Uncertainty Quantification

p2-26 January 22 14:40-15:40

Inversion of landslide parameters for the 1945 Makran tsunami

M. Mamajiwala(UCL), D. Gopinathan(UCL), S. Guillas(UCL), M. Heiderzadeh(Brunel University London), D. Salmanidou(UCL), D. Roy(IISc) and K. Rajendran(IISc)

Abstract

The Makran subduction zone can generate tsunamis, such as the 1945 devastating event over the Arabian Sea (around 3000 deaths). Measurements are scarce and some have large uncertainties. Nervertheless, it was shown that a combination of an earthquake and a submarine landslide is the most likely source for 1945. In this work, we go a step further and infer the landslide characteristics and their probability distributions by carrying out a Bayesian calibration, i.e. an inversion, against a new combination of field measurements. The forward tsunami numerical model VOLNA is run over a design of computer experiments using both the landslide dynamics and the sea-bed uplift due to earthquake. VOLNA is a non-linear shallow water wave equations solver using a finite volume method. The measurements are reported as run-ups and splash values, for different locations, so we explore the use of Bayesian hierarchical models to deal with the complex nature of the variations in uncertainties across such measurements. Finally, we also explore the applicability of history matching to reduce the parameter search space before performing the calibration. This allows us to initially rule out large portions of the input space of landslide characteristics in order to reduce the overall computational costs.

Contact information

Ms. Mariya Mamajiwala University College London