Scholarlink

<p>This study presents a unified framework for quantitative uncertainty analysis in process-based models, applied to runoff and soil loss simulations using tRIBS-VEGGIE-FEaST (Triangulated irregular network - based Real time Integrated Basin Simulator- VEGetation Generator for Interactive Evolution -Flow Erosion and Sediment Transport). The framework integrates Generalized Likelihood Uncertainty Estimation (GLUE), global sensitivity analysis, and Polynomial Chaos Kriging (PCK) to efficiently quantify parameter uncertainty and its propagation. Results show that PCK effectively emulates the original model, accurately capturing uncertainty while reducing computation time from approximately 20,000 hours (for 10,000 runs) to seconds. The study reveals substantial model uncertainty, with volume error ranges up to 300% for runoff and 800% for soil loss. Global sensitivity analysis based on Sobol’ indices identifies hydraulic conductivity and clay fraction as the most influential parameters, with temporal variations in sensitivity. High equifinality among parameters is observed, except for hydraulic conductivity. This framework is expected to be applicable to various regions and models, enhancing understanding of parameter interactions and facilitating rapid uncertainty quantification in predictions using computationally intensive, physics-rich process-based models.</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

Robust uncertainty analysis of a process-based model for runoff and soil erosion simulations using surrogate modeling: a synthetic study

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年4月30日

查看全文
  • Nguyen Hao Quang,
  • Vinh Ngoc Tran

摘要

This study presents a unified framework for quantitative uncertainty analysis in process-based models, applied to runoff and soil loss simulations using tRIBS-VEGGIE-FEaST (Triangulated irregular network - based Real time Integrated Basin Simulator- VEGetation Generator for Interactive Evolution -Flow Erosion and Sediment Transport). The framework integrates Generalized Likelihood Uncertainty Estimation (GLUE), global sensitivity analysis, and Polynomial Chaos Kriging (PCK) to efficiently quantify parameter uncertainty and its propagation. Results show that PCK effectively emulates the original model, accurately capturing uncertainty while reducing computation time from approximately 20,000 hours (for 10,000 runs) to seconds. The study reveals substantial model uncertainty, with volume error ranges up to 300% for runoff and 800% for soil loss. Global sensitivity analysis based on Sobol’ indices identifies hydraulic conductivity and clay fraction as the most influential parameters, with temporal variations in sensitivity. High equifinality among parameters is observed, except for hydraulic conductivity. This framework is expected to be applicable to various regions and models, enhancing understanding of parameter interactions and facilitating rapid uncertainty quantification in predictions using computationally intensive, physics-rich process-based models.