The objective of this study was to present the improvement of land cover mapping of Sanjiang Plain using MODIS data.
A filter method based on time series was applied to remove EVI noise. A set of variables that were derived from MODIS
data and DEM consisted of:(i)phenological feature variables derived from EVI temporal profile;(ii)geographical
environment variables;(iii)spectral feature variable obtained from MODIS surface reflectance data;(iv)several feature
band variables. Then the land cover map was generated using a simple but reasonable decision tree. In addition, our
classification result was compared with land use map derived from Landsat ETM+ images using a confusion matrix, as
well as a comparison with MODIS-IGBP land cover product was discussed. The comparison demonstrated the good
behavior of the multivariable approach and technical processing used in the research, as well as indicated the promise of
MODIS data for land cover mapping at regional scale.