Predicting stream pathogen loading from livestock using a geographical information system-based delivery model

Recent cases of drinking water contamination by pathogens have underscored the importance of preventing livestock waste from entering surface waters. To this end, analytical tools are needed that can identify subwatersheds or livestock operations that contribute disproportionately to contamination. This paper presents a geographical information system (GIS)-based transport model (SEDMOD) that provides an index of pathogen loading potential to streams by characterizing five key transport parameters: flow-path hydraulic roughness, gradient, and slope shape, stream proximity, acid a normalized soil moisture index. We applied SEDMOD to 12 subwatersheds (10 agricultural, 2 forested controls) of the Saw Kill, a tributary of the Hudson River, New York, and compared model predictions with measured fecal coliform (FC) levels. The transport model, combined with a livestock density GIS layer, could explain 50% of the variation in average FC discharge among the subwatersheds (r = 0.71, P = 0.01, n = 12), By contrast, neither total livestock FC output nor predicted FC transport were correlated with geometric mean FC concentration (P > 0.05). In a multiple regression, predicted FC transport, mean water temperature, and mean turbidity could account for 80% of the observed variation in FC discharge (r = 0.90, P = 0.001, n = 12). We conclude that, although more field work and algorithm development is needed to yield more accurate quantitative predictions, the model Is useful for predicting the relative contribution of diverse livestock operations within a varied landscape. This provides watershed managers and regulators with a rating method to prioritize sites for nonpoint source pollution control.