Remote Sensing Techniques to the Detection and Mapping of Flooding Dynamicswithin the Pantanal, Mato Grosso Do Sul State, Brazil: Preliminary Results

摘要

The Pantanal, an extensive floodplain region in Central Western Brazil, isannually flooded. Thematic Mapper scenes (bands 3, 4, and 5) from the dry and flood seasons, were analyzed. The objective is to evaluate the performance of classification algorithms (Maximum Likelihood, K-means, Parallelepiped, Slicing) in order to discriminate four thematic classes (water, flooded areas, wet areas, and dry areas) of interest. An analysis from the results of these classifications is made. The classifier on Maximum Likelihood (MAXVER) showed that in the scene of the dry season, the results were the most consistent, followed by the Density Slicing Classifier at band 4. Recommendations are made for further works on flooding dynamics within this huge ecosystem.