Abstract

Sudden avulsions, unexpected channel migrations and backfilling phenomena are autogenic phenomena that can considerably change the propagation patterns of sediment-laden flows on alluvial fans. Once the initial and boundary conditions of the hazard scenario with a given return period are determined, the assessment of the associated exposed areas is based on one numerical, essentially deterministic, process simulation which may not adequately capture the underlying process variability. We generated sediment-laden flows on an experimental alluvial fan by following a "similarity-of-process concept". Specifically, we considered a convexly shaped alluvial fan model layout featuring a curved guiding channel. As loading conditions, we defined a reference, an increased and a reduced level for the released water volume and the predisposed solid fraction, respectively. Further, we imposed two different stream power regimes and accomplished, for each factor combination, eight experimental runs. The associated exposure areas were recorded by video and mapped in a GIS. We then analysed exposure data and determined exposure probability maps superposing the footprints of the eight repetitions associated with each experimental loading condition. The patterns of exposure referred to the specific loading conditions showed a noticeable variability related to the main effects of the total event volume, the solid fraction, the interactions between them, and the imposed stream power in the feeding channel. Our research suggests that adopting a probabilistic notion of exposure in risk assessment and mitigation is advisable. Further, a major challenge consists in adapting numerical codes to better reflect the stochastics of process propagation for more reliable flood hazard assessments.