Integrating MIKE model simulations with CNNs for rapid and accurate urban flood prediction

Rapid prediction of urban pluvial flooding is an important tool for mitigating current urban flooding disasters. This paper constructs a fast prediction model for urban flooding based on a machine learning approach. Firstly, MIKE numerical model simulations with high accuracy results are used as the data driver, and then the convolutional neural network (CNN) urban pluvial flooding model is structured based on CNN principles. An empirical study was conducted in the urban area of Zhoukou City to validate the proposed model. Results show that the proposed 1D-CNN model achieves a mean prediction error of 5.74% for inundation depth at the waterlogging-prone locations, and completes inference for a 3 h rainfall scenario in approximately 8 s, demonstrating both high accuracy and near-real-time computational efficiency for emergency urban pluvial flooding prediction. Therefore, the CNN urban pluvial flooding model trained by learning can quickly predict results with high accuracy and can support emergency response.