S53A-4473:
Landslide caracteristics determination using bayesian inversion and seismic recording

Friday, 19 December 2014
Laurent Moretti1, Anne Mangeney1, Yann Capdeville2, Eleonore Stutzmann1, Thomas Bodin3 and Francois Bouchut4, (1)Institut de Physique du Globe de Paris, Paris, France, (2)CNRS, Paris Cedex 16, France, (3)University of California Berkeley, Berkeley, CA, United States, (4)Université Marne la Vallée, CNRS, Marne la Vallée, France
Abstract:
Gravitational instabilities, such as landslides, avalanches, or debris flows, play a key role in erosional processes and represent one of the major natural hazards in mountainous, coastal, and volcanic regions. Despite the great amount of field, experimental and numerical work devoted to this problem, the understanding of the physical processes at work in gravitational flows is still an open issue, in particular due to the lack of observations relevant to their dynamics. In this context, the seismic signal generated by gravitational flows is a unique opportunity to obtain information on their dynamics and characteristics. Here we present the study of the 1997 Boxing Day landslide that occurred in Montserrat. We accessed the force applied by the landslide to the ground surface responsible of the seismic waves by inverting the seismic waveform recorded (force-time function). This force was then used as a constraint in a bayesian inversion problem where the forward problem is the force-time function calculation obtained by simulating the landslide with the SHALTOP model (mangeney et al., 2007). With this method, we are able to give an estimate of the rheology (friction coefficient) and the initial shape of the collapsing mass. The volume retrieved is very similar to that obtained by field observations. The friction coefficient determined is also similar to that constrained by former studies or to that predicted by empirical laws (Lucas et al., 2014). Furthermore the method permits to give an estimate of the error made on these parameters.