Satellite Images (CBERS-2B and SPOT-5) Processing used to Map Mangrove Forests and the Coastal Landscape in the Sao Francisco River Estuary, Northeast Brazil

摘要

The increasing mangrove loss requires mapping studies through cost-effective remote sensing tools. This paper evaluates different image processing techniques using satellite images from SPOT 5 HRG2 and free images from CBERS 2B CCD1 in a free GIS to map mangroves and landscape units in the Sao Francisco River Estuary (Brazilian Northeast). False colour composites, principal components analysis (PC), IHS transformation to merge the images, supervised classification, visual interpretation and ground-truth were accomplished, and a land use and cover map produced. Our results show that the use of data from PC analysis increased the classification performance from 83.57percent (CBERS images) to 91.77percent (merged images and PC) and 96.59percent (CBERS images and PC). A false colour composite of PC and merged data with 2.5 m spatial resolution was the best result for visual interpretation. The map reveals that 78.1percent of the study area is occupied by coastal plain (153 km~(2)) whereas mangroves occupy 14.9percent (30 km~(2)), but are fragmented by aquaculture (4.5 km~(2) - 2.2percent) and agriculture (9 km~(2) - 4.4percent). We concluded that the integration of CBERS and SPOT images and the application of important image processing techniques were more effective in discriminating the study area landscape units than the information provided by individual images.