Water quality in surface bodies remains a pressing issue worldwide. While some regions have rich water quality data, less attention is given to areas that lack sufficient data. Therefore, it is crucial to explore novel ways of managing source-oriented surface water pollution in scenarios with infrequent data collection such as weekly or monthly. Here we showed sparse-dataset-based prediction of water pollution using machine learning. We investigated the efficacy of a traditional Recurrent Neural Network alongside three Long Short-Term Memory (LSTM) models, integrated with the Load Estimator (LOADEST). The research was conducted at a river-lake confluence, an area with intricate hydrological patterns. We found that the Self-Attentive LSTM (SA-LSTM) model outperformed the other three machine learning models in predicting water quality, achieving Nash-Sutcliffe Efficiency (NSE) scores of 0.71 for COD_Mn and 0.57 for NH₃N when utilizing LOADEST-augmented water quality data (referred to as the SA-LSTM-LOADEST model). The SA-LSTM-LOADEST model improved upon the standalone SA-LSTM model by reducing the Root Mean Square Error (RMSE) by 24.6% for COD_Mn and 21.3% for NH₃N. Furthermore, the model maintained its predictive accuracy when data collection intervals were extended from weekly to monthly. Additionally, the SA-LSTM-LOADEST model demonstrated the capability to forecast pollution loads up to ten days in advance. This study shows promise for improving water quality modeling in regions with limited monitoring capabilities.

Urban flooding is a global water disaster resulting from the expansion of urban impervious surfaces and the strengthening of extreme precipitation events, especially in China. Nonetheless, few studies have focused on the spatial distributions of urban flooding characteristics and their variations in the context of climate change. In this study, eight critical metrics (i.e., maximum flooding volume, total overloaded manholes with different flooding volumes or durations, total flooding volume, mean and maximum flooding durations, maximum inundation area, and depth) are adopted to characterize the urban flood events. The impacts of climate change on these metrics are assessed for two periods, the 2030s (2020–2049) and 2070s (2060–2089), and compared with those in the baseline period (1976–2005). The Future Science City Park in Beijing, China, is selected as our study area. The results show that all four flood events are well simulated, with both efficiency coefficients and correlation coefficients being over 0.8. The number of overloaded manholes and the total flooding volume are projected to increase 19.3%–44.8% and 171%–716% under 20‐year rainfall events due to climate change in the two future periods. The spatial distribution of overloaded manholes with different increased flooding volumes is projected to expand to almost the whole area from the region with lowland and limited drainage capacity. Furthermore, the maximum inundation area and depth are projected to increase obviously. This study will be helpful for designing and improving the drainage system, controlling urban flooding, and adapting to climate change.