Assessment of VIIRS 375m active fire detection product for direct burned area mapping

a r t i c l e i n f o The use of active fire detections for direct burned area mapping has been limited by the coarse spatial resolution and long revisit cycles of previous sensors. However, the recently developed VIIRS 375 m active fire detection product offers enhanced spatial resolution and temporal revisit cycle, and high sensitivity to small active fires. In this study, we apply that product and assess its performance for direct burned area estimation. Consecutive VIIRS 375 m active fire detections were aggregated to produce the VIIRS 375 m burned area (BA) estimation over ten ecologically diverse study areas. The accuracy of the burned area estimations was assessed by comparison with Landsat-8 supervised burned area classification. The performance of the VIIRS 375 m BA estimates was dependent on the ecosystem characteristics and fire behavior. Higher accuracy was observed in forested areas characterized by large long-duration fires, while grasslands, savannas and agricultural areas showed the highest omission and commission errors. VIIRS 375 m data proved to have higher sensitivity to small fires and higher fire detection rates compared to previous studies, indicating good potential as a complementary dataset for use in routine burned area mapping systems based on cumulative fire count or seeding of spectral signature approaches. Results from the application of VIIRS 375 m fire data to directly monitor the growth of long-lasting wildfires using the 2013 Rim Fire example in California showed high agreement with available fine resolution airborne data, therefore serving as an alternative source of information for near real-time mapping of fire-affected areas. Every year, more than 300 Mha of land burn globally, causing significant ecological and economic consequences, and associated climatolog-ical effects as a result of fire emissions (Giglio, Randerson, & Werf, 2013; van der Werf et al., 2010). In recent decades, burned area estimates generated from satellite data have provided systematic global information for ecological analysis of fire impacts, climate and carbon cycle models, and fire regime studies, among many others A variety of sensors and techniques have been employed for burned area mapping. These techniques are traditionally grouped into three types of approaches: (i) burned area estimation using active fire counts (ii) image classification using the spectral properties of burned residues Currently, active fire detection products are not considered an adequate data source to directly map burned areas due to limitations related to their coarse spatial resolution, high omission …