Adapting CASCADE for braided rivers: A 1.5D sediment transport approach with variable substrate and width

Braided rivers are dynamic systems that support diverse habitats associated with their shifting mosaic of anabranches, backwaters, bars and islands. These units are characterized by chaotic, but not random, distributions of substrate, elevation and hydraulics at multiple scales. Modelling sediment supply or flow-driven changes in riverbed composition is notoriously difficult given the inherent dynamism and multiple scales defining large braided rivers.

In this research, we present a new data-rich modelling framework which combines census-scale [1 m] substrate classification with detailed 2D hydraulic models. The resulting transport capacity estimates can, for a static bed, be quickly applied to any transient flow scenario while retaining spatial detail. We then use the information gathered during the 2D modelling to parameterise a 1.5-dimensional transport solution in the time-evolving CASCADE sediment routing framework.

The 2D model uses a substrate map of a 56-km reach of the Rangitata [Rakitata] River, Aotearoa New Zealand, derived by machine learning based on high-fidelity helicopter lidar and orthophotography. A library of 2D steady-state hydraulic models is then run over the substrate map to predict the spatial and temporal capacity for sediment transport.

The adapted 1.5D-CASCADE model captures a defining feature of braided rivers, width variability, with a reach-specific hypsometric solver, tested against the 2D results, that predicts flow and sediment transport. The half-dimension is width, discretised against height using the 2D predictions of inundation and active area. The 1.5D model can then evolve bed composition both laterally across the braidplain and longitudinally down the river, within hydraulic geometry set by the template survey, including storage and remobilisation in side channels and floodplains.

The models are tested and applied to simulate the effects of flow regulation on bed composition in the Rangitata River. The model’s longitudinal consistency was only possible when using the spatial substrate data, and predictions are corroborated by lidar change detection. Results demonstrate that subtle changes in flow regime can alter where sediment is stored across the braidplain, with sedimentation impacts focused on side channels. Transfer of the model to other rivers indicates that width-varying solvers produce more stable sediment routing predictions than any single width, while remaining computationally efficient.