What controls bank-erosion-driven lateral river migration? Insights from a global synthesis

Rivers migrate across floodplains, which cover roughly 10% of Earth’s continental surface, through either catastrophic avulsions or gradual bank erosion. These modes of movement operate on markedly different timescales, ranging from days to weeks in the case of avulsion to decades or centuries for bank erosion-driven migration. Lateral river channel mobility poses significant socioeconomic risks, while also playing a key role in regulating global biogeochemical cycles through sediment deposition and re-mobilisation.
While river avulsions are comparatively well understood, the specific controls on lateral migration through gradual bank erosion are not yet fully resolved. Previous studies have explored individual factors influencing lateral migration rates, including the roles of water discharge, sediment supply, channel planform curvature, bank height, riverbank vegetation, and bank erodibility.
However, the combined and interacting effects of such factors on lateral migration rates have not yet been systematically evaluated. Furthermore, limited work has assessed how these controls may vary when migration rates are measured across multiple spatial scales (e.g., individual meander bends versus reach-length averages), observation periods, and river planform types (i.e., single-thread versus braided systems).
We curate and analyse a global compilation of lateral migration rates comprising approximately 500 measurements from 300 rivers and streams reported across 88 published English-language studies. For each measurement site, we independently quantify the previously mentioned (potential) predictor variables and key morphometric parameters, including channel gradient, upstream drainage area, active channel width, multi-decadal lateral migration extent, and channel belt area.
We apply principal component analysis and multivariate regression to quantify the relative importance of these predictors, identify a minimal set of dominant controls, and derive a predictive relationship linking migration rates to their governing parameters.
Preliminary analyses show robust power law relationships between bank erosion rate and both active channel width and channel belt width (derived from Sentinel-2 optical imagery), indicating that channel-scale geometry exerts a first-order control on lateral migration rates. Normalizing bank erosion rates by channel width is, therefore, necessary to isolate and evaluate secondary controls and reveal additional trends in the data. In contrast, we find no significant relationship between bank erosion rate (either raw or normalized by channel width) and a vegetation metric, such as canopy height (derived from Sentinel-2 and GEDI spaceborne LiDAR data), across the compiled dataset. This lack of correlation is consistent across all river planform types. Despite the common assumption that bank erosion increases with planform curvature, reach-averaged sinuosity shows no systematic relationship with either raw or channel-width-normalized bank erosion rates. This suggests that bank erosion may be more closely linked to local curvature than to reach-scale planform geometry.
Further inferences would provide a basis for predicting lateral migration rates under changing climate conditions and can be integrated into existing numerical landscape evolution models. Because such models rarely incorporate lateral river migration and therefore often fail to reproduce the wide river valleys observed in nature, our results offer a means to enhance their ability to simulate realistic patterns of long-term (10³-10⁴ years) fluvial widening and floodplain development.