Scholarlink

Persistent and intense rainfall resulting from cyclones leads to flooding and inundation of low-lying regions, resulting in fatalities and damage to assets. Precise modelling of rainfall and runoff during cyclones is crucial to mitigating the impact of natural disasters. The main objective of this study was to develop a hybrid model for the Nagavali River Basin which integrates physically based hydrologic modelling system (HEC-HMS) into machine learning models (SVM, ANN, RF) to predict daily runoff discharges in the Nagavali Basin, Andhra Pradesh. In the past 10 years, the river basin has been more vulnerable to cyclones. Five flood events that are caused by cyclones (with a total of 70 data sets) are used for model calibration and validation. Five statistical indices (Mean Absolute Error, Root Mean Square Error, Correlation Coefficient, Error of Peak Discharge and R2) are employed to assess prediction performance. The overall superiority of the present approach is revealed through systemic comparison among physically based hydrologic model (HEC-HMS) and three hybrid combinations (HEC-ANN, HEC-SVR, HEC-RF). Overall results show hybrid models give better results as they reduce the uncerrtainties that occur in a physical-based model and improve their predicting capability. For this basin there is an overall improvement of 16- 79% of Mean Absolute Error, 36–86% of Root Mean Square Error and 7–42% of R2 from physical based model to hybrid models. Out of the three hybrid models, the results obtained from the HEC-SVR model are much better and provides the most accurate runoff discharge predictions.

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

Simulation of Floods During Cyclone Events for a Coastal Basin Using HEC-HMS and Machine Learning Algorithms

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Buridi Sravya,
  • Venkata Reddy Keesara,
  • Eswar Sai Buri

摘要

Persistent and intense rainfall resulting from cyclones leads to flooding and inundation of low-lying regions, resulting in fatalities and damage to assets. Precise modelling of rainfall and runoff during cyclones is crucial to mitigating the impact of natural disasters. The main objective of this study was to develop a hybrid model for the Nagavali River Basin which integrates physically based hydrologic modelling system (HEC-HMS) into machine learning models (SVM, ANN, RF) to predict daily runoff discharges in the Nagavali Basin, Andhra Pradesh. In the past 10 years, the river basin has been more vulnerable to cyclones. Five flood events that are caused by cyclones (with a total of 70 data sets) are used for model calibration and validation. Five statistical indices (Mean Absolute Error, Root Mean Square Error, Correlation Coefficient, Error of Peak Discharge and R2) are employed to assess prediction performance. The overall superiority of the present approach is revealed through systemic comparison among physically based hydrologic model (HEC-HMS) and three hybrid combinations (HEC-ANN, HEC-SVR, HEC-RF). Overall results show hybrid models give better results as they reduce the uncerrtainties that occur in a physical-based model and improve their predicting capability. For this basin there is an overall improvement of 16- 79% of Mean Absolute Error, 36–86% of Root Mean Square Error and 7–42% of R2 from physical based model to hybrid models. Out of the three hybrid models, the results obtained from the HEC-SVR model are much better and provides the most accurate runoff discharge predictions.