Ground-motion dependency on seasonal variations; observations at the ARGONET Array, Cephalonia, Greece

We use 964 earthquake records from the ARGONET, Cephalonia, Greece, array database to investigate the possible seasonal variation of ground motion at two surface and four borehole stations. We first apply the standard spectral ratio (SSR) method to compute the spectral site amplification factor, SAF(f), using as reference the deepest borehole station of the array, which is located at the engineering bedrock. We identify clear manifestations of a seasonal variation at both surface stations, even though one of them is installed on rock. At the deeper stations, the seasonal variation is only detectable in the high-frequency part of the vertical component at shallow depth, or is not detectable at all. This implies that the driving mechanism of the phenomenon operates at shallow depth. The identified good correlation of the observed cyclic variation with meteorological parameters, that is, rainfall and air temperature, suggests that the atmosphere-soil interaction and the hydrological system contribute to this mechanism. The seasonal variation at the two surface stations is observed at high frequencies (generally > 20 Hz). We show that correcting for this variation in SSR through appropriate modeling leads to more reliable estimates of the mean SAF at high frequencies, by reducing the associated standard deviation. As a second step, we compute horizontal-to-vertical spectral ratios (HVSRs) using the same dataset and verify that the dominant features of the identified seasonal variation (or its absence at deeper stations) are also traceable by this method. This suggests that at sites with abundant data, HVSRs could provide direct information on the importance of the seasonal variation in ground motion, without the need for a reference station.