The seismic network of Ischia island from 1993 to 2021 Open Access

Ischia is a volcanic island belonging to the Phlegraean volcanic district, SW of the city of Naples. Its history is studded with eruptions, the first one being the AD 1302 Arso eruption (Iacono 1996). After that moment, a long period of volcanic quiescence began. In the last thousand years, the centre of the island has been affected by a significant resurgence (c. 1000 m), creating the resurgent block of Mt Epomeo (Acocella and Funiciello 2006).

On Ischia island, the volcanic hazard pairs with the seismic one because the island was shaken by several earthquakes that caused considerable damage and casualties (e.g. Selva et al. 2021; Giudicepietro et al. 2021; D'Auria et al. 2018). Seismicity is mainly concentrated north of the Mt Epomeo structure. The last damaging earthquake (M_w 3.9) occurred on the 21 of August 2017, causing two casualties and extensive damage in the village of Casamicciola, in the north part of the island (Azzaro et al. 2017).

In this volcanological and seismological context, a review of the history of the seismic network used to feed the seismic catalogue is important to understand the performance of the network and possibly to justify an increase in the seismicity.

After the Ischia earthquakes of 1881 and 1883 (M_w = 4.1 and 4.8, respectively) (Carlino et al. 2021), the concern of the Italian Government about the geological characteristics of this volcanic island increased. In late nineteenth century, this interest was led by scientist M. S. de Rossi who proposed the creation of a meteorological and geodynamical observatory to monitor seismic, volcanic, hydrothermal and hydrological phenomena on the island. In 1884, the Italian Ministry of Agriculture appointed Giulio Grablovitz, a pioneer geophysicist from Trieste, to undertake the project and develop the observatory. Grablovitz was the scientist who conceived the seismic tank, an innovative instrument at that time, capable of recording seismic events at great distance from the epicentre. The observatory was built in the area of ‘Gran Sentinella’, where Grablovitz installed a seismic tank (Fig. 1). Other seismic and meteorological instruments were installed in this observatory and in another building that was located at the Ischia port. The first diagram recorded by the seismic tank dates back to the 18 of October 1895. Thanks to this instrument, Grablovitz recorded some of the most powerful earthquakes of the late nineteenth century; e.g. the one registered on the 15 of June 1896 in Sanriku, Japan (M_w = 8.5) (Ferrari 2009; Luongo et al. 2012). The observatory, called the ‘Osservatorio Geofisico’, worked until 1923.

Seventy years later, the first modern seismic station installed by the Osservatorio Vesuviano (OV) on the island was OC9 (Fig. 2), set within the Osservatorio di Casamicciola. In the following 3 years, other two seismic stations were installed on the island, FO9 in the Forio lighthouse and CAI in the Aragonese Castle (Fig. 2). The seismic network was further updated by researchers from the OV, who installed digital stations along with the analogic short-period ones, and consequently changed the names of the stations from OC9 to IOCA, and from FO9 to IFOR. The upgrade of these sites was done by installing two three-component broadband sensors (60 s lower corner frequency) both coupled with GILDA digitizers developed at the OV (Orazi et al. 2006). The analogue station CAI was improved by upgrading it from a single to a three-component sensor.

In 2011, the IOCA site was upgraded: an accelerometer was installed in the Osservatorio di Casamicciola, next to the analogue short-period and to the digital broadband velocimeter, with the aim to record possible earthquakes with magnitudes comparable to the ones of 1881 and 1883 (Capello et al. 2011). As for the velocimeter, the accelerometer was paired with a GILDA digitizer as well. In 2014, the fourth seismic site, IMTC, was installed in the inner part of the island, on the west side of Mt Epomeo (Fig. 2). The network was considerably improved by the OV and the Istituto Nazionale di Geofisica e Vulcanologia (INGV) after the earthquake of the 21 of August 2017 (M_d = 4.0) that struck the town of Casamicciola (0.25 g measured less than 800 m from the epicentre; De Novellis et al. 2018) and caused two fatalities.

