GNGTS 2013 - Atti del 32° Convegno Nazionale

(Crampin, 1977). The S-phase identification is approached by exploring the characteristic of an S-wave on three-components recording (3C) (Cichowicz, 1993). The product of different polarization filters applied to a 3C recording combines the major characteristics of an S-wave arrival into one single characteristic function (CF) (Diehl et al . 2009, Amoroso et al . 2012). Significant improvements in the quality of hypocentre location have been achieved through the use of correlation-based phase repicking (Satriano et al. , 2008). In particular, the technique of Rowe et al. (2002) is based on cross-correlation and clustering of similar waveforms. In this study we have reconstructed a detailed three-dimensional image of the shallow crustal volume embedding the active normal fault system in the Campania–Lucania Apennines (southern Italy) which were causative of moderate to large earthquakes during the past centuries, e.g. the Ms 6.9, 1980 Irpinia earthquake. It is obtained by the joint inversion of P and S first arrival times from microearthquakes recorded by a dense, high dynamic range, multi-component seismic network recenty deployed in the area. We addressed the issues of data quality and the implementation of a reliable and robust tomographic inversion strategy in order to improve the resolution of the seismic image and accuracy on earthquake locations. As concerns the phase arrival time measurement, techniques based on polarization filtering and refined re-picking by waveform cross-correlation have been applied to enhance the accuracy of first P- and S-wave readings. Data inversion has been performed by using an iterative, linearized, delay-time 3D tomographic method for the joint determination of source location and medium velocity parameters. We found that P- and S-wave velocity models follow the geological structure, with high Vp/Vs and low VpxVs values reflecting the occurrence of significant fluid accumulation within a ~15 km wide rock volume characterized by intense micro-seismicity. We suggest that significant concentration of background seismicity is controlled by high pore fluid pressure. Data and processing techniques. The used dataset consists of 1311 events with 0.1≤ML ≤3.2, recorded in the period August, 2005- April, 2011 by a total of 42 stations owned and operated by the research consortiumAMRA scarl IstitutoNazionale di Geofisica e Vulcanologia (INGV) (Fig. 1). The P- and S- phases have been initially hand-picked on three-component ground velocity recordings. The seismic events were preliminary located in a 1D reference velocity model, recently developed for the area (De Matteis et al. , 2011), by using the NLLoc code, based on a probabilistic, non-linear, global-search Earthquake Location method in 3D media (Lomax et al. , 2000). For our analysis, a first data selection has been performed on the basis of the event location quality: only the events for which at least 5 P- and 2 S- picked arrival times were available, with an azimuthal gap smaller than 200 degrees and an RMS smaller than 0.5 s have been selected. The application of these selection criteria reduced the number of events to 634, to be used for further tomographic analysis. For the analyzed dataset, we observed that co-localised events recorded by the same station, showed inconsistent S minus P times. This inconsistency may be due to changes in the signal to noise ratio or to the presence of multiple arrivals (due to near-surface wave type conver- sion or multi-path) in the S-wave arrival time windows, causing large uncertainties both on hypocentral location and on wave velocities variations in the subsurface. To overcome this problem, we first performed a specific analysis aimed at the optimal dentification and picking of S-phases, using a technique which combines the polarization analysis of single, three-com- ponents recordings (3C) of an event with the analysis of lateral waveform coherence across the network (Amoroso et al. , 2012). A new time series, called weighted characteristic function for the S phase (CFsw), is constructed upon the original three-component recordings that should enhance the polarization characteristics of the S-wave. For a given station, the 3C seismogram has been rotated into the ray coordinate system assuming the incidence angle as measured in a 0.3 sec window after the first-P arrival and the theoretical back-azimuth obtained from the preliminary earthquake location. Then the polarization attributes, e.g. the directivity, the 6 GNGTS 2013 S essione 1.1