Empirical and Experimental Validation of Channel Dynamics Models

Significant advances in understanding the morphodynamics of alluvial rivers have been achieved through a combination of field monitoring, laboratory modelling and morphometric analysis. While each has shed significant insight, all have limitations: monitoring is typically localized and short-term; scaling difficult to achieve in the laboratory; and morphometric studies confounded by the complexity of linking form and process. Coupled numerical modelling of flow and sediment transport is seen as a way to relax these constraints and provide tools for investigation the evolution of alluvial systems. However, the complexity of solving mutually adjusting flow and topographic fields severely restricts the application of conventional hydraulic models at appropriate reach and decadal space and timescales. Reduced complexity approaches, such as cellular automaton (CA) and raster modelling offer alternative frameworks to explore interactions at these scales and combine computational efficiency with simplified physical theory. While such models appear capable of qualitatively reproducing complex alluvial morphologies, confidence in their predictions requires further robust model testing. This research presents an analysis of the hydraulic engines of popular CA approaches applied to braided rivers. In these, sediment flux is driven by simple raster flow routing algorithms that predict the distribution of discharge by apportioning flow between cells on basis of local elevation differences. Variants on this theme incorporate additional sophistication by allowing flow routing at high angles and applying uniform flow formulae to predict pathways on the basis of water surface slope. These schemes are applied to a high-resolution terrain and distributed flow dataset for the braided river Feshie, surveyed in July 2003 and 2004. Observations of the distributed pattern of flow depth, cross-sectional flow and allocation between anabranches are used to test predictions and Monte Carlo simulation used to assess sensitivity to terrain model errors. CA frameworks are found to perform well, with schemes based on water slope superior those driven directly by the topography. All models were, however, found to be highly sensitive to minor terrain model errors, limiting their application to low precision airborne datasets