In the following months, five sites (IBRN, IMNT, IPSM, IVLC, PTMR) distributed on Ischia island and on the adjacent island of Procida (D'Auria et al. 2018) were equipped with broadband velocimeters and/or accelerometers. Some mobile seismic stations were also installed by the SISMIKO group (an operational group of INGV) some days after the earthquake of the 21 of August 2017 (Galluzzo et al. 2017), but most of them have been lately removed. Only two mobile seismic stations are currently working on the island (red dots in Fig. 2). These two stations will be removed in the near future because their housing sites are temporary. The improved network allowed to record the few and small (M_max = 2.5) aftershocks of the August 2017 earthquake, and has been fundamental to constrain its lack of aftershocks.

We analysed the signals recorded by the permanent seismic network deployed since 1993 on the islands of Ischia and Procida and transmitted in real time to the monitoring room of the OV and the two mobile seismic stations that are currently working on Ischia. We analysed the time evolution of the sensitivity of the seismic network to earthquake detection and location in order to evidence its improved performance.

The seismic network of Ischia is currently set up with digital stations equipped with two different data loggers: GILDA, developed by the OV (Orazi et al. 2006) or Guralp DM24 (Güralp Systems Ltd., UK). On the island of Ischia, three analogue seismic stations are still working: IOCA (formerly OC9) and IFOR (formerly FO9) co-placed with digital stations for redundancy and comparison with old data, and the station CAI currently located at the Aragonese Castle, on the eastern part of the island. An improvement of this site was not considered given that it is highly noisy. The seismic sensors installed on the island are three-component short-period velocimeters, broadband 60 s or broadband 30 s. In addition, four accelerometers Episensor ES-T (Kinemetrics,Pasadena, USA) are installed close to some of the velocimeters, with different full-scale amplitudes: 0.5 g on PTMR; 1 g on IVLC and IPSM; and 0.25 g on IOCA and IBCM. The full scale of the station IOCA, equal to 0.25 g, has not been modified since its installation in 2011. The other ones were chosen taking into account the site quality and the distance from the source of all the highest magnitude earthquakes recorded on the island, i.e. Casamicciola town. The configuration of the seismic network is shown in Figure 2 and the characteristics are reported in Table 1. The selection of the station sites was conditioned by various factors: site accessibility, anthropic noise, site effects and possibility of using pre-existing and/or electricity-supplied structures.

The computing of the power spectral density (PSD) of seismic noise is the standard method for quantifying seismic background noise and, consequently, a way to define the quality of a site.

The PSDs and their probability density functions (PDF) were computed from the waveform of the noise deconvolved by the corresponding instrumental response functions using ObsPy, a Python toolbox for seismology (Krischer et al. 2015). The computation of PSD and PDF is based on the algorithm described by McNamara and Buland (2004). For each seismic channel, the software calculates the PDF of the distribution of the PSD values at each spectral interval. PDFs provide the occurrence probability of a seismic signal level in a frequency interval.

We plotted (see Figure 3) the PDF of the PSDs of some of the digital stations installed in different places on the island to provide information on the site quality of the seismic network of Ischia.

Ground shaking largely depends on local geological conditions, which typically modify ground motion in terms of amplitude, frequency, wave composition and duration. Therefore, the characterization of the local ground conditions of a station site should be the first part of any site description. The horizontal/vertical (H/V) spectral ratio is the ratio between the Fourier amplitude spectra of the horizontal and the vertical components of noise. The H/V peak frequency is well correlated with the fundamental resonance frequency (f₀) of the site (Nakamura 1989) and, for this reason, is used as an indicator of the underground structure features. Scientific discussions are still under way about the fundamental-mode surface wave contribution to H/V ratios when performed on unconsolidated sediments (e.g. Fäh et al. 2001, 2003; Tramelli et al. 2010). Nakamura (2000) asserted that the peak in the H/V is the result of vertically incident SH waves (horizontally polarized shear-waves), without the contribution of the fundamental-mode Rayleigh wave. Nevertheless, caution must be exercised when considering H/V curves (see SESAME (Seismotectonic and Seismic Hazard Assessment of the Mediterranean Basin) manual for details; Ohrnberger 2005; Acerra et al. 2004). Several authors have highlighted that the fundamental resonance frequency (f₀) represents one of the most relevant indicators for site characterization (e.g. Parolai et al. 2002; Thabet 2019; Cultrera et al. 2021). In order to conduct the characterization of the subsoil at the recording sites of the Ischia seismic monitoring network, we applied the H/V spectral ratio technique to at least 3 hours of seismic noise, recorded during the night of the 1 of August 2021. To remove the effect of seismic transient in the signals (non-stationary anthropogenic sources located near the station sites), a ‘detrigger’ algorithm (Wathelet et al. 2005) was applied on 120 s signal sliding windows. The Fourier spectra were calculated on the windows selected with the ‘detrigger’ algorithm, then they were smoothed using the Konno–Ohmachi algorithm (Konno and Ohmachi 1998). The H/V curves were then calculated using the geometric mean of the spectra of the horizontal components divided by the spectrum of the vertical one (Fig. 4). Figure 5 shows the average value of the spectral ratios for each station, calculated for the individual windows according to the procedure described above.

