By coupling leaf with canopy radiation transfer model, a modified vegetation index was explored to extract the vegetation water content in order to monitor the vegetation water content from remote sensing data. Some vegetation indices for the vegetation leaf water content, such as NDVI, WI, NDWI, MSI, CSI and NDII, were analyzed and compared based on the hyperspectral data simulated by PROSAIL. The results show that the water content can not be estimated by NDVI, and CSI is not an ideal tool. The rest four indices WI, NDWI, NDII, and MSI can generate relatively good results with the correlation coefficients square between the estimation and real value being all about 0.76. Then, sensitivity factors were analyzed for NDWI extracting the water content, and the results show that NDWI and MCARI have a good linear relationship. This method can better remove the interference of LAI. The correlation coefficient square between the estimation and real value has been improved to 0.97, which is increased by 27% compared to the foregoing value. The estimated residual is reduced from 0.0156 to 0.00535, a reduction of 65% compared to NDWI method.