An efficient global sensitivity analysis approach for distributed hydrological model  

作　　者：SONG Xiaomeng ZHAN Chesheng XIA Jun KONG Fanzhe 

机构地区：[1]School of Resource and Earth Science,China University of Mining&Technology,Xuzhou 221008,Jiangsu,China [2]Key Laboratory of Water Cycle&Related Land Surface Processes,Institute of Geographic Sciences and Natural Resources Research,CAS,Beijing 100101,China

出　　处：《Journal of Geographical Sciences》2012年第2期209-222,共14页地理学报（英文版）

基　　金：National Key Basic Research Program of China,No.2010CB428403;National Grand Science and Technology Special Project of Water Pollution Control and Improvement,No.2009ZX07210-006

摘　　要：Sensitivity analysis of hydrological model is the key for model uncertainty quantification. However, how to effectively validate model and identify the dominant parameters for distributed hydrological models is a bottle-neck to achieve parameters optimization. For this reason, a new approach was proposed in this paper, in which the support vector machine was used to construct the response surface at first. Then it integrates the SVM-based response surface with the Sobol＇ method, i.e. the RSMSoboI＇ method, to quantify the parameter sensi- tivities. In this work, the distributed time-variant gain model （DTVGM） was applied to the Huaihe River Basin, which was used as a case to verify its validity and feasibility. We selected three objective functions （i.e. water balance coefficient WB, Nash-Sutcliffe efficiency coefficient NS, and correlation coefficient RC） to assess the model performance as the output responses for sensitivity analysis. The results show that the parameters gl and g2 are most important for all the objective functions, and they are almost the same to that of the classical approach. Furthermore, the RSMSobol method can not only achieve the quantification of the sensitivity, and also reduce the computational cost, with good accuracy compared to the classical approach. And this approach will be effective and reliable in the global sensitivity analysis for a complex modelling system.Sensitivity analysis of hydrological model is the key for model uncertainty quantification. However, how to effectively validate model and identify the dominant parameters for distributed hydrological models is a bottle-neck to achieve parameters optimization. For this reason, a new approach was proposed in this paper, in which the support vector machine was used to construct the response surface at first. Then it integrates the SVM-based response surface with the Sobol＇ method, i.e. the RSMSoboI＇ method, to quantify the parameter sensi- tivities. In this work, the distributed time-variant gain model （DTVGM） was applied to the Huaihe River Basin, which was used as a case to verify its validity and feasibility. We selected three objective functions （i.e. water balance coefficient WB, Nash-Sutcliffe efficiency coefficient NS, and correlation coefficient RC） to assess the model performance as the output responses for sensitivity analysis. The results show that the parameters gl and g2 are most important for all the objective functions, and they are almost the same to that of the classical approach. Furthermore, the RSMSobol method can not only achieve the quantification of the sensitivity, and also reduce the computational cost, with good accuracy compared to the classical approach. And this approach will be effective and reliable in the global sensitivity analysis for a complex modelling system.

关 键 词：response surface methodology sensitivity analysis support vector machines RSMSobol method Huaihe River Basin 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置] P334.92[自动化与计算机技术—控制科学与工程]