Research on developing a strategic model for warning and forecasting floods and inundation in the Quang Nam - Da Nang region

Tóm tắt

Early warning is crucial for saving lives and property from natural disasters in general, and for providing information to support evacuation from flood-prone areas in particular. This study aims to develop a simplified model by using an Artificial Neural Network (ANN) to predict floods in the lower land of the Vu Gia - Thu Bon rivers system. The results demonstrate strong performance, with R² coefficient values of 0.968, 0.958, and 0.964 for the training, validation, and testing processes, respectively. The 12-hour and 24-hour lead-time models at Cam Le station on the Vu Gia system achieved a correlation coefficient (η) of 0.917 and a forecast accuracy coefficient (P) of 88.94%. Similarly, the models at Hoi An station on the Thu Bon system achieved an η of 0.94 and a P of 95.85%. Additionally, these models do not rely on forecasted data during the prediction process, which helps reduce uncertainty stemming from forecast input inaccuracies. The computation time is only a few seconds compared to the hours required for physically based models. Therefore, the use of artificial neural networks (ANN) is highly recommended for testing in other case studies. Keyword: Flood, hydrological forecast, simplified model, ANN, Vu Gia - Thu Bon