Large-Depth Ground-Penetrating Radar for Investigating Active Faults: The Case of the 2017 Casamicciola Fault System, Ischia Island (Italy)

Abstract

We conducted large-depth Ground-Penetrating Radar investigations of the seismogenic Casamicciola fault system at the volcanic island of Ischia, with the aim of constraining the source characteristics of this active and capable fault system. On 21 August 2017, a shallow (hypocentral depth of 1.2 km), moderate (M_d = 4.0) earthquake hit the island, causing severe damage and two fatalities. This was the first damaging earthquake recorded on the volcanic island of Ischia from the beginning of the instrumental era. Our survey was performed using the Loza low-frequency (15–25 MHz) GPR system calibrated by TDEM results. The data highlighted variations in the electromagnetic signal due to the presence of contacts, i.e., faults down to a depth larger than 100 m below the surface. These signal variations match with the position of the synthetic and antithetic active fault system bordering the Casamicciola Holocene graben. Our study highlights the importance of employing large-depth Ground-Penetrating Radar geophysical techniques for investigating active fault systems not only in their shallower parts, but also down to a few hundred meters’ depth, providing a contribution to the knowledge of seismic hazard studies on the island of Ischia and elsewhere.

1. Introduction

In this paper, we focus on the identification at depth of the synthetic fault system of the Casamicciola Holocene graben, which was responsible for the 2017 earthquake, by using a large-depth GPR system (hereinafter LDGPR). This was already employed successfully to study buried faults at depth in different geological frameworks [1]. Our LDGPR system was designed for high-conductivity soils at the Pushkov Institute of Terrestrial Magnetism, Ionosphere, and Radio Wave Propagation (IZMIRAN) for a planned space mission to Mars [2]. A characteristic feature of our LDGPR is the use of low frequencies (tens of MHz) combined with high energy potential, which allows for probing highly conductive soils such as wet clay that are inaccessible for standard GPR.

The seismotectonic setting of the Ischia Island in the Bay of Naples, Southern Italy is characterized by an active resurgent caldera that has experienced several historical catastrophic earthquakes, making Ischia an area of high seismic risk. The island is densely inhabited and a major tourist destination, further increasing the importance of understanding the seismotectonics of the Ischia earthquake source (Figure 1).

On 21 August 2017, a main shock occurred at 20:57 CEST (18:57 UTC), rated Md 4.0, reactivating the Casamicciola Terme capable fault system. Despite the moderate magnitude, several buildings and a church collapsed, causing two deaths and 42 injuries and leaving 2600 people homeless. This earthquake is important because it is the first well-documented, significant instrumental earthquake at Ischia, occurring 134 years after the catastrophic 28 July 1883 I° XI MCS event, which killed more than 2300 people. For the first time, detailed mapping of the primary co-seismic ruptures of the 2017 earthquake was carried out, showing extensional surface faulting of the synthetic fault system of the Casamicciola Holocene graben [3,4,5] (Figure 1). Other very strong damaging earthquakes (occurring in 1796, 1828, 1881, and 1883) hit Ischia Island during the last three centuries, with similar seismotectonic characteristics [6,7,8,9].

The epicentral area of the 2017 earthquake is confined within just a few square kilometers in the Casamicciola Terme area at the base of the N flank of Mt. Epomeo (787 m a.s.l.). Here, a clear stratigraphic and morphological trace highlighting the effects of the earthquakes is represented by a graben structure, which originated as the result of Holocene extensional tectonic deformation [10,11]. Mt. Epomeo is the main morphological structure created by the Ischia caldera resurgence, with an average uplift rate of ca. 3 cm/yr in the last 30 kyr [11].

Figure 1. Location of our geophysical survey, Ischia Island (Italy): ISH1, ISH2, ISH3, ISH4 are the GPR profiles (black lines); TDEM1-7 are the time-domain electromagnetic measurements (black squares) across the synthetic and antithetic active faults system of Holocene Casamicciola graben (red lines, marks on down-thrown side). Geology and active faults are from Vezzoli (1988) [10] and Tibaldi and Vezzoli (1998) [11]. The co-seismic ruptures of the 2017 earthquake (yellow lines) are from Nappi et al. (2018) [3]. The 21 August 2017 mainshock (the yellow large star) is from https://terremoti.ov.ingv.it/gossip/ischia/2017/index.html (accessed 1 July 2024) [12]; historical earthquakes (red circles) are from Selva et al. (2021) [13].

