Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.668494
Title: Emulators in the investigation of sensitivities and uncertainties in tsunami models
Author: Sarri, A.
ISNI:       0000 0004 5367 368X
Awarding Body: University College London (University of London)
Current Institution: University College London (University of London)
Date of Award: 2015
Availability of Full Text:
Access through EThOS:
Full text unavailable from EThOS. Please try the link below.
Access through Institution:
Abstract:
This thesis consists of three parts with the unifying theme of applying emulators and other statistical methods to the investigation of sensitivities and uncertainties in tsunami models. The first part contains the statistical emulation of the wave time series at fixed locations that are obtained using an analytical landslide-generated tsunami computer model. Leave-one-out diagnostics are used to validate the emulator, showing excellent agreement in predictions and model evaluations. The emulator is used for sensitivity and uncertainty analyses, where many evaluations are necessary. The computational time required for both analyses is at least 20,000 times lower using the emulator instead of the computationally expensive model. The second part investigates the significance of the seabed deformation representation on tsunami wave predictions of the advanced VOLNA tsunami model. A novel realistic representation of the tsunami source deformation using quadratic curves is compared with simpler representations used in existing tsunami models: single block, four blocks, sixteen blocks and “piano keys”. Factor-of-two differences in wave elevations obtained for different representations of seabed deformation. The last part presents an investigation of tsunamis produced by different fault rupture scenarios on the Cascadia Subduction Zone. Polygons are used to model the source geometry and quadratic curves the deformation in different source event scenarios that are chosen using experimental design. The VOLNA model is evaluated for these scenarios to explore how the source characteristics affect the wave amplitude and coastal inundation predictions. Statistical emulation is applied to a large number of locations in order to investigate the use of emulators in place of expensive tsunami simulators in practical applications. The leave-one-out diagnostics show that the application of Registration and Functional Principal Components techniques to the emulation process leads to improved predictions. Finally, a limiting event is investigated in order to obtain the worst-case scenario inundation along the whole margin in a single simulation.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.668494  DOI: Not available
Share: