RAP-to-FBFM selection process, produced most accurate reconstruction nearby 2016 Cherry Road wildfire. Results Parameterizing FARSITE FBFMs LANDFIRE resulted low levels agreement between observed area burned, maximum Sorensen’s coefficient (SC) Cohen’s kappa ( K ) values 0.38 0.36, respectively. In contrast, derived unsupervised RAP led much greater simulated-to-observed (SC = 0.70, 0.68). map generated greatest was crosswalk another (2016 Road), high 0.80, 0.79). Conclusions Using inform increased compared parameterization standard maps, steppe. Additionally, method appeared have regional generalizability. Flanking backfires were primary source disagreements FARSITE, sources error may require modeling lateral heterogeneity processes at finer scales than here.