Alluvial open channel beds often exhibit statistically periodic irregularities, known as dunes. Dunes have considerable effects on sediment transport and flow resistance. When growing during a flood, the dunes create more resistance and flood levels may rise significantly. Accurate prediction of dune properties therefore contributes to effective flood risk management. Recently, significant progress has been made in understanding bedform dynamics, thanks to significant advances in monitoring flow and bedform morphology in laboratory and field, as well as in their numerical modelling. Nowadays, numerical modelling captures not only the characteristics of the mean flow field, but also those of turbulence, including coherent flow structures above non-flat beds. These advances enable radical progress in modelling and understanding the behaviour of alluvial bedforms. Sediment motion can nowadays be measured and simulated in detail. The effect of turbulent flow on sediment particles can be understood in a more physics-based way. Sediment transport can be linked to bed topography and the bed deformations can be studied by considering the motion of sediment particles in the flow. This gives insight into the evolution and migration of ripples and dunes under turbulent flow and the effects they have on floods. Generation and migration of dunes are determined by sediment transport, which in its turn is influenced by the turbulence structures above and behind the dunes. This kind of physical phenomena involves a wide range of time scales. Often, the measurements cannot capture the smallest time scales, whence many experimental studies consider only statistical properties under homogenous steady-state conditions. Consequently, models based on such measurements can only be empirical. They usually fall short in computing the physical nature of the phenomena in unsteady flows. Therefore, a rigorous physics-based numerical model is needed. This study concentrates on detailed simulations of flow, sediment transport and bedform morphodynamics. Based on these simulations the governing physics behind these phenomena are studied. This is achieved by developing a detailed three-dimensional numerical model for hydrodynamics, sediment transport and morphodynamics. The model simulates the time- dependent water flow by Large Eddy Simulation (LES) on a locally refined Cartesian grid. The sediment is considered as rigid spherical particles moving in the water under gravity and flow-induced forces. The change of bed (morphodynamics) is the net result of pick-up and deposition of sediment on each portion of the bed. The model is validated against theoretical and experimental results of previous studies published in the literature. The resulting model is complex and time-consuming, especially in time-varying flow conditions, such as a flood wave. Therefore, the insights and data obtained with it were used to develop parametric models that can be used operationally at larger spatial and temporal scales. In a number of three-dimensional simulations over two-dimensional bedforms the form drag resulting from the bedforms is compared with existing theoretical, empirical and semi-empirical formulae. It is found that the numerical results agree very well with these formulae. Bedforms in nature, however, are usually three-dimensional. When comparing simulated flows over three-dimensional dunes and with the two-dimensional case, the form drag on three-dimensional dunes turns out to be very different. Based on this finding, the form drag is parameterized for two- and three-dimensional dunes. Furthermore, the generation and migration of dunes under steady flow conditions is studied, and the results are compared with former experimental studies. This comparison shows a very good agreement between the numerical and experimental findings. Dune evolution during floods often shows a hysteresis, with different dune heights at the same discharge during the rising and falling stage of the flood wave. As flow resistance in the main channel of a river is mainly controlled by dune dimensions, the hysteresis in dune height is reflected in the time-evolution of the flow resistance during floods, thus yielding a dynamic roughness. Limited knowledge on this phenomenon, combined with computational limitations, usually keep dynamic roughness behaviour from being included in flood simulation models. To understand the physics behind this hysteresis effect, channels with different discharges and different grain sizes are simulated. It is shown that the hysteresis in the form drag is a function of both variables. Extension of the simulations to the upper flat bed regime shows that the model captures most of the physical phenomena in this regime and yields a flat bed as observed in experiments and in the field.