—Mark Parr, U.S. Fish and Wildlife Service
A common method in image classification is the practice of removing pixels the analyst wishes to exclude from analysis through the use of a mask. This method is used to focus the subsequent analysis on pixels, or ground features, that are of interest to the analyst and to reduce the variability of the remaining pixels, thus (hopefully) increasing the accuracy of the classification. The Gulf Coast Joint Venture developed a permanent water mask using National Wetlands Inventory data, the 2001 National Landcover Dataset, and National Agricultural Inventory Program aerial photography. A combination of ArcGIS and ERDAS IMAGINE was used in the analysis. The design, development, and accuracy assessment is described along with examples of how this mask will be used to estimate bird habitat along the western US Gulf Coast.