Climate change and rapid urbanization have made it difficult to predict the risk of pollution in cities under different types of rainfall. In this study, a data-driven approach to quantify the effects of rainfall characteristics on river pollution was proposed and applied in a case study of Shiyan River, Shenzhen, China. The results indicate that the most important factor affecting river pollution is the dry period followed by average rainfall intensity, maximum rainfall in 10 min, total amount of rainfall, and initial runoff intensity. In addition, an artificial neural network model was developed to predict the event mean concentration (EMC) of COD in the river based on the correlations between rainfall characteristics and EMC. Compared to under light rain (< 10 mm/day), the predicted EMC was five times lower under heavy rain (25–49.9 mm/day) and two times lower under moderate rain (10–24.9 mm/day). By converting the EMC to chemical oxygen demand in the river, the pollution load under non-point-source runoff was estimated to be 497.6 t/year (with an accuracy of 95.98%) in Shiyan River under typical rainfall characteristics. The results of this study can be used to guide urban rainwater utilization and engineering design in Shenzhen. The findings also provide insights for predicting the risk of rainfall-runoff pollution and developing related policies in other cities.@en

Efficient and accurate streamflow predictions are important for urban water management. Data-driven models, especially neural network (NN) models can predict streamflow fast, while the results are uncertain in some complex river systems. Physically based models can reveal the underlying physics, but it is relatively slow and computationally costly. This work focuses on evaluating the reliability of three NN models (artificial neural networks (ANN), long short-term memory networks (LSTM), adaptive neuro-fuzzy inference system (ANFIS)) and one physically based model (SOBEK) in terms of efficiency and accuracy for average and peak streamflow simulation. All the models are applied for a tidal river and a mountainous river in Shenzhen. The results show that, the ANN model calculates fastest since the hidden layer's structure is simple. The LSTM model is reliable in average streamflow simulation in tidal river with the lowest bias while the ANFIS model has the best accuracy for peak streamflow simulation. Furthermore, the SOBEK model shows reliability in simulating average and peak streamflow in mountainous river due to its ability to capture uneven spatial rainfall in the area. Overall, the results indicate that the LSTM model can be a helpful supplementary to physically based models in streamflow simulation of complex urban river systems, by giving fast streamflow predictions with usually acceptable accuracy. Our results can provide helpful information for hydrological engineers in the application of flooding early warning and emergency preparedness in the context of flooding risk management.@en