Optimal Landsat TM band combinations and vegetation indices for discrimination of six grassland types in eastern Kansas

Hyperspectral sensors can make narrow-band measurements for several hundred regions of the electromagnetic spectrum, and with increasing frequency, multi-dates of remotely sensed data are being used for Earth observation purposes. The use of more spectral bands is creating greater demand for larger computer storage capacity and faster data processors. This study evaluates the use of raw Thematic Mapper (TM) band combinations and several derived vegetation indices to determine optimal vegetation indices and band combinations for discriminating among six grassland management practices in eastern Kansas. Results showed that among the transformed dataset, the Greenness Vegetation Index was found to be the best for discriminating among grassland management types. When evaluating the raw TM bands, TM4 (NIR) was always selected in Discriminate Analysis as the best discriminating variable. There is no significant improvement in grassland discrimination by using a combination of the raw TM bands and the vegetation indices. Increasing the number of TM bands by using multiple dates of imagery will improve discrimination accuracy up to a point, but the use of too many bands (greater than 10 or 12) can actually decrease discrimination accuracy.