Location and detection sensitivity analysis is an important tool to define the performances of a seismic network and to correctly interpret a seismic catalogue. It is a fact that a lack of resolution of the seismic network in an area cannot be interpreted as a lack of seismicity in the same area. We applied the method developed by Tramelli et al. (2013) to estimate the time variation of the network sensitivity.

We computed the theoretical amplitudes of simulated events with a Brune's (1970) source model, assuming a fixed value of 0.1 MPa for the stress drop, a very low value but compatible with the results obtained by Carlino et al. (2021), who studied the source and attenuation parameters of the seismicity of Ischia. An average rock density of 2200 kg m⁻³ and a mean quality factor Q = 200 were assumed for the analysed area (Nardone et al. 2020). The assumed rock density and stress drop affect the low-frequency spectral amplitude, while elastic and anelastic attenuation directly affect the amplitude of the waves recorded by the receiver. Event–station distance, quality factor Q (inverse of attenuation), density and stress drop are thus the parameters that influence the signal-to-noise ratio (SNR) at the recording station. We assumed a SNR of 1.5 for an earthquake to be detected by a station, and a minimum number of three stations detecting the earthquake to allow hypocentral location. Three detecting stations are considered enough for locating an event, because it is reasonable that the S-wave be identifiable at least on one of them, being all three-component stations. With these parameters, we obtained the results shown in Figure 6 for detection sensitivity, and in Figure 7 for location sensitivity. The contour maps of location sensitivity indicate the minimum magnitude that an earthquake with hypocentre at certain depth (1000 m in this case) must have to be located using the available seismic network; in our case, to be detected by at least three seismic stations.

We compared the high (NHNM—new high noise model) and low noise models (NLNM—new low noise model) of Peterson (1993) with the PSDs obtained for the island of Ischia. For direct comparison to the NLNM and NHNM, the PSDs are converted into decibels (dB) with respect to acceleration (m s⁻²)² / Hz (Fig. 3). Most of the stations of the seismic network are located in highly urbanized areas and their PSDs lie close to the NHNM curve, especially for periods below 0.1–2 s.

The H/V curves show (Figs 4 & 5) as a common feature a resonance pick below 1 Hz. In particular, a pick around 0.3–0.6 Hz is usually evident for all the seismic stations located on the island (see vertical grey band in Fig. 4 and Table 2). The amplitude of the peak is variable (from a minimum of 2.0 at the IMTC station to a maximum of 6.4 at the IBCM station). The sites placed near the north coast, such as IPSM and IVLC, exhibit the minimum peak frequency (0.35 Hz), while the stations placed in the central and southern part of the island, such as IMTC and IBCM, show the highest frequency peaks (0.6 Hz; Table 2). Nardone et al. (2020), through a joint inversion of dispersion data and the average of the H/V curves below 1 Hz, obtained the mean structure of S-waves velocity (Vs) and attenuation of Ischia. They showed a deep impedance contrast at about 1.1 km depth and associated it to an interface, which extends below the whole island, probably composed of laccolithic crystalline rocks (Nardone et al. 2020). These results support the assumption that, at large scale, the structure below the whole island can be approximated as a medium with 1D geometry.

