Uncertainty of Predicted In-stream Fecal Coliform Concentration for TMDL Development using HSPF: a Two-Phase Monte Carlo Approach

Abstract

Water quality modeling is increasingly used to develop watershed management plans and Total Maximum Daily Loads (TMDLs). Although there is an inherent uncertainty present in water quality modeling, limited scientific guidance is available related to estimation of uncertainty in TMDLs. A two-phase Monte Carlo (MC) approach was used to estimate uncertainty in in-stream fecal coliform (FC) bacteria concentration attributable to both knowledge uncertainty and stochastic variability. In-stream FC bacteria concentrations were simulated using a Hydrologic Simulation Program – FORTRAN (HSPF) model developed for the Mossy Creek watershed in Virginia. The two-phase MC simulation was conducted for two different pollutant allocation scenarios suggested in the Mossy Creek bacteria TMDL. The allocation scenario requiring 99% reduction in cattle direct-deposit indicated similar contribution of knowledge uncertainty and stochastic variability, while the allocation scenario requiring 94% reduction in cattle direct deposit indicated higher contribution of knowledge uncertainty than stochastic variability. Cattle direct-deposit was a significant source of knowledge uncertainty in the prediction of in-stream FC concentrations. Overall, the allocation scenario requiring 99% reduction in cattle direct-deposit violated the single-sample FC criterion (400 cfu/100ml) 7 times (days), out of 1218 days with a 95% confidence interval of 0 to 16 days. The allocation scenario requiring a 94% reduction in cattle direct-deposit violated the single-sample FC criterion 12 times (days) and produced a 95% confidence interval of 3 to 51 days. This information can be used by decision makers and stakeholders to choose their level of confidence in achieving a particular water quality standard and the associated level of needed pollutant reduction to achieve that confidence level.