Comparison of the Autocalibration Tools Available Within SWAT2005 and Manual Calibration

Abstract

A Soil and Water Assessment Tool (SWAT) watershed loading and water quality model was constructed to support the Total Maximum Daily Load (TMDL) process and evaluate management and land use scenarios within the nearly 7,000 km2 Bitterroot River watershed. The Bitterroot River is located in western Montana and flows north from near the Idaho-Montana border to Missoula, Montana, where it joins the Clark Fork River. A SWAT model was developed using SWAT2000. A rigorous calibration effort was imperative because the model will be used as part of point and non-point source loading analyses, sediment and nutrients allocations, and formulation of water quality restoration plans. The initial model calibration followed a manual or iterative process that consisted of first reviewing and refining the hydrology and then the water quality coefficients. Subsequently, AVSWAT-X, a beta version of the SWAT2005 model release, became available and included auto-calibration tools. These auto-calibration tools were utilized as a comparison test of both the manually calibrated coefficients and the results. The calibration was evaluated qualitatively by graphing annual and monthly results and quantitatively by calculating Nash-Sutcliffe Coefficient of Efficiency (COE) values, percent bias, and comparison to seasonal criteria. A combination of manual and auto-calibration values were used in the calibrated Bitterroot River SWAT Model to provide predicted results within the ranges of literature-reported model calibration statistics. The calibrated model provided Nash-Sutcliffe COE values of 0.78 or greater for monthly values and percent bias less than 10% at three calibration locations.