Site response function evaluation at Campi Flegrei accelerometric stations

Campi Flegrei is a nested caldera in Southern Italy well known to the geoscience community for its complex geology and recurrent bradyseismic events. Recently, the interest in the region renewed as a consequence of an ongoing bradyseismic crisis with a particular focus on the associated increment in seismicity. Being one of the most densely populated volcanic areas in the world, with a population of over 500K people, the high social impact that strong seismic events could have makes the monitoring of the area not only of interest for the seismological community but also a priority for the Department of Civil Defence. Consequently, the Department of Civil Defence recently installed several new accelerometric stations in the area, as part of the Italian accelerometric network (RAN), to improve the monitoring operations.

The main goal of this study was to characterize the site response functions at these stations. The analysis has been performed on the recordings from 16 stations for 40 events taking place between 2020 and 2024 with a magnitude between 2.5 and 4.4. A parametric implementation of the generalized inversion technique (GIT) based on a mixed-effects model has been used to model the Fourier amplitude spectra obtained from the recorded waveforms as the contribution of source, path, and site effects, with the inclusion of random effect terms to handle systematic biases related to specific events. The model has been constrained by fixing the frequency-independent site amplification at a reference station, which has been chosen based on a weighted scheme based on geophysical and geomorphological proxies, and then fitted to the data by minimizing the least squares errors. The site response functions (SRFs) have been then obtained by combining the frequency-independent site amplification and the high-frequency correction, obtained during the inversion, with the frequency-dependent site amplification obtained from the analysis of the residuals.

The attenuation parameters obtained from the inversion are consistent with the estimates available in the literature while the magnitudes estimated are in agreement with the ones reported in third-party databases. Furthermore, the results have been validated using a second semi-parametric GIT implementation, in which both source and site effects are treated non-parametrically. The high station-to-station variability of the SRFs at the 16 stations reflects the high geological complexity of the area. The SRFs obtained have been also compared to the corresponding HVSR results: the dominant frequencies are correctly determined using either the GIT approach or the HVSR analysis although the amplitudes from the latter method are generally inaccurate, as well known from the literature. No relevant correlation has been found between the SRFs and the results from (first level) seismic microzonation suggesting that a more detailed analysis is required to provide a correct characterization of the seismic response in the area. The obtained SRFs and seismic spectral model have multiple applications, ranging from Civil Defence purposes to seismic engineering.