Mean velocity and turbulence measurements are made in a shear layer generated in a short test-section wind tunnel. This shear layer is developed by the spire-roughness technique and simulates as closely as possible the lower part of the neutral atmospheric surface layer for the study of pressures on 1:70 scale models of low-rise buildings. In Ref. [1] and in a companion paper [2], to be published at a later date, the statistical quantities describing the fluctuating pressures acting on a low-rise building in model and in full scale are compared. Comparison of the generated shear flow with measurements obtained in the atmospheric surface layer indicates that simultaneous simulation of mean-velocity power index, surface-roughness length, turbulence intensity and turbulence integral scale is extremely difficult to achieve. On the basis of results obtained to date, it is tentatively concluded that close simulation of the turbulence intensity and development of a turbulence integral scale at least as large as the largest model dimension are required for proper simulation of the fluctuating wind pressures. In order to satisfy these requirements, the scaling of the upstream roughness elements must be exaggerated.