Seafloor and water-column observatories have proved to be excellent tools in long-term monitoring to study processes occurring in hard-to-access abyssal areas [1
]. Real-time data recorded by seafloor seismic stations have been used to both improve the location of earthquakes occurring in offshore areas and to evaluate the accuracy of eruption mechanisms of both seamounts and sub-aerial volcanoes [3
]. Since the late 2000s, many programs have begun in Canada, the USA, Japan, Europe, and many other countries to establish permanent underwater infrastructures. In Canada, Ocean Networks Canada (ONC) manages, among other infrastructures, the deep-sea cabled network North East Pacific Time-series Underwater Networked Experiments (NEPTUNE) in the Northeast Pacific Ocean and the shallow-water cabled Victoria Experimental Network Under the Sea (VENUS) in the Salish Sea between Vancouver Island and the continent [5
]. In the USA, a network of multiple element hydrophone arrays (NOAA) was deployed for monitoring the seismicity occurring at the Juan de Fuca and Gorda Ridges [6
]. The installation of the Ocean Observatories Initiative Cabled Array supports near-continuous geophysical monitoring [8
]. OOI allows the monitoring of both the volcanic eruption of the Axial Seamount, focusing on the magma recharge beneath the southeastern part of the summit caldera [9
], and the offshore earthquakes that are important for understanding Cascadia interplate dynamics [11
In Europe, the European Multidisciplinary Seafloor and water column Observatory European Research Infrastructure Consortium (EMSO-ERIC) aims to explore the seas around Europe to gain a better understanding of geophysical, geochemical, and environmental phenomena [12
]. The NEMO-SN1 multidisciplinary observatory, a node of EMSO ERIC, was deployed in the Western Ionian Sea first in stand-alone configuration and then in cabled mode [13
]. The NEMO-SN1 was deployed to monitor the offshore seismicity linked to the seismogenic structures that cause the most destructive earthquakes occurring in Italy [16
], and to focus on the deep volcanic structures of Mt. Etna [16
Sea dynamics are the source of a typical background seismic signal: oceanographic noise. The origin of this signal may be associated with different causes: the nonlinear interaction of the sea waves with the seafloor bathymetry [18
], local atmospheric conditions [19
], wave height amplitudes caused by local atmospheric perturbations [20
], and resonance effects of sedimentary layers [22
]. Waves and currents generate seabed disturbances [23
] and tides act as trigger of microearthquakes from normal faults [24
]. Seismic data recorded by seafloor seismometers are affected by daily and seasonal variations, including ocean loading, bottom currents, and marine organisms, as well as instrumental problems [23
A typical low-frequency seismic signal recorded at active volcano sites around the world is the volcanic tremor, which is associated with the movement of magma, fluids, or gas [31
]. Rising gas pockets within the magma column interact with the conduit walls, causing a vibration that produces a tremor signal. This vibration precedes and accompanies the explosive eruption; volcanic tremor plays a significant role in real-time volcano monitoring as the pattern of this signal strongly reflects the evolution of the eruptive activity [32
]. In recent years, ocean bottom seismometers (OBSs) and seafloor observatories have been used to monitor from the seafloor the volcanic activity of Stromboli [35
] and Etna [36
] volcanoes. In the case of Mt. Etna, some authors reported that the south-eastern flank of this volcano slides into the Ionian Sea at a rates of centimetres per year. These results are based on the data of seafloor displacement collected during a 2016–2017 monitoring campaign [37
In this paper, we analyse the background seismic signal recorded by NEMO-SN1 to determine the possible use of this signal in the volcano monitoring of Mt. Etna. Most amplitude increases of the background seismic signal are related to oceanographic processes. In the Western Ionian Sea, the influence of oceanographic dynamics on the background seismic signal is evident when comparing seismic data and wave heights recorded at a buoy located off Catania city harbour [38
]. Due to their close correspondence, the gravity waves induced by local winds can be considered the main cause of oceanographic noise generation.
