Two Model Performance Comparisons with Multisite Observations Based on Uncertainty Methods for Modeling Hydrologic Dynamics

Abstract

The HSPF [Hydrologic Simulation Program—FORTRAN (Computer software). EPA National Exposure Research Laboratory, Athens, Georgia] and SWAT [Soil and Water Assessment Tool (Computer software). USDA Agricultural Research Service and Texas A&M AgriLife Research, Temple, Texas] models are often used to assess hydrological processes in a shallow water table setting. In the paper, these two models are evaluated and compared for accuracy in simulating runoff to the Peace River Basin in Central Florida, based on the generalized likelihood uncertainty estimation (GLUE) method. Both models performed reasonably well in predicting daily streamflow in three gauges with acceptable R2 values (.66–.85) and Nash-Sutcliffe’s Ens (.6–.76) with reasonable P and R factors. The parameter sensitivity was quantified, and the GLUE method involved Monte Carlo sampling over feasible spaces of calibration parameters. The parameter pairwise correlation is investigated, as well as the uncertainties along with equifinality. The results indicate that HSPF and SWAT models performed satisfactorily after the calibration and the parameters of both models are identified with considerable uncertainty.