GNGTS 2015 - Atti del 34° Convegno Nazionale

three modeled shapes. The plateau value decreases from a value of 2.2 up to 1 (approximately at the midpoint of the geometries). The PGD peak is largest for the half-circle (black curve in Fig. 3 top panel) with the maximum amplification occurring exactly at its uppermost vertex, whereas the triangle is mostly characterized by deamplification of the ground motion (green curve in Fig 3 top panel). The slope geometry (red curve in Fig. 3 top panel), in case of x component and S- source polarization, shows a PGD increase (from 2.2 in the plateau part to 2.9) with the largest amplification occurring in a narrow zone immediately after the slope (between 2800 and 2900 m), similar to a “damage belt zone” effect. For z component and P- source polarization (middle panel of Fig. 3), the largest PGD peak is observed for the uppermost vertex of the triangle (green curve in Fig. 3 middle panel) with value from 1.4 (plateau part of the figure) to 1.9 (vertex). The uppermost vertex of the half- circle (black curve) in this case does not modify the PGD with respect to the plateau value, whereas rapid variations of amplification and deamplification appear near the base of the half- circle geometry. The slope (red curve in Fig. 3 middle panel) shows lower deviation between PGD maxima and PGD minima than the other geometries. The effect of the topographic shapes on PGD, as observed in the Fig. 3 for both the polarization, starts from the progressive 2200 up to 3200 m (x axis), i.e. 200 m before and after the irregular shapes. This measurement (200 m) is comparable to the half-width (300 m) of the modeled shapes, therefore their effect seems to spatially extend for a length of the same order of the half-width of the topographical irregularities. Case-study: Mount Ocre. After the 2-D simulation of P-SV waves propagating in simple irregular shapes, the realistic topographic profile of Mount Ocre (courtesy of Antonio Avallone) was modeled. Mount Ocre (about 9 km in SE direction from L’Aquila downtown) is interesting because it is available at this site a record of the Mw 6.1 L’Aquila mainshock (on April 6, 2009) acquired by a high-rate GPS station (site named CADO with a 10 Hz as sampling rate; Avallone et al., 2014). CADO is on the crest of a narrow ridge, which is elongated in the NW-SE direction bounding the Aterno river valley. Following Avallone et al. (2014), the vertical component of L’Aquila mainshock recorded at CADO shows a maximum subsidence of about 18 cm, and the horizontal components show a strong nearly-harmonic high-amplitude 1 Hz phase. This harmonic phase shows a maximum peak-to-peak amplitude of 36.4 cm in the east component and 22.6 cm in the north component. The particle motion in the horizontal plane results in a maximum displacement of 42.8 cm in the N+60° clockwise direction, approximately normal both to the fault strike of the region and to the ridge elongation. The horizontal polarization at 1 Hz is assumed to be a site property because is independent from the nature of the source signal. Indeed using directional H/V ratios, the same polarization resulting from the mainshock was observed from noise measurements surrounding CADO, and also from local small-magnitude earthquakes recorded by a co-located seismic station (Avallone et al., 2014). The outcropping rock at CADO is fractured and weathered limestone, but the observed strong site effect at 1 Hz is not consistent with the results of array experiments conducted within the aim of the NERA Project nearby CADO (Rovelli et al., 2012). The possible role of the topography at CADO has been investigated using a 2-D model with the following parameters: Vs=1000 and Vp=2000 m/s; Qs=100 and Qp=200; mass density fixed to 2.5 g/cm 3 . For simplicity and because the analysis is focused on the amplification induced by the topographic profile, the relief is modeled as having uniform rock properties and the adjacent Aterno basin is not included in the model. The ground motion amplification at Mount Ocre was estimated using the SSR approach for synthetic waveforms with respect to a reference site. The reference site was the synthetic seismogram of the same model without topography. A P-SV computation, using a different code than specfem2D , was also performed at this site by means of a finite-element approach in time-domain (Caserta 1998; Caserta et al. 2002) which adopts a triangle mesh generator (http://www.cs.cmu.edu/~quake/triangle.html) . However the SSR results of the two codes ( specfem2d and Caserta code) were very similar up to 8 Hz. GNGTS 2015 S essione 2.2 91