Uncertainty Analysis of Hydrological Parameters of Sarbaz Watershed Using SWAT Model and SUFI-2 Algorithm

Abstract

The SWAT model was used to simulate the monthly runoff of the Sarbaz watershed (6324 Km2) located in Sistan-va- Baluchestan province of Iran. The main objective of the study was to evaluate the accuracy of the SWAT model simulation in the calibration and validation stages and to analyze the uncertainty of input parameters of the model. A hydrological modeling approach was used to identify the sensitive hydrological input parameters of the watershed through the SUFI-2 algorithm. The important parameters of the model were determined by the general sensitivity analysis, and the model was, then, calibrated for 1997-2005 with two years as the warm-up (1997–1998) by a multi-objective optimization approach using the SWAT-CUP program. Subsequently, the model was validated on a dataset of five years (2006–2010). To reduce the uncertainty, different components of the water balance were considered in the calibration stage, and data from the Pishin station were used. Based on the p-value and t-stat, the most sensitive parameters were the curve number and the parameter of alpha in return flow, and the parameter with the lowest sensitivity was the percentage of deep aquifer feeding on the shallow aquifer. The performance of the model was tested using two objective functions of Nash–Sutcliffe efficiency (NSE) and coefficient of determination (R2). R2 and NSE were estimated at 0.83 and 0.80 in calibration and 0.53 and 0.38 in validation, respectively. Acceptable values ​​were also obtained for the uncertainty assessment indicators. The p-factor and r-factor were calculated to be 0.76 and 1.53 for the calibration stage and 0.72 and 1.23 for the validation stage, respectively. In addition to confirming the goodness of fit for the conditions of the Sarbaz watershed, these results indicate that the model can positively affect the simulated runoff in future projects due to the dynamics of the model parameters in the specified uncertainty interval.