Hydrological and hydrodynamic coupling simulation under composite underlying surfaces in urban polder areas

The Nansha District of Guangzhou, China, is a typical example of a plain river network polder area. Polder areas, located in coastal lowlands, are frequently subjected to tidal influences and heavy rainfall, resulting in significant flood risks. The diverse land surface types and complex runoff generation and confluence mechanisms pose challenges to flood forecasting in these areas, particularly in the absence of high-resolution topographic data. This study proposed a runoff and confluence model incorporating various land surfaces, including paddy fields, drylands, forests, urban areas, and water surfaces. The MIKE 11 model was used to simulate river flood evolution. The model was applied in flood forecasting and risk assessment. A BP neural network (BPNN) was employed for error correction to reduce model uncertainty in forecasting. Five different rainfall scenarios with varying return periods (RP) were used to assess the region's drainage capacity and associated risks. The results showed that the model can simulate the river water level process well. The average Nash-Sutcliffe efficiency coefficients ( NSE ) reached 0.86 and 0.91 in the calibration and validation periods, respectively. Following the application of the BPNN correction, the NSE s of all floods were greater than 0.9. The drainage capacity assessment revealed that the current drainage facilities in the study area are at risk of urban flooding when confronted with a design storm with an RP= 50. The model can be used for flood forecasting and risk assessment in polder areas without the need for high-resolution data. • Employing high-resolution remote sensing imagery to obtain precise land use data. • Developing a composite underlying surface runoff generation model. • BP neural network enhanced the forecasting accuracy of the model. • Considering the facilitative or hindering effects of tidal levels on drainage.