Uncertainty Analysis as a First Step of Developing a Risk-Based Approach to Nonpoint Source Modeling of Fecal Coliform Pollution for Total Maximum Daily Load Estimates

Salado Creek in Bexar County, Texas is one of 65 streams listed as impaired water bodies for its high concentration of fecal coliform bacteria in the Clean Water Act’s 303(d) list. The Hydrological Simulation Program-FORTRAN (HSPF) available in the Environmental Protection Agency’s (EPA) Better Assessment Science Integrating point and Non-point Sources (BASINS) computer model was applied to the Salado Creek watershed for studying its applicability as a prediction tool for in-stream fecal coliform bacterial concentration from both point and nonpoint sources associated with different types of landuses in the watershed. In addition, the sensitivity of simulated peak values of in-stream fecal coliform concentrations to changes in parameters associated with the bacterial simulation was evaluated. The hydrology of the watershed was calibrated for a period from 1990 January 1 to 1993 December 31. The model was validated for hydrology for the year of 1995. The simulated peak value of in-stream fecal coliform concentrations was found to be most sensitive to parameters that represent the maximum storage of fecal coliform on the pervious land surface and surface runoff that removes 90 percent of fecal coliform from the pervious land surface. In-stream fecal coliform concentrations were also sensitive to stream water temperature, first-order decay rate of fecal coliform and a temperature correction coefficient for the first order decay rate. A First Order Analysis (FOA) was conducted to determine the fraction of the variance of the simulated peak in-stream fecal coliform concentration due to the uncertainty in these most sensitive parameters. The result of the FOA showed that the major portion of the variance in simulated in-stream peak fecal coliform concentration was attributed to the maximum storage of fecal coliform on the pervious land surface. Thus, the current study emphasizes the fact that small errors in parameterizing the maximum storage of fecal coliform over a given landuse class may result in large errors in predicted coliform counts