Ambient noise data analysis for the reconstruction of the deep subsurface model in the Roman area

Abstract

During the last years the use of passive seismic exploration techniques as recording microtremors or microseism with an array of 3-component sensors become an attractive tool for in-situ measurements of shear-wave velocity Vs due to the easiness to use passive sources as well as to the possibility of investigating deep sedimentary structures at a very low cost compared to crosshole or downhole techniques. The applications of this method are not limited to earthquake engineering but may also be extended to general soil characterization. The main assumptions behind noise-array analysis are that ambient vibrations are mostly composed of surface waves and that the ground structure is approximately horizontally stratified as one-dimensional heterogeneous media; surface waves are dispersive and show a variation of the apparent velocity with the frequency. Another method that is based on ambient vibrations is the HVSR method of Nakamura (1989), which has proven to be a convenient technique to estimate the fundamental frequency of soft deposits (Lachez and Bard, 1994; Lermo and Chavez-Garcia, 1994). This procedure has the potential to allow estimates of shear-wave velocity of the unconsolidated sediments. In one-dimensional structures, average HVSR can be assumed to measure the Rayleigh waves ellipticity (defined as the ratio between the horizontal and vertical displacement eigen-functions in the P-SV case, at the free surface of the fundamental mode Rayleigh wave); hence the shape of HVSR can be used to estimate the shear wave velocity profile. Array methods, established by Horike (1985) after the pioneering work done by Aki (1957), make use of the dispersive character of surface waves, and allow to determine shear-wave velocity profiles from the inversions of dispersion curves; Fäh et al. (2001) proposed an inversion scheme to observed HVSR ratios from ambient noise in order to retrieve the S-velocity structure from a single ambient vibration record. In the present work, experimental ambient noise data was collected in 4 selected sites nearby the city of Rome, properly chosen because of the good knowledge of geometry of soft sediment and bedrock, improved by seismic reflection lines, deep and shallow borehole from literature by using 2D Array with variable geometry and aperture; were then analyzed with 3 different methods M-SPAC or Modified Spatial Autocorrelation method, f-k or frequency wavenumber method and Cross Correlation for the extraction of the characteristic dispersion curves of each site and also for each noise measure, the HVSR (ratio between horizontal and vertical components of the natural soil shaking) spectral ratio was calculated with the consequent identification of the fundamental resonance frequency for each site. The dispersion curves were extracted with the three different methods, well defined in different frequency ranges. Since the velocity of the S waves with the depth is closely related to the wavelengths and frequencies of phase velocity and that the depth of investigation is a function of the frequency, the three curves have been joined to broaden the frequency band which was then used in the inversion process along with the HVSR peak to obtain deep underground models. The results obtained showing how the use of different array analysis methods or the f-k methods, the M SPAC method and Cross Correlation has allowed to characterize the coherent part of the wavelength in terms of propagation properties and the implementation of the methodologies outlined in this paper has allowed to highlight the potentialities of the individual methods, their applicability limits and the stability of results and the joint use of these different analysis method help to enlarge and obtain a reliable large frequency range; furthermore the joined inversion of array dispersion curves and HVSR data in the inversion process allows to define a velocity deep subsurface model down to about 600 meters.