Variance-based global sensitivity analysis for multiple scenarios and models with implementation using sparse grid collocation

JOURNAL OF HYDROLOGY

卷:528

Article

Dai, Heng¹  Ye, Ming¹ 

[1] Florida State Univ, Dept Comp Sci, Tallahassee, FL 32306 USA

关键词: Model uncertainty;    Model averaging;    Scenario uncertainty;    Scenario averaging;    Variance decomposition;    Sparse grid collocation;   

DOI  :  10.1016/j.jhydrol.2015.06.034

来源: Elsevier

【 摘 要 】

Sensitivity analysis is a vital tool in hydrological modeling to identify influential parameters for inverse modeling and uncertainty analysis, and variance-based global sensitivity analysis has gained popularity. However, the conventional global sensitivity indices are defined with consideration of only parametric uncertainty. Based on a hierarchical structure of parameter, model, and scenario uncertainties and on recently developed techniques of model- and scenario-averaging, this study derives new global sensitivity indices for multiple models and multiple scenarios. To reduce computational cost of variance-based global sensitivity analysis, sparse grid collocation method is used to evaluate the mean and variance terms involved in the variance-based global sensitivity analysis. In a simple synthetic case of groundwater flow and reactive transport, it is demonstrated that the global sensitivity indices vary substantially between the four models and three scenarios. Not considering the model and scenario uncertainties, might result in biased identification of important model parameters. This problem is resolved by using the new indices defined for multiple models and/or multiple scenarios. This is particularly true when the sensitivity indices and model/scenario probabilities vary substantially. The sparse grid collocation method dramatically reduces the computational cost, in comparison with the popular quasi-random sampling method. The new framework of global sensitivity analysis is mathematically general, and can be applied to a wide range of hydrologic and environmental problems. (C) 2015 Elsevier B.V. All rights reserved.

【 授权许可】

Free