Evaluation of the soil and water assessment tool (SWAT) for simulating E. coli concentrations at the watershed-scale

Water quality management at the watershed level requires a framework to identify sources, apportion water quality risks and develop mitigation strategies to reduce health risks. Watershed-scale models have been used as a support tool to understand the sources, fate and transport of fecal bacteria and pathogens in the environment. The Soil and Water Assessment Tool (SWAT) model was applied in this study to understand the sources and drivers of microbial water quality in the Clouds Creek watershed in Georgia, USA. A criterion to evaluate the performance of the SWAT bacterial model was also developed in this study using the Nash-Sutcliffe Efficiency (NSE) performance measure. The SWAT model was successfully calibrated and validated for flow with Nash-Sutcliffe Efficiency (NSE) of 0.81 and 0.55, respectively. Escherichia coli (E. coli) predictions were good with NSE of 0.32 and 0.34 for the calibration and validation timeframes, respectively. Based on the criteria developed in this study, SWAT bacterial model for E. coli and fecal coliform can be judged as “satisfactory” when NSE > 0.20. The contribution of sources followed this order: in-stream cattle manure deposition > cattle manure application > poultry manure application > septic systems > wildlife manure, suggesting that a reduction in livestock access to streams would be the most effective approach to reduce fecal bacterial loads in this watershed and others impacted by fecal contamination. Finally, our results suggest that the SWAT model is capable of simulating E. coli dynamics in the Clouds Creek watershed and can provide insights into source impacts for risk management. [Display omitted] •Nash-Sutcliffe Efficiency performance criteria were developed for bacterial models.•E. coli predictions were satisfactory with Nash-Sutcliffe Efficiency ≥ 0.32.•In-stream manure deposition was the most significant source of fecal bacteria.•Bacteria release from manure and temperature were the most sensitive parameters.•Flow predictions were satisfactory with average Nash-Sutcliffe Efficiency = 0.68.