Figure 1. Location of our geophysical survey, Ischia Island (Italy): ISH1, ISH2, ISH3, ISH4 are the GPR profiles (black lines); TDEM1-7 are the time-domain electromagnetic measurements (black squares) across the synthetic and antithetic active faults system of Holocene Casamicciola graben (red lines, marks on down-thrown side). Geology and active faults are from Vezzoli (1988) [10] and Tibaldi and Vezzoli (1998) [11]. The co-seismic ruptures of the 2017 earthquake (yellow lines) are from Nappi et al. (2018) [3]. The 21 August 2017 mainshock (the yellow large star) is from https://terremoti.ov.ingv.it/gossip/ischia/2017/index.html (accessed 1 July 2024) [12]; historical earthquakes (red circles) are from Selva et al. (2021) [13].

The interpretation of the Casamicciola structure as a Holocene graben due to the uplift of Mt. Epomeo was initially proposed by Rittmann in 1930 [14]. This interpretation was later substantiated and extensively documented by field geological data provided by Vezzoli (1988) [10], which included the first modern 1:10.000 scale geological map of Ischia Island. Subsequent studies by Tibaldi and Vezzoli (1998) [11] and Sbrana et al. (2018) [15], among others, further developed and confirmed this understanding of the geological structure of the area.

After the 2017 mainshock in Casamicciola, various models of the seismogenic sources of the earthquake have been proposed based on different data sets, leading to an ongoing scientific debate. Seismological studies by Braun et al. (2018) [16], GPS data from De Novellis et al. (2018) [17] and Devoti et al. (2018) [18], InSAR measurements by Albano et al. (2018) [19], Montuori et al. (2018) [20], Trasatti et al. (2019) [21], and Carlino et al. (2022) [22], and geological-macroseismic data sets analyzed by Nappi et al. (2018, 2020, 2021) [3,4,5] have contributed to these interpretations. These diverse analyses, utilizing different data sets, sometimes conflicting, have fueled the ongoing scientific discussion regarding the nature of the structure responsible for the recurrent earthquakes in Casamicciola. Considering its seismological characteristics, we infer that that the source of the 2017 Casamicciola earthquake is the same as that of all the strong historical earthquakes at the island.

As regards previous applied geophysics studies of the island, they involve gravity data [23], aeromagnetic data [24] and references therein, and their combined analysis [25,26], performed on the whole island. The central-western sector of the island was investigated by a magnetotelluric (MT) survey carried out along two main profiles that provided an electrical resistivity map to a depth of 3 km [27]. A characterization of the shallow structure of the southwestern Ischia’s hydrothermal system was carried out by Di Napoli et al. (2011) [28] with a multidisciplinary survey including time-domain electromagnetic (TDEM) and resistivity tomography studies. A seismic tomography study was carried out all over the island by Capuano et al. (2015) [23]; more recently, Nardone et al. (2023) [29] investigated the elastic and electric characteristics of the shallower underground portion of the northern sector of Ischia Island using the Horizontal-to-Vertical Spectral Ratio (HVSR), seismic array techniques (f-k), polarization analysis, and TDEM data. None of the studies carried out so far at Ischia involved Ground-Penetrating Radar (GPR) data to characterize its active structures.

GPR is a geophysical electromagnetic (EM) technique that uses microwaves in the frequency range of 10 MHz to 3 GHz to investigate shallow buried structures. The choice of operating frequency is a compromise between the need to investigate deeper regions using lower frequencies and the need for higher resolution using higher frequencies. The dielectric constant improves the achievable resolution, but electrical conductivity leads to soil loss and limits the depth of investigation. GPR is commonly used in engineering studies to identify shallow subsurface structures and sources. However, its usefulness in geological studies is compromised by the limited exploration depth that can be achieved. Previous studies have involved only shallow parts of fault systems through 2D and 3D surveys, e.g., [30] and references therein. Furthermore, GPR has also been successfully used to identify normal faults in volcanic environments, e.g., [31]. Here, we exploit the effectiveness of a low-frequency, large-depth GPR system (LDGPR), the Loza 2N system (Figure 2). The Loza 2N GPR operates in the tens of MHz frequency band, capable of investigating large depths and therefore effective for the study of buried faults.

