National and Kapodistrian University of Athens
A composite model of shear-wave splitting and occurrence rate towards earthquake forecasting
Abstract
dc:descriptionSince the antiquity, earthquakes played an important role in the development of society, destroying civilizations and opening the way for the development of news. Humanity tried to understand and explain this phenomenon by attributing the powerful shaking to divine intervention. Ever since, the feat of earthquake prediction has eluded human knowledge. Today, although we know that no chthonic entities shake the ground so violently, it still escapes us. Considered the holy grail of seismology, many theories and experiments have been developed towards estimating the occurrence of the next catastrophic seismic event. In 1998, a M 5.0 earthquake was successfully forecasted in real-time, by stress precursors studied through shear-wave splitting. As a shear-wave enters an anisotropic medium, it splits in two, travelling with different velocities and having perpendicular polarizations. The faster one is aligned to the main anisotropic feature, and the slow arrives at the station with a time-delay accumulated along the path. One such feature is the existence of fluid-filled microcracks in the upper crust, which are aligned according to the maximum compressive horizontal stress component, according to the Anisotropic Poro-Elasticiy (APE) model. The “stress-forecast” of 1998 was achieved by considering the implications of APE. However, other factors have been shown to affect splitting, such as faults, leading to structurally-controlled anisotropy. Another approach to estimating the occurrence of an impending event is earthquake forecasting, which utilizes empirically optimized statistical models to determine the occurrence rate for different threshold magnitudes. Earthquake forecasting has shown promise and has applications in different timescales. In the current thesis, I investigate the possibility of developing a composite model that uses estimates from shear-wave splitting constrained by statistical modelling – or vice versa. To this end, the Pytheas software was updated and upgraded to facilitate the faster and more effective processing of waveform data using modern concepts, such as the direct acquisition of data from available FDSN webservices (e.g. EIDA nodes) and the integration of parallel processing. In the context of the thesis, a new workflow was developed to homogenize and systematize the analysis of data from various datasets. Shear-wave splitting was determined automatically with the eigenvalues method and cluster analysis. A large volume of seismic recordings from earthquake sequences in Greece was analyzed. Specifically, data were analyzed from Thiva Basin (2020), Samos Island (2020), Styra in Euboea (2022-2023), the Western (2020-2021) and Eastern (2020) Gulf of Corinth. Analysis led to the determination of 18,400 splitting results, out of which 8,031 were considered of adequate quality, after grading with an automatic algorithm. Quality assurance was conducted manually on a sample of results. Out of the 20 temporal variations of normalized time-delays extracted after visual inspection of temporal plots, 10 presented satisfying correlation and statistical significance. No strong association was observed between variations and the occurrence of strong earthquakes. However, variations linked to changes in seismicity were identified, especially in the cases of Styra and the Eastern Gulf of Corinth. Moreover, I investigated the application of an Epidemic-Type Aftershock Sequence (ETAS) model, to produce daily forecasts. The modified ETAS algorithm (integrating Bayesian statistics and Markov Chain Monte Carlo sampling) was applied in the Western Gulf of Corinth, for a catalogue comprising of events related to the December 2020 to February 2021 activity. Two sub-sequences were utilized. The first included swarm-like characteristics, while the second followed a more traditional aftershock sequence. The forecasts were produced for four 24-hour periods, following the initial training 24-hour-long interval right after the initiation of each sub-sequence. Four magnitude thresholds were considered (equal to the magnitude of completeness, 2.5, 3.0 and 4.0). A good agreement was observed between estimates and the second sub-sequence, while, probably due to the swarm-like features, the first one deviated. Unfortunately, it was not possible to extract coherent temporal patterns of shear-wave splitting and, subsequently, model their response to impending events. The results showed that the source of splitting is not easily understood, with varying influences of fluid processes, regional stress and local structures. Nevertheless, there is promise for associating variations of splitting to the development of earthquake sequences. On the other hand, short-term forecasts with the ETAS model were satisfying, even though its application was hindered by the effect of swarm-like activity. Currently, a composite model of shear-wave splitting and earthquake occurrence is not possible. Advances in large data processing and our theoretical understanding of splitting are the key for investigating this prospect.
Degree
thesis:*- Grantor dc:publisher
- National and Kapodistrian University of Athens
- Year dc:date
- 2023
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Spingos, Ioannis
- Σπίγγος, Ιωάννης
Subjects
dc:subject × 12Rights
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/55618