Comparison of Two Versions of SWAT Models in Predicting the Streamflow in the Xuanmiaoguan Reservoir Catchment

Abstract

Correct streamflow prediction is critical for determining the availability and efficiency of watershed spatial plans and water resource management. In the Xuanmiaoguan (XMG) Reservoir Catchment, two different versions of the Soil and Water Assessment Tool (SWAT) model are compared to discharge predictions. One version is the Topo-SWAT, in which the overland flow is generated by saturation excess (Dunne) runoff mechanism, while the other is driven by infiltration excess runoff mechanism, i.e., the Regular-SWAT. These SWAT models were calibrated and validated with discharge at daily and monthly steps, and then, the annual runoff volume and spatial distribution of runoff generation areas were also discussed. At the monthly scale, the un-calibrated Topo-SWAT model outperformed the un-calibrated Regular-SWAT model throughout the whole time (2010-2016). The Nash-Sutcliffe efficiency coefficients (NSE) using Topo-SWAT and Regular-SWAT were 0.59, 0.58 for calibration and 0.69, 0.72 for validation for daily streamflow, and 0.69, 0.65 for calibration and 0.73, 0.88 for validation for monthly streamflow, respectively, based on the parameter sensitivity analysis results. There was a 5-year understatement for yearly runoff volume using Regular-SWAT, but a 4-year underestimation using Topo-SWAT, which had a different year in 2015. Regular-SWAT and Topo-SWAT have significantly different geographical distributions of runoff generating locations within the watershed for one occurrence (greater rainfall). The findings reveal the most accurate contributing regions for runoff generation in the research catchment, allowing for more effective implementation of best management techniques (BMPs).

Keywords

• swat model, streamflow, saturation excess runoff, runoff generation areas