Abstract: To set up NDVI remote sensing estimation model of different grassland types in Tibet, a regression estimation model was established by using use of multivariate linear regression, which was between fresh grass biomass and relative variables such as SPOT/VEGETATION mean annual maximum normalized difference vegetation index (NDVI), annual rainfall and annual accumulated temperature.And it analyzed the correlated the relationship between the average amount of fresh grass biomass and the average annual maximum NDVI, mean annual rainfall and other factors for all grassland types.The results showed that:Rainfall was the most important affecting factor to fresh grass growing.The multivariate linear regression model of fresh biomass based on NDVI fitted the grass (R=0.964), alpine meadow (R=0.959), alpine desert steppe (R=0.772), temperate grasslands (R=0.892) and alpine grasslands (R=0.797) well.