Impact of model structure on the accuracy of hydrological modeling of a Canadian Prairie watershed

Abstract

Study region: Prairie Pothole Region of Canada. Study focus: The Prairie region spans across approximately 870,000 km2 of the Great Plains region of Canada (80%) and the United States (20%). The presence of a large number of depressional wetlands (potholes) results in dynamic surface-water and stream connectivity during wet and dry year necessitating an improved understanding of watershed-scale interactions of the Prairie Potholes. The Soil Water Assessment Tool (SWAT) hydrological model with three structural variants is utilized to assess the degree of accuracy associated with increasing model complexity and its impact on the model calibration of the Upper Assiniboine River Basin at Kamsack. New hydrologic insight for the region: The SWAT model was calibrated and verified with three different structural arrangements in 1) lumped pothole, 2) semi discretized pothole, 3) and fully discretized pothole representation. The fully discretized pothole version of the SWAT reflected streamflow best (KGE of 0.78) but with greater uncertainty, larger data and computational resource requirements. The fully discretized (modified) model, however, was able to capture the high flow and the fill-and-spill processes, which is a defining characteristic of the Prairie Pothole Region (PPR). Significant improvements to the predictive ability of SWAT in the case of the modified model was observed, thus allowing an enhanced understanding of the aggregate effect of potholes in this watershed. Keywords: Prairie potholes, Hydrologic connectivity, Soil and Water Assessment Tool (SWAT), Multi-model comparison, Parameter uncertainty