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

Braided rivers undergo continuous reorganization of their channel network, producing highly dynamic planform evolution and strongly intermittent sediment transport. Despite this apparent complexity, recent research suggests that braided rivers repeatedly occupy a finite number of characteristic morphological configurations, referred to as morphological states. Identifying these recurrent configurations provides a compact framework for describing and modeling river morphodynamics. This repository contains the complete set of binary water masks extracted from a long-duration laboratory experiment conducted at the Laboratory of Environmental Hydraulics (LHE), École Polytechnique Fédérale de Lausanne (EPFL), Switzerland. The experiment consisted of a 3.3 m × 1.0 m mobile-bed flume operated under constant water discharge (0.15 L s⁻¹) and constant sediment feed rate (1.5 g s⁻¹) for approximately 1,200 hours. Overhead images of the braided river were acquired every minute and processed using an automated image-processing workflow based on HSV color segmentation to extract the wetted area of the channel as binary masks. The masks constitute the primary dataset used in the companion studies Braided Rivers as Morphological States: A Data-Driven Identification of Recurrent Planform Configurations and Morphological State Transitions and Sediment Transport Variability in Braided Rivers. In the first study, the masks were used to identify recurrent morphological states through image similarity analysis, dimensionality reduction, and density-based clustering. In the second study, these identified states formed the basis for a stochastic description of braided-river evolution and a probabilistic model of sediment transport variability. The dataset is intended to support research in river morphodynamics, geomorphology, sediment transport, computer vision, machine learning, and image analysis. By providing the complete sequence of binary masks, the repository enables reproducibility of the published analyses and facilitates the development and benchmarking of new methods for image-based characterization of river morphology.