Supino, Mariano (2018) A probabilistic approach for the estimation of earthquake source parameters from spectral inversion. [Tesi di dottorato]

Preview

Text supino_mariano_31.pdf Download (11MB) | Preview

Collection description

The characterization of the mechanisms of earthquake generation and propagation is a major challenge in understanding the Earth engine. Although the seismic rupture non-linearly combines several space and time scales, some macroscopic parameters can provide insights in its evolution, such as the earthquake size and the stress drop released during a seismic event. However, the estimation of these parameters is very uncertain (Cotton et al., 2013), owing to uncertainties in data and models and to the strong coupling between source effects and wave propagation up to the observation sites. The objective of this thesis is the characterization of the seismic source parameters using the amplitude spectrum of the displacement records and assuming that the earthquake behaves as a circular crack (Keilis-Borok, 1959). Several methods for the characterization of the source using a spectral analysis have been proposed in literature. Systematic comparison between different methodologies highlighted the dependence of the results on the fitting model, due to the high correlation between the parameters, especially comparing EGF and TGF based techniques (Ide et al. 2003; Oye et al., 2005). A probabilistic approach can allow to investigate such a correlation, defining a probability density function (PDF) in the parameter space and allowing for a consistent estimate of the uncertainties. Using the probabilistic framework developed by Tarantola (2005), and specifically the notion of conjunction of states of information, I developed a probabilistic approach to retrieve the source parameters seismic moment (through the low-frequency spectral level), the corner frequency (that is a proxy of the rupture length) and the high-frequency decay parameter. Information on the source of an earthquake requires the modeling of the wave propagation too; I choose to use in this work a theoretical Green’s function, adding one parameter to invert related to the propagation (a frequency-independent Q-factor) beyond the three source parameter that I want to retrieve. I model the observations with an operator, defined on these four parameters, which is non-linear; thus, a global exploration of the model space is required in order to find the best solution to describe the data. Additionally, the joint a-posteriori probability density function (PDF) is computed around the best model, to extract the correlation matrix of the parameters. This allows to obtain estimates and uncertainties from the PDF, that are taking into account the correlations. The global exploration relies on the building of a Markov chain in the parameter space and on the combination of a deterministic minimization with a random exploration of the space (Basin-Hopping method, Wales and Doye, 1997; Wales, 2003). The main advantages of this new methodology are the following : • A fully probabilistic approach associated with a global exploration method can provide a robust information about the “best-fit” model, with correct estimation of uncertainties and parameter correlation. • The shape of the estimated PDF can assess the quality of the solutions, allowing to rule out noisy data and thus enabling the use of the method for automatic processing of large datasets. I performed three applications of the method. In Chapter 4, I analyzed the Central Italy 2016-2017 sequence, characterizing the source of all the earthquakes with Ml > 4 (56 events); in Chapter 5 I characterized the source of more than 10000 LFEs occurred in the Nankai region (Japan) during the period 2012-2016; in Chapter 6 I analyzed the micro-seismicity (Ml between 0 and 4.5, 1061 events) occurred from 2016 to 2017 in the Northern Ibaraki region (Japan).

Downloads

Actions (login required)

View Item