Improving thematic mapper based classification of wildfire induced vegetation mortality

Abstract

In many areas suppression of wildfire has produced fuel accumulations that pre‐dispose forests to undesirable fire behavior. Image processing techniques can be used to combine different elements of terrain data into a single composite image. This composite terrain image is used to improve the accuracy of a supervised classification of expected vegetation mortality in a 20, 000 acre forest fire in the Cibola National Forest in New Mexico. Error matrices are produced that indicate that combining TM and terrain data provides a 40% improvement in accuracy compared to TM data alone. Computer‐assisted mapping of observed and potential patterns of wildfire can provide forest managers cost‐effective tools for wildfire planning and ecosystem management.