Study of Southern Tyrrhenian and Sicilian regions by a sequential procedure to integrate WAM seismic tomographies and Bouguer anomaly data

A procedure to obtain 3D velocity-density models and earthquake relocation by integrated inversion of P and S wave traveltimes and Bouguer anomaly distribution was applied to a large dataset concerning the Southern Tyrrhenian and Sicilian areas. The seismic dataset was subdivided into two subsets for separate inversions, whose results were later on joined by the WAM (Weighted Average Model) technique. This is a post-processing technique proposed by Calò et al. (2009) by which preliminary tomographic models are unified in a common 3D grid. The first dataset concerns 28873 P and 9990 S arrival times of 1800 earthquakes located in the area 14°30′ E - 17°E, 37°N - 41°N while the second dataset contains 31250 P and 13588 S arrival-times related to 1951 events located in the area 11° E - 15°48′ E, 36°30′N - 39°N. The selected events were recorded at least by 10 stations in the period 1981-2005 and marked by RMS < 0.50 s. The second dataset was integrated with P-wave traveltimes picked in several sesmic profiles carried out in the study region. The Bouguer anomaly measurements were interpolated in the nodes of a 8x8 km regular grid covering the area 12° E - 16°01′ E, 36°13′ N - 38°31′ N. The proposed procedure allows to invert seismic and gravimetric data with a sequential technique to avoid the problematic optimization of the relative weights to assign to the different type of data. A first WAM provides a preliminary Vp, Vs and Vp/Vs models and a first ipocentral relocation. Since the obtained Vs model seems poorly constrained by the S wave arrival times, the Vp model is converted in a new Vs model, through a Vs-Vp correlation law proposed by T.M. Brocher (2005), and used, jointly to the Vp model, as input for a second WAM. The results of this second step are used to derive, by the empirical Brocher’s equations, 2 density distributions associated to the Vp and Vs models. These density models are statistically compared and the distribution of their average value is determined. The gravimetric forward problem for this density model was solved using the 3GRAINS code implemented by Snopek and Casten (2006). The program uses the Nagy formula (1966) to calculate the gravity attraction of right rectangular prisms. After the analysis of the gravimetric residues distribution, we compute a minimum norm correction vector for the density values. The correction distribution is used both to improve the density-velocity correlation equations (ρ-Vp, ρ-Vs) and determine 2 correction vectors for Vp and Vs models respectively. Using the corrected Vp and Vs distributions as input for a new tomographic inversion the velocity and density models are iteratively upgraded. The proposed procedure underlined the necessity of the different data integration although the seismic problem seemed to be a priori well constrained. Furthermore it allowed to highlight some velocity and density features that could play a crucial rule for the reconstruction of the geodynamic evolution of the study area.