GNGTS 2014 - Atti del 33° Convegno Nazionale

GNGTS 2014 S essione 3.1 35 L-shaped array refractions microtremors (LeMi) J. Boaga 1 , C. Strobbia 2 , G. Cassiani 1 1 Dipartimento di Geoscienze, Università degli Studi di Padova, Italy 2 Total S.A., Pau, France Introduction. Surface wave methods are nowadays the main testing tools for the site characterization concerning engineering applications. Their use has been growing in past years especially due to the increasing interest in shear wave velocity measurement, which is essential in geotechnical earthquake engineering applications (Beavers, 2002; Chaillat et al. 2009). One of the reasons is the progression of several internationals a-seismic building codes that propose simplified seismic scenario based on Vs classification (e.g. Vs30 parameter as in Moss, 2008; EC8, 2003). Surface wave methods based on dispersion properties studies are nowadays widely adopted in local subsoil Vs characterization: different frequencies involve different soil thicknesses, and consequently travel at different velocities. Dispersion properties of surface waves are then used to define vertically heterogeneous media (Thomson, 1950; Tokimatsu, 1995; Foti, 2003, 2011; Socco and Strobbia, 2004; Strobbia and Cassiani, 2011), and they represent by now the most diffused techniques for Vs modeling. Surface wave methods are free from many practical and theoretical limitations of the classical body-wave analyses and they are free from the logistical effort of drilling (Boaga et al. , 2010, 2011; Vignoli et al. 2010, 2012; Foti et al., 2011). The technique requires an accurately recorded Rayleigh/Love wavefields to be analyzed for its dispersive properties, and the consequent inversion of the dispersion curve (e.g. phase velocity versus frequency). Surface wave methods can be divided in active methods, relying on the use of controlled sources, and passive methods, basing on the analysis of ambient noise, or microtremors. The active (controlled source) surface wave methods retrieve dispersion properties using several procedures: linear array methods as MASW (Park et al., 1999), coupled receivers methods as SASW (Nazarian et al., 1983) or single receiver methods as FTAN (Levshin et al., 1972; Nunziata et al., 1999; Boaga et al., 2010). Active methods basing on the use of controlled sources are accurate, but have limited exploration depth linked the ability of generating low frequencies with adequate sources. Passive methods have no control of the sources but, with the same array geometry, can reach deeper investigation depth thanks to the low frequency content of seismic noise and microtremors. The standard approaches for passive surface wave methods in shallow engineering applications derive for the seismological array processing. In seismological applications microtremors techniques use mainly 2D arrays of low frequency receivers in various and irregular geometries (Aki, 1957; Frosh and Green 1966; Tokimatsu et al. , 1992; Okada, 2003). 2D arrays deployment identifies in fact the source direction and it is required in order to identify the velocity and the direction of a wavefield which is not controlled and comes from unknown directions (Bonnefoy-Claudet et al., 2006). In exploration seismology common passive arrays techniques are: i) beamforming methods as F-K, ii) spatial autocorrelation SPAC (Aki, 1957), iii) extended Spatial autocorrelation ESAC (Ohori et al. , 2002), iv) high resolution arrays like MUSIC (Schmidt, 1981) and v) cross-correlation methods (Sabra et al., 2005; Shapiro et al., 2005). The logistic constraints of typical engineering applications often do not allow deploying large 2D arrays. Moreover, the required processing techniques, despite their relative simplicity, are often not available to the ‘practitioners’ community. This is the main reason for the success of the simple linear array passive method called ReMi (Refraction Microtremors: Louie, 2001). A linear array in a diffused wavefield has a response which is not function of its direction, but only of its size (length and receiver spacing). An averaged kinematics spectrum, such as the spectral density in the frequency wavenumber domain f-k, can be used to estimate the local propagation properties. The most important limitation of the ReMi approach is related to the basic assumption that the recorded data consist of a uniform wavefield. When this is not the case, the ReMi spectrum depends on the unknown source distribution, and its interpretation or