Abstract: The authors studied the regional land cover classification based on MODIS time-series data. The study area is located in Northeastern China, in which there are relative homogeneous land cover types. Savitzky-Golay filter was used to reduce the effect on cloud overlay, loss of data and abnormal data. Classification results proved that NDVI time-series data could be used to differentiate better the woody(perennial) cover from the herbaceous(annual) cover, and vegetation from non-vegetation types depending on the seasonal differences. Grassland and cropland (one crop per-year), needle-leaf deciduous forest and broadleaf deciduous forest had similar phenological characteristics which were easy to be confused. However LST(land surface temperature) data was added to resolve this problem. The overall land cover classification accuracies using NDVI and TVI(temperature-vegetation index) were 62.26% and 71.63% respectively according to the validated results with 363 ground truth survey samples. The results show that the TVI includes more information and is more sensitive to land cover than NDVI, and MODIS data have their own advantages in the regional land cover mapping.