Integrating DEM and machine learning to estimate monthly water level variability on the Qiangtang Plateau from 2000 to 2021

To improve the accuracy and consistency of QP lake water level monitoring, we account for key parameters that affect DEM-derived water level errors, including the accuracy of lake boundary extraction, DEM quality, and the reliability of the area-elevation relationship. XGBoost model is employed to correct these discrepancies, leveraging altimetry-derived water level as a reference for calibration, thereby facilitating more accurate water level estimations. Based on the corrected results, we reconstruct monthly lake water level for QP lakes from 2000 to 2021, revealing the spatiotemporal dynamics of water level.为提升QP湖泊水位监测的精度与一致性，本研究针对影响基于数字高程模型（DEM）反演水位误差的关键参数展开分析，涵盖湖泊边界提取精度、DEM质量以及面积-高程关系的可靠性。本研究采用极限梯度提升（XGBoost）模型校正此类误差，以卫星测高反演水位作为校准基准，进而实现更为精准的水位估算。基于校正后的结果，本研究重建了2000至2021年间QP湖泊的逐月水位序列，揭示了水位的时空动态变化规律。