Land Cover Classification Based on the MODIS-EVI Time-Series Using Decision Tree Method

摘要

The MODIS data has high temporal resolution but rather coarse spatial resolution, therefore, the MODIS-EVI data, which was more sensitive towards the phonological information than the MODIS-NDVI data, was chosen to build up the time series of studying area, in order to monitor and depict the original phonological characteristics of land cover. Moreover, the DEM, SLOPE, Homogeneity data, which all represent the differences of geographical distribution, and LST data, which represents the differences of earth-atmosphere interaction, were combined as ancillary data together with MODIS-EVI to build a decision tree. After the classification validation, the overall accuracy attainted to 71.9% and Kappa coefficient is 0.66. Therefore, it is proved that the land cover classification with high accuracy but low cost in regional scale is possible.