The application of computational fluid dynamics to natural river channels: three-dimensional versus two-dimensional approaches

This paper addresses the recent trend in fluvial geomorphology towards using computational fluid dynamics to explore the adjustment between flow processes, sediment transport and river channel morphology. It aims to evaluate the extent to which three-dimensional (3D) models improve predictive ability and prediction utility compared to two-dimensional (2D) applications. This is achieved through comparing the predictions of both 3D and 2D models with high-quality field data. Identical boundary conditions, obtained from a confluence within a gravel-bed river system with high relative roughness, are defined for each model. Evaluation of the 3D model suggests that there is a fundamental limitation upon model predictive ability due to problems of specifying topographic complexity. However, comparison with the 2D model shows that the 3D model has a higher predictive ability, particularly if the 2D model is not corrected for the effects on flow structure of secondary circulation. Further, the 3D model provides more reliable estimates of bed shear stress and other more useful information, such as the three-dimensional flow field important for mixing processes. This suggests that there is significant merit in the move towards 3D models, but that research is required to incorporate methods developed in other fields for dealing with boundary condition uncertainties.