Comparative analyses of East Texas forest cover maps generated from Landsat and AVHRR data

Abstract

Comparing satellite data derived map products are affected by differences in data characteristics, image acquisition dates, processing techniques, and classification schemes used for assigning pixels to a thematic class. By comparing two forest maps generated from Landsat Enhanced Thematic Mapper Plus (ETM+) and Advanced Very High Resolution Radiometer (AVHRR) images acquired on the same day, and processed using identical classification scheme and methods these differences were minimized. The ETM+ derived map had higher classification accuracy values and more precise area estimates than the AVHRR derived map. In the ETM+ derived map, 87 of the 599 verification data were misclassified, whereas in the AVHRR derived map, 155 of the 469 verification data were misclassified. Detailed error analyses by land cover class revealed that a land use based definition of forest accounted for 74% (64 out of 87) and 57% (89 out of 155) of the classification errors in ETM+ and AVHRR derived maps, respectively.