SEARCH

SEARCH BY CITATION

Keywords:

  • dynamic stress transfer;
  • Mount Etna volcano;
  • volcano seismology

Abstract

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Seismic Networks
  5. 3. Data Analysis
  6. 4. Discussion and Conclusions
  7. Acknowledgments
  8. References
  9. Supporting Information

[1] Influences of distant earthquakes on volcanic systems by dynamic stress transfer are well documented. We analyzed seismic signals and volcanic activity at Mount Etna during two periods, January 2006 and May 2008, that clearly showed variations coincident with distant earthquakes. In the first period, characterized by mild volcano activity, the effect of the dynamic stress transfer, caused by an earthquake in Greece (M = 6.8), was twofold: (1) banded tremor activity changed its features and almost disappeared; (2) a swarm of volcano-tectonic (VT) earthquakes took place. The changes of the banded tremor were likely due to variations in rock permeability, caused by fluid flows driven by dynamic strain. The VT earthquake swarm probably developed as a secondary process, promoted by the dynamically triggered activation of magmatic fluids. The second period, May 2008, showed an intense explosive activity. During this interval, the dynamic stress transfer, associated with the arrival of the seismic waves of the Sichuan earthquake (M = 7.9), affected the character of the seismo-volcanic signals and on the following day triggered an eruption. In particular, we observed changes in volcanic tremor and increases of both occurrence rate and energy of long period events. In this case, we suggest that dynamic stress transfer caused nucleation of new bubbles in volatile-rich magma bodies with consequent buildup of pressure, highlighted by the increase of long period activity, followed by the occurrence of an eruption. We conclude that stresses from distant earthquakes are capable of modifying the state of the volcano.

1. Introduction

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Seismic Networks
  5. 3. Data Analysis
  6. 4. Discussion and Conclusions
  7. Acknowledgments
  8. References
  9. Supporting Information

[2] Various processes and events external to volcanoes are known to modulate volcanic activity [Manga and Brodsky, 2006]. Processes proposed in the literature for variations in volcanic activity driven by short- and long-term stress fluctuations include earthquakes [e.g., Linde and Sacks, 1998; Eggert and Walter, 2009], Earth tides [e.g., Sparks, 1981], variations in meteorological parameters [e.g., Neuberg, 2000; Patanè et al., 2007], changes in sea level [McGuire et al., 1997], and ice loading [Jellinek et al., 2004].

[3] More generally, stress changes produced by earthquakes are classified as (1) permanent (static) and (2) transient (dynamic) stress associated with the passage of seismic waves [Manga et al., 2009]. Unlike the amplitude of static stress, which decays with distance as ∼Δ−3, where Δ is the distance from the epicenter, the amplitude of the dynamic stress, which propagates as seismic waves, decreases more slowly as ∼Δ−2 and Δ−3/2 for body and surface waves, respectively [Hill and Prejean, 2007]. A third stress transfer mode, called quasi-static stress change, is associated with slow viscous relaxation of the lower crust and upper mantle beneath the epicenter of a large earthquake [Hill et al., 2002]. Both static and dynamic stresses may be significant in the near field (within up to a few fault lengths), but only dynamic stresses are significant in the far field [Manga and Brodsky, 2006].

[4] The variations of the volcano activity, temporally related to distant earthquakes, may consist in both eruption occurrences and changes in seismicity. In the former case, Linde and Sacks [1998] and Manga and Brodsky [2006] showed that the eruptions occur in the spatial and temporal vicinity of large earthquakes more often than would be expected by chance. There are several examples in the literature of clustered occurrences of eruptions and earthquakes at both short and very large distances. Large tectonic earthquakes may induce volcanic activity at distances ranging from several fault lengths to thousands of kilometers [Hill et al., 2002; Manga and Brodsky, 2006]. Focusing on the changes of seismicity on volcanoes, seismic signals are generally considered an indicator of their internal state of activity [Neuberg, 2000]. According to Chouet [1996], in volcanic areas we can distinguish seismic signals belonging to two different groups: (1) seismic events associated with shear failures in the volcanic edifice which are called volcano-tectonic (VT) or high-frequency earthquakes [McNutt, 2005] and (2) so-called seismo-volcanic signals that are generated by processes involving fluid excitation and include long period (LP) events, very long period (VLP) events, and volcanic tremor. Many papers have clearly demonstrated that earthquakes at regional and teleseismic distances can modify occurrence rates and characteristics of both VT earthquakes and seismo-volcanic signals. For example, small VT earthquakes occurred at the Katmai volcanic cluster within the coda of the 2002 Denali earthquake [Moran et al., 2004]. The daily rate of deep LP events at Mauna Loa sharply changed in coincidence with the arrival of seismic waves from the great Sumatra earthquake on 26 December 2004 [Okubo and Wolfe, 2008]. Teleseismic waves from the 1999 Chi-Chi earthquake (Taiwan) dynamically triggered tremor and micro-earthquakes at Aso volcano [Miyazawa et al., 2005]. Finally, the character of tremor has been observed to change at the same time as earthquakes occurred at some volcanoes, such as Stromboli [Carniel and Tarraga, 2006], Ambrym [Carniel et al., 2003], and Shishaldin [Moran et al., 2004].

