Prediction of urban flood inundation using Bayesian convolutional neural networks

Abstract

Urban flood risk management has been a hot issue worldwide due to the increased frequency and severity of floods occurring in cities. In this paper, an innovative modelling approach based on the Bayesian convolutional neural network (BCNN) was proposed to simulate the urban flood inundation, and to provide a reliable prediction of specific water depth. To develop the model, a series of historical rainfall data during the last 20 years were collected in Rushan China and the responding flood events were reproduced using physically based hydraulic model. The flood condition factors used in modeling include spacial factors and precipitation factors. The results showed that the BCNN model not only inherits the powerful ability of aggregating spacial information from CNNs to perform high level of accuracy and computational efficiency in predicting 2D urban flood inundation maps, but also offers a measure of uncertainty in the form of predictive variance, providing insights into the confidence and reliability of its predictions. The proposed BCNN method offered a new perspective for the analysis of surrogate model regarding real-time forecasting of flood inundation.