Objective To select an optimum time series in remote sensing data for evaluating land cover changes and their impacts on southern China.

Method Three types of remote sensing NDVI data from AVHRR, SPOT-VGT and MODIS were analysed to compare their similarities and differences in different vegetation types over Hanjiang River basin from 2001 to 2006, and their correlations were analysed using linear regression method.

Result and conclusion These data had generally similar distribution patterns in space. MODIS and SPOT-VGT NDVI matched well. MODIS could recognize objects more clearly on the earth surface due to its spectrally narrow sensors with high spatial resolution. In MODIS NDVI, values varied within a wide range, therefore more vegetation types could be detected. Three types of NDVI changed seasonally similar to the same amplitude. MODIS NDVI reflected vegetations seasonally change more accurately. In AVHRR NDVI, some vegetation types change did not match with the actual situation. Although different vegetation types seasonally changed in the same way in three types of NDVI, changes in a more temporally identical pace could be found in MODIS NDVI and SPOT-VGT NDVI compared with AVHRR NDVI. MODIS NDVI could distinguish seasonal changes of croplands, closed shrublands and savannas from other vegetation types. It showed that MODIS NDVI had a higher resolution on reflecting vegetation types and their development compared to SPOT-VGT NDVI and AVHRR NDVI. Throughout the whole basin and on a variety of vegetation types, there were some linear relationships among three NDVIs, while the linear relationship between MODIS NDVI and SPOT-VGT NDVI was the strongest. Based on the linear regression of MODIS NDVI and AVHRR NDVI, MODIS NDVI was estimated accurately consistent with the measured values of NDVI of Hanjiang River basin in 2000. MODIS NDVI time series can provide developing ideas of the historical years.