Concerns for microbial safety of surface water facilitate development of predictive models that estimate concentrations and total numbers of pathogen and indicator organisms leaving manure-fertilized fields in overland flow during runoff events. Spatial variability of bacterial concentrations in applied manure introduces high uncertainty in the model predictions. The objective of this work was to evaluate the uncertainty in model predictions of the manure-borne bacteria overland transport caused by limited information on the spatial distribution of bacteria in surface-applied manure. Experiments were carried out at the ARS Beltsville experimental watershed site (OPE3) in Maryland. Dairy bovine manure was applied at a 59·3 t/ha rate on the 3·55 hectare experimental field. Faecal coliform (FC) concentrations in manure measured in 2004, 2005, 2007, and 2009 varied by 4 orders of magnitude each year. Both runoff volume and FC concentrations in runoff water were monitored using a runoff flume equipped with a refrigerated pump sampler. Two runoff events occurred before the manure was incorporated into the soil. A bacteria transport add-on module simulator of transport with infiltration and runoff (STWIR) was linked with the event-based kinematic runoff and erosion model (KINEROS2) to simulate convective-dispersive overland transport, bacteria release from manure, reversible attachment-detachment to soil, and surface straining of infiltrating bacteria. The model was successfully calibrated with the field experiment data. Monte Carlo simulations were carried out to account for the spatial variation in FC in applied manure and uncertainty in the FC distribution in manure caused by the small number of samples. A tenfold and twofold variation in FC concentrations in the runoff were obtained within the 90% probability interval when initial FC spatial distributions in the manure were represented by 5 and 29 samples, respectively.