We constrained our LDGPR interpretations thanks to four TDEM soundings. The TDEM method involves sending a transient electromagnetic pulse into the ground and measuring the resulting electromagnetic response connected to the variations in soil electric conductivity. The method induces electric current flow within the subsurface and, after the transmitted signal is shut off, measures a voltage signal that is returned from the underground materials. The returning signal is measured as a function of time, and data are inverted to recover a layered earth model of resistivity, e.g., [32,33] and references therein. In volcanic areas, TDEM might be effective at detecting the changes in electrical resistivity characterizing volcanic rocks with respect to the surroundings and at mapping subsurface hydrothermal systems by detecting variations in electrical conductivity associated with fluid pathways, e.g., [34,35,36] and references therein. At Ischia, we carried out four TDEM surveys (in a square central loop configuration with a side of 50 m) next to the LDGPR profiles. This allowed for obtaining 1D models of the underground through the identification of several electro-layers, in addition to yielding the calibration of the depth of the high-resistivity bedrock.

2. Survey and Data Analysis

The Ground-Penetrating Radar (GPR) survey consists of sending electromagnetic microwaves under the subsurface, using a frequency range between 10 MHz and 2.5 GHz, with the aim of identifying possible permittivity contrasts in the underground by transmitting high-frequency radar pulses and recording the amplitude and delay time of the reflected pulses, e.g., [37]. Permittivity, along with magnetic permeability and conductivity, are the main parameters influencing the radiation of electromagnetic waves in the underground. Ground-Penetrating Radar is a highly valuable tool for shallow geophysical exploration, in terms of spatial resolution, portability, and costs. Volcanic deposits are suitable for the use of GPR, as they are typically thin (~100 m), shallow, and compositionally homogeneous and can be electrically resistive. These features may allow for good transmission of radar energy, e.g., [38].

The low-frequency, large-depth Loza GPR system (LDGPR) is equipped with two unshielded monostatic antennas, one Tx and one Rx, placed at the mutual distance of 6 and 10 m, respectively, for generating radar pulses at the central frequencies of 100, 25, and 15 MHz, e.g., [39]. The two antennas can be moved along the profiles using a parallel and/or an in-line array (Figure 2).

More specifically, our LDGPR systems consists of (i) a high-power transmitter unit of 15 kV, 50 MW; (ii) a receiving unit to record broadband pulses with direct signal digitization, without stroboscopic conversion and 120 dB dynamic energy; (iii) two 6–15 m long Wu–King antennas without cable connection and with the radio trigger placed on foils made of high-density polyethylene resistant to rubbing (Figure 2) that are resistively loaded dipoles covered by a dielectric layer that avoids the dispersion of the transmitted signal; (iv) a controller connected to the receiver unit through a coaxial shielded multipolar cable.

This system allows for digitizing signals with no requirement for multiple pulse repetition and waveform storage. The capability of direct digitizing allows the Loza instrument to be moved rapidly, with no loss in the quality of measurements with respect to traditional GPR systems.

The transmitting (Tx) antenna radiates an electromagnetic signal into the probed medium, and a receiving (Rx) antenna gathers the portion of energy reflected/backscattered by the electromagnetic anomaly structures. A 2D image of the underground is obtained by moving Tx and Rx antennas along a line and by collecting, at each measurement point, the backscattered field as a function of the signal propagation time along the survey profile. This 2D image is also referred to as radargram or a B-scan and is composed of several A-scans, with uniform steps. The radargram provides a representation of the subsurface features, wherein localized objects appear as hyperbola, and material interfaces appear as constant signals. B-scans allow one to detect the levels of sharp variations in the received signal and significantly simplify the geological interpretation.

