| Tropical floodplain wetlands are important from various perspectives, including hydrology, biogeochemistry and conservation, while facing imminent threats aggravated by insufficient baseline information. Recent advances in remote sensing and image analysis can address this problem. The integration of radar imagery (L-HH) with topographic datasets (elevation, slope and waterbodies) in an object-oriented analysis was tested as a method of rapid wetland mapping in the Equateur Province, Democratic Republic of Congo. Three classification schemes at different aggregation levels were produced to test thematic detail against accuracy. The highest level classifications include upland, lake, river, flooded forest, nonflooded forest, shrub and herbaceous vegetation. The maps range from 47% accuracy for the 7-class wetland map to 78% accuracy for the 3-class floodplain map compared to a reference map (FAO Africover). The method shows promise for developing inventories and monitoring programs to support wetland management in the central Congo River basin and other tropical riverine environments. |
