The natural variability of surface characteristics determines the location of high surface shear stress ('[tau]S') zones where wind erosion potential is enhanced; as a result, a clearer understanding of the spatial distribution of '[tau]S' on complex surfaces is critical for wind erosion modelling. This study examines the spatial variability of '[tau]S' on regular and complex arrays surfaces in a wind tunnel, using a dense network of Irwin sensors. Roughness arrays consisted of film canisters in three different spatial configurations and three roughness densities. Results suggest that roughness width and porosity influenced shelter area shape and size. Staggered arrays did not have expected flow regimes, and complex arrays had multiple flow regimes. 'R' values compared well with those found in previous studies; however the 'm' parameter was low, and significantly differed between roughness configurations. Finally, '[tau]S' was not spatially uniform, therefore ' [lambda]', which is a spatial average, does not effectively describe complex surfaces. Therefore a better way to describe surface cover is required for wind erosion modelling.