The Uncertainty In Runoff Simulation Modeled By Swat

The water quality model is an important tool in watershed management by simulating the migration and transformation of water pollutants. However, water quality model normaly describes the complicated process of pollutants migration and transformation by mathematical formula, which results in much uncertainty of the prediction, especailly for the variation of input parameters which may cause different impact on the simulation result. The deviation between the simulated results and the observed value may restrict the development and application of the water quality model. Therefore the widely used Soil and Water Assessment Tool (SWAT) was selected to analyse the parameters uncertainly in the study. In this suty, we first established the hydrological processes model of SWAT for the upstream catchment of Tiaoxi Watershed and the uncertainty of input parameters were discussed. The results were showed as follows:(1) The hydrological processes model of SWAT for the upstream catchment of Tiaoxi Watershed was constructed on the basis of parameter calibration and validation. Four parameters including CN2. ESCO、EPCO and REVAPMN were chosen as the sensitivity parameters for the runoff simulation. The results showed that the Ens and Er values of survey region were0.88and0.06respectively for model calibration and were0.81and0.04respectively for model validation, which indicated that the model was applicable to thestudy area.(2) The GLUE method was used to analyze the uncertainty of model input parameters. The variation of parameter could lead to variation of prediction uncertainty. The result showed that CN2made the most important contribution to the model uncertainty, followed by ESCO, EPCO and REVAPMN. The95PPU analysis showed that the observed values of runoff were approximately distributed in the range of95PPU. The model uncertainty increased with the increase of the runoff volume, therefore the uncertainty showed significant difference among different hydrological years. In addition, the "equifinality for different parameters" phenomenon in GLUE method has been further validated in this paper. The results of multi-parameters analysis showed that the synergistic effect of double-parameter group of CN2/ESCO, CN2/EPCO, and CN2/REVAPMN made great influence on the model uncertainty. And the value of parameter group with ESCO/EPCO may result in the best or worst model simulation, while another two parameter groups of ESCO/REVAPMN and EPCO/REVAPMN may make little impact to the model uncertainty.(3) The SUFI-2method was also used to analyze the uncertainty of model input parameters. The analysis showed that the uncertainty of model parameters was well simulated by SUFI-2method with higher values of Ens and Er in the likelihood objective function. The result of uncertainly analysis by SUFI-2was better than that of GLUE method. According to the results of multi-parameter uncertainty analysis, parameter values of group ESCO/EPCO, ESCO/REVAPMN and EPCO/REVAPMN may also make influence on the model simulation and lead a better model simulation. Moreover, the uncertainty analysis of typical year suggested that the uncertainty of SWAT model simulation in dry year was higher than that of wet year and median year, which may be induced by the mechanisms of hydrological processes of SWAT model.(4) The comparison between the GLUE and SUFI-2method in uncertainly analysis showed that the two methods could obtain satisfactory simulated results and come to coincident conclusion on the uncertainty analysis. The GLUE method has several merits in light of counting process and the description of parameter distribution, while the simulation effects and uncertainty analysis of SUFI-2method was more stable than that of GLUE method.