A precipiton-based approach to model hydro-sedimentary hazards induced by large sediment supplies in alluvial fans
Résumé
Mountain ranges are frequently subjected to mass wasting events triggered by storms or earthquakes and supply large volumes of sediment into river networks. Besides altering river dynamics, large sediment deliveries to alluvial fans are known to cause hydro-sedimentary hazards such as flooding and river avulsion. Here we explore how the sediment supply history affects hydro-sedimentary river and fan hazards, and how well can it be predicted given the uncertainties on boundary conditions. We use the 2D morphodynamic model Eros with a new 2D hydrodynamic model driven by a sequence of flood, a sediment entrainment/transport/deposition model and a bank erosion law. We first evaluate the model against a natural case: the 1999 Mount Adams rock avalanche and subsequent avulsion on the Poerua river fan (West Coast, New Zealand). By adjusting for the unknown sediment supply history, Eros predicts the evolution of the alluvial riverbed during the first post-landslide stages within 30 cm. The model is subsequently used to infer how the sediment supply volume and rate control the fan aggradation patterns and associated hazards. Our results show that the total injected volume controls the overall levels of aggradation, but supply rates have a major control on the location of preferential deposition, avulsion and increased flooding risk. Fan re-incision following exhaustion of the landslide-derived sediment supply leads to sediment transfer and deposition downstream and poses similar, but delayed, hydro-sedimentary hazards. Our results demonstrate that 2D morphodynamics models are able to capture the full range of hazards occurring in alluvial fans including river avulsion aggradation and floods. However, only ensemble simulations accounting for uncertainties in boundary conditions (e.g., discharge history, initial topography, grain size) as well as model realization (e.g., non-linearities in hydro-sedimentary processes) can be used to produce probabilistic hazards maps relevant for decision making.