9:15 AM - 9:30 AM
[SSS11-08] Consideration of Uncertainty in Seismic Hazard Assessment Based on Strong Motion Observation Record Database
Keywords:Seismic Hazard Assessment, Uncertainty, Strong Motion Database, Ground Motion Prediction Equation (GMPE)
In order to properly consider the epistemic uncertainties related to ground motion equations (GMPEs) in seismic hazard assessment, GMPEs based on various ideas for application of observation records to extrapolated regions. It is necessary to construct a framework for deriving the above and evaluating the performance of those models. Morikawa et al. (2020; JpGU) have prototyped a strong motion observation record database, which is the basis for this.
In order to evaluate the performance of GMPEs, it is necessary not only to use a common database in deriving the models, but also to be able to calculate ground motions under common conditions for earthquakes and site. In this study, from the viewpoint of applying to the seismic hazard assessment in Japan, the applicable conditions that can be the target of performance evaluation of the GMPE were determined as a tentative "specifications." It includes the condition that it is applied to the site very near source fault where the fault distance is within 1 km and to a mega-earthquake of magnitude 9. The analyst’s concept regarding the modeling of extrapolation for which these observation records are not sufficiently obtained is clarified. Based on this prototype strong motion database and tentative specifications, we start to construct multiple GMPEs, including improvements to existing ones.
On the other hand, it is quite hard to verify the strong motion prediction results for a mega-earthquake along the Nankai Trough or near the source fault based on observation records. As the first step in the study to evaluate the reliability of such strong motion prediction results, we investigated strong motion intensity distribution for discretized magnitude-fault distance space in the strong motion observation record database. In the case with many records, it is close to the lognormal distribution. In the case with few records, however, it is close to the uniform distribution. We plan to further analyze the temporal changes in the distribution due to the accumulation of records. Such characteristic analysis of the database itself not only present the insufficiency of observation records, but also aims to formulate a method for evaluating the reliability of the added simulation data by changes in the distribution.
In order to evaluate the performance of GMPEs, it is necessary not only to use a common database in deriving the models, but also to be able to calculate ground motions under common conditions for earthquakes and site. In this study, from the viewpoint of applying to the seismic hazard assessment in Japan, the applicable conditions that can be the target of performance evaluation of the GMPE were determined as a tentative "specifications." It includes the condition that it is applied to the site very near source fault where the fault distance is within 1 km and to a mega-earthquake of magnitude 9. The analyst’s concept regarding the modeling of extrapolation for which these observation records are not sufficiently obtained is clarified. Based on this prototype strong motion database and tentative specifications, we start to construct multiple GMPEs, including improvements to existing ones.
On the other hand, it is quite hard to verify the strong motion prediction results for a mega-earthquake along the Nankai Trough or near the source fault based on observation records. As the first step in the study to evaluate the reliability of such strong motion prediction results, we investigated strong motion intensity distribution for discretized magnitude-fault distance space in the strong motion observation record database. In the case with many records, it is close to the lognormal distribution. In the case with few records, however, it is close to the uniform distribution. We plan to further analyze the temporal changes in the distribution due to the accumulation of records. Such characteristic analysis of the database itself not only present the insufficiency of observation records, but also aims to formulate a method for evaluating the reliability of the added simulation data by changes in the distribution.