Toward operational monitoring of forest cover change in California using multitemporal remote sensing data

This paper presents preliminary results of research to improve upon an existing operational forest change detection monitoring strategy in California. Comparisons were made between Landsat 5 TM and Landsat 7 ETM scene normalization techniques (absolute versus, relative). Prior to normalization, scenes containing wildfire smoke plumes were successfully corrected using a space-varying haze equalization algorithm. Simple dark object subtraction provided improved performance over relative (pseudo-invariant feature) approaches. A decision tree classifier produced high change map overall accuracy (86%) for five categories of forest cover change.