Abstract: Utilizing satellite remote sensing technology for the retrieval of water depth is a crucial method for quickly acquiring water depth information over large areas. However, traditional optical satellite images often suffer from low inversion accuracy due to various factors, including the presence of sediments, suspended particles, and chlorophyll in the water. These constituents alter the color and turbidity of the water, making it challenging to accurately estimate water depth through remote sensing techniques. To mitigate the influence of these water color materials on the accuracy of water depth inversion, a novel approach was proposed, centered around a mathematical model and the use of Sentinel-2 remote sensing imagery. This approach involves analyzing the mathematical relationships between water depth and the surface area of different types of water channels. This method deduces the river surface depth model by generalizing and classifying the river section.It combines the Sentinel-2 remote sensing images to determine the relevant parameters in the mathematical model, and derives an expression that can directly calculate the water depth from the water surface area, which can minimize the influence of water color on water depth retrieval from remote sensing images. To validate the effectiveness of this approach, it was applied to the Xiaoqing River, and the results were compared with ground-truth data. The average absolute error in water depth inversion at two hydrographic stations, Huangtaiqiao and Shicun, was found to be 0.03 meters and 0.23 meters, respectively. These results demonstrate a significant improvement in the accuracy of water depth estimation. The method proves to be more reliable in areas with varying water color and turbidity. It can quickly obtain water body depth information based on satellite remote sensing technology and has promotion and application value.