Some sites show a second peak in the 1–10 Hz frequency band (IFOR, IMNT, IVLC and PTMR) (Figs 4 & 5a). It is plausible that the peaks in the 1–10 Hz frequency band could be related to an impedance contrast caused by a surface layer. An extended amplification band in a wide frequency interval (1–20 Hz) and with rather low amplitudes is present at the ICVJ site (Fig. 4). This case may reflect the presence of a complex geological environment close to the area of the epicentres of the main earthquakes that occurred on the island (1881, 1883 and 2017; Carlino et al. 2021).

Particularly in the ‘Gran Sentinella’ zone, the natural hill where the station IOCA is located, the H/V curves resulting from IOCA records show a very narrow peak, centred at 1.6 Hz, reaching the maximum amplitude of 7.22. As described in Vassallo et al. (2018), after the 2017 earthquake (M_d = 4.0), the Emersito++ Working Group (an emergency operating group of INGV) carried out several geophysical surveys aimed to study the site effects in the town of Casamicciola, which was the most damaged area. They installed a 2D seismic array on the ‘Gran Sentinella’ hill, near the IOCA site; by analysing the data recorded with this array, Nardone et al. (2017) identified the presence of a strong heterogeneity in the subsoil, highlighted also by the H/V peak, most likely due to lava domes.

The low frequency peak, present in all the H/V characterizing the island, influences the seismic response in a frequency band comparable to the source spectrum one. The estimate of the corner frequency and of the amplitude of ground displacement spectra at low frequency, and consequently of the moment magnitude of an earthquake, can be influenced by site amplification below 10 Hz, especially if all the seismic sites have the same amplification band. A detailed inversion of the site amplification curves of the stations present on the island and of the medium attenuation should be performed as soon as the number of recorded earthquakes makes it possible. This would allow to deconvolve the site amplification from the recorded signal and to have a more accurate estimate of the source spectrum.

Analysis of the sensitivity of the seismic network of Ischia over time shows the improvement from the impossibility to locate earthquakes in 1993 to the ability to locate even M_d = 0 earthquakes with a hypocentral depth of less than 1 km. Note that earthquakes with higher hypocentral depth are not expected in the island, being the brittle/ductile transition located at almost 2 km depth (Castaldo et al. 2017).

The detection capability in the north of the island can be considered relatively high also in historical periods after the 1883 Casamicciola earthquake, scientists have tested innovative seismic instruments in this location (Luongo et al. 2012). Since 1993, instrumental earthquakes have been detected thanks to the installation of the seismic station OC9 in the Osservatorio di Casamicciola. The earthquake catalogue starting from 688 before Christ was reconstructed by Selva et al. (2021). This catalogue is divided into historical and instrumental data. The epicentres of the earthquakes from the historical part of the catalogue were calculated calculating the centre of gravity of the sites with the highest intensities. For each epicentre, a reliability index associated to the data quality is reported in Selva et al. (2021). In that article, the authors describe the intensity–magnitude relationship used to estimate the moment magnitude of the historical seismicity, and the relation between moment and duration magnitude that was used for the instrumental catalogue.

The historical catalogue can be compared with the images of the location sensitivity and detection sensitivity in Figures 6 and 7. The analysis performed by Carlino et al. (2021) and Calderoni et al. (2019) on the source parameters gave new information on the stress drop of the 2017 earthquake of Casamicciola, which is very low (c. 0.1 MPa). We used this value for the simulated earthquakes in the sensitivity analysis. The current detection sensitivity is very high (M_d > −1) all over the island, including the neighbouring island of Procida. This result is corroborated by data from the seismic catalogue plotted in Figure 8 that has been reconstructed by Selva et al. (2021). Merging information coming from macroseismic and instrumental data, these authors analysed the spatial, temporal and magnitude distribution of earthquakes, showing that their distribution is significantly non-stationary and characterized by a b-value larger than 1. Their analysis evidenced that the seismicity of the island is characterized by isolated events, small swarms with few low-energy events concentrated in few months and sequences with a destructive mainshock accompanied by a low number of aftershocks. Thanks to the improvement of the seismic network described here, it was possible to clearly understand that the seismicity of the island is not only concentrated in the area of Casamicciola, but it is also located in the central western part of the island, especially along the faults surrounding the most uplifted part of the resurgent block of Mt Epomeo. Analysis of the sensitivity shows that, in the past (before 2015), the seismic network was blind to low duration magnitude events (less than 0) in the area of Monte Corvo, but this gap was resolved thanks to the installation of the IMTC station. No earthquakes were recorded in the east and SE part of the island or in the island of Procida, but the sensitivity of the network was low. Thanks to the network currently installed, we are able to detect and locate even low magnitude earthquakes in this area.

