Scholarlink

<p>Seasonal streamflow forecasts are essential given climate-driven extremes that breach stationarity in traditional methods. The complex hydrology and competing demands necessitate improved forecasting in the Upper Feather River Basin (UFRB), a key California State Water Project source upstream of Oroville Dam. We introduce a hybrid framework combining dynamical downscaling via WRF and the WEHY-HCM snow-hydrology model with a lead-time–dependent exponential-smoothing filter that adaptively corrects bias and quantifies uncertainty. Applied to December–July ensemble forecasts for water year 2024 using hindcast error training (2018–2023), this approach reduced RMSE by 8.7–318.3&#xa0;million m³ across eight initialization months and eliminated systematic bias. The resulting 10–90% exceedance bands captured ~ 80% of observed flows, offering reliable confidence intervals. This hybrid method delivers accurate, low-bias streamflow forecasts for reservoir operations, flood mitigation, and irrigation planning in the UFRB and provides a transferable template for other basins facing hydroclimatic variability.</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

Hybrid physical–statistical framework for seasonal streamflow forecasting in the Upper Feather River Basin, California

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年8月30日

查看全文
  • Z. Ozcan,
  • Y. Iseri,
  • F. Ulloa,
  • N. Imbulana,
  • E. Snider,
  • M. Mure-Ravaud,
  • M. L. Anderson,
  • M. L. Kavvas

摘要

Seasonal streamflow forecasts are essential given climate-driven extremes that breach stationarity in traditional methods. The complex hydrology and competing demands necessitate improved forecasting in the Upper Feather River Basin (UFRB), a key California State Water Project source upstream of Oroville Dam. We introduce a hybrid framework combining dynamical downscaling via WRF and the WEHY-HCM snow-hydrology model with a lead-time–dependent exponential-smoothing filter that adaptively corrects bias and quantifies uncertainty. Applied to December–July ensemble forecasts for water year 2024 using hindcast error training (2018–2023), this approach reduced RMSE by 8.7–318.3 million m³ across eight initialization months and eliminated systematic bias. The resulting 10–90% exceedance bands captured ~ 80% of observed flows, offering reliable confidence intervals. This hybrid method delivers accurate, low-bias streamflow forecasts for reservoir operations, flood mitigation, and irrigation planning in the UFRB and provides a transferable template for other basins facing hydroclimatic variability.