GNGTS 2013 - Atti del 32° Convegno Nazionale

with cross-correlation coefficient greater than a minimum threshold, defined by trial and error in order to recognize the minimum number of families that allows us to correctly reproduce the spatial distribution of earthquakes. In order to overcome possible dissimilarities between events differing from each other by more than one order of magnitude, the bridging technique, which is based on the Equivalence Class approach (Press et al. , 1988), is applied. Finally, differential times derived from earthquake cross-correlation (computed for all available station within 50km from the epicenters) are used in conjunction with travel-time differences from manually picked P- and S-phases to relocate the events using the double-difference (DD) algorithm “HypoDD” (Waldhauser and Ellsworth, 2000; Waldhauser, 2001). Results are discussed in the next section. Results and discussion. The right panel of Fig. 1 shows the hypocenter distribution obtained after the relocation of the earthquakes belonging to the Sampeyre swarm. The figure indicates the existence of two nearby sources, recognized as two distinct families by the waveform similarity analysis; one (Family 1) is shallower and collects 12 events characterized by impulsive P-wave arrivals and one (Family 2) is slightly deeper and collects a larger number of earthquakes (72). As evident from the seismic cross section in the bottom panel of Fig. 1, both sources would present high-angle dipping planes directed towards S-SE, compatible with the geological setting of the area where a SE-dipping shear zone occurs in the lower part of the Maira Valley and along the southern side of the Varaita Valley (Balestro et al. , 1995). In order to prove the hypothesis about the activation of two distinct sources, we compare the b -value of the Gutenberg and Richter (1944) relationship for the two earthquake families. To improve the completeness of the two data sets at lower magnitudes, an automatic procedure for detecting micro-seismicity is applied to the stream of waveforms recorded by DOI and PZZ stations. The procedure uses the STA/LTA triggering method with parameters (e.g., filter band, length of the STA and LTA windows, STA/LTA threshold) calibrated for the two stations on the grounds of the ambient noise and micro-earthquake duration. Specifically, the algorithm implements a coincidence system which detects a potential earthquake whether both DOI and PZZ signals exceed the STA/LTA threshold (= 3) within a common 10s window. Approximately 2800 micro-earthquakes are identified and then separated into distinct clusters via cross- correlation analysis. In particular, 40% of micro-earthquakes are joined into families: 280 of them (those characterized by a very impulsive P-wave arrival) are associated to Family 1 while 592 to Family 2. b -values are calculated by applying the maximum likelihoodmethod proposed by Weichert (1980). A b -value equal to 0.98 (+/- 0.05) is obtained for Family 1 while a b = 0.81 (+/- 0.03) for Family 2. The two magnitude-frequency distributions are shown in Fig. 3. Comparing the b -values indicates that Family 1 and Family 2 are characterized by a different proportion of small and larger earthquakes, thus suggesting that they do not come from the same population. To verify this, the Utsu’s p -test (Utsu, 1992) is applied. The test confirms our hypothesis that the 2010 Sampeyre swarm is the consequence of the activation of two distinct interacting fractures having different seismic productivity. Given the small difference Fig. 3 – Cumulative magnitude-frequency distributions for Family 1 and Family 2. 21 GNGTS 2013 S essione 1.1