Self-formed braid bars in a numerical model

Braided rivers have highly variable bar dimensions, bar shapes and bar dynamics, as can be seen in the Brahmaputra-Jamuna in Bangladesh and the Waimakariri in New Zealand. Our objective is to understand the necessary conditions for predicting detailed braid bar morphology and dynamics, and to determine the effects of physical and non-physical numerical settings. We used the two-dimensional morphological model Delft3D to produce braid bar patterns. This model solves quasi-3D flow including near-bed secondary flow due to streamline curvature, sediment transport including effects of transverse slope, and mass conservation of sediment. We consider our physics-based nonlinear numerical model complementary to field studies and experiments. We search for the simplest possible initial situation and least amount of boundary conditions that reproduce braid bars, in order to identify what physics and boundary conditions are required. We used different values for grid resolution, transverse bed slope effect and perturbation, and different sediment transport formulas. Also, we identified the difference in braid bar shape and dynamics between 2D depth averaged flow and 3D flow in a sigma grid. Model runs were started from a plane bed with dimensions and constant boundary conditions based on an empirical channel pattern stability diagram. A small random perturbation is added to the upstream discharge partitioning across the inflow boundary (1%) and on the initial bed level (5 mm). Our results show the reproduction of the essential features of real braid bars including planform shape, length/width ratio, interaction between compound and unit bars, and bartails (seen as 'wings' on both downstream sides). The conditions are sufficient to produce realistic bar patterns. Regardless of the settings, initially, relative small sized bars with a high mode (mode 7-8) develop in the entire reach, which advect out of the model. At the same time, relatively high bars (80% of the mean waterdepth) are initiated upstream, migrating downstream and gradually occupying the entire reach, whilst bars grow and mode reduces. Eventually a braid bar pattern with steady statistics is reached. The results show that a 2D or 3D grid affects how dominating the upstream formed bars are on the final bar pattern. Moreover, the braid bars remain more dynamic in the 3D computations. The grid resolution has a major effect on the detailed bar shapes and dynamics but not on the general dimensions. Furthermore, the transverse bed slope strongly affects the bar height. The formative time scale of bars is strongly related to the nonlinearity of the sediment transport equation. The discharge and bed level perturbations have minor effects. The braid bar patterns as well as channel dimensions and other statistics are in good agreement with natural data. We conclude that self-formed braided bar patterns in our numerical model are realistically reproduced by using the simplest possible initial and boundary conditions. The formation of a braid bar pattern in general strongly depends on the width-depth ratio and the detailed braid bar shape and dynamics are affected by the model settings.