Hyperconcentrated flow/flood is a water-driven sediment transport phenomenon, which is characterized by high sediment concentrations between normal sediment-laden flow and debris/mud flow. In a hyperconcentrated flow, strong interactions exist between water flow, sediment, and river bed, which may not only change the flow rheological properties, but also affect the characteristics of sediment transport and bed deformation. In the lower reach of the Yellow River (China), hyperconcentrated flows/floods are frequently observed (with sediment concentrations > 200 kg/m3) and primarily Newtonian, turbulent flows (van Maren et al., 2009a). This reach is characterized by two special, and possibly unique, hydrodynamic and sediment transport phenomena that are associated with hyperconcentrated flow: a downstream increasing peak discharge (at a rate far exceeding the contribution from tributaries) during hyperconcentrated floods, and a downstream decreasing runoff due to water diversions. Assuming spatially continuous diversions along a constant-width channel, previous studies suggest a longitudinally convex bed at the equilibrium state. However, the validity of a convex bed profile, for discrete diversions in natural channels of longitudinally varying width, remains to be justified. Also, such equilibrium analysis does not reveal the morphological time scale (MTS) associated with water diversions. Moreover, though many explanations have previously been proposed for the peak discharge increase, they have focused on only one possible mechanism (e.g., bed roughness change, bed erosion, floodplain influences) and no consensus has been achieved. The underlying physics still remain largely unknown. Research efforts are therefore needed to further investigate these two issues, which comprise the main work of the present PhD research and this thesis. For the study of the downstream peak discharge increase phenomenon, mathematical modelling is the main research method, together with the field data analysis. High-resolution morphodynamic modelling of complex fluvial processes, such as in a hyperconcentrated flood, has so far been limited by model accuracy or computational efficiency. In order to account for the strong interactions during the hyperconcentrated flood and to acquire accurate and efficient solutions in the field scale, a fully coupled morphodynamic model has first been developed using the finite volume method for structured grids. Physically, this model is based on the concept of non-capacity sediment transport, and it incorporates the effects of sediment density and bed deformation on the flow (both in mass and momentum), as well as the influences of turbulence and sediment diffusions. Numerically, this model combines the high accuracy of high-order upwind schemes and the efficiency of centered schemes by the extension of a recent upwind-biased centered (UFORCE) scheme (Stecca et al., 2010) originally developed for clear flow and scalar transport over a fixed bed, to sediment-laden flows over an erodible bed. For stability, a two-stage splitting approach together with a second order Runge-Kutta method is used for the source terms. Moreover, the full set of governing equations is solved at one time to obtain synchronous solutions in mathematics. The model is verified in a number of dam-break tests, covering a wide range of complex (sediment-laden) flows. It is demonstrated to accurately simulate shock waves and reflection waves, as well as rapid bed deformations at high sediment transport rates. Using this model, the relative role of bed roughness change and bed erosion on the downstream peak discharge increase is then investigated in schematized 1-D channels for two hyperconcentrated floods. The results reveal that although erosion effects may contribute to the downstream discharge increase (especially in case of extreme erosion), for most cases the increase is mainly due to a reduction in bed roughness during peak discharge conditions. Additionally, based on the concept of channel storage reduction, the effects of decreasing bed roughness and (very strong) bed erosion can be integrated in the explanation of the peak discharge increase. Later, this model is also applied to reveal the floodplain influences on the peak discharge increase in schematized 2-D channel-floodplain reaches. The results indicate that the cross-sectional changes of channel erosion and floodplain deposition during hyperconcentrated floods are often limited and that it is difficult to drive a peak discharge increase in the downstream direction. For the study of the water diversion impact, a general theoretical framework is proposed to predict the equilibrium state of the fluvial system, which is applicable to both continuous and discrete water diversions in a longitudinally width-varying channel. Numerical experiments by the SOBEK-RE software (version 2.52.005, Delft Hydraulics, 2005) complement the MTS studies for water diversions. The effects of diversion intensity, diversion placement (discrete and continuous) and diversion schemes (pure water and water-sediment mixture) are also systematically studied. The present work confirms the previous findings that water diversions lead to a decrease of the equilibrium depth with respect to natural conditions and a convex bed in a constant-width channel. Moreover, it reveals that in a widening channel a convex bed also develops under conditions of water diversions, while convex, concave or quasi-linear beds may occur in a narrowing channel. Non-monotonic beds may develop in a strongly narrowing channel, depending on the diversion schemes. On a large spatial scale, diversion placement is less important for the equilibrium development. The MTS for water diversions and natural development are very similar and large, indicating considerable influences of water diversions on river morphology. The present thesis advances our understanding of the long-term impact of water diversions on the evolution of a river.