Large-scale experimental study of how gravel-bed rivers erode banks and reshape planform

River meandering is a fundamental process that shapes landscapes and poses risks to infrastructure, yet predicting its dynamics remains a challenge, especially in gravel-bed rivers where heterogeneous sediment mixtures complicate the underlying physics. This knowledge gap persists because traditional research approaches–field studies, small-scale laboratory experiments, and numerical models–face limitations in isolating the feedbacks between flow, sediment transport, and bank erosion at the necessary scale and resolution. This dissertation bridges these methodological gaps by leveraging a one of a kind, largescale experimental facility to deconstruct the mechanics of meander migration in gravel-bed rivers. The research employs a series of controlled experiments at the Outdoor Experimental River Facility (OERF), designed to achieve field-like hydraulic conditions. It progresses from foundational facility validation to an investigation of bulk morphodynamic response and, finally, to the underlying grain-scale drivers. The initial work defined the narrow operational window for inducing morphodynamic change, identifying rapid bed armouring and bank resistance as key constraints. Subsequent experiments in a sinuous channel under cyclic floods revealed that meander migration is governed by a dynamic, stage-dependent dialogue between inner-bank deposition (bar-push) and outer-bank erosion (bank-pull). A key finding is that brief flood peaks, though representing only 3% of the experiment’s duration, set the trajectory for planform change and were responsible for the majority of lateral expansion. Furthermore, the dominance of bar-push or bank-pull was found to be intensely context-dependent, with individual bends under identical forcing exhibiting divergent stability trajectories. The dissertation links these macroscopic patterns to their grain-scale causes. Using high-resolution imagery and a deep-learning model to map grain-size distributions, it demonstrates that the initial flood peak establishes a persistent, coarse gravel framework that imparts a strong channel memory. Meander dynamics are shown to be governed by a two-mode partitioning: a stable, immobile coarse skeleton that steers flow, and a highly mobile sand matrix that enables deposition. The sediment sorting pattern within a bend creates a self-reinforcing feedback loop that dictates its migration trajectory and stability. The findings provide a process-based explanation for river meandering, demonstrating that macroscopic patterns are an emergent property of grain-scale processes. This research advances our understanding by clarifying what limits and what drives meander migration at a scale that closely approximates natural conditions. The rich, high-resolution dataset offers a benchmark for validating next-generation numerical models, while the conclusions provide actionable insights for river management.