Horizontal-to-Vertical Spectral Ratios (HVSR) of NW Italy seismic stations elaborated during the "FOCUS-HVNEA" NOA-ILGE project

This data set consists of Horizontal-to-Vertical Spectral Ratios (HVSR) resulting from the application of the software package HVNEA (HV Noise and Earthquake Automatic Analysis) with the aim of comparing them with those resulting from the application of another method, namely STATION (Seismic sTATion and sIte amplificatiON). The results, relative to more than 24,000 HVSR, derive from the processing of 700,000 seismograms recorded over different time periods by 8 stations of the networks IV (Italian Seismic Network), GU (Regional Seismic Network of North Western Italy) and GV (Mobile RSNI). To compare the results of the two methods as accurately as possible, the waveforms were subjected to the same preprocessing already used to elaborate the results stored in the STATION database. To this end, the methodological workflow applied with HVNEA for station IV.MURB involved the selection of segments from continuous recordings for each event reported in the INGV catalogue located within a radius of 120 kilometres from the station. Starting from the automatically picked S-wave onsets, 12-second windows were then extracted and used for the analysis of earthquake recordings. Regarding the noise analysis, it should be noted that STATION again considers 12-second windows selected before the P-wave onset, while HVNEA requires the use of a signal window of at least 60 seconds. A window of 3,600 seconds was used for the analysis. The comparison of the HVSR was performed in the frequency band 0.1–15 Hz. All analysed curves, for both earthquake and noise recordings, show generally similar shapes and identify significant peaks in correspondence of the same frequency ranges, although the amplitudes obtained with STATION are systematically higher than those obtained with HVNEA. To obtain a quantitative comparison, various statistical metrics commonly used to measure the discrepancy between data sets were applied, namely the Mean Squared Error, the Mean Absolute Error and the Pearson Correlation Coefficient. This publication results from work conducted under the transnational access/national open access action at the Site effects Laboratory – INGV L’Aquila supported by WP3 ILGE–MEET project, PNRR–EU Next Generation Europe program, MUR grant number D53C22001400005.

The HVSR method is widely used for assessing site effects and seismic microzonation applications. It involves calculating the ratio of horizontal to vertical ground motion spectra at a single station to estimate the fundamental resonant frequency of the site. With the ever-increasing number of seismic stations and data, the need for automated tools for efficient routine analysis of both environmental vibrations and earthquake data is growing. To compare the results of the two methods as accurately as possible, the waveforms were subjected to the same preprocessing already used to elaborate the results stored in the STATION database. To this end, the methodological workflow applied with HVNEA for station IV.MURB involved the selection of segments from continuous recordings for each event reported in the INGV catalogue located within a radius of 120 kilometres from the station. Starting from the automatically picked S-wave onsets, 12-second windows were then extracted and used for the analysis of earthquake recordings. Regarding the noise analysis, it should be noted that STATION again considers 12-second windows selected before the P-wave onset, while HVNEA requires the use of a signal window of at least 60 seconds. A window of 3,600 seconds was used for the analysis. The comparison of the HVSR was performed in the frequency band 0.1–15 Hz.