SWAT Model Prediction of Phosphorus Loading in a South Carolina Karst Watershed with a Downstream Embayment
The SWAT model was used to predict total phosphorus (TP) loadings for a 1555-ha karst watershed—Chapel Branch Creek (CBC)—which drains to a lake via a reservoir-like embayment (R-E). The model was first tested for monthly streamflow predictions from tributaries draining three potential source areas as well as the downstream R-E, followed by TP loadings using data collected March 2007-October 2009. Source areas included 1) a golf course that received applied wastewater, 2) urban areas, highway, and some agricultural lands, and 3) a cave spring draining a second golf course along with agricultural and forested areas, including a substantial contribution of subsurface water via karst connectivity. SWAT predictions of mean monthly TP loadings at the first two source outlets were deemed reasonable. However, the predictions at the cave spring outlet were somewhat poorer, likely due to diffuse variable groundwater flow from an unknown drainage area larger than the actual surface watershed, for which monthly subsurface flow was represented as a point source during simulations. Further testing of the SWAT model to predict monthly TP loadings at the R-E, modeled as a completely mixed system, resulted in their over-predictions most of the months, except when high lake water levels occurred. The mean monthly and annual flows were calibrated to acceptable limits with the exception of flow over-prediction when lake levels were low and surface water from tributaries disappeared into karst connections. The discrepancy in TP load predictions was attributed primarily to the use of limited monthly TP data collected during baseflow in the embayment. However, for the 22-month period, over-prediction of mean monthly TP load (34.6 kg/mo) by 13% compared to measured load (30.6 kg/mo) in the embayment was deemed acceptable. Simulated results showed a 42% reduction in TP load due to settling in the embayment.