Fine Identification of Lake Water Bodies and Near-Water Land Using Multi-Source Remote Sensing Fusion: A Case Study of Weishan Lake, China

Lakes play a crucial role in maintaining agricultural irrigation water sources, regulating climate, and supporting the long-term resilience of regional ecosystems. However, accurately delineating the boundaries between lakes and land remains challenging due to seasonal hydrological fluctuations, spectral obfuscation with farmland, and the limitations of single-sensor methods. This study constructs a multi-source remote sensing framework integrating Sentinel-1 SAR, Sentinel-2 optical data, DEM, and key environmental variables to identify the water body, near-water body, and non-water surface of Weishan Lake, a major irrigation source in northern China. The study systematically compares various methods, including the optical index method, SAR-based threshold segmentation, and machine learning classifiers. The results show that the random forest model has higher accuracy and temporal robustness. Introducing the “near-water body” category allows for more accurate characterization of transitional areas sensitive to seasonal hydrological and agricultural processes. Migration tests of the model in three external lake systems demonstrate its strong generalization ability, while correlation analysis and SHAP-based analysis indicate that NDVI and elevation are the main factors influencing the spatial pattern of water and land. The proposed framework supports sustainable irrigation management by enabling accurate water boundary monitoring and enhancing the understanding of agricultural hydrological interactions.