Abstract

Despite the importance of bank erosion in rivers, most computational fluid dynamics (CFD) models have limited capacity to examine bank retreat and channel-floodplain interactions, as they lack bank stability algorithms and ignore vegetation effects. This research seeks to develop a numerical model to improve our understanding of key properties of bank material and vegetation cover with respect to lateral erosion in river meanders at intermediate spatial (5−10 meander bends) and temporal (2−3 years) scales. Following a comparison of six different morphodynamic models for three sinuous laboratory configurations, the CFD model TELEMAC-2D was chosen to receive a newly developed bank retreat module that respects geotechnical principles and integrates spatial analysis concepts. It was tested against morphological datasets from two contrasted river reaches, the semi-alluvial Medway Creek (Ontario) and alluvial St. François River (Quebec). Statistical analysis, combined with the use of machine learning algorithms, demonstrate that the coupled model is able to fit observed bank retreat location and extent. Some local disagreement with observations along Medway Creek seems associated with the heterogeneity of soil material and stratigraphy, and in vegetation cover present at the field site. The coupled model was also used to identify key geotechnical parameters and optimal parameter values for the studied reaches. An epistemological reflection on the purpose of modelling in fluvial geomorphology leads to the conclusion that the primary model strength lies in its ability to provide explanations on bank retreat mechanisms. Further research should seek to test more thoroughly morphodynamic modelling in complex geomorphological environments.