Multi-temporal Landsat TM data were utilized to classify the cultivated land located in the middle area of the Nagara River basin, Gifu prefecture. Color images of the vegetation indices including DVI, NDVI, SAVI and GVI were composed from respective multi-temporal VI images obtained from original Landsat TM data. The newly composed images clearly demonstrated the cultivated land such as paddy field, upland field and orchard. The distribution of the cultivated land related to the total river basin was also easily discriminated. However, by comparing the change profiles of the vegetation indices, SAVI was confirmed to be the most suitable index to reflect the vegetative conditions of major land-cover categories at different times, and the composed color image based on this index was thus applied for further classification analysis. As compared to the original TM images, classification results based on the composed SAVI image agreed much well with the related census data, with the classification accuracy for the paddy field and the upland field being especially high. The corresponding classification errors for these two land-cover categories were generally below 10%.
View full abstract