Dissertation committee:
Γεράσιμος Τσελέντης, Καθηγητής, Τμήμα Γεωλογίας και Γεωπεριβάλλοντος ΕΚΠΑ, Διευθυντής του Γεωδυναμικού Ινστιτούτου του Εθνικού Αστεροσκοπείου Αθηνών (Επιβλέπων)
Βασίλειος Καραστάθης, Διευθυντής Ερευνών του Γεωδυναμικού Ινστιτούτου του Εθνικού Αστεροσκοπείου Αθηνών
Φίλιππος Βαλλιανάτος, Καθηγητής, Τμήμα Γεωλογίας και Γεωπεριβάλλοντος ΕΚΠΑ
Υπόλοιπα μέλη της Επταμελούς Επιτροπής
Νικόλαος Σαρλής, Καθηγητής, Τμήμα Φυσικής ΕΚΠΑ
Γεώργιος Καβύρης, Αναπληρωτής Καθηγητής, Τμήμα Γεωλογίας και Γεωπεριβάλλοντος ΕΚΠΑ
Ευθύμιος Σκορδάς, Αναπληρωτής Καθηγητής, Τμήμα Φυσικής ΕΚΠΑ
Χρήστος Ευαγγελίδης, Κύριος Ερευνητής του Γεωδυναμικού Ινστιτούτου του Εθνικού Αστεροσκοπείου Αθηνών
Summary:
In this thesis, the application of advanced high-speed passive seismic tomography techniques was investigated in the area of the west coast of central Greece, with the aim of retrieving a three-dimensional shear velocity model. In short, to retrieve this model, we followed the following procedures:
1. We first derived the Green's functions between pairs of stations, i.e., the impulse response of the Earth as if one station were the transmitter and the other the receiver. This was achieved by cross-correlating the noise recordings of the two stations, after these recordings had undergone appropriate pre-processing.
2. Next, provided we obtained the Green's function surface waves and since they are known to exhibit dispersion, for each Green's function we obtained the group velocity dispersion relation of the fundamental mode under appropriate quality control conditions. This was achieved by applying a sequence of bandpass filters with different center frequencies. In each such application, by considering the travel time spent by the fundamental mode envelope to travel the distance between the two stations, we derived the group velocity value for each center frequency.
3. We then inverted the mean dispersion relation of all pairs of stations using Monte Carlo methods so as to recover a model of horizontal and homogeneous layers with elastic properties, so that when solving the equations of linear elasticity, we obtain a dispersion relation close to that we observed from the data.
4. Finally, this recovered model was used as the initial model during the inversion in three dimensions, which was done by the method of iteratively reweighted least squares. With this method we iteratively perturbed the original model, and thus the values of the shear velocities below each point on the surface. At each perturbation and for each point on the surface, we can also identify a dispersion relation by solving the equations of linear elasticity. So, for each frequency of each dispersion relation and for all points on the surface we also have a value of the group velocity of the surface wave. Given this, we can solve the eikonal equation to find the theoretical travel time of the rays between pairs of stations for each frequency. These iterative perturbations are made until the theoretical travel times converge to the experimental ones.
The recovered model agrees in terms of absolute velocity values with previous geophysical studies carried out in the wider area while also agreeing with the already known geology of the area. By performing a checkerboard test, we can also verify that given the spatial distribution of stations in the survey area and our initial model, we can successfully recover a priori unknown but existing velocity anomalies. Finally, by applying Monte Carlo error propagation techniques, we can safely conclude that our final model derivation process is robust to measurement errors attributable to chance.
