Examining the potential of using remotely sensed fire data to predict areas of rapid forest change in South America

Abstract

Major advances have been made in the last decades estimating deforestation by using satellite data. This has been mostly done by detecting differences in a time series of high resolution data. However, the interpretation of the images and the detection of change is cumbersome and expensive. Furthermore, the area covered by a high resolution image is relatively small and only available on a monthly bases. This work investigates if detected fire of the low resolution (AVHRR) sensor can be used as a quick and inexpensive way of locating areas of rapid forest cover change. It tests the hypothesis that vegetation fires, as detected by earth orbiting satellites, are a reliable indicator of “hot spots” of deforestation in tropical South America. Fine spatial resolution satellite data were used to measure the deforestation that occurred in the mid 1990s in 41 sites spread across the tropical forests of South America. These data were then compared to the number of fires occurring in these sites, as detected by the NOAA-AVHRR satellite between April 1992 and December 1993. At the individual site level, 25 of the test sites showed a significant statistical relationship between the fires and the deforestation occurring in the following years. A further 6 sites showed neither fire activity nor significant deforestation. When aggregated to the regional level, only the set of sites lying in relatively dry areas with the same land use practices (located in Brazil) showed a significant correlation between fire and deforestation. This correlation enabled us to produce a map of high deforestation risk occurring in 1992 in the Brazilian Amazon, which showed close correspondence to the actual deforestation fronts of the year 1992.