Watershed simulation models are popularly used for developing Total Maximum Daily Loads (TMDL) for different water quality parameters. The TMDL modeling process typically involves four stages: hydrologic calibration, water quality calibration, TMDL development, and load reduction analysis. The success of the whole process initially depends on successful hydrology calibration. This paper presents the results of an iterative methodology that was developed for use in calibrating HSPF for a karst watershed. This study focused on the development of a pathogen TMDL for Cane Run Watershed in central Kentucky. The headwaters of the watershed originate within the urban services boundary of Lexington, Kentucky and then flow through a predominately rural area which contains several horse farms. Due to the area's karst topography, this basin contains several large sink holes that divert flow from the part of the upper watershed to an adjacent watershed where it emerges at Royal Spring which serves as a primary water supply for the adjacent city of Georgetown Kentucky. As a consequence, the streamflow in much of the upper part of the watershed is essentially zero for many months of the year. This phenomenon presented several challenges in trying to characterize the watershed loadings as well as how to model the system so as to develop an effective TMDL.