As regards GPR resolution versus the depth of investigation, there is a trade-off between GPR vertical resolution and depth penetration. Higher-frequency antennas offer better vertical resolution but can only reach shallower depths. Achieving the optimal balance between resolution and depth is crucial for successful prospections using GPR. The GPR horizontal resolution is directly related to the size of this footprint. At shallow depths, the footprint is smaller, resulting in better horizontal resolution. As the depth increases, the footprint expands, leading to a decrease in horizontal resolution, e.g., [40].

The depth, z, of the targets is estimated by converting the round-trip travel time, t, according to the following relation:

$$z={\displaystyle \frac{c\cdot t}{2\sqrt{\epsilon }}}$$

(1)

wherein $c=3\times {10}^8$ m/s represents the light propagation velocity, and ε is the average relative dielectric permittivity of the investigated medium, e.g., [41].

We collected four LDGPR profiles at Ischia Is., three parallel-trending N-S profiles approximately 1300 m long (namely ISH1, ISH2, ISH4), spaced about 600 m from each other and orthogonal to the expected direction of the fault. The fourth profile (ISH3) trends NW-SE and is about 1 km long (Figure 1).

Table 1. Acquisition parameters of the LDGPR survey at Ischia Island.

ID Data	Length (m)	Spatial Step (m)	Frequency (MHz)	Dielectric Permittivity Average	Max. Estimated Depth of Signal (m)	Easting Start (UTM—m)	Northing Start (UTM—m)	Easting End (UTM—m)	Northing End (UTM—m)
ISH1	862	0.9	15	6	125	406,698	4,510,675	407,103	4,511,304
ISH2	1330	1.5	100, 25	13, 5	190	407,324	4,510,451	407,273	4,511,211
ISH3	509	0.9	25	5	95	407,177	4,510,726	406,757	4,510,879
ISH4	814	0.8	25	5	180	406,015	4,510,572	406,182	4,511,279

reports the acquisition parameters adopted for each profile in terms of length, spatial step, working frequency, and maximum depth of the investigation. Figure 3 shows the row data obtained along our four profiles for frequencies of 15, 25, and 100 MHz.

We employed different antennas with frequencies for our profiles for two reasons. The first reasoning is logistical. As the 15 MHz antenna array has a length of approximately 13 m (including the Tx and Rx offset), its use is mostly not feasible for the steep and narrow roads on the slopes of Mt. Epomeo, except for the ISH1 profile located on wider roads. For profiles ISH2, 3, and 4, we could instead easily use the 25 MHz frequency, given that the overall length of the array is 7 m. The second reasoning for using antenna with different frequencies is our need to have radargrams with different resolutions at variable depths. As the ISH2 profile crosses over the 2017 earthquake co-seismic ruptures [3], and we reported signs of breakages already in the first meters below the surface, we used a 100 MHz frequency antenna there as well, to focus on the shallowest 20–30 m. As regards the polarization mode of the radar, we note the 100 MHz antenna has a co-polarized polarization and an orthogonal configuration of Tx and Rx, whereas the 25 MHz antenna has a co-polarized polarization with a parallel configuration of Tx and Rx.

In our study, the max observation time window is 2044 ns, discretized by 512 time samples. To improve the interpretability of the considered LDGPR data, we performed the following data corrections: (i) a band-pass filter, which allows for removing noise from the radargram and improving the results of later procedures; (ii) zero timing, which defines the actual starting time, t_0, of the observation time window. In this survey, we set t_0 = 1.5 ns; (iii) time gating, which selects the portion of the observation time window where the target response occurs and allows for eliminating the direct antenna coupling, t_TG, as well as the clutter signal. In this survey, we set t_TG = 10 ns; (iv) topographic correction, to insert the real altitude of each acquired A-scan and georeference in the radargrams; (v) two-way time–depth conversion by means of Equation (1) by assuming an average relative dielectric permittivity ε = 5–6. This permittivity was inferred based on the analysis of stratigraphic information of boreholes we drilled in the area. Furthermore, as mentioned, we carried out four 1D TDEM surveys in the area that allowed us to infer the electrical characteristics of the underground down to a maximum of 100 m and calibrate the depth of the high-resistivity bedrock (see Figure 4, Figure 5, Figure 6 and Figure 7).

