Overcoming spatial resolution limitations: fine-scale monitoring of total phosphorus in small water bodies using airborne hyperspectral remote sensing

Eutrophication, primarily driven by elevated total phosphorus (TP) concentrations, poses a persistent threat to water quality in rapidly urbanizing regions such as the Yangtze River Delta. Although remote sensing has been widely applied for regional water quality monitoring, conventional satellite observations are fundamentally constrained by their spatial resolution, which limits their ability to characterize small and fragmented water bodies that play a critical role in local nutrient dynamics. To overcome these limitations, we present a fine-scale TP monitoring framework for small water bodies based on high-resolution (0.75 m) airborne hyperspectral remote sensing imagery. A deep learning model for water body extraction was developed using the RS-Mamba architecture, which captures global contextual information to effectively identify small and fragmented water bodies. This is followed by a LightGBM regression model to retrieve TP concentrations from hyperspectral reflectance data. Validation against simultaneous in situ measurements yielded a coefficient of determination (R2) of 0.7394 and a root mean square error (RMSE) of 0.0203 mg/L. Additional cross-validation using an independent dataset further confirmed the robustness and generalization capability of the proposed approach. The resulting TP maps reveal that some small water bodies exhibit higher TP concentrations than major lakes and rivers, indicating pollution hotspots that are often overlooked in satellite-based research. These findings demonstrate that integrating high-resolution airborne hyperspectral observations with machine learning techniques enables effective fine-scale monitoring of TP in small water bodies, providing valuable support for refined water environment management and sustainable development in highly urbanized regions.