Constraining the near-surface Vs structure by the joint inversion of <i>SH</i> wave transfer function, Rayleigh wave dispersion and ellipticity information

SUMMARY We examine the possibility to improve near-surface VS models determined by Rayleigh wave dispersion and ellipticity inversions by incorporating the SH-wave transfer function in joint inversion schemes. We propose an optimal inversion approach by comparing such joint-inversion results against standard dispersion curve (DC) inversions, to also investigate the possibility to determine VS models that can be simultaneously compatible with both surface wave information, as well as with the local 1-D site-response. For the investigation we employ synthetic data consisting of Rayleigh wave dispersion and ellipticity, as well as the horizontally polarized shear wave (SH) transfer function for three 1-D test models. We introduce different levels of random errors in the model S-wave velocities and layer thicknesses, in an attempt to replicate typical deviations from 1-D structure and mimic their effect on real data. Seven types of joint inversions are explored with the use of the Neighbourhood Algorithm and compared to the corresponding inversion of only DC data. The results indicate that while the joint inversion of DC data with Rayleigh wave ellipticity or fundamental frequency, f0, information partially improves the estimated VS models (compared to DC-only inversions), a significant model improvement is observed when also including site ground response in the joint inversion, even for large random deviations (i.e. 20 per cent) from the 1-D starting models. This suggests that the incorporation of local site response in the inversion of Rayleigh wave data (DC and/or ellipticity) can improve near-surface geophysical modelling, while also providing VS models that are compatible with the observed (hence possibly also future) earthquake site response. We finally demonstrate the applicability of the proposed joint-inversion approach on earthquake and noise data from the ARGOstoli NETwork (ARGONET) array (Cephalonia, Greece).