There are many problems for the current water resource estimation methods. For the traditional water resources assessment method, the runoff reduction cannot be accurately estimated, because the unmeasured terms of runoff reduction are either neglected or statistically estimated according to the local population and water use ration. Also, the hydrogeological parameters used to calculate the water replenishment to aquifer are empirical. Furthermore, the traditional water resources assessment method is labor consuming and fund consuming. For the hydrological model, model processes are complicated, and parameters should be calibrated with a lot of observation data. In order to solve the above-mentioned problems, this paper proposes an innovative water resources estimation model based on remote sensing evapotranspiration (ET). There, the water resources are defined as the difference between precipitation and uncontrollable ET. The uncontrollable ET is the natural ET from natural surface and doesn't include the ET from water consumption of humans. By combining the remote sensing ET and land use according to the natural characteristic of land uses, the uncontrollable ET is obtained. While the cropland uncontrollable ET is obtained by extracting the soil evaporation using satellite based Normalized Difference Vegetation Index (NDVI). A case study result shows that calculated water resources are relatively high correlated with that of obtaining from water resources reports. And the deviation of the calculated water resources can be eliminated by improving the accuracy of the precipitation interpolation and land use classification.