A variety of algorithms are available for parameterizing the hydrodynamic bottom roughness associated with grain size, saltation, bedforms, and wave–current interaction in coastal ocean models. These parameterizations give rise to spatially and temporally variable bottom-drag coefficients that ostensibly provide better representations of physical processes than uniform and constant coefficients. However, few studies have been performed to determine whether improved representation of these variable bottom roughness components translates into measurable improvements in model skill. We test the hypothesis that improved representation of variable bottom roughness improves performance with respect to near-bed circulation, bottom stresses, or turbulence dissipation. The inner shelf south of Martha’s Vineyard, Massachusetts, is the site of sorted grain-size features which exhibit sharp alongshore variations in grain size and ripple geometry over gentle bathymetric relief; this area provides a suitable testing ground for roughness parameterizations. We first establish the skill of a nested regional model for currents, waves, stresses, and turbulent quantities using a uniform and constant roughness; we then gauge model skill with various parameterization of roughness, which account for the influence of the wave-boundary layer, grain size, saltation, and rippled bedforms. We find that commonly used representations of ripple-induced roughness, when combined with a wave–current interaction routine, do not significantly improve skill for circulation, and significantly decrease skill with respect to stresses and turbulence dissipation. Ripple orientation with respect to dominant currents and ripple shape may be responsible for complicating a straightforward estimate of the roughness contribution from ripples. In addition, sediment-induced stratification may be responsible for lower stresses than predicted by the wave–current interaction model.