During the deployment of NEMO-SN1 seafloor observatory, a vigorous explosive event occurred on Mt. Etna, generating the powerful 2002–2003 eruption. Analysing the background seismic signal recorded by NEMO-SN1, we noted that some amplitude increases and frequencies are linked to Mt. Etna’s activity, and that this signal has the same characteristics of the volcanic tremor recorded by land seismometers. We performed continuous monitoring of the eruptive activity of Mt. Etna from the seafloor, which is an unusual point of view. Some important evidence showed that the oceanographic noise was recorded clearly when Mt. Etna explosive activity was absent. Instead, oceanographic dynamics are hidden when explosive activity occurred (mainly as Strombolian explosions or lava fountains). Recognizing that the volcanic and oceanographic dynamic processes coexist may help to better comprehend the behaviour of Mt. Etna during its explosive activity, offering us the opportunity to study the submerged flank of the Mt. Etna volcano and to monitor its volcanic activity.
4. Background Seismic Signals from NEMO-SN1
We computed the RMS (Root Mean Square) amplitude of the seismic signal (Z component) recorded by NEMO-SN1 to discriminate the contributions of the sea dynamics and volcanic activity of Mt. Etna. We reconstructed the time pattern of the background seismic signal filtering in two frequency bands: the first ranging between 0.1 and 5 Hz (Figure 2
a), and the second between 0.5 and 5 Hz (Figure 3
a). The signals in these two frequency bands appeared very differently. Figure 2
a shows the signal filtered in 0.1–5 Hz and shows several background increases with high amplitudes that lasted several days.
These increases were not correlated with the explosive activity of Mt. Etna but were related to the sea dynamics. Figure 2
b shows the Sea Wave Height (SWH) recorded at the buoy located off Catania harbour [38
]. A correlation between the occurrence of these noise increases associated with SWH data is observed. This highlights that the noise is mainly associated with the local sea wave conditions affecting the seismological signal in the microseismic band (0.1–0.5 Hz; [25
In Figure 3
a, we reconstructed the time pattern of the background seismic signal filtered in the frequency range between 0.5 and 5 Hz, as tremors mainly reflect the volcanic activity in this range [31
]. The correspondence between the beginning and evolution of the explosive activity and the pattern of the background seismic signal recorded by NEMO-SN1 (Figure 3
a) was checked using reports describing the volcanic activity [51
The increases in background signal are strongly linked to the explosive activity of Mt. Etna: the highest amplitudes were visible during the eruption onset (27 October 2002), whereas slight amplitude increases were only visible during the occurrence of lava fountains and fracture opening episodes, corresponding to the main phases of volcanic activity (Table 1
). These increases in the background signal showed evident similarities with the low-frequency signal identified as volcanic tremor, typically observed on land; in the time period of November 7 to December 9 (Figure 3
b), two lava fountain episodes and the opening of new effusive vents occurred, accompanied by effusive and Strombolian activity [47
5. Time, Spectral, and Polarization Analyses
Time, spectral, polarization, and particle-motion analyses were performed on continuous seismic signals to characterize volcanic tremors and oceanographic noise. In particular, we compared two episodes of continuous signal amplitude increase to discriminate between the source of the two signals. The continuous Z-component signals shown in Figure 4
a,b were 0.1–5 Hz filtered. A background signal increase was recorded simultaneously with the occurrence of a lava fountain episode (4–5 November 2002; Figure 4
a). Other increases in the background seismic signal were not associated with explosive activity. One of these episodes (Figure 4
b) occurred on 21–22 November 2002.
The amplitude trends of both signals are rather similar (Figure 4
a,b), except for the roughly higher values of oceanographic noise amplitude.
In order to estimate the quantitative relationship between SWH and RMS computed in both the 0.1–5 Hz and the 0.5–5 Hz frequency band, we performed a cross-correlation analysis between the two signals. We considered the two episodes represented in Figure 4
as well as the whole period shown in Figure 2
a,b (Figure 5
The results show that the correlation between volcanic tremor and SWH is very poor, in particular for the seismic signal filtered in the 0.1–5 Hz frequency band [25
] (cc = 0.37; Figure 5
a). On the other hand, a very high correlation between oceanographic noise and SWH is found (correlation coefficients cc = 0.97 and cc = 0.73 for 0.1–5 Hz and 0.5–5 Hz, respectively; see Figure 5
b). Such findings suggest that the oceanographic noise recorded by NEMO-SN1 is probably due to gravity waves generated on the air-sea interface.
