Logo Logo
Help
Contact
Switch language to German
Efficient treatment and quantification of uncertainty in probabilistic seismic hazard and risk analysis
Efficient treatment and quantification of uncertainty in probabilistic seismic hazard and risk analysis
The main goals of this thesis are the development of a computationally efficient framework for stochastic treatment of various important uncertainties in probabilistic seismic hazard and risk assessment, its application to a newly created seismic risk model of Indonesia, and the analysis and quantification of the impact of these uncertainties on the distribution of estimated seismic losses for a large number of synthetic portfolios modeled after real-world counterparts. The treatment and quantification of uncertainty in probabilistic seismic hazard and risk analysis has already been identified as an area that could benefit from increased research attention. Furthermore, it has become evident that the lack of research considering the development and application of suitable sampling schemes to increase the computational efficiency of the stochastic simulation represents a bottleneck for applications where model runtime is an important factor. In this research study, the development and state of the art of probabilistic seismic hazard and risk analysis is first reviewed and opportunities for improved treatment of uncertainties are identified. A newly developed framework for the stochastic treatment of portfolio location uncertainty as well as ground motion and damage uncertainty is presented. The framework is then optimized with respect to computational efficiency. Amongst other techniques, a novel variance reduction scheme for portfolio location uncertainty is developed. Furthermore, in this thesis, some well-known variance reduction schemes such as Quasi Monte Carlo, Latin Hypercube Sampling and MISER (locally adaptive recursive stratified sampling) are applied for the first time to seismic hazard and risk assessment. The effectiveness and applicability of all used schemes is analyzed. Several chapters of this monograph describe the theory, implementation and some exemplary applications of the framework. To conduct these exemplary applications, a seismic hazard model for Indonesia was developed and used for the analysis and quantification of loss uncertainty for a large collection of synthetic portfolios. As part of this work, the new framework was integrated into a probabilistic seismic hazard and risk assessment software suite developed and used by Munich Reinsurance Group. Furthermore, those parts of the framework that deal with location and damage uncertainties are also used by the flood and storm natural catastrophe model development groups at Munich Reinsurance for their risk models.
Seismic Hazard, Seismic Risk, Uncertainty, Monte Carlo, Simulation, Variance Reduction, Sensitivity Analysis, Insurance, Location Uncertainty, Munich Re
Scheingraber, Christoph
2019
English
Universitätsbibliothek der Ludwig-Maximilians-Universität München
Scheingraber, Christoph (2019): Efficient treatment and quantification of uncertainty in probabilistic seismic hazard and risk analysis. Dissertation, LMU München: Faculty of Geosciences
[thumbnail of Scheingraber_Christoph.pdf]
Preview
PDF
Scheingraber_Christoph.pdf

37MB

Abstract

The main goals of this thesis are the development of a computationally efficient framework for stochastic treatment of various important uncertainties in probabilistic seismic hazard and risk assessment, its application to a newly created seismic risk model of Indonesia, and the analysis and quantification of the impact of these uncertainties on the distribution of estimated seismic losses for a large number of synthetic portfolios modeled after real-world counterparts. The treatment and quantification of uncertainty in probabilistic seismic hazard and risk analysis has already been identified as an area that could benefit from increased research attention. Furthermore, it has become evident that the lack of research considering the development and application of suitable sampling schemes to increase the computational efficiency of the stochastic simulation represents a bottleneck for applications where model runtime is an important factor. In this research study, the development and state of the art of probabilistic seismic hazard and risk analysis is first reviewed and opportunities for improved treatment of uncertainties are identified. A newly developed framework for the stochastic treatment of portfolio location uncertainty as well as ground motion and damage uncertainty is presented. The framework is then optimized with respect to computational efficiency. Amongst other techniques, a novel variance reduction scheme for portfolio location uncertainty is developed. Furthermore, in this thesis, some well-known variance reduction schemes such as Quasi Monte Carlo, Latin Hypercube Sampling and MISER (locally adaptive recursive stratified sampling) are applied for the first time to seismic hazard and risk assessment. The effectiveness and applicability of all used schemes is analyzed. Several chapters of this monograph describe the theory, implementation and some exemplary applications of the framework. To conduct these exemplary applications, a seismic hazard model for Indonesia was developed and used for the analysis and quantification of loss uncertainty for a large collection of synthetic portfolios. As part of this work, the new framework was integrated into a probabilistic seismic hazard and risk assessment software suite developed and used by Munich Reinsurance Group. Furthermore, those parts of the framework that deal with location and damage uncertainties are also used by the flood and storm natural catastrophe model development groups at Munich Reinsurance for their risk models.