Parameter estimation of SWAT and quantification of consequent confidence bands of model simulations

Soil and Water Assessment Tool (SWAT) is a river basin scale model widely used to study the impact of land management practices in large, complex watersheds. Even though model output uncertainties are generally recognized to affect watershed management decisions, those uncertainties are largely ignored in model applications. The uncertainties of SWAT simulations are quantified using various methods, but simultaneous attempt to calibrate a model so as to reduce the uncertainty are seldom done. This study aims to use an uncertainty reduction procedure that helps calibrate the SWAT model. The shuffled complex evolutionary metropolis algorithm for uncertainty analysis is employed for this purpose, and is demonstrated using the data from the St. Joseph River basin, USA. The values of the performance indices, the r2 and the Nash–Sutcliffe efficiency (NSE) for the simulations during calibration period was found to be 0.81 (same for r2 and NSE) and 0.79 for validation period indicating a good simulation by the model. The results also indicate that the algorithm helps reduce the uncertainty (percentage of coverage = 62% and average width = 19.2 m3/s), and also identifies the plausible range of parameters that simulate the processes with less uncertainty. The confidence bands of simulations are obtained that can be employed in making uncertainty-based decisions on watershed management practices.