Remote sensing of forest fires in boreal ecosystem from space

This paper presents a comprehensive investigation of Canadian boreal forest fires using satellite measurements. Algorithms were developed for detecting active fires (hotspots), burned areas, and smoke plumes using single-day NOAA-AVHRR images and 10-day AVHRR NDVI composites. The algorithms were rigorously validated using conventional fire survey data. The hotspot algorithm identified almost all fire events, but cumulative hotspot area was significantly smaller (similar to 30%) than burned area reported by fire agencies. The hybrid, burn mapping technique provided estimates of Canada-wide burned area that were within 5 percent of official statistics. A neural-network classifier was also developed that allows smoke plumes to be effectively separated from cloud cover at a regional scale.