Epistemic Uncertainty Associated to Parametric and Non-Parametric Git Results Related to Initial Parametrization and Target Region Dataset: Application on the Epos-France Database.
Description:
Understanding ground motion behaviour and dependencies is essential for Seismic Hazard Assessment, and several approaches have been implemented such as the Generalized Inversion Techniques (GITs). They allow decomposing the Fourier amplitude spectrum of the seismic signal into the three main contributions of ground-motion: the source, the attenuation and the site response. However, these techniques rely on several initial assumptions, such as defining a geographical area where crustal attenuation is considered homogeneous or setting parameters necessary to run the inversions. These choices are not straightforward and can have non-negligible impact on the results, particularly in the case of a sparse dataset. Hence, it is necessary to assess epistemic uncertainty associated to results obtained by such approaches. To this aim, we perform sensitivity tests for two different GIT algorithms: one parametric (Grendas et al. 2022), where source and attenuation terms are constrained a priori using physical models, and one non-parametric (Oth et al. 2011), for which the physical models are fitted in post-inversion phase. These approaches are applied on the Epos-France database (Buscetti et al. in prep) containing recordings for M_l≥3 earthquakes that occurred in France and neighbouring countries during the 1996-2021 period. First, plural inversions are performed in the Alps region (area with the largest number of recordings), where initial parameters vary. The results of these inversions allow us quantifying the epistemic uncertainties related to both, the inversion algorithm and the inversion parametrization on source parameters (M_w,f_c), attenuation parameters (Q,γ) and site terms, as well as to identify potential biases related to some inconsistent parametrizations. In a second step, the most coherent parametrization is used to perform inversions in different geographical areas of France, in order to discuss uncertainty relative to the chosen region and to illustrate the impact of the amount and quality of available regional data on the epistemic uncertainties associated to the GIT results.
Session: From Earthquake Recordings to Empirical Ground-Motion Modelling - II
Type: Oral
Date: 5/2/2024
Presentation Time: 11:00 AM (local time)
Presenting Author: Margaux
Student Presenter: Yes
Invited Presentation:
Authors
Margaux Buscetti Presenting Author Corresponding Author margaux.buscetti@cea.fr French Alternative Energies and Atomic Energy Commission |
Paola Traversa paola.traversa@edf.fr Electricité de France |
Fabrice Hollender fabrice.hollender@cea.fr French Alternative Energies and Atomic Energy Commission |
Vincent Perron Vincent.PERRON@cea.fr French Alternative Energies and Atomic Energy Commission |
|
|
|
|
|
Epistemic Uncertainty Associated to Parametric and Non-Parametric Git Results Related to Initial Parametrization and Target Region Dataset: Application on the Epos-France Database.
Category
From Earthquake Recordings to Empirical Ground-Motion Modelling