GNGTS 2015 - Atti del 34° Convegno Nazionale

and location of receivers (Paolucci et al., 2002; Pagliaroli, 2006; Hailemikael, 2010 for an overview). Topographic effect on rock sites is difficult to decouple from other amplification effects, such as surface layering/weathering, cracking of the bedrock, presence of faults, and directivity of the source. For these reasons, seismic and geological experiments at topographic sites are important because can provide information on the local velocity structure. Numerical simulations available in literature agree in showing complex amplification/ deamplification patterns, caused by interaction of elastic waves with the surface curvature. The maximum level of amplification occurs when the incident wavelength is comparable to the horizontal dimension of the topographic irregularities. As first approximation, the amplification increases as the average slope of the modeled geometry becomes steeper; for incidence not vertical, the area of largest amplification shifts in opposite direction with comparison to the propagation direction. Experimental studies on topographic sites are mainly based on analysis of earthquake and noise data collected by several receivers, aimed at following ground-motion variation along the topography. The methods of analysis include mostly standard spectral ratio using a reference- site (SSR), single-station method (horizontal-to-vertical spectral ratio, HVSR), polarization and array analysis (Spudich et al., 1986 among many others). Topographic amplification is often associated to a ground-motion polarization, observed both on noise and earthquake recordings (Marzorati et al., 2011; Burjánek et al. 2012a, 2012b; Pischiutta et al., 2013). Although the cause of polarization is still debated, largest amplification seems to appear in transverse direction to the main topography elongation (see also the recent NERA-JRA1 project, WP11, Waveform modeling and site coefficients for basin response and topography, Responsible activity leader P.Y. Bard, ftp://www.orfeus-eu.org/pub/NERA/Deliverables/) . The amplification caused by topography is usually expressed by the topographic aggravation factor ( TAF ) in the time or frequency domain. TAF is defined as the ratio between some ground motion parameters (PGA, PGV, PGD, Fourier amplitude spectra, response acceleration spectra etc) of a receiver along the topographic profile and the same parameter measured at an “ideal” free-field site. TAF is equivalently expressed as the ratio between ground motion amplification ( X ) related to 2-D (and/or 3-D) effects, and the stratigraphic amplification at the same site related to 1-D effects: TAF = X (2- D )/ X (1- D ) Numerical and experimental studies indicate the topographic effect as depending on the frequency, but European seismic design code (EC8; CEN2008) provides a topographic coefficient ( St ) frequency-independent. St is only depending from the steepness of the slope and the geometry of the topographic sites. The same approach follows the Italian building code (NTC08), which takes into account four classes of topographic categories (T1, T2, T3 and T4) with St ranging from 1 to 1.4; design elastic spectra are uniformly multiplied by St coefficient to include the topographic effect. 2-D SEM Modeling of “ basic geometries ”. The spectral-element method (SEM) (as implemented in the specfem2d code https://geodynamics.org/cig/software/specfem2d/ ; Komatitsch and Vilotte, 1998; Chaljub et al., 2007) was used for studying the effect of three basic 2-D geometries. The specific of the SEM is in the combined use of high-order Legendre polynomials and of the Gauss-Lobatto-Legendre (GLL) quadrature rule to evaluate the integrals within the weak (or variational) formulation of the equation of motion (see Fichtner 2011 for an extensive overview). P-SV models were computed using the following irregular geometries: an half-circular hill (diameter 600m), a triangular ridge (with dimension 600·300mfor base and height, respectively), and a slope with dip of 45° (dimension 300·300 m for base and height, respectively). The 2-D models were subjected to a vertical incidence of planar waves (Fig. 1); the seismic input is a Ricker pulse with a central frequency of 5 Hz exciting the model approximately from GNGTS 2015 S essione 2.2 87