With the aim of supporting these results, a cross-correlation analysis was performed on the whole period (Figure 5
c). We started the cross-correlation analysis on November 1, 2002, excluding the peak linked to the seismic swarm that preceded and accompanied the eruption onset. The cross-correlation confirms a higher value for the RMS computed in the 0.1–5 Hz band (cc = 0.74) compared to the one in the 0.5–5 Hz band (cc = 0.57). This indicates that the seismic signal recorded by NEMO-SN1 mainly consists of oceanographic noise, well detected in the 0.1–0.5 Hz frequency band [25
]. Further, the results suggest that the local wind could be the main source in oceanographic noise generation.
The spectrogram and spectra (Figure 6
a,c and Figure 6
b,d, respectively) computed on the two signals are different mainly in the 0.01–0.5 Hz frequency band. To better appreciate the spectral difference between volcanic tremor and oceanographic noise, we compared spectrograms computed on the two signals related to the lava fountain episode and a background signal increase (Figure 6
a,c), both calculated in a 30-minute time window (grey striped in Figure 4
a,b). We observed that the volcanic tremor (Figure 6
a) was characterized by a variable spectral amplitude content in the 0.1–0.5 Hz frequency band, whereas oceanographic noise (Figure 6
c) had one well-defined near-monochromatic frequency around 0.3 Hz.
A more detailed spectral analysis of both oceanographic noise and volcanic tremor signals was performed on the 0.0–0.5 Hz frequency band (Figure 6
b,d) on 14 30-minute segments. The volcanic tremor showed a rough frequency peak close to 0.1 Hz, whereas spectral amplitudes remained almost constant at frequencies ranging between 0.2 and 0.5 Hz (Figure 6
b). Spectra computed on oceanographic noise (Figure 6
d) show the two peaks at frequencies close to 0.1 and 0.3 Hz. These frequencies correspond to the two typical microseismic peaks caused by ocean wave energy coupling into the motion of the earth: the Single Frequency (SF) peak (~0.1 Hz) and the much stronger Double Frequency (DF) peak (~0.3 Hz) [25
]. The SF peak is generated by direct ocean wave pressure fluctuations at the ocean bottom; the amplitude of wave-induced pressure fluctuations decreases exponentially with depth from the sea surface depending on the bathymetry. The DF peak is produced by the interaction of sea gravity waves with the seafloor [19
]. The SF peak has been associated with high-amplitude storm waves impacting long stretches of coastline nearly simultaneously, whereas the DF peak is related to wind speed and direction, implying that the energy reaching the ocean floor is generated locally by ocean gravity waves [52
]. In the Ionian Sea, the SF peak is considered characteristic of deep-sea monitoring sites that are not too far from the coastlines, which is more easily observed during favourable weather conditions. The presence of the energetic DF peak suggests the influence of local winds [53
]. Although spectra of oceanographic noise showed well-defined SF and DF peaks (Figure 6
d), in the volcanic tremor signal spectra, the SF peaks are not clearly delineated (Figure 6
b). The spectral amplitudes of SF and DF peaks computed for the volcanic tremor signal were both about −130 dB, whereas the SF and DF peaks of oceanographic noise signal had more pronounced spectral amplitudes (about −120 dB and −90 dB, respectively). This means that spectral amplitudes linked to volcanic activity hide the effects linked to the sea in this low-frequency band.