[5] The effects of dynamic stress changes on volcanoes depend on seismic moment, directivity, radiation pattern, crustal structure, and especially the initial state of the magmatic system prior to the earthquake [Hill et al., 2002; Manga and Brodsky, 2006]. A volcanic eruption is preceded by the buildup of pressure in magma bodies beneath the volcano and by the ascent of magma to shallower levels. When a system has evolved to a critical state, it may be particularly sensitive to small perturbations of stress field [McLeod and Tait, 1999]. Small stress changes can trigger a sequence of events that result in an increase of the magma pressure and eventually lead to an eruption [Walter and Amelung, 2006]. Numerous studies have also shown that hydrothermal systems and mud volcanoes are very sensitive to dynamic stresses caused by large earthquakes [e.g., Husen et al., 2004; Manga and Wang, 2007; Mellors et al., 2007]. A variety of mechanisms have been proposed to explain how dynamic stress can alter volcanic and hydrothermal systems but considerable uncertainties remain [e.g., Husen et al., 2004; Manga and Brodsky, 2006].

[6] Mount Etna is a Quaternary stratovolcano located along the eastern coast of Sicily in a regional complex geodynamics, where major regional structural lineaments play an important role in the dynamic processes of the volcano [e.g., Gresta et al., 1998]. According to Murray [1990], recent studies confirm that the shallow reservoirs at Mount Etna are temporary and are occupied by magma only during short periods preceding single eruption or eruptive cycles. Shallow magma reservoirs under Mount Etna have been recognized using seismic data. For instance, Murru et al. [1999] used the b-values of the frequency-magnitude relation to map magma chambers under Mount Etna at depths of 10.5 ± 3 and 3.5 ± 2 km below sea level (bsl). De Gori et al. [2005] by attenuation tomography evidenced the presence of a low Qp body located at shallow depth (0–3 km bsl) beneath the south and southwestern sides of the edifice, where the magma was likely stored during 1994–2001. Moreover, Patanè et al. [2002, 2006] detected well-defined anomalous low P to S wave velocity ratio volumes both during the 2001 and 2002–2003 eruptions, interpreted as intrusion of volatile-rich basaltic magma at depth <5 km. The summit area of Mount Etna is currently characterized by four active craters: Voragine, Bocca Nuova, southeast crater, and northeast crater (hereafter referred to as VOR, BN, SEC, and NEC, respectively; see Figure 1). These craters are characterized by persistent activity that can be of different and sometimes coexistent types: degassing, lava filling or collapses, low rate lava emissions, phreatic, phreato-magmatic, or Strombolian explosions, and lava fountains [e.g., Cannata et al., 2008]. Some recent papers have investigated the relationship between Mount Etna activity and seismo-volcanic signals, highlighting how these signals are very useful in tracking the volcano dynamics [e.g., Patanè et al., 2008; Cannata et al., 2009; Di Grazia et al., 2009].

image

Figure 1. Shaded relief map of Mount Etna with location of the seismic stations equipped with short-period (diamonds) and broadband sensors (triangles) used in this work. The top left inset shows the distribution of the four summit craters (VOR, Voragine; BN, Bocca Nuova; SEC, southeast crater; NEC, northeast crater) and the eruptive fissure (EF), opened on 13 May 2008.

Download figure to PowerPoint

[7] The volcano-earthquake interaction at Mount Etna has been investigated in several papers that on the whole suggest the existence of a strict connection between seismicity and volcanic activity. For instance, Cardaci et al. [1993] showed a relation between seismic events and flank eruptions. Gresta et al. [1994] found a statistically significant link between destructive local earthquakes and the end of eruptions. De Rubeis et al. [1997] in studying the clustering properties of earthquakes found long-term, mid-term, and short-term variations related to different volcanic processes. Relations between seismicity and eruptive activity from 1980 at Mount Etna were also documented by Patanè et al. [2004]. Gresta et al. [2005] highlighted that the changes in Coulomb failure stress due to the 1981 and 2001 eruptions are compatible with the observed seismicity on the east slope of Mount Etna. The comparison between large magnitude historical earthquakes in eastern Sicily and eruptions at Mount Etna suggests the existence of their two-way mechanical coupling [Feuillet et al., 2006]. Moreover, Walter et al. [2009] hypothesized a dynamic trigger mechanism to explain three contemporary volcanic phenomena, taking place in 2002 (Etna eruption, submarine degassing near Panarea, and an eruption of Stromboli), following an earthquake occurring in northern Sicily. Finally, Carbone et al. [2009] found gravity steps at Mount Etna likely related to dynamical stress transfer between tectonic and magmatic system at a local scale.

