Author_Institution :
Dept. de Engenharia Florestal, Inst. Superior de Agronomia, Lisbon, Portugal
Abstract :
The only vegetation index (VI) used in all Advanced Very High Resolution Radiometer (AVHRR)-based burned area studies performed so far is the normalized difference vegetation index (NDVI), in spite of serious known deficiencies caused by sensitivity to atmospheric conditions and soil background. In this study, the ability of various VI to discriminate between burned and unburned surfaces is compared, using burn maps derived from classification of Landsat thematic mapper (TM) imagery as reference data. After assessing the discriminatory potential of the AVHRR channels 1-3, the following indexes were compared: the NDVI, the VI 3 (VI3), the global environmental monitoring index (GEMI), and a modified version of GEMI, called GEMI3, VI performance was assessed in two ways: with a signal-to-noise ratio measure and a measure of the commission error resulting from segmentation of the burned area histograms at a fixed level of omission error. Maps of commission errors were generated for each VI and cross tabulated with a land cover map to identify the land cover types more prone to confusion with burns. The newly proposed index GEMI3 performs best, followed by the GEMI, VI3, and finally NDVI, the latter being clearly out performed, even by channel 2 used individually
Keywords :
geophysical techniques; pollution measurement; remote sensing; vegetation mapping; AVHRR; GEMI; IR; Landsat thematic mapper; NDVI; NOAA; burn map; burned surface detection; burnt area; fire; geophysical measurement technique; global environmental monitoring index; infrared; land cover type; multispectral remote sensing; normalized difference vegetation index; optical imaging; pollution; remote sensing; satellite remote sensing; vegetation index; vegetation mapping; visible; Area measurement; Fires; Image analysis; Monitoring; Radiometry; Remote sensing; Satellites; Soil; Spatial resolution; Vegetation mapping;