Due to the availability of continuous broadband three-component signals, we performed polarization and particle-motion analyses with the aim of identifying the direction of the locations of the sources. We used the standard covariance matrix method [54
], which works in the time domain by bandpass filtering the signal around the frequency band of interest. The broadband data from the different components were converted to ground velocity by deconvolving the instrument response. We computed the polarization on the three-component signals by applying the Covariance Matrix Decomposition (CMD) developed in the SEISMPOL code [56
The background seismic signal recorded by NEMO-SN1 has a complex waveform, where different aspects act concurrently with the polarization of a signal in the same frequency band. We attempted polarization analyses using a total of 110 10-s-long windows of 0.1–2 Hz-filtered signal. The results from polarization were compared with those derived from particle motion analysis, performed using 110 4-s-long windows of 0.1–2 Hz filtered signal. Polarization and particle motion computed on volcanic tremor and oceanographic noise showed high values of rectilinearity (>0.8) and high incidence angles (>60°) (Figure 7
b). Results from these analyses indicate that most energy in the 0.1–2 Hz frequency band is due to the generation of converted waves prevalently from P–waves to S–waves (P-to-SV) at the seafloor boundary for both signals. Our results concur with those of Imanishi et al. [57
], who found that low-frequency noise is composed of nearly vertical incident S-waves. In the case of volcanic tremors, our results agree with those recorded at Mt. Etna and other volcanic areas. A wavefield consisting of P-waves radiated by a vertical extended source was observed at Mt. Etna [58
]; body waves radiated by an extended source characterized the wavefield recorded some kilometres away from the summit craters. Dominant P-SV converted waves have been identified in the Phlegrean Caldera [60
], whereas S-waves were identified as the dominant wave type at Arenal volcano [61
The most relevant difference between volcanic tremor and oceanographic noise is observed in the strong directivity of wave propagation, shown in the two rose diagrams in Figure 7
a,b. Volcanic tremor propagates from NW to SE, compatible with a wave front coming from Mt. Etna, with azimuth values ranging from 290° to 320° (Figure 7
a). The particle motion of oceanographic noise (Figure 7
b) shows an E–W direction, corresponding to the sea wave direction perpendicular to the coast (azimuth ranging 85–94°).
6. Comparison between Volcanic Tremor and Oceanographic Noise
We computed the Power Spectral Density (PSD) by applying Fast Fourier Transform to characterize the background seismic signal. Figure 8
a–d show the PSD computed over two-hour-long segments chosen in four different periods: (1) October 17 refers to the non-eruptive period preceding the eruption onset, (2) October 27 coincides with the eruption onset, (3) November 22 corresponds to a background seismic increase not associated with explosive activity, and (4) December 8 is the eruptive period.
The comparison of the computed spectra with the reference Low-Noise Model (LNM) and High-Noise Model (HNM) curves [62
] shows that during the eruption onset (Figure 8
b), the spectral amplitude exceeded the HNM curves, being about 40 dB higher than the PSD computed in the eruptive period (Figure 8
d). Both spectra show that the energy of the seismic data is concentrated in the 0.1–4 Hz frequency band, with a higher energy content in the 0.5–2 Hz frequency band (Figure 8
b), which was typical of the Mt. Etna volcanic tremor during the occurrence of the 13 lava fountains in 2013 [36
]. The non-eruptive period (Figure 8
a) and the background seismic increase period (Figure 8
c) show a different pattern, with the presence of two peaks in the typical microseismic band: the SF peak (~0.1 Hz) and the stronger DF peak (~0.3 Hz) [25
]. In Figure 8
a,c, the pattern of microseismic spectra in the 0.01-0.5 Hz band shows the typical spectral features of deep water, as in the Cascadian area [26
7. Discussion and Conclusions
The typical background seismic signal recorded by seafloor stations is a continuous oceanographic noise that is associated with tidal variation and currents and is due to ocean loading, which causes local crustal stresses [27
]. In order to focus on the possible generation of oceanographic noise in the Ionian Sea, we computed the RMS amplitude pattern of the seismic signal filtered in the frequency range between 0.1 and 5 Hz, and compared it with the wave height recorded at the buoy located off Catania harbour [38
] (Figure 2
b). The frequency of oceanographic noise recorded by NEMO-SN1 is primarily below 0.5 Hz and we found two distinctive peaks at ~0.1 and ~0.3 Hz (SF and DF microseismic peaks, respectively) in the spectrum of the oceanographic noise signal (Figure 6
d). The sea gravity waves induced by local winds are the main cause of the oceanographic noise generation.
