When a boundary layer develops over a bed that is hydrodynamically rough at a length scale or scales larger than the grain size (a macrorough bed), as is usually the case where bed forms are present, it is necessary to distinguish among total boundary shear stress and its components, form drag and spatially averaged skin friction. It is known that the mean-velocity field reflects the composite boundary shear stress. Above about one roughness height above the tops of the roughness elements, the velocity does not vary horizontally. Its vertical profile is semilogarithmic and scales with the total friction velocity u*t and total roughness length zot. This region is here called the integrated logarithmic layer (ILL). Below the ILL the velocity varies horizontally in response to the irregular boundary; this region is called the surface layer.In the first of two sets of experiments reported here, skin-friction measurements were made with an array of flush-mounted hot films at four points on the stoss slope of one of a field of two-dimensional immobile current ripples. Total boundary shear stress was also measured, as were mean-velocity profiles in the ILL and the surface layer. The ILL behaves as described above. Although surface-layer velocity profiles are semilogarithmic, their semilogarithmic slope is not proportional to the local skin-friction velocity, so they do not locally obey the law of the wall. Rather, the velocity field can be decomposed into a spatially averaged rotational component and a local inviscid perturbation. The measured skin-friction field is consistent with a simple model for sediment transport over the bed forms except near reattachment, where the fluctuating skin friction is important. The data are also consistent with the drag-partition theories of Einstein and Barbarossa (1952) and Engelund (1966). Normalized skin-friction spectra do not vary with streamwise position but do vary with Reynolds number; skin-friction probability density functions show significant increases in skewness and kurtosis near reattachment but do not vary strongly with Reynolds number.In the second set of experiments the skin-friction vector field was measured around isolated hemispheres, with model sedimentary tails one and four obstacle heights long and without tails. The measured skin-friction fields are not consistent with deposition along the obstacle-flow centerline downstream of reattachment, which occurs about two obstacle heights downstream of the trailing edge of the hemisphere. This applies for local bed-load erosion and deposition and for general deflation of the bed, and is not substantially altered by the presence of either tail. Measurements were also made of skin friction, total boundary shear stress and ILL velocity profiles over h,B-rough arrays of hemispheres with and without tails four roughness heights long, at two areal densities. The skin-friction field in the denser array is significantly distorted from that around an isolated element. The measured skin friction in both arrays is signif...