The parameterisation of surface roughness effects on aeolian sediment transport is a key source of uncertainty in wind erosion models. Roughness effects are typically represented by bulk drag-partitioning schemes that scale the threshold friction velocity (u*t) for soil entrainment by the ratio of shear stress on roughness elements to that on the vegetated land surface. However, the approach does not explicitly account for the effects of roughness configuration, which may be important for sediment flux. In this study we investigate (1) the significance of roughness configuration for aeolian sediment transport, (2) the ability of drag partitioning approaches to represent roughness configuration effects, and (3) the implications of our findings for model applications. We use wind tunnel measurements of surface shear stress distributions to calculate sediment flux for a suite of roughness configurations, roughness densities and wind velocities. Our results show that roughness configuration has a significant effect on sediment flux, influencing estimates by more than one order-of-magnitude. Measured and modelled drag partitioning approaches overestimate the predicted flux by two to three orders-of-magnitude. The drag partition is sensitive to roughness configuration, but current models cannot effectively represent this sensitivity. The effectiveness of drag partitioning approaches is also affected by estimates of the aerodynamic roughness height (z0) used to calculate wind shear velocity (u*). Unless the roughness height is consistent with the drag partition and u*t, resulting fluxes can show physically implausible patterns. These results should make us question current assessments of the magnitude of vegetated dryland dust emissions. Representing roughness effects on surface shear stress distributions will reduce uncertainty in quantifying wind erosion, enabling better assessment of its impacts and management solutions.