Previous studies of surface particle structures in gravel bed rivers, in particular coarse grain clusters, conclude that these bedforms increase bed stability and reduce sediment transport rates. However, few studies have quantitatively analysed the effect of the surface distribution of coarse grains, in isolation or arranged into distinct bedforms, on particle entrainment, focussing on descriptions of bedform formation and mechanical behaviour. A novel image analysis technique is described that allowed recognition of coarse grains and calculation of particle cluster areas. Under laboratory conditions, increasing discharges were applied to a unimodal sand-gravel bed. The size class containing the D90 size was painted in ultraviolet (UV) paint. Image analysis software was used to identify this size class on digital images of the bed surface taken under UV light. Both the total surface area occupied by UV grains and the area of accumulations of connected grains were related to the threshold shear stress for entrainment. As discharge magnitude increased, coarse grains actively grouped together forming fewer, larger accumulations of coarse particles. Lower values of critical bed shear stress occurred for beds with higher degrees of clustering; beds are more stable when coarse grains are more evenly distributed over the bed surface. This implies that using bed surface composition as a descriptor of bed resistance to entrainment may need to be complemented by a descriptor of the spatial distribution of the coarsest grains.