[8] In this paper, we investigate variations of activity of Mount Etna in terms of changes in both VT earthquakes and seismo-volcanic signals coincident with large distant earthquakes that were likely triggered by dynamic stress transfer. In particular, we analyze two periods characterized by M = 6.8 and M = 7.9 earthquakes, occurring in southern Greece on 8 January 2006 and in Sichuan province (China) on 12 May 2008, respectively (see Table 1).

Table 1. Details of the Two Earthquakes Whose Dynamic Stress Effects on Mount Etna Volcano Are Investigated in This Papera
Date (UTC)LocationDepth (km)RegionMagnitudeDistance (km)Azimuth (deg)
  • a

    The information about time, location and magnitude were provided by U.S. Geological Survey, National Earthquake Information Center (http://earthquake.usgs.gov/). The distance and azimuth values were calculated with respect to Mount Etna.

2006 January 8 (1134:55)36.300°N, 23.358°E66southern Greece6.8760100
2008 May 12 (0628:01)39.986°N, 103.365°E19eastern Sichuan, China7.9783065

2. Seismic Networks

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Seismic Networks
  5. 3. Data Analysis
  6. 4. Discussion and Conclusions
  7. Acknowledgments
  8. References
  9. Supporting Information

[9] The permanent seismic network at Mount Etna is run by Sezione di Catania, Istituto Nazionale di Geofisica e Vulcanologia (INGV), and is composed of both short-period and broadband stations (Figure 1). The former are equipped with Lennartz seismometers with a 1 s cutoff period, while the latter are equipped with Nanometrics TRILLIUM seismometers, with flat velocity response within the 40–0.01 s period range. The number of available stations used for the analyses performed in this work varies depending on their deployment time and quality of recordings.

3. Data Analysis

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Seismic Networks
  5. 3. Data Analysis
  6. 4. Discussion and Conclusions
  7. Acknowledgments
  8. References
  9. Supporting Information

[10] Effects of large and distant earthquakes on the state of the volcano were investigated by performing different analyses on the recorded seismic signals. First, to get a preliminary idea of the changes of seismic energy released from the volcano, we calculated the root-mean-square (RMS) time series of the seismic recordings using a moving window of fixed length. Obviously, this takes into account the energy contribution of both volcanic tremor and seismic transients, such as LP and VLP events and earthquakes. For a given time series, the pth percentile can be defined as the value such that at most (100p)% of the measurements are less than this value and [100(1 − p)]% are greater. In light of this, the estimation of percentile enables us to efficiently remove seismic transients and estimate the background signal level. Then, in order to calculate the percentile envelope, we apply the aforementioned percentile definition to moving windows of the RMS envelope. The percentage threshold should be chosen on the basis of the number of transients in the signal that must be included or excluded in our calculations. Thus, two percentile envelopes, calculated by using low and high threshold values, can provide complementary information to evaluate the seismic state of the volcano. Further, to detect prospective changes in such a state before and after the passage of large earthquake seismic waves, we used the β-statistic approach [Matthews and Reasenberg, 1988]. This parameter has been widely applied to evaluate variations in seismicity, using the rates of occurrence of earthquakes, and is defined as [Miyazawa and Mori, 2005]

  • equation image

where n is the total number of earthquakes in the postseismic term, E(n) is the expected value estimated from the activity in the preseismic period, and var(n) is the variance of the postseismicity. Since we deal not only with seismic transients, such as earthquakes, but also with continuous signal, namely volcanic tremor, we substitute the mean or the percentile values of the RMS envelope for the number of earthquakes. This alternative definition of β-statistic was applied by Miyazawa and Mori [2005] to detect the changes in the activity of deep low-frequency tremors along the Nankai subduction zone immediately following the Tokachi-oki earthquake (Mw = 8.1). One of the most useful applications of the β-statistic to dynamic triggering involves drawing maps showing the spatial distribution of likely instances of dynamic triggering. Such maps can be developed with the traditional definition of the β-statistic based on variations in seismicity rates [e.g., Kilb et al., 2000, 2002; Gomberg et al., 2001], as well as with the alternative definition dealing with seismic amplitudes [Miyazawa and Mori, 2005]. Finally, we apply short time Fourier transform (STFT) to the seismic signals to find prospective changes in the spectral content concomitant with dynamic stress due to distant large earthquakes.

[11] We applied the above-described analyses to the seismic signals preceding, accompanying, and following the M = 6.8 earthquake of 8 January 2006 in southern Greece and the M = 7.9 earthquake of 12 May 2008 in Sichuan province (China) (see Table 1). We summarize the results in sections 3.1 and 3.2.

