Author(s): Wang Fan; Chen Chang

Linked Author(s):

Keywords: Flood forecasting; Neural networks; Piedmont plain cities; Urban rivers

Abstract: The flood in the piedmont plain city exhibits characteristics of both mountain and urban floods, posing challenges for hydrological simulation and flood forecasting. This study utilizes two different modelling approaches to make flood forecasting and evaluate the performance of models in Xiaoqing River and Futuan River basins in China. The findings demonstrate that: 1) the integrated model, based on BiGRU, exhibits flexible capabilities in predicting discharge and water level processes. It is suitable for forecasting both single flood events and providing continuous predictions for long series of processes, while maintaining a high level of prediction accuracy within a specific forecast step. 2) the hybrid model, combining the LSTM and the two-dimensional hydrodynamic model, demonstrates remarkable accuracy in predicting the water level process of the target section. In this model, the LSTM is employed to simulate flood processes in hilly areas and provide boundary conditions for the two-dimensional hydrodynamic model. Through comprehensive case assessment and analysis, we contend that both modeling methods can be effectively utilized as innovative approaches for predicting river floods in piedmont plain cities.

DOI: https://doi.org/10.3850/iahr-hic2483430201-289

Year: 2024