Hillslope vegetated buffers are recommended to prevent water pollution from agricultural runoff. However, models to predict the efficacy of different grass buffer designs are lacking. The objective of this work was to develop and test a mechanistic model of coupled surface and subsurface flow and transport of bacteria and a conservative tracer on hillslopes. The testing should indicate what level of complexity and observation density might be needed to capture essential processes in the model. We combined the three-dimensional FEMWATER model of saturated-unsaturated subsurface flow with the Saint-Venant model for runoff. The model was tested with data on rainfall-induced fecal coliforms (FC) and bromide (Br) transport from manure applied at vegetated and bare 6-m long plots. The calibration of water retention parameters was unnecessary, and the same manure release parameters could be used both for simulations of Br and FC. Surface straining rates were similar for Br and bacteria. Simulations of Br and FC concentrations were least successful for the funnels closest to the source. This could be related to the finger-like flow of the manure from the strip along the bare slopes, to the transport of Br and FC with manure colloids that became strained at the grass slope, and to the presence of micro-ponds at the grassed slope. The two-dimensional model abstraction of the actual 3D transport worked well for flux-averaged concentrations. The model developed in this work is suitable to simulate surface and subsurface transport of agricultural contaminants on hillslopes and to evaluate efficiency of grass strip buffers, especially when lateral subsurface flow is important.