A VIIRS direct broadcast algorithm for rapid response mapping of wildfire burned area in the western United States

Abstract

We present a direct broadcast (DB) rapid response burned area mapping algorithm for Visible Infrared Imaging Radiometer Suite (VIIRS) data that combines products driven by the spectral signal of fire-affected areas from both emissive and reflective spectral bands. The algorithm processes VIIRS infrared M-bands (750 m) using spectral ratios of the top of atmosphere reflectance from a single satellite scene to identify pixels exhibiting surface properties consistent with burn scars. Next, this collection of candidate burn scar pixels is screened using a contextual filter based on VIIRS I-band (375 m) active fire detections (AFD) which removes erroneously classified pixels and provides burn scar detections (BSD). The AFD and BSD are then resampled to a 375 m grid and reported jointly as VIIRS burned area (VBA). The accuracy of the VBA was assessed for 390 wildfires (11–114,500 ha in size) in the western United States. The spatial accuracy was assessed by comparison with a validation dataset of Monitoring Trends in Burn Severity (MTBS) burned area and incident fire perimeter polygons. The VBA temporal accuracy was evaluated using a time series of daily fire perimeter polygons derived from high resolution airborne infrared imagery. The algorithm's burned area mapping accuracy is 59%. The algorithm detected 60% of burned area on the initial day of burning and 73% within 24 h.