Starts: Apr 30, 2014 5:00:00 PM
Starts: Jun 18, 2014 9:00:00 AM
Second Academic Conference on Research, Teaching and Service in Disaster Risk Reduction and Resilience
Starts: Jun 19, 2014 9:00:00 AM
IRDR Special Report on Transitional Recovery and Reconstruction in the Eastern Philippines after Typhoon Yolanda
Published: May 23, 2014 11:07:18 AM
Published: Feb 24, 2014 3:30:38 PM
Published: Jan 8, 2014 2:45:50 PM
Office location: Rm 38, 2nd floor, South Wing, UCL Main Quadrangle
Title: Statistical emulation of tsunami models for more efficient uncertainty and sensitivity analyses
Supervisors: Serge Guillas, Statistical Science
Source(s) of funding:IRDR and Statistical Science, UCL
Address: UCL Institute for Risk and Disaster Reduction, University College London, Gower Street, London, WC1E 6BT
Phone No: 020 3108 1107
Project description: This research project is focused on the statistical analysis of tsunami models. The high computational demand of these models motivated the development of statistical surrogate models, known as emulators, that save computational time and make feasible expensive sensitivity and uncertainty analyses. Statistical emulation is applied to a landslide-generated tsunami model looking at the time series of the wave elevation at specific locations along the shoreline. It is shown that the emulator predictions are very accurate, even when only few runs of the expensive computer model are used. The emulator is further used for uncertainty and sensitivity analyses of the computer model. Impressively the computational time required for performing these analyses is at least 20,000 times lower using the emulator in place of the computer model. Time-space correlations are also investigated to obtain a spatio-temporal emulator that can accurately predict the wave elevations at any time and location.
Parallel to this, a parametric model is developed to describe the seabed displacement and real geometry in the case of hypothetical earthquake events, for the Cascadia subduction zone, utilising existing data. This is used to represent various earthquake scenarios and extensively investigate the induced tsunami waves propagation and coastal runup. This will enable risk analysis of such tsunamis in this region.