3.1. Southern Greece Earthquake

[12] From the end of the 2004–2005 eruption (on 8 March 2005) to the onset of the 2006 eruption (on 14 July 2006), the volcanic activity at Mount Etna mainly consisted in degassing, sporadic ash emissions, and a strong explosion, taking place on 12 January 2006. The pattern of the RMS envelope of seismic recordings, calculated from 1 December 2005 to 31 January 2006, is shown in Figure 2a, and provides a useful characterization of the volcanic activity. On 17 December, banded tremor activity, consisting of intermittent spindle-shaped packets of volcanic tremor (Figure 2b), started at the same time as RMS sharply increased. The banded tremor activity gradually became less evident and ended on 12 January coincident with a strong explosion at VOR or BN crater, which was in turn followed by a sharp decrease of tremor RMS amplitude (Figure 2c).

image

Figure 2. (a) RMS envelope of seismic signal recorded at the vertical component of ECPN station from 1 December 2005 to 31 January 2006 and filtered in the frequency band 0.5–1.5 Hz. The RMS was calculated by using nonoverlapping 5-min-long time windows. The top grey area indicates the period characterized by banded tremor activity, which gradually became less evident. (b, c) The 3-day-long time windows of RMS envelope, showing in detail the banded tremor activity and the strong explosion of 12 January, respectively.

Download figure to PowerPoint

[13] We investigated in detail the patterns of the RMS envelope calculated during 7–9 January 2006. The RMS was calculated at all available stations by use of 5-s-long moving windows overlapped by 2 s. In Figure 3 the RMS (blue dots) computed at the vertical component of ECPN and EMPL is shown. Then, we calculated the 5th and 95th percentile time series of the RMS envelope within 5-min-long moving windows (red and green lines in Figure 3, respectively). We also obtained the normalized STFT and the frequency peaks using 40.96-s-long windows at ECPN station (Figure 4).

image

Figure 3. RMS envelopes of seismic signal recorded at the vertical component of (a) ECPN and (b) EMPL stations from 7 January 2006 at 1640 to 9 January 2006 at 1200 and filtered in the frequency band 1–10 Hz. The envelopes were calculated by using 5-s-long moving windows overlapped by 2 s (blue dots). The red and green lines indicate the 5th and 95th percentile time series of the RMS envelope within 5-min-long moving windows. The top thick horizontal black lines represent the preseismic and postseismic windows, used for β-statistic calculation. The vertical dashed black line indicates the arrival time of the seismic waves of the Greek earthquake.

Download figure to PowerPoint

image

Figure 4. Normalized short time Fourier transform (STFT) and frequency of the spectral peak with the highest amplitude (red dots) of the seismic signal recorded from 7 January 2006 at 1640 to 9 January 2006 at 1200 at the three components of ECPN station. The spectral analysis was performed by 40.96-s-long moving windows. The vertical dashed black line indicates the arrival time of the seismic waves of the Greek earthquake.

Download figure to PowerPoint

[14] On 8 January 2006 one of the strongest Mediterranean earthquakes of the past decade took place in southern Greece (Table 1). The arrival of seismic waves of the Greek earthquake at Mount Etna at 1136 (all times reported in this paper are in UTC) on 8 January, clearly evident in the seismogram of EMPL shown in Figure 5, coincides with a sharp increase in signal amplitude (evident in the RMS envelope and 95th percentile time series in Figure 3) and a clear decrease of the dominant frequency (Figure 4). We calculated the spectrum of a 40-min-long window of seismic signal comprising most of the earthquake wave train. It shows high spectral amplitude values in a wide frequency range 0.02–10 Hz with the highest peak at 0.03 Hz (Figure 6a). In order to calculate the peak dynamic stress [Hill and Prejean, 2007], the peak particle velocity was computed by using the signals recorded by 10 broadband stations. Therefore, multiplying such a value by the shear modulus ((1–2) × 104 MPa [Trasatti et al., 2008]) and dividing by the S wave velocity (∼2000 m/s [Patanè et al., 2002]), a value of peak dynamic stress of 13 ± 3 kPa was obtained.

image

Figure 5. The 24-h helicorder plot from the vertical component of EMPL station for 8 January 2006. The red arrow indicates the arrival time of seismic waves of the Greek earthquake.

Download figure to PowerPoint

image

Figure 6. Spectra of seismic signals recorded at the vertical component of EMNR station and comprising most of the wave train of the (a) Greek and (b) Sichuan earthquakes. The former was calculated by a 40-min-long seismic signal window, and the latter was calculated by a 80-min-long window.

Download figure to PowerPoint

[15] The following variations in seismic signals recorded at Mount Etna immediately after the Greece earthquake wave passage were observed:

[16] 1. The amplitude of volcanic tremor, highlighted by the 5th percentile time series, slightly increased.

[17] 2. The periodic changes of the seismic amplitude, evident in the 5th percentile time series before the arrival of seismic waves of the Greek earthquake due to the banded tremor activity, temporarily disappeared.

[18] 3. Spectral content changes, particularly evident in the vertical and N–S components, took place.

