SWAT模型参数敏感性分析及应用

地理信息系统(GIS)支持下的SWAT(Soil and Water Assessment Tool)分布式水文模型以流域离散化空间参数来描述流域水文变化特性,从物理意义上表达流域内的水文过程,但众多不确定的参数影响了模型的应用效果,因此有必要对参数进行敏感性分析。将SWAT模型应用到祁连山黑河上游山区流域,进行了11年(1990-2000年)逐日径流模拟,通过一个简便的敏感性分析方法,将模型影响水文过程的参数分成4类敏感级别,最后确定模型的参数。在11年的逐日模拟中,1990-1995年为参数敏感性分析期和模型率定期,1996-2000年为模型的检验期,模拟结果显示,在黑河山区流域,丰水年逐日出山径流的模型效率系数R2达到0.8以上,平水年和枯水年R2在0.51~0.79之间。

Application and parameters sensitivity analysis of SWAT model

SWAT (Soil and Water Assessment Tool), a distributed hydrologic model integrated with geographic infor-mation system (GIS) to simulate the hydrological processes and runoff generations in mountainous catchments , was applied to the Heihe River basin located in the Qilian Mountains of northwestern China. Simulations at present were preliminarily focused on predicting the continuous daily runoff in a high altitude, cold, semi-arid 10 009 km~2 basin with the total control hydrological station of Yingluoxia. Because the sensitivity analysis and calibration is a key and difficult issue of the application and development of the distributed hydrological model and spatial variability, access difficulties model input parameters always contain uncertainty. In this paper, model is calibrated with observed data from four principal hydrologic stations in the Heihe River basin. A simple approach of sensitivity analysis on the pri-or-determined parameters was used to assess the sensitivity of those parameters to the final performance of the model, the calculated sensitivity indices are ranked into four classes. The laws are discovered that change some important pa-rameters how to affect model results and thus to determine the most sensitive parameter values. The calibration proce-dures have considerably been simplified, the blindness on the determination of those pre-determined factors was de-creased. The runoff generation processes and their calculation principles were described. The basic GIS data layers include the Digital Elevation Model (DEM), soil and land use/cover digital maps generated from the digitalization of the field survey results. Four meteorological stations within or near the study basin were selected to derive the distrib-uted input meteorological parameters. Based on the topographic features, soil and land use/cover distribution pat-terns, SWAT - GIS delineated the basin into 157 homogeneous hydrologic respondent units (HRU), each of which represents a particular feature of soil and land use/cover combinations within the sub-watershed and a group of param-eters was assigned to them for charactering its hydrological responses. The model simulation'period was divided into calibration (1990-1995) and validation (1996-2000) periods. Results from the 11 year's simulation suggested a good performance of the model demonstrated by Nath - sutcliffe coefficient R~2 of over 0.80 in the rain-rich years and the difference between the observed and simulated runoff was less than 16% in general.