Testing forecasts for meandering river migration using a rescued dataset of rivers across the continental US

Abstract:

Many alluvial meandering rivers migrate laterally as they erode their banks and deposit sediment along point bars. Quantitative predictions for river migration are critical for forecasting sediment transport, landscape evolution, and the stability of engineered structures. Yet this task poses significant challenges, as the spatial distribution of migration varies along individual rivers and between rivers of different scales. The migration process is described by a number of models that have been applied to hindcast individual meander trajectories in limited cases. However, we currently lack comprehensive frameworks to test meandering models, which limits our ability to forecast future migration. To address this gap, we evaluated the performance of a simple and commonly used model that predicts bank migration along the channel using a weighted aggregate of its local and upstream curvatures. To test the model, we used bank position observations previously compiled by the US National Cooperative Highway Research Program (NCHRP). This rich dataset, comprising 1,503 meander bends on 89 migrating rivers in the U.S. over six decades, has posed challenges for analysis due to proprietary data formats and incomplete documentation. We improved the usability of the NCHRP dataset by reconstructing missing information on spatial referencing, observation time, and the geometry of mapped river segments; and by converting channel bank coordinates from their legacy format. After this processing, 81% of the original dataset was determined to be useable for analysis. The resulting dataset captures a range of channel migration rates and geometries. We will present results of comparisons between the predicted and actual bank positions, focusing on relationships between channel geometry and model misfit. The revised NCHRP dataset will provide a unique complement to existing datasets for more rapidly migrating rivers by including dozens of examples of rivers in varied geomorphic settings across the continental US. We will describe ongoing work to further integrate these data with numerical models to improve forecasts for meandering river migration.