[19] 4. Finally, some peaks of the RMS and 95th percentile time series (Figure 3) and the seismogram in Figure 5 displayed the occurrences of energetic amplitude transients, which are local VT earthquakes with hypocenters beneath the southern flank of the volcano (see black dots in Figure 7) at depths ranging between 10 and 15 km bsl. The significant time relationship between the arrival of seismic waves of the Greek earthquake and the local earthquake swarm is supported by both a histogram of the earthquake daily number and a strain release curve during the 6-month-long period from 1 November 2005 to 30 April 2006 (Figure 8): the highest daily strain release and the maximum daily number of earthquakes reflect the swarm that developed just 2 h after the arrival of the Greek earthquake waves.

image

Figure 7. The β-statistic maps developed calculating (a) 5th percentile, (b) 95th percentile, and (c) mean values of the preseismic and postseismic RMS windows (see top thick horizontal black lines in Figure 3). The blue circles indicate the seismic stations used to calculate the maps. The black dots represent the epicenter of the earthquake swarm occurring on 8 January 2006, after the arrival time of the seismic waves of the Greek earthquake. The black curves represent elevation contours at 1000 m intervals.

Download figure to PowerPoint

image

Figure 8. Daily earthquake number (histogram and left vertical axis) and cumulative strain release curve (blue line and right vertical axis) for earthquakes occurring at Mount Etna during November 2005 to April 2006. The seismic energy (E) was computed using the relationship log E (erg) = 9.9 + 1.9M – 0.024M2 [Richter, 1958].

Download figure to PowerPoint

[20] Finally, β-statistic maps were drawn by using time windows of 14 h. The preseismic window is up to ∼1 h before the arrival of P waves of the Greek earthquake, while the postseismic window starts ∼2 h after (see black horizontal top line in Figure 3). Three maps were developed by using 5th percentile, 95th percentile, and mean values of RMS windows (Figure 7). The space distribution of colors is slightly different in the three cases. In fact, the 5th percentile map, mostly affected by the changes in the volcanic tremor, exhibits positive β-statistic values, and then increases of seismic activity, mainly located in the summit area (Figure 7a). This is consistent with results of previous works showing volcanic tremor centroids generally located beneath the summit area [e.g., Patanè et al., 2008]. Conversely, in the 95th percentile map, more influenced by amplitude transients, the region with the highest β-statistic values was the southern flank, where the epicenters of the earthquake swarm are clustered (Figure 7b). The map with mean values is intermediate between the 5th percentile and the 95th percentile maps (Figure 7c).

3.2. Sichuan Earthquake

[21] In May 2008, Mount Etna was characterized by intense eruptive activity. On May 10, a lava fountain took place at SEC, the most active summit crater during the past decade. Three days after, an eruption occurred from an eruptive fissure, opened east of the summit area (EF in top left inset of Figure 1). This eruption, lasting until 6 July 2009, was characterized at its beginning by spectacular Hawaiian fountaining activity and during the following months by phases of variable intensity lava flows and Strombolian activity. The RMS envelope of the seismic signal, recorded from 15 April to 15 June 2008 (Figure 9), clearly reflects this volcanic activity. In particular, the two main amplitude peaks coincide with the lava fountain at SEC and with the onset of the eruption at EF.

image

Figure 9. RMS envelope of seismic signal recorded at the vertical component of ECPN station from 15 April to 15 June 2006 and filtered in the frequency band 0.5–1.5 Hz. The RMS was calculated by using nonoverlapping 5-min-long time windows.

Download figure to PowerPoint

[22] Similarly to the study described in section 3.1, we investigated the patterns of RMS, 5th and 95th percentile time series (Figure 10), and the spectral content (Figure 11) during 11–13 May 2008. The arrival of seismic waves of the Sichuan earthquake on 12 May at 0639, a day before the onset of the eruption, well evident in the EPDN seismogram shown in Figure 12, is clearly also highlighted by the sharp change of the spectral content, consisting in a decrease of the dominant frequencies (Figure 11). The spectrum of a 80-min-long window of the earthquake wave train exhibits high spectral amplitude below 0.08 Hz (Figure 6b). The peak dynamic stress, calculated by using the method explained in section 3.1, was 4 ± 0.7 kPa. The RMS amplitude did not show any increases due to the seismic wave passage because the frequency band considered (1–10 Hz) excluded almost the entire energy contribution of the teleseismic waves (Figure 10).

image

Figure 10. RMS envelopes of seismic signal recorded at the vertical component of (a) ECPN and (b) EPDN stations from 11 May 2008 at 0000 to 14 May 2008 at 0000 and filtered in the frequency band 1–10 Hz. The envelope was calculated by using 5-s-long moving windows overlapped by 2 s (blue dots). The red and green lines indicate the 5th and 95th percentile time series of the RMS envelope within 5-min-long moving windows. The top thick horizontal black lines represent the preseismic and postseismic windows, used for β-statistic calculation. The vertical dashed black lines indicate the arrival time of the seismic waves of the Sichuan earthquake and the eruption onset.

