Scholarlink

<p>Compound flooding caused by the combination of rainfall and tidal level has resulted in severe losses. Rapid and accurate prediction of compound flooding is critical for mitigating disasters in coastal urban areas. Numerous studies have integrated physics-based models with deep learning models to achieve fast flood prediction. Nevertheless, few studies focused on predicting the evolution process of compound floods, which hinders the development of emergency management measures for coastal cities. In this study, we propose a hybrid approach for rapid prediction of compound flood process by coupling the hydrological-hydrodynamic model with the convolutional neural network (CNN). Firstly, an urban flood simulation model based on the Personal Computer Storm Water Management Model (PCSWMM) is utilized to generate flood depth data under various scenarios. Subsequently, a CNN model is constructed for predicting compound flood processes using rainfall, tidal level, and flood depth data. Taking Haidian Island as an example, the flood depth predicted by the constructed CNN model closely matches that simulated by the PCSWMM model, indicating that the CNN model reliably reflects the flood depth variation trends within the study area, with an average absolute error (MAE) of 0.012&#xa0;m, a root mean square error (RMSE) of 0.021&#xa0;m, and a Pearson correlation coefficient (PCC) of 0.977. Furthermore, the CNN model significantly reduces computational time, operating approximately 450 times faster than the PCSWMM simulations. The proposed approach is applicable for rapid prediction of compound floods in coastal cities, providing reference for decision makers to take action to mitigate flooding.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Rapid Prediction of Compound Flood Based on Hydrological-Hydrodynamic Model and Convolution Neural Network

  • 期刊论文

  • 发表日期：

    2026年2月25日

查看全文
  • Kui Xu,
  • Yizhuang Tian,
  • Lingling Bin,
  • Chengguang Lai,
  • Weichao Yang

摘要

Compound flooding caused by the combination of rainfall and tidal level has resulted in severe losses. Rapid and accurate prediction of compound flooding is critical for mitigating disasters in coastal urban areas. Numerous studies have integrated physics-based models with deep learning models to achieve fast flood prediction. Nevertheless, few studies focused on predicting the evolution process of compound floods, which hinders the development of emergency management measures for coastal cities. In this study, we propose a hybrid approach for rapid prediction of compound flood process by coupling the hydrological-hydrodynamic model with the convolutional neural network (CNN). Firstly, an urban flood simulation model based on the Personal Computer Storm Water Management Model (PCSWMM) is utilized to generate flood depth data under various scenarios. Subsequently, a CNN model is constructed for predicting compound flood processes using rainfall, tidal level, and flood depth data. Taking Haidian Island as an example, the flood depth predicted by the constructed CNN model closely matches that simulated by the PCSWMM model, indicating that the CNN model reliably reflects the flood depth variation trends within the study area, with an average absolute error (MAE) of 0.012 m, a root mean square error (RMSE) of 0.021 m, and a Pearson correlation coefficient (PCC) of 0.977. Furthermore, the CNN model significantly reduces computational time, operating approximately 450 times faster than the PCSWMM simulations. The proposed approach is applicable for rapid prediction of compound floods in coastal cities, providing reference for decision makers to take action to mitigate flooding.