Microbial contamination of surface waters frequently occurs on permanent natural grasslands subject to extensive grazing. Management of these problems requires developing methods to identify critical source areas that are responsible of significant losses of fecal microorganisms. In this study, GIS analysis of watersheds was used to calculate the flow of fecal bacteria (Escherichia coli) to the outflow of a watershed by summing bacterial flows in runoff from pixels containing cowpats. Calculations were performed in two steps: (i) identification of pixels with bacteria and runoff by modeling the distribution of cowpats and variable sources of surface runoff, and (ii) parameterization by inverse analysis of deterministic and stochastic functions for bacterial emission from cowpats and for retention during their transmission to the watershed outflow. During bacterial transport in water flow, bacterial retention on the soil surface has a large influence. Despite this effect, bacterial concentration in runoff remains high. In general, cowpat age, runoff volumes and the location and proportions of bacteria-emitting and non-emitting surfaces determine critical source areas and bacterial f