Scholarlink

<p>This study proposes a structured stochastic calibration and decision framework for semi-distributed rainfall–runoff models in transboundary river basins. The framework is designed to support objective-function selection and decision-level interpretation under spatially heterogeneous and data-scarce conditions, rather than simultaneous multi-objective optimization. Four optimization algorithms, namely Particle Swarm Optimization, Genetic Algorithm, Bayesian Markov Chain Monte Carlo, and Grid Search, are independently combined with four objective functions: Nash–Sutcliffe Efficiency, Root Mean Square Error, Kling–Gupta Efficiency, and the Correlation Coefficient. Each algorithm–objective pair is optimized separately. The resulting solutions are subsequently examined using post-calibration Pareto-based analysis. The framework is applied to three hydrological stations in the transboundary Imjin River Basin, shared between South and North Korea, which exhibit contrasting degrees of transboundary exposure. The results reveal pronounced spatial differences in calibration behavior associated with upstream data availability. Stations influenced by data-inaccessible upstream regions show wider parameter dispersion and less stable Pareto structures. In contrast, stations with more reliable observations exhibit stronger convergence and clearer dominance patterns. Adaptive and probabilistic algorithms generally demonstrate more stable calibration behavior than deterministic methods under data-scarce and hydrologically complex conditions. Overall, the proposed framework reframes model calibration as a post-calibration, multi-criteria decision problem. In this formulation, Pareto analysis supports station-specific objective-function interpretation and selection under heterogeneous data conditions, rather than optimizing an aggregated performance metric. By explicitly separating optimization from decision-making, this perspective provides a stochastic, diagnostic, and transferable basis for improving flood modeling and water-resource assessment in transboundary basins characterized by asymmetric data availability.</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

A stochastic multi-objective calibration framework for semi-distributed rainfall–runoff models in transboundary river basins

  • 期刊论文

  • 发表日期：

    2026年7月31日

查看全文
  • Thi-Duyen Nguyen,
  • Hyun-Han Kwon,
  • Deg-Hyo Bae

摘要

This study proposes a structured stochastic calibration and decision framework for semi-distributed rainfall–runoff models in transboundary river basins. The framework is designed to support objective-function selection and decision-level interpretation under spatially heterogeneous and data-scarce conditions, rather than simultaneous multi-objective optimization. Four optimization algorithms, namely Particle Swarm Optimization, Genetic Algorithm, Bayesian Markov Chain Monte Carlo, and Grid Search, are independently combined with four objective functions: Nash–Sutcliffe Efficiency, Root Mean Square Error, Kling–Gupta Efficiency, and the Correlation Coefficient. Each algorithm–objective pair is optimized separately. The resulting solutions are subsequently examined using post-calibration Pareto-based analysis. The framework is applied to three hydrological stations in the transboundary Imjin River Basin, shared between South and North Korea, which exhibit contrasting degrees of transboundary exposure. The results reveal pronounced spatial differences in calibration behavior associated with upstream data availability. Stations influenced by data-inaccessible upstream regions show wider parameter dispersion and less stable Pareto structures. In contrast, stations with more reliable observations exhibit stronger convergence and clearer dominance patterns. Adaptive and probabilistic algorithms generally demonstrate more stable calibration behavior than deterministic methods under data-scarce and hydrologically complex conditions. Overall, the proposed framework reframes model calibration as a post-calibration, multi-criteria decision problem. In this formulation, Pareto analysis supports station-specific objective-function interpretation and selection under heterogeneous data conditions, rather than optimizing an aggregated performance metric. By explicitly separating optimization from decision-making, this perspective provides a stochastic, diagnostic, and transferable basis for improving flood modeling and water-resource assessment in transboundary basins characterized by asymmetric data availability.