3. Results

Figure 4, Figure 5, Figure 6 and Figure 7 show the most interesting radargrams derived from a measuring frequency of 15 and 25 MHz, obtained by means of the data processing chain described in the previous section. The processed data allow for identifying several areas (highlighted by dark blue rectangles) with strong reflections due to variations in the dielectric permittivity of the subsoil, likely related to the presence of buried and/or outcropping faults. These areas are well correlated with the positions of the 2017 earthquake’s co-seismic ruptures from Nappi et al. (2018) [3] and of the active fault system from Vezzoli (1988) [10] and Tibaldi and Vezzoli (1998) [11] (see Figure 9). We also note that the layers detected by our GPR data (shown by white solid lines in Figure 4, Figure 5, Figure 6 and Figure 7) are compatible with the geological sections in Mancini et al. (2022) [42] and Nardone et al. (2023) [29]. Furthermore, the layers from GPR data match with the outcome of our TDEM loops (TDEM 1–4, light blue lines) and the electro-layers from Nardone et al. (2023) [29] (black lines), especially for ISH2 and ISH4 (Figure 5 and Figure 7).

One of the main parameters of the EM wave is its attenuation, which manifests itself in the absorption of the wave energy and in the dispersion of the speed, depending on the crossed medium. Knowledge of the attenuation behavior of GPR waves in the medium allows rock types to be distinguished. The Q-factor is a frequency attribute of the wave field expressed by the ratio between the central frequency of the signal spectrum and its width. The Q-factor is inversely proportional to the absorption coefficient and represents its attenuation behavior (Green, 1955). It is given by Q = ϖ/((2δ), where ϖ = 2πf (with f antenna frequency), and δ = vπ is the damping rate (with v the phase speed and α the attenuation coefficient). Previous studies demonstrated how some GPR attributes were successful for highlighting normal faults within volcanic environments, e.g., [31].

We computed the Q-factor on a portion of the ISH2 section acquired using a 100 MHz antenna to infer the fault system dip (Figure 8). The reasoning for this analysis is that the ISH2 profile crosses over the 2017 earthquake’s co-seismic ruptures (Nappi et al., 2018) [3]. This antenna choice allowed for the identification of an anomaly between the progressives 0.84 and 0.93 km in the depth range of 5–20 m. Our Q-factor analysis focused on this portion of ISH2 and highlighted a large step that may be the fault displacement of the rupture. Thus, we infer a subvertical contact dipping towards the north that may represent the synthetic fault system of the Casamicciola Holocene graben, activated in 2017.

4. Conclusive Remarks

Our study of the Casamicciola fault system at the volcanic island of Ischia with a large-depth Ground-Penetrating Radar system aimed at constraining the source characteristics of this active and capable fault system. Our survey, performed by the Loza low-frequency (15–25 MHz) GPR system calibrated by TDEM results, highlighted variations in the electromagnetic signal due to the presence of contacts, i.e., buried and outcropping faults down to a depth larger than 100 m below the surface. As it is known, GPR techniques are mostly used to study the shallowest portions of fault systems.

Our study demonstrates the optimal potential of combined TDEM and deep GPR surveys for investigating active buried faults at depths in volcanic areas.

None of the geophysical studies carried out so far at Ischia involved such a high-resolution investigation (comparable only with what can be obtained through field surveys) of its active fault system at depths of about 100 m.

The main results of our research highlighted the following (Figure 9): (1) a good correlation of the subsoil dielectric permittivity variations detected by LDGPR with the position of (i) the 2017 earthquake primary co-seismic ruptures [3], (ii) the synthetic and antithetic fault system bordering the Casamicciola Holocene graben inferred by geological studies [10,11] and morphological analysis from LIDAR data [43]; (2) a northward dip displacement of the synthetic fault system inferred by the Q-factor GPR attribute.

GPR represents an effective tool in terms of spatial resolution, portability, and costs and may be efficiently used for volcanological and palaeoseismological studies by employing a variety of GPR frequencies and attributes. Within palaeoseismology, GPR has been used for different purposes: (1) preliminary investigations; (2) identification of trenching sites; (3) parallel analysis of GPR and trench data. Our study suggests that the application of LDGPR, possibly combined with other geophysical techniques, may offer a quick fault detection method for paleoseismological trench sites for future works.