Abstract: To solve the problem of saturation and plant type dependence of vegetation index composed of two feature bands, this study used analytical spectra devices(ASD) and Sentinel 2 multispectral satellite(S2) data to establish a new leaf equivalent water thickness(LEWT) inversion model by using the improved distance projection ratio index(DPRI2),and then inverted the canopy equivalent water thickness(CEWT).The results show that:(1) Random forest(RF) algorithm of R language can reduce the human factors in the selection of feature bands.(2) The structure of DPRI2 is simple, which can fully reflects the combination of angle, distance and projection between feature bands, and is conducive to improving the inversion accuracy of LEWT and CEWT.(3) DPRI2 cannot be limited by vegetation types and is suitable for CEWT inversion in the study area with multiple vegetation types.Monitoring CEWT by remote sensing is helpful to exploring the regulation mechanism of vegetation in regional hydrological cycle, and to scientifically evaluating regional water resources reserves.