An assessment of several linear change detection techniques for mapping forest mortality using multitemporal landsat TM data

Forest canopy changes can be detected by a variety of methods of analysis of multitemporal satellite images. Issues surrounding the use of remote sensing in operational forest monitoring include which change detection method is most appropriate, and to what extent scenes should be preprocessed for the minimization of irrelevant interdate differences. Results indicate better performance for principal component analysis and a multitemporal Kauth-Thomas transformation as compared to the Gramm-Schmidt orthogonalization process. There is little evidence to suggest that preprocessing beyond simple DN matching methods improves results. Relationships between change components and mortality are found to be specific to the particular image data being used and the particular forest type under study. Canopy change can be detected reliably, but precise estimates of mortality levels require calibration using field data in all new situations. Change in Kauth-Thomas wetness is the most reliable single indicator of forest change.