Identification of rainfall-activated multi-indicator pollution pulses and lagged responses in Erhai Lake based on continuous water quality monitoring

Rainfall runoff can activate non-point source pollution and generate short-term water-quality anomalies in plateau lake basins, but event-scale multi-indicator responses remain difficult to identify using conventional mean-based assessments. Here, water-quality records available between June 2021 and December 2024 from the Lake Center, Jiangwei, Xiaoguanyi, and Jinhe monitoring sections of Erhai Lake were combined with daily precipitation data to identify 127 rainfall events. An event-scale framework was developed to quantify post-event responses, classify single-indicator and compound pollution pulses, assess lagged rainfall correlations, and interpret potential drivers using random forest models. Total phosphorus, ammonium nitrogen, and turbidity showed relatively high rainfall-associated pulse proportions, while synchronous total phosphorus–turbidity anomalies were strongly associated with rainfall and were consistent with potential particulate phosphorus mobilization. Cumulative rainfall over 15–30 days was especially relevant to total phosphorus, turbidity, chlorophyll-a, and algal density. Temporally blocked validation showed limited cross-year transferability of several random forest models, particularly for ammonium nitrogen, turbidity, and chlorophyll-a. The framework is therefore intended for event-scale screening of rainfall-associated water-quality risks and requires local recalibration before operational forecasting or application to other lake basins.