Braided Rivers as Morphological States: A Data‐Driven Identification of Recurrent Planform Configurations

Abstract

Braided rivers reorganize their channels from time to time, featuring complex morphological dynamics and intermittent sediment transport that are difficult to describe using scalar metrics or single representative configurations. To better capture this complexity, we introduce a stochastic approach, which involves tracking the planform dynamics of channels and the associated bedload transport rates. In this approach, each state represents a given planform channel geometry, and the system can randomly transition from one state to another over time. We conducted a 1‐month laboratory experiment. We identified morphological states from plan view images using an automated workflow combining several techniques (image comparison metrics, dimensionality reduction, and density clustering). The sequence of states revealed long periods of morphological stability, separated by short periods of channel reorganization. Bed morphology also showed great diversity, ranging from narrow, single‐thread channels to large, braiding networks prone to sediment storage. While earlier studies have revealed the chaotic nature of modeled braided system evolution, which implies the unpredictability of exact configurations, we found that: (a) there was a limited number of recurrent states; (b) the lifetime of each morphological state was associated with a well‐defined timescale; (c) there were preferential transitions from one state to another; and (d) bed morphological states were related to the variability of the sediment transport rate.