A contribution for a global nurned area map

Abstract

The goal of this work was to develop methodologies for burned area mapping at 1 km resolution using SPOT-VEGETATION (VGT) images from tropical (Southeastern Africa and Brazil), temperate (Iberian Peninsula) and boreal (Eastern Siberia / Northeastern China) regions. For each study area seven months of daily images were used in order to map the areas burned during the entire fire season. Linear discriminant analysis or classification trees were applied, depending on the study area, to monthly composite images derived from the daily images, and monthly burned area maps were produced. The final VGT 1 km burned area maps were validated with burned area maps derived from 30 m Landsat imagery, using linear regression. Twenty-four Landsat scenes were used in the validation of the maps produced for the four study areas. The accuracy of the VGT maps was variable, dependent on vegetation type and on the spatial pattern of the burned areas.