The seismic network deployed on Ischia island has been highly improved over the years, leading to a dense network. The configuration of a new seismic network requires its characterization in terms of site quality and network sensibility.

The first seismic station was installed in 1993, but it was only until 1996, thanks to the installation of two additional stations, that the network allowed to locate events on the island. Before 2017, the study of the historical seismicity of Ischia led researchers from the OV to implement the seismic network of the island by installing an accelerometer at the Osservatorio di Casamicciola site (IOCA accelerometer, in 2011; Capello et al. 2011) and the new IMTC site (2015) in the central part of the island. As it happened after the 1883 Casamicciola earthquake, the 2017 Casamicciola earthquake (M_d = 4.0) drew attention to the seismic potential of the island and a substantial upgrade of the seismic network was conducted. Immediately after the earthquake, several temporary seismic stations were installed on the island (Galluzzo et al. 2017). Most of them were installed in houses or hotels abandoned due to the earthquake. The permanent seismic stations were installed in the following months, reaching the current configuration in 2019.

Usually, islands and highly populated areas are characterized by low signal-to-noise ratio due to natural (sea waves) and anthropic noise. Ischia has both these drawbacks. In addition, the seismic network maintenance depends on the possibility to reach the island. Despite this, the best compromise between low-noise and reachability of sites led to a network that is currently able to locate shallow (≤1 km) earthquakes with duration magnitudes higher or equal than 0, especially in the northern part of the island, and to detect events with a duration magnitude higher than −1 close to the station sites on both Ischia and Procida. These values are confirmed by the seismic catalogue of the detected events on the island of Ischia (https://www.ov.ingv.it/index.php/cataloghi-sismici-vulcani-napoletani/933-ct-isc-1999-2021/file). The stations IOCA and IFOR are the most used to detect earthquakes as they are close to the seismogenetic area and have a noise level above 1 Hz, which is lower with respect to the other stations. Almost all the stations show a high noise level, especially below 1 Hz. This can invalidate the estimate of the amplitude of ground displacement spectrum at low frequencies and, consequently, the estimate of the seismic moment magnitude for higher magnitude earthquakes. For low magnitude earthquakes, with corner frequencies higher than 1, this noise does n't affect the detection capability of the station.

The authors have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

AT: conceptualization (lead), data curation (equal), formal analysis (lead), software (equal), supervision (lead), writing – original draft (lead), writing – review & editing (lead); MO: conceptualization (equal), data curation (lead), validation (equal), visualization (equal); LN: data curation (equal), formal analysis (equal), writing – original draft (equal); ABo: data curation (equal), formal analysis (equal), validation (supporting), writing – original draft (supporting); ABe: data curation (supporting); CB: data curation (supporting); MCap: data curation (supporting); AC: data curation (supporting); MCas: data curation (supporting), writing – original draft (supporting); LD: data curation (equal); WDC: data curation (supporting); ADF: data curation (supporting); DG: data curation (supporting), writing – original draft (supporting); GG: data curation (supporting); FG: data curation (supporting), writing – original draft (supporting); FL: data curation (supporting); DLB: data curation (supporting); MM: data curation (supporting); CM: data curation (supporting); RP: data curation (equal); PR: data curation (supporting); GS: data curation (supporting); VT: data curation (supporting); FB: writing – original draft (supporting).

The Italian Presidenza del Consiglio dei Ministri-Dipartimento della Protezione Civile (DPC) supported the monitoring activities at Ischia and the studies performed for this article.

Seismic waveforms, locations and magnitudes of Ischia earthquakes since 2018 are available at https://terremoti.ov.ingv.it/gossip/ischia/index.html.

Continuous traces are available at http://www.orfeus-eu.org/data/eida/.