Incluye 45 referencias bibliográficas en las páginas 874-875

The Atlantic rainforest of southern Bahia is one of the last remnants of the lowland forest of eastern Brazil that once covered the entire coastal area from Rio Grande do Norte to Rio Grande do Sul ( lat 8°–28° S) and has been deforested to a small fraction of its original cover (1–12%). All recent vegetation surveys have been based on optical satellite data, which is hampered by cloud cover and by southern Bahia's intricate mix of forest patches with other tree crops, especially cocoa. We describe the application of radar remote-sensing data to distinguish forest patches from cocoa planted in the shade of natural-forest trees. Radar, unlike optical sensors, is not obstructed by cloud cover and can acquire information about forest structure by penetrating into the vegetation canopy. The vegetation map generated from radar data clearly separates forest patches based on the degree of structural disturbance such as the density of shaded trees, the openness of the canopy, and the density of the monodominant Erythyrina shaded trees. The structural classification based on the radar data, and shown on the map, can help researchers assess the degree of fragmentation of the original Atlantic coastal forest and delineate areas of less disturbance with higher potential for conservation of biodiversity. This information can then be applied to conservation planning, especially the design and monitoring of nature reserves and the modeling of biological corridors.