Site effects using methods based on lateral homogeneity and laterally heterogeneous media: An impossible marriage?

Site effect studies are currently mostly based on seismic noise measurements, processed using methods like horizontal-to-vertical spectral ratios (HVSR) to estimate dominant frequencies and maximum amplifications, or methods based on cross-correlation between stations like spatial autocorrelation (SPAC) and seismic interferometry. However, HVSR and SPAC assume that the underlying medium has a 1D structure and that seismic noise consists mostly of surface Rayleigh waves. We present a detailed analysis of seismic noise records at two sites in Euroseistest (Mygdonian basin, northern Greece) where two normal faults have created lateral heterogeneity in the subsoil structure. Site GRA is located near fault F3, with a larger throw than fault F2, close to site FRM. Our results show the effects of these irregularities. Dominant frequency results for both sites become spatially incoherent close to the fault. In the case of site GRA, a range of wavelengths is missing in the recorded seismic noise field. These effects extend up to some fraction of the affected wavelengths in the case of fault F3. The effect of lateral variations on seismic noise measurements will be lost in gross spatial averaging of the results or simply the failure of HVSR or SPAC methods will be observed.