Statistical analysis on soil response of the Italian Seismic Network from the CRISP database

The characterization of the soil properties at seismic stations is extremely important for all the studies related to seismic network data. On the other hand, the availability of such a large dataset of site indicators offers the possibility to look for statistical correlations between different site indicators. In the framework of the 2022-25 PRIN-SERENA project (“Mapping seismic site effects at regional and national scale”, granted by Italian Ministry of University and Research), we use the information archived in the CRISP database relating to the site characterization of more than 400 stations belonging to the Italian Seismic Network (http://crisp.ingv.it/). 

We first analyze the distribution of the most significant indicators with large sample size: Horizontal-to-Vertical spectral ratio on both noise and earthquakes (HVSR), lithological classification, site and topography classes from national and european building codes. We then consider the HVSR as a reference proxy of site effect estimation at the stations sites and we look for relations with the others indicators. The cluster analysis of the HVSR curves highlights that about half of the stations have amplitudes reaching, on average, values of 4. They can also be grouped in four different shapes: flat curve (sites without HV amplification), amplification in the low-to-intermediate (f<2-3 Hz) and in the high (f>2-3 Hz) frequency ranges, large amplification for frequencies from about 1 to 3 Hz. Moreover,  the mean HV curves from noise maintain values lower or similar to those from earthquakes, whereas single noise peaks have greater amplitudes.

A not straightforward correlation with the other proxies is clearly recognizable, except a weak but significant dependence with the lithology and soil classes: the resonance frequency decreases as the soil characteristics deteriorate, and its amplitude slightly increases as the site characteristics degrade. The comparison of the results with the site correction term of the Local Magnitude shows that the combination of these two conditions causes an overestimation of the magnitude computation at about 10% of the seismic stations.