Dongting Lake (DTL) and Poyang Lake (PYL) are two Yangtze River-connected floodplain lakes in China. This study investigates the influence of river–lake connectivity on lake water levels and droughts from 1980 to 2023. An explainable machine learning framework, combining Extreme Gradient Boosting (XGBoost) and Shapley Additive Explanations (SHAP) is employed to quantify both the independent and interactive effects of Yangtze River discharge and catchment inflows on lake droughts. Compared with PYL, Yangtze River discharge plays a stronger regulatory effect on the water level of DTL. Autumn was identified as the primary season for drought occurrence because of significant water level declines. Two main mechanisms driving these droughts were identified: the C–/Y– type, caused by both reduced inflows from the catchment and a weakened support effect from the Yangtze River, and the C–/Y+ type, where droughts result from reduced catchment inflows, but the support effect from the Yangtze River is insufficient to offset the water deficit. Consequently, maintaining river–lake connectivity is essential for drought mitigation in these coupled systems, but different strategies are needed. For DTL, the priority should be maintaining the connectivity of the inflow channels from the Yangtze River, whereas for PYL, enhancing the support effect from the Yangtze River is critical for suppressing excessive lake outflow during dry periods. • A comparative analysis was performed for two river-connected lakes with a distinct connectivity. • XGBoost–SHAP model reveals different drought drivers in river-connected lakes. • Yangtze river-dominant events were 40 % more frequent in Dongting lake than in Poyang lake. • Drought periods amplified the influence of connectivity differences on lake-level responses.
Xue et al. (Tue,) studied this question.