In October 2002, a vigorous eruptive and explosive event started on Mt. Etna. The eruption provided an occasion to test the capacity of NEMO-SN1 to record the low-frequency seismic signals associated with volcanic activity (volcanic tremor) and to discriminate the low-frequency signals from those associated with sea dynamics (oceanographic noise). During the eruption, we observed a strict correspondence between amplitude increases of the background seismic signal recorded at NEMO-SN1 and the occurrence of explosive activity (Figure 3
and Figure 4
a). Spectral analysis performed on the background seismic signal demonstrated a frequency range between 0.1 and 4 Hz, with higher energy in the 0.5-2 Hz frequency band (Figure 8
b), which is comparable to the predominant frequencies closer to 2 Hz recorded by land stations located on Mt. Etna [34
] and by ocean bottom seismometers (OBSs) located close to Etna and Stromboli volcanoes [35
]. Changes in spectral amplitudes during the two different phases of the eruption (from about –60 dB to –120 dB in Figure 8
b,d) depend on the intensity of volcanic activity [34
], but also on the tremor source position [63
The cross-correlation analysis between the seismic signal (RMS) recorded by NEMO-SN1 and the SWH suggests that the oceanographic noise probably originated from local winds perturbing the sea surface. Such noise dominates the seismic signal over the whole studied period and it is mainly concentrated in the 0.1-5 Hz frequency band [25
]. On the other hand, the seismic signal generated by the volcano (volcanic tremor) highlights a very low correlation with the SWH and, accordingly, with the oceanographic noise. Consequently, the cross-correlation analysis between RMS and SWH allowed us to effectively discriminate the volcanic tremor from the oceanographic noise.
The wave field compositions resulting from polarization and particle motion analyses computed for volcanic tremor and oceanographic noise are rather similar, showing a complex overlapping pattern that consists mainly of converted waves (mainly P to SV waves). A source mechanism linked to pressure variations due to fluid movements is the driving force for both volcanic tremor and oceanographic noise generation. In the case of volcanic tremor, this mechanism is compatible with the presence of a storage zone for Mt. Etna that extends under the seafloor, linked to the dynamics of the magmatic fluids [64
]. A compressive mechanism linked to pressure variations may explain the wave field composition of oceanographic noise. We inferred that oceanographic noise is largely due to the pressure applied by the sea gravity waves on the seafloor, which deform the underlying sediment and crust.
The most important difference between volcanic tremor and oceanographic noise is the strong directivity of the two signals. Volcanic tremors show a preferential direction from NW to SE (Figure 7
a). This confirms that the 0.5–2 Hz bandpass filtered signals come directly from Mt. Etna. The oceanographic noise shows a marked directivity from polarization and particle motion analyses, having a predominant wave direction from E to W (rose diagram with a narrow range of azimuth values ranging from 85° to 94° in Figure 7
b) and corresponding to the sea wave direction roughly perpendicular to the coast.
The spectra computed on volcanic signals recorded during the occurrence of explosive activity of Mt. Etna and non-eruptive periods provide meaningful results. Seismic data show the maximum energy corresponding to the eruption onset and eruptive period when their patterns are representative of the volcanic activity. During the non-eruptive period, increases in background signal are observed but spectra have a different trend with two visible microseismic spectral peaks at ~0.1 and ~0.3 Hz (Figure 8
a,c, respectively) associated with the sea dynamics.
In the absence of explosive activity at Mt. Etna, sea dynamics prevail. When explosive activity increases, the energy associated with the volcanic tremor linked to Mt. Etna volcanism becomes the main contribution, prevailing on the oceanographic dynamics. Since volcanic tremors, as recorded by land seismic stations around Mt. Etna, were also observed on the seafloor, we inferred that explosive activity is generated by processes that are efficiently detected by the seafloor station.
The monitoring of the volcanic activity from a seafloor observatory can help to pinpoint possible hazard sources in the Western Ionian Sea. Understanding the similarities and differences of seismic signals containing the contribution of both volcanic and oceanographic processes may help assessing the behaviour during explosive activity at Mt. Etna, which is crucial to minimizing volcanic hazards.