Download figure to PowerPoint

image

Figure 11. Normalized short time Fourier transform (STFT) and frequency of the spectral peak with the highest amplitude (red dots) of the seismic signal recorded from 11 May 2008 at 0000 to 14 May 2008 at 0000 at the three components of ECPN station. The spectral analysis was performed by 40.96-s-ong moving windows. The vertical dashed black lines indicate the arrival time of the seismic waves of the Sichuan earthquake and the eruption onset.

Download figure to PowerPoint

image

Figure 12. (a) The 24-h helicorder plot from the vertical component of EPDN station for 12 May 2008. (b) Hourly number of LP events. The red arrow in Figure 12a indicates the arrival time of seismic waves of the Sichuan earthquake.

Download figure to PowerPoint

[23] However, a variation in RMS trend is evident immediately after the teleseismic waves: the 5th percentile time series exhibited a gradually decreasing trend, lasting up to roughly 2 h before the onset of the eruption (at ∼0930 on 13 May). Conversely, from ∼1500 the 95th percentile time series increased, because of an increase in both number and energy of amplitude transients (well evident also in the seismogram in Figure 12), mainly consisting in LP events. One of the most energetic events, highlighted by a sharp peak in the 95th percentile time series, occurred at 0424 on 13 May. Normalized STFT and frequency peak series showed no significant variations following the teleseismic wave arrival. On the other hand, strong spectral changes preceded and accompanied the onset of the eruption.

[24] Also in this case, β-statistic maps were drawn by using mean, 5th percentile, and 95th percentile values of 14-h-long RMS windows. The preseismic window is up to ∼1 h before the arrival of the teleseismic waves, while the postseismic window starts ∼2 h after (see black horizontal top line in Figure 10). The distribution of colors is very different in all three cases. The 5th percentile map, highlighting the changes of the background volcanic tremor, exhibits negative β-statistic values (Figure 13a), as expected, accounting for the decreasing amplitude trend of the 5th percentile series (Figure 10). Conversely, positive values were observed in the 95th percentile map, mainly influenced by LP events (Figure 13b). The maximum values occur in the northern part of the summit area, where LP events and infrasound signals were mainly located just before the onset of the 13 May eruption [Di Grazia et al., 2009]. The map with mean values is complex due to the contrasting behaviors of volcanic tremor and amplitude transients (Figure 13c).

image

Figure 13. The β-statistic maps developed calculating (a) 5th percentile, (b) 95th percentile, and (c) mean values of the preseismic and postseismic RMS windows (see top thick horizontal black lines in Figure 10). The blue circles indicate the seismic stations used to calculate the maps. The black curves represent elevation contours at 1000 m intervals.

Download figure to PowerPoint

[25] In order to understand whether the gradual decrease of the tremor amplitude was due to energy decrease of the source or to its deepening, we looked at tremor centroid locations during 11–13 May 2008. The location information was provided by Di Grazia et al. [2009] and plotted in Figure 14. A deepening and northward shift can be noted in both used frequency bands 0.5–2.5 and 0.5–5.0 Hz from the time of teleseismic wave arrival to the eruption onset, coinciding with the tremor amplitude decrease. Consequently, we inferred that the observed tremor amplitude variation was mainly related to the source shift, rather than to source energy change. At ∼0930 on 13 May a sharp shallowing and eastward shift of the tremor centroids highlighted the opening of the fissure and the onset of the eruption.

image

Figure 14. Digital elevation model and section of Mount Etna showing source centroids of volcanic tremor locations, filtered in the bands 0.5–2.5 and 0.5–5.0 Hz, computed during the period 11–13 May 2008 (see the bottom time color bar). The black thick line in the map indicates the eruptive fissure opened on 13 May 2008.

Download figure to PowerPoint

4. Discussion and Conclusions

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Seismic Networks
  5. 3. Data Analysis
  6. 4. Discussion and Conclusions
  7. Acknowledgments
  8. References
  9. Supporting Information

