Extraction of damaged areas of windfall trees by typhoons using Landsat TM data

Two typhoons attacked, at an interval of about 2 weeks, the north part of Kyushu Island, Japan, in September 1991. Many trees, especially in artificial forests (main tree species are cedar and cypress) were felled by the typhoons. Landsat TM data taken before and after damage were collected and registered in order to extract the damaged areas. Typical damaged points were selected on the registered images, referring to aerial photographs taken immediately after damage, and the change characteristics of TM bands 1-5, 7 and Normalized Difference Vegetation Index (NDVI) due to the damage were examined. It showed that bands 5 and 7 of the middle-infrared increased more than other bands and that NDVI decreased. Bands 5, 7 and NDVI of each temporal TM data were merged into a single six-band dataset, and the damaged areas were extracted by a maximum likelihood classification using the merged dataset. The damaged areas extracted were evaluated using the aerial photographs. The damaged areas of windfall trees could be extracted with an accuracy of 90% using temporal Landsat TM data acquired before and after damage.