Large-scale habitat mapping of the Brazilian Pantanal wetland: A synthetic aperture radar approach

Abstract

This study uses 50m spatial resolution, dual-season, HH and HV L-band ALOS/PALSAR, and HH and HV C-band RADARSAT-2 data, as well as a comprehensive set of ground reference points, to map the diverse habitats of the hydrologically variant subregions of the Pantanal by using a hierarchical object based image analysis approach. First, mean and standard deviation values of image object training sites were evaluated, and used as the basis for forming preliminary land cover class thresholds for each subregion. Then, a combination of additional refined thresholds, hierarchical rules, and a supervised nearest neighbor algorithm (eCognition Feature Space Optimization) employing several features as primary inputs (mean, standard deviation, seasonal change detection, brightness, maximum difference, area, roundness, brightness, compactness, shape index, and length/width) was utilized, resulting in the definition and classification of ten habitat classes: Forest/Woodland, Riparian Forest, Open Wood Savanna, Open Wood Savanna subject to prolonged flooding, Open Grass Savanna, Agriculture, Swampy Grassland, Swampy Mixed Savanna, Vazantes, and Water. This classification was achieved with an overall accuracy of 80% for the entire Pantanal. The produced habitat spatial distribution maps will provide vital information for determining refuge zones for terrestrial species, and connectivity of aquatic habitats during the dry season, as well as providing crucial baseline data to aid in monitoring changes in the region, and to help define conservation strategies for habitat in this wetland.