[26] Interactions between strong earthquakes and volcanoes are now well documented around the world. Understanding whether the activity of a certain volcano is affected by stress transfer of a distant earthquake is of great importance not only for research but also for monitoring purposes. In this paper we investigated the effects of two strong distant earthquakes on Mount Etna volcano: one occurred in southern Greece on 8 January 2006 and the other occurred in Sichuan province (China) on 12 May 2008. Previous studies suggested that distant earthquakes are more effective at triggering local seismicity than regional seismicity [e.g., Anderson et al., 1994]. Therefore, as also suggested by other authors [e.g., Gomberg et al., 2001; Hill and Prejean, 2007], the ability of an earthquake to trigger seismicity and/or affect the state of geothermal and volcanic systems at first approximation could depend on the peak dynamic stress. In view of this, we calculated the values of peak dynamic stress at Mount Etna, which were equal to 13 ± 3 and 4 ± 0.7 kPa for the earthquakes in Greece and Sichuan, respectively. According to previous studies, similar values of peak dynamic stress in geothermal and volcanic areas are able to trigger earthquakes [Brodsky and Prejean, 2005], tremor [Miyazawa et al., 2005], or more generally modify the state of geothermal and volcanic systems [e.g., Walter et al., 2009]. Brodsky and Prejean [2005] stated that the effects of dynamic stress depend also on the frequency of the oscillations: long-period waves, similar to the ones observed during these two earthquakes (dominant frequencies lower than 0.08 Hz), are considered more effective at triggering earthquakes and affecting the fluids in the rocks than short-period waves of comparable amplitude. On the basis of these considerations, the dynamic stresses from the analyzed earthquakes were potentially able to affect the Mount Etna volcano system. However, for a volcanic system to be susceptible to change in response to small dynamic stresses, it is also necessary that the volcano be in a critical state before the wave arrival [e.g., Hill et al., 2002]. In light of this, we studied in detail the variations of activity of Mount Etna in terms of changes in both VT earthquakes and seismo-volcanic signals during two periods, January 2006 and May 2008, when the two earthquakes took place.

[27] During December 2005 to January 2006, Mount Etna was characterized by banded tremor activity (Figure 2). This particular kind of tremor has sometimes been observed at Mount Etna [Gresta et al., 1996; Cannata et al., 2010], and at other volcanoes and geysers [e.g., McNutt, 1992, 1994; Baptie and Thompson, 2003], and its source has recently been modeled as a two-phase hydrothermal instability flow model [Fujita, 2008]. Such activity gradually became less evident and ended on 12 January 2006 at the same time as a strong explosion took place (Figure 2). This explosion did not involve juvenile magma and was considered phreatic [Andronico and Cristaldi, 2006; Andronico et al., 2006], supporting the hypothesis of the hydrothermal nature of the banded tremor preceding the explosion. Therefore, the volcanic activity at Mount Etna during the studied period can be considered similar to that of a geyser, where the presence of magma in the upper part of the volcano played only a passive role as heat source for shallow groundwater [Cannata et al., 2010]. Husen et al. [2004] showed that geysers are very sensitive to the dynamic stress transfer of distant earthquakes. In fact, according to Ingebritsen and Rojstaczer [1996], geysers behave as chaotic systems and thus their activity is strongly affected by the boundary conditions. Also similar to geyser systems, the behavior of banded tremor may be considered chaotic [Cannata et al., 2010] and thus very small changes of parameters, such as heat and steam flows, permeability and porosity of the rocks, and so on, are able to modulate banded tremor or even to stop it. Brodsky et al. [2003] proposed a model to explain how the dynamic stress transfer can modify the permeability of the rocks: fractures clogged by precipitation of minerals can be reopened by fluid flows driven by dynamic strain. According to this model, we speculate that the variations of banded tremor activity, observed after the Greek earthquake, were due to a permeability change of the rocks, caused by dynamic stress transfer (Figures 3, 4, and 7).

[28] In the same way, the occurrence of the swarm of VT earthquakes 2 h after the wave arrival of the Greek earthquake may have been dynamically triggered (Figures 5 and 7). Although the mechanism of dynamic triggering of earthquakes is yet to be fully understood, it has been clearly shown that such phenomena are ubiquitous [e.g., Hill and Prejean, 2007; Velasco et al., 2008] and that volcanic and geothermal areas are particularly susceptible to dynamic triggering [e.g., Moran et al., 2004; Brodsky and Prejean, 2005; West et al., 2005]. Mechanisms that have been proposed by other investigators to explain remote triggering of earthquakes can be divided into two groups: (1) direct triggering by various modes of frictional failure on critically stressed faults; (2) activation of crustal fluids and/or transient, aseismic deformation by the passing dynamic stresses, with triggered seismicity developing as a secondary process [Hill, 2008]. Frictional failure models are consistent with the prompt onset of triggered seismicity during or shortly after the cessation of dynamic stressing associated with passing seismic waves. Fluid activation and transient deformation models admit the possibility of significant delays in the onset of triggered energy release, with the dominant energy release developing hours to a few days after the dynamic stresses have passed [Hill, 2008]. In our case, the swarm of VT earthquakes took place 2 h after the wave passage of the Greek earthquake, suggesting that the second group of models is more likely. Moreover, since the focal depth of the earthquake swarm was 10–15 km bsl, we infer that magmatic fluids, releasing at a depth of ∼12 km bsl the largest amount of carbon dioxide (CO2) [Caracausi et al., 2003], promoted the VT seismicity. Therefore, this triggered seismicity can be considered a secondary result of locally triggered deformation associated with the change in state of a nearby magma body.

