This study investigates turbulence models for application to boundary layer flows. Firstly, steady channel flow and transient pipe flows are considered. Calculations of a low-Reynolds-number k-epsilon model, a k-epsilon-S model (a strain parameter model which has not been applied to unsteady flows previously) and a Reynolds Stress Transport model are compared with experimental and DNS data. The eddy viscosity turbulence models (k-epsilon, k-epsilon-S) satisfactorily predict the mean flow parameters of steady channel flow. However the k-epsilon-S model proves superior in comparison with turbulence quantities. Near to the pipe wall, the k-epsilon-S model best captures the details of periodic pipe flow detail, whereas in the outer flow region the RSTM gives closest agreement with the experimental data. The high-Reynolds-number k-epsilon and k-l eddy viscosity turbulence models are examined in a separate study of oscillatory flows over smooth and rough beds. The computations are considered over a wider range of experimental parameters than previously investigated. The turbulence models are assessed by comparison with field measurements and laboratory data sets including a new set of experimental measurements. Both models predict the bed shear stress and velocity adequately, but the k-epsilon model emerges as the superior scheme when considering turbulence quantities. An attempt is made to quantify the uncertainty in the Reynolds shear stress and eddy viscosity experimental data. The k-epsilon model calculations more frequently lie within the experimental uncertainty bands. However this uncertainty range is wide; any improvement would require a corresponding improvement in the experimental resolution of rough bed flows.