ABSTRACT In the context of ongoing climate warming, understanding how alpine lake‐basin systems respond to climate variability is crucial for ensuring regional water security and assessing hazard risks. However, lake water level (LWL) dynamics often exhibit pronounced non‐stationarity and scale‐dependent characteristics, and the driving mechanisms constrained by basin hydrological processes remain insufficiently understood. This study used Sarez Lake in the Pamir Plateau as a case study and developed a multiscale attribution framework that combined time‐series decomposition, wavelet coherence analysis, and explainable machine learning based on continuous daily LWL observations from 2005 to 2023. The framework was used to characterise the frequency‐dependent responses of lake levels to climatic forcing and their non‐linear driving mechanisms. The results showed that the LWL in Sarez Lake exhibited pronounced regime shifts and multiscale oscillations. Significant coupling between LWL and climatic factors was primarily concentrated within the range of 294–417 day periods, indicating clear scale‐dependent relationships. Shapley additive explanation (SHAP) evaluation further indicated that runoff and soil moisture were the predominant drivers, with LWL exhibiting non‐linear responses to these variables and robust interactions with snowmelt‐related processes. Overall, the LWL dynamics in Sarez Lake are not controlled by a single hydrological variable but emerge from scale‐dependent non‐linear responses governed jointly by basin water supply and thermal conditions. These findings offer novel insights into the hydrological responses of alpine lake‐basin systems to climate change, thereby contributing to an enhanced understanding of water resource management and hazard assessment in high‐mountain regions.
Kong et al. (Mon,) studied this question.