Spatial Filtering of Radar Data (RADARSAT) for Wetlands (Brackish Marshes) Classification

One of the main wetland areas for migratory bird breeding in Mexico is Laguna Madre, at State of Tamaulipas. It is imperative to have an adequate mapping of these ecosystems for management and protection. The objective of this work is to determine the most effective spatial filter treatment applied to a two-angle RADARSAT data set to reduce speckle effects and improve marsh classification performance of Tamaulipas wetlands. Eighteen spatial filter treatments were applied to the data based on four algorithms: Maximum a Posteriori, Lee-Sigma, Local Region, and Median. A supervised classification was applied to the filtered images with five resulting general classes. Overall classification performances were compared to determine statistical differences between filter treatments. An edge/boundary retention analysis and a visual assessment study were also applied to the data. It was determined that the Lee-Sigma algorithm with three iterations and square window sizes of three, five, and seven pixels, respectively, produced the best overall results with the 8-m pixel RADARSAT data.