Resource information
Tropical coastal wetlands form complex and dynamic ecosystems based on a mixture of vegetation, soil, and water components. Optical remotely sensed data have often been used to characterize and monitor these ecosystems, which are among the environments most threatened by climate change and anthropogenic activity worldwide. The present study sought to evaluate the spectral response of Landsat-5 Thematic Mapper (TM) images for the interpretation of different wetlands and associated environments at the mouth of the Amazon River, including mangroves, saltmarshes, beaches, and dunes, as well as secondary vegetation, water with different levels of sediment suspension, and human occupation. A Spectral Angle Mapper (SAM) classifier was applied to the analysis of Landsat-5 TMsatellite imagery to evaluate the potential for the mapping of these coastal wetland land cover classes. The characterization and comparison of the different spectral classes were obtained through the collection of at least 20 polygonal samples (5� � 5 pixels) for each class, with a total of 4,544 points. Spectral separability indices for each pair of classes were based on an Analysis of Variance, with Tukey post-test. The results indicated that most land cover classes could be separated spectrally with Landsat-5 TM. The overall accuracy and Kappa indices for the results of the classification were 86.1 and 0.84� %, respectively. The results of this spectral analysis demonstrated the potential of the SAM classifier for the classification of the different tropical wetlands in a typical Amazon coastal setting from optical remotely sensed data.
