River dynamics: extreme events, machine learning and the role of biology

River dynamics occurs across a large range of timescales ranging from sediment transport and bank erosion during individual flood events to long-term change in river profile and geometry in response to change in climate or tectonic forcing. Bridging the gap between these two timescales, and predicting how rivers evolve at intermediate timescales (100-1000 years) remain a major challenge in fluvial geomorphology, with direct implication for a variety of societal issues such as hazards management following large earthquakes or prediction of integrated river response in the context of climate change. Knowledge gaps exists, for instance, on how quickly river geometry responds to external perturbation (dams, faults,…), or how vegetation and biology impact river morphology and dynamics. I illustrate recent work pertaining to the above issues, using 2D/3D numerical modelling and high resolution topographic data acquired from LiDAR and processed with 3D machine learning approaches.