The observable decay with frequency (f) of Fourier amplitude spectra for ground-motion recordings is controlled by a parameter κ: exp(-πκf). We analyze data from the KiK-net network, which is composed of stations with paired ground-motion sensors, one at the surface and one in a borehole. This study estimates κ for all KiK-net stations, using earthquakes recorded between 1998 and 2006. An inversion scheme is used to separate site, source, and path contributions to κ.
We correlate the site component of κ (κ0), estimated at the surface and the depth, with different S-wave velocity measures and with the fundamental resonant frequency of the site. The results show that the best correlations involve shallow soil S-wave velocity measures. The superficial layers of the soil predominantly influence κ0, but a remaining component with a deep origin is also observed. The source component of κ is small and presents a clear regional dependence, while no correlations with magnitude or depth of the earthquake are observed.
Data from the Next Generation of Attenuation (NGA) database are also used to estimate κ0 (surface site component) at the stations with a measured VS30 value. The VS30-κ couples estimated from both KiK-net and NGA data are compared with results from the literature, and a new correlation is established.
Most of the ground-motion prediction equations (GMPEs) for stable continental regions are derived for very hard rock sites (VS30>2000 m/s). In order to use these equations for standard rock sites (VS30 around 800 m/s), conversion factors are required. Using the new correlation VS30-κ and the host-to-target adjustment method of Campbell (2003), we compute amplification factors from very hard rock to rock sites. We show that in these conversions, the effects of both VS30 and κ have to be taken into account.