Flooding disasters, reproduction of salmonids, and development of Earth landscapes are all heavily impacted by the quantity and the quality of sediment transported by rivers. Yet after more than a century of work we have no satisfactory theory for sediment transport. Therefore empirical sediment transport formulas often poorly compare with field measurements, by sometimes order of magnitudes. Such poor formulas are used in engineering softwares based on two-phase shallow water equations, giving therefore overall unreliable results. Hence it is difficult to assess, for example, the impact upon a stream of extreme sediment-laden floods, which is an issue for public safety, management of water resources, and environmental sustainability within the critical zone. An important reason for our limited ability to predict sediment transport, is due to the very wide range of grain sizes leading to size segregation, also named grain size sorting. This phenomenon largely modifies fluxes and results in patterns that can be seen ubiquitously in nature, such as armoring (Fig. 1). The overall objective of this proposal is to conduct a multi-scale and multi-disciplinary study of size segregation in sediment transport to ultimately improve sediment transport modeling. The specific objectives of the project focused on bedload transport are to (i) improve our understanding of the physical processes in size segregation, especially particle-particle interactions and the feedback with the transporting fluid; (ii) develop a hierarchy of process-based models at different scales for understanding, predicting and upscaling (iii) incorporate validated elementary segregation formulations in classical sediment transport models to ultimately improve the predictive capability of civil and bioengineering tools. The project will be organized in 4 multidisciplinary work packages: experimental, numerical, image analysis and management and multiscale analysis, investigating at interconnected scales (Fig. 4). WP1 is devoted to original experimental investigations. Starting from the inherent complexity of field measurements and analysis, the structure follows a downscaling path proposing 3D experiments with natural material, quasi 2D experiments with spherical glass beads and finally 2D dry granular experiments. Deliverables of WP1 include high resolution spatio-temporal datasets for understanding, theorizing and validating models. WP2 follows an upscaling path from Euler/Lagrange models to shallow water engineering models including Euler/Euler and mixture models. Once validated on experimental results, process-based bedload transport models (deliverables) will also help interpret experimental findings at different scales. WP3 is entirely devoted to image analysis since most experimental measurements (including in the field) will rely on image analysis, especially particle tracking and segmenting algorithms. Deliverables include open source sediment tracking algorithms and ground-truth datasets (true trajectories). Finally WP0 will be devoted to project management but also to promoting ’the multiscale attitude’. A good part of requested funding is devoted to two three-year PhDs and two 12-month post-docs. The consortium includes all necessary competencies to study sediment transport on a large range of scales: physical geography/fluvial geomorphology (Irstea, UBC, SFU), civil engineering (Irstea, 3SR), granular physics (IPR), fluid (Legi, Irstea, NRL) and soil mechanics (3SR), image analysis (laHC) and applied mathematics (UM). The coordination by Irstea, an organization which has a long experience of disseminating multidisciplinary research, of research networking and transfer to public and private end-users, guarantees that this project, potentially transformative, will be much more than the juxtaposition of isolated research.