Land cover classification in Ibaraki-prefecture area using Landsat-5-TM data

In order to extract some parameter data for atmosphere-soil-vegetation simulation model from Landsat-5-TM data, as a first step, the land cover classification was carried out by supervised maximum likelihood method for Landsat-5-TM data. To this method, training area selection is critical for the accuracy of classification. Since histograms even at same class sampling points have some discrepancies, we select many training areas for each class. Also the selection of training area was determined to approach classified land cover area to statistics of agriculture and forestry. The 85% of classification from the Landsat-5-TM data in 1987 agreed with the measurements that are basic categories classified by field surveys. (author)