Scholarlink

<p>Accurately modeling stage‒discharge relationships has become a critical task in flood risk management due to the increasing uncertainty of river systems responses to extreme events under climate change. Despite advancements in real-time water level monitoring and forecasting, traditional methods struggle to accurately forecast flood water levels, particularly during high-flow events. This study addresses this gap by applying Bayesian optimization (BO) to improve the accuracy of the stage‒discharge modeling and enhance efficiency of flood risk management. Various acquisition functions within the BO framework were employed, and their performance was systematically compared across different return periods (2, 5, 10, 25, 50, and 100 years) and flow percentages (1% to 60%). Additionally, return periods were calculated using different probability distributions, including generalized extreme value, lognormal, Weibull, and Gumbel Max, to conduct a comprehensive analysis. The results demonstrated that the expected improvement per second plus (EIPSP) acquisition function provided the lowest relative errors among the evaluated approaches, with errors as low as 2.90% for the 2-year return period and 3.52% for the 100-year return period, compared with errors ranging from 18.95% to 26.16% for the traditional SRC method. The EIPSP function also showed consistently lower errors when maximum flow percentages were analyzed. These findings suggest that Bayesian optimization can improve the accuracy of stage–discharge modeling and may support flood risk assessment and water resource management under high-flow conditions.</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

Advancing Stage‒Discharge Relationship Modeling via Bayesian Optimization: A Case Study of the Upper Gatineau Watershed

  • 期刊论文

  • 发表日期：

    2026年6月29日

查看全文
  • Firdaws Abdelhafiz,
  • Isa Ebtehaj,
  • Afshin Amiri,
  • Silvio Jose Gumiere,
  • Bahram Gharabaghi,
  • Hossein Bonakdari

摘要

Accurately modeling stage‒discharge relationships has become a critical task in flood risk management due to the increasing uncertainty of river systems responses to extreme events under climate change. Despite advancements in real-time water level monitoring and forecasting, traditional methods struggle to accurately forecast flood water levels, particularly during high-flow events. This study addresses this gap by applying Bayesian optimization (BO) to improve the accuracy of the stage‒discharge modeling and enhance efficiency of flood risk management. Various acquisition functions within the BO framework were employed, and their performance was systematically compared across different return periods (2, 5, 10, 25, 50, and 100 years) and flow percentages (1% to 60%). Additionally, return periods were calculated using different probability distributions, including generalized extreme value, lognormal, Weibull, and Gumbel Max, to conduct a comprehensive analysis. The results demonstrated that the expected improvement per second plus (EIPSP) acquisition function provided the lowest relative errors among the evaluated approaches, with errors as low as 2.90% for the 2-year return period and 3.52% for the 100-year return period, compared with errors ranging from 18.95% to 26.16% for the traditional SRC method. The EIPSP function also showed consistently lower errors when maximum flow percentages were analyzed. These findings suggest that Bayesian optimization can improve the accuracy of stage–discharge modeling and may support flood risk assessment and water resource management under high-flow conditions.