[29] During May 2008, Mount Etna was characterized by a lava fountain on 10 May at SEC, and, 3 days later, by an eruption lasting ∼14 months. Therefore, this second period, which was characterized by more intense volcanic activity, involving fresh gas-rich magma, was quite different from the first. The arrival of seismic waves of the Sichuan earthquake on 12 May at 0639 was followed by significant variations in seismo-volcanic signals. In particular, we observed a gradual decrease of the volcanic tremor amplitude, starting just after the teleseismic wave arrival, mainly related to a deepening and northward shift of the source (Figure 14), and an increase of number and energy of LP events, taking place roughly 9 h later (Figures 10, 12, and 13). We infer that the volcanic tremor source shift was likely triggered by dynamic stress transfer. In fact, magmatic fluids play a fundamental role in generating volcanic tremor and are particularly susceptible to dynamic triggering [Hill and Prejean, 2007]. Variations of the nature of volcanic tremor at the same time as distant earthquakes were also observed at other volcanoes, such as Ambrym [Carniel et al., 2003] and Shishaldin [Moran et al., 2004].

[30] According to the widely accepted model of Chouet [1996], LP events are the result of the acoustic resonance of fluid-filled cracks triggered by pressure transients [Chouet, 1996]. On the basis of this model, the observed enhanced LP activity seems to suggest an increasing overpressure, likely triggered by dynamic stress transfer, that led on the following day to an eruption. Several examples in the literature describe increases in LP activity preceding eruptions, such as at Redoubt in 1989 [Chouet et al., 1994], at Pinatubo in 1991 [White et al., 1992], and at Etna in 2007 [Patanè et al., 2008]. Many models have been developed to explain how dynamic stress can cause buildup of pressure in magma bodies. According to Manga and Brodsky [2006], under proper conditions, small pressure changes, associated with dynamic stresses in a crystallizing magma that is close to critical supersaturation, should be capable of triggering a significant increase in bubble nucleation leading to a marked pressure increase. Another model, called “falling roofs” [Manga and Brodsky, 2006], suggests that passing seismic waves may dislodge dense crystals from the ceiling and walls of a magma chamber. These crystals, sinking due to gravity, may stimulate convection in the chamber. Bubbles would nucleate and grow in the ascending volatile-rich magma, thereby increasing pressure in the magma body and deforming the overlying magma crust [Hill et al., 2002]. Under both of these models magma bubbles, which are also involved in the generation of LP seismicity [e.g., Chouet et al., 2003; James et al., 2004], play a fundamental role in phenomena of earthquake-magmatic interactions [Hill and Prejean, 2007]. Moreover, both models require that magma bodies must be volatile saturated to be effective. We suggest that at Etna such increasing overpressure within a shallow magma batch, triggered by the Sichuan teleseismic waves and highlighted by the enhanced LP activity, led to the eruption on the following day. It is worth noting that, as suggested by many variations of geophysical parameters taking place days and months before the eruption [Di Grazia et al., 2009], the volcano system had to be in a critical state before the arrival of the teleseismic waves. For this reason, it was so susceptible to the dynamic stress transfer. Finally, that such a distant earthquake was able to affect the Etna volcano system at a distance of ∼7800 km should not be surprising. The same earthquake also triggered deep tremor both in southwest Japan at ∼2800 km [Miyazawa et al., 2008] and in central California at ∼11,000 km [Peng et al., 2009].

[31] In conclusion, we have presented evidence that phenomena of dynamic stress transfer took place at Mount Etna and are able to modify the state of the volcano, and even lead to eruptions. This finding is very important for both future monitoring and research purposes. However, uncertainties still remain on the precise mechanisms by which dynamic stress alters the volcano system. This study, together with previous works [e.g., Gresta et al., 2005; Walter et al., 2009], shows that Mount Etna can be considered a laboratory to study phenomena of earthquake-volcano interaction.

Acknowledgments

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Seismic Networks
  5. 3. Data Analysis
  6. 4. Discussion and Conclusions
  7. Acknowledgments
  8. References
  9. Supporting Information

[32] We are indebted to David Hill for his very important and kind help in improving our manuscript. We also thank Vincenzo Milluzzo for his useful help in developing the β-statistic map software. We acknowledge Ferruccio Ferrari, Salvo Gambino, and Eugenio Privitera for fruitful discussions and “Gruppo di Analisi Dati Sismici”—INGV-Catania for earthquake data. We thank an anonymous reviewer for very useful suggestions.

References

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Seismic Networks
  5. 3. Data Analysis
  6. 4. Discussion and Conclusions
  7. Acknowledgments
  8. References
  9. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Seismic Networks
  5. 3. Data Analysis
  6. 4. Discussion and Conclusions
  7. Acknowledgments
  8. References
  9. Supporting Information
FilenameFormatSizeDescription
jgrb16540-sup-0001-t01.txtplain text document1KTab-delimited Table 1.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.