Water conservation is an important part of the ecosystem service function. As a complex ecosystem, forests have different contributions to water conservation. The forest litter plays an important role in the water conservation function, for it directly covers the ground surface. Remote sensing and hyperspectral technology provide a solution for the long-distance detection of water conservation in planar areas. Especially in plateau areas, remote sensing is the most effective way to obtain surface information. In this paper, the hyperspectral data of main tree species were measured by ASD spectrometer in the Grand Canyon research area of Yarlung Zangbo. The litter samples were obtained by sampling on the plot and the water holding capacity of the samples was calculated. Use leaf spectrum information to construct vegetation index related to litter water conservation capacity, then establish multiple regression model of vegetation index and effective retention capacity, and invert the water conservation capacity distribution of main tree species in the Grand Canyon based on Sentinel-2 images. In the end, the accuracy of the inversion model is evaluated based on the verification points. The results showed that: (1) The reflectance of Quercusaqui folioides was the highest, the lowest was the Picea Linzhi, and the total reflectance trend of the three species was similar; (2) The effective interception amount of litter is sorted from large to small as follows: Picea Linzhi (48.36 t·ha^-1)>Quercusaqui folioides (39.24 t·ha^-1)>Pinus densata (32.32 t·ha^-1). Picea Linzhi leaves are easy to decompose and store, Quercusaqui folioides’ leathery leaves and Pinus densata’s oily leaves are not conducive to decomposition. Therefore, the litter of Picea Linzhi has the highest effective interception amount. (3) Through the Person correlation coefficient analysis and the multiple linear regression model, it is found that the higher the wax parameters and attenuation degree of leaves, the weaker the water conservation capacity of litter; the better the vegetation growth trend, the higher the pigment and the water content of leaves, the stronger the water conservation capacity. (4) The results of the accuracy evaluation of the inversion model of the water conservation capacity of litter are good. R² of the test points of Pinus densata, Picea Linzhi and Quercusaqui folioides is 0.943, 0.815 and 0.812, and RMSE is 1.597, 2.270 and 1.953. It shows that the model can be used for the prediction and distribution of the water conservation capacity of forest litter in Grand Canyon.