Sediment transport was studied in non-submerged overland flow over grass in a laboratory. Artificial turf (astro-turf) was used to simulate natural grass and no infiltration was allowed at this stage of the investigation. Experiments were conducted for different grass densities, flow rates, sediments inflows, and sediment types. It was observed that concentration of sediment in runoff decreases exponentially with the distance and reaches asymptotically a constant value. Measured sediment deposition was compered with the results calculated by the Kentucky model (tollner et al., 1976). The model over-predicted grossly the trapping efficiency of all particle fractions, but it is unreliable particularly for small particles. A new simplified relationship was established between particle fall number, Nf, and percentage of particles trapped in the grass. The relationship should be verified on natural grass before it is used in practice. Finally, infiltration of water and particles should be studied on natural turf, as well as, influence of grass blades bending, before the complex model of sediment behaviour in grass is established.