Abstract: Leaf water content of vegetation not only affects the growth of vegetation itself, but also is an important factor of water conservation in this area.Especially in the western Sichuan Plateau, the application of remote sensing technology to estimate the water content of vegetation leaves has an important role in the ecological environment and water environment of this area.Based on Sentinel-2 satellite image data, six vegetation indexes(FVC、NDVI、NDVI705、NDWI1、NDWI2、TVI) are adopted in this study.Combined with field measured equivalent water thickness of leaves, we build a multiple stepwise regression model, random forest model and BP neural network model, and the inversion process and results are cross-verified by ten folds.The optimal model is used to obtain the water content distribution of vegetation leaves in the Songpan County test area of the western Sichuan Plateau.The results show that the root mean square error(RMSE) and the Mean Absolute Percentage Error(MAPE) of the BP neural network model is the smallest, and the model accuracy is the highest.The model can effectively invert the water content of vegetation leaves.