Use of Multi-Sensor Data to Identify and Map Tropical Coastal Wetlands in the Amazon of Northern Brazil

Abstract

Recognizing and mapping wetlands in the Amazon coastal from optical and synthetic aperture radar (SAR) data is critical to understand coastal evolution. Multispectral optical images are obtained only during the dry season, while SAR images can be acquired throughout the year, but present low spectral resolution. The aim of this paper was to investigate the use of remote sensing images, which allowed the accurate identification and mapping of coastal environments based on the complementary information provided by the synergism of multisensory data and supervised classifications with statistical validation from the overall accuracy and Kappa index. The mapping of these environments was based on the supervised classification of Landsat ETM+ images and ETM+-SAR product, which permitted the identification of eight classes: coastal plateau, mangroves, floodplain + freshwater marshes, tidal sandflats + sandbars + ebb-tidal delta, macrotidal beaches, water, frontal dunes + paleodunes + interdunes + mobile dunes, and salt marshes. Overall accuracy and Kappa indices for the classification maps of the wetland study area were 93.3% and 0.909 for Landsat ETM+ and 93.8% and 0.917 for the ETM+-SAR product. This indicates that the integrated product provides additional information, which permits the more efficient identification and mapping of tropical coastal wetlands.