Location: Southeast Watershed Research
Title: Calibration of Uncertainty Analysis of the SWAT Model Using Genetic Algorithms and Bayesian Model Averaging Authors
|Zhang, Xuesong - TEXAS A&M UNIV|
|Srinivasan, Raghavan - TEXAS A&M UNIV|
Submitted to: Journal of Hydrology
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: June 15, 2009
Publication Date: August 15, 2009
Citation: Zhang, X., Srinivasan, R., Bosch, D.D. 2009. Calibration of Uncertainty Analysis of the SWAT Model Using Genetic Algorithms and Bayesian Model Averaging. Journal of Hydrology. 374:307-317. Interpretive Summary: Computer based hydrologic and water quality models have become a common tool for examining field and watershed scale processes. The simulation results are used for a wide variety of applications including water resource and urban planning. However, simulation results are frequently presented as precise values, implicitly omitting any uncertainty that may be related to the estimates provided by the model. A method for improving modeling estimates and quantifying simulation uncertainty is presented here. The methods presented provide a useful tool for attaining reliable simulation and uncertainty results with the SWAT computer simulation model. These results will provide water resource managers with more useful results to guide their decision process.
Technical Abstract: In this paper, the Genetic Algorithms (GA) and Bayesian model averaging (BMA) were combined to simultaneously conduct calibration and uncertainty analysis for the Soil and Water Assessment Tool (SWAT). In this hybrid method, several SWAT models with different structures are first selected; next GA is used to calibrate each model using observed streamflow data; finally, BMA was applied to combine the ensemble predictions and provide uncertainty interval estimation. This method was tested in two contrasting basins, the Little River Experimental Basin in Georgia, USA, and the Yellow River Headwater Basin in China. The results show that this hybrid method can provide better deterministic predictions compared with the best calibrated model using GA. Uncertainty intervals estimated by this method at 33.3%, 66.7%, and 90% coverage levels were analyzed. The difference between the predicted percentage of coverage values and their corresponding coverage levels is within 10% for both calibration and validation periods in these two test basins. This hybrid methodology holds the promise to be an effective tool to attain reliable deterministic simulation and uncertainty analysis of SWAT.