Swarms of microearthquakes associated with the 2005 Vulcanian ...

21
Geophys. J. Int. (2010) 182, 808–828 doi: 10.1111/j.1365-246X.2010.04647.x GJI Mineral physics, rheology, heat flow and volcanology Swarms of microearthquakes associated with the 2005 Vulcanian explosion sequence at Volc´ an de Colima, M´ exico Vyacheslav M. Zobin, 1 Oleg E. Melnik, 2 Miguel Gonz´ alez, 1 Orlando Macedo 3 and Mauricio Bret ´ on 1 1 Observatorio Vulcanol´ ogico, Universidad de Colima, Colima 28045, M´ exico. E-mail:[email protected] 2 Institute of Mechanics, Moscow State University, Moscow, Russia 3 Instituto Geof´ ısico del Per´ u, Arequipa, Per´ u Accepted 2010 May 1. Received 2010 April 27; in original form 2009 September 2 SUMMARY The swarms of microearthquakes, that appeared at Volc´ an de Colima, M´ exico before and after its large 2005 Vulcanian explosions, are discussed. The study of 966 microearthquakes is based on the seismic records of short-period seismic station EZV4 situated at a distance of 1.7 km from the crater. Three samples of microearthquakes were selected: the seismic records associated with a single 2005 September 16 large explosion, and the seismic records associated with two sequences of large explosions, the 2005 March 10 and 13 explosions and the 2005 May 30 and June 2, 5 and 7 explosions. These explosions were seven of 15 large explosions (E 10 11 J) that occurred during the 2005 explosive sequence. The microearthquake waveforms were identified as the records of rockfalls and microexplosions. The explosive microearthquakes represent from 84 to 89 per cent of the total number of microearthquakes with the readable waveforms. The dominant frequencies of the explosive microearthquakes were 2.8 Hz for the 2005 March explosion sequence and 2005 September 16 explosion, and 1.3, 2.0 and 2.8 Hz for the 2005 May-June explosion sequence. Energy of microexplosions that generated the microearthquake waveforms ranged from 10 4 to 7 × 10 7 J. The force component, corresponding to the largest microexplosive events, was estimated at a level from 3.1 × 10 7 to 3.6 × 10 8 N. The appearance of microearthquakes before large volcanic explosions and the sharp increase in their rate of appearance some hours before an explosion makes them a useful instrument for volcano monitoring. Key words: Volcano seismology; Explosive volcanism; Eruption mechanism and flow emplacement; Volcano monitoring. 1 INTRODUCTION The swarms of microearthquakes (magnitudes about 0 to 2) were recorded on the different volcanoes of the world (N´ nez-Corn´ u et al. 1994; Orozco Rojas 1994; Neuberg et al. 1998; Carniel et al. 2006; Iverson et al. 2006; Macedo et al. 2008; Waite et al. 2008). They are associated with effusive-explosive eruptions. At Volc´ an de Colima, M´ exico they were named as ‘hiccups’ or ‘pulgas’, at other volcanoes ‘drumbeats’, ‘low-frequency (LF)’, ‘long-period (LP)’, ‘very-long-period (VLP)’ or ‘hybrid’ earthquakes. These seismic events were observed at different types of volca- noes: basaltic (Stromboli), andesitic (Volc´ an de Colima, Ubinas) and dacitic (Mount St Helens). 1.1 Hybrid microearthquakes at Stromboli volcano, Italy The 924-m high basaltic stratovolcano Stromboli, situated in South- ern Tyrrhenian Sea on the island Stromboli, is an open-conduit volcano characterized by the permanent emission of a gas plume from the craters and by a persistent activity lasting since more than 1000 yr ago (Rosi et al. 2000). Two recent significant effusive eruptions interrupted the persistent explosive activity during 2002 December–2003 March and beginning from 2007 February 27 to April 2. During the effusive phases, the visible explosions were absent, but with decreasing rate of effusive activity, the large ex- plosions were recorded again. The microearthquakes were recorded before, during and after these effusive episodes (Martini et al. 2007; Bonaccorso et al. 2008). Martini et al. (2007) defined microearthquakes at Stromboli as ‘hybrid’ according to their waveforms that begin with a LF pulse followed by a high-frequency pulse. The main swarms of these events were recorded on 2007 March 6–8 and March 20, before and after the 2007 March 15 explosion. Their foci, located by using a probabilistic approach, were clustered within a small volume close to the summit craters of the volcano at a depth of about 400–500 m beneath the crater. The study of other sequence of similar events recorded at Stromboli before the 2003 April 5 explosion (Carniel 808 C 2010 The Authors Journal compilation C 2010 RAS Geophysical Journal International Downloaded from https://academic.oup.com/gji/article-abstract/182/2/808/571345 by guest on 28 March 2018

Transcript of Swarms of microearthquakes associated with the 2005 Vulcanian ...

Page 1: Swarms of microearthquakes associated with the 2005 Vulcanian ...

Geophys. J. Int. (2010) 182, 808–828 doi: 10.1111/j.1365-246X.2010.04647.x

GJI

Min

eral

phys

ics,

rheo

logy

,he

atflow

and

volc

anol

ogy

Swarms of microearthquakes associated with the 2005 Vulcanianexplosion sequence at Volcan de Colima, Mexico

Vyacheslav M. Zobin,1 Oleg E. Melnik,2 Miguel Gonzalez,1 Orlando Macedo3

and Mauricio Breton1

1Observatorio Vulcanologico, Universidad de Colima, Colima 28045, Mexico. E-mail:[email protected] of Mechanics, Moscow State University, Moscow, Russia3Instituto Geofısico del Peru, Arequipa, Peru

Accepted 2010 May 1. Received 2010 April 27; in original form 2009 September 2

S U M M A R YThe swarms of microearthquakes, that appeared at Volcan de Colima, Mexico before andafter its large 2005 Vulcanian explosions, are discussed. The study of 966 microearthquakesis based on the seismic records of short-period seismic station EZV4 situated at a distanceof 1.7 km from the crater. Three samples of microearthquakes were selected: the seismicrecords associated with a single 2005 September 16 large explosion, and the seismic recordsassociated with two sequences of large explosions, the 2005 March 10 and 13 explosionsand the 2005 May 30 and June 2, 5 and 7 explosions. These explosions were seven of15 large explosions (E ≥ 1011 J) that occurred during the 2005 explosive sequence. Themicroearthquake waveforms were identified as the records of rockfalls and microexplosions.The explosive microearthquakes represent from 84 to 89 per cent of the total number ofmicroearthquakes with the readable waveforms. The dominant frequencies of the explosivemicroearthquakes were 2.8 Hz for the 2005 March explosion sequence and 2005 September16 explosion, and 1.3, 2.0 and 2.8 Hz for the 2005 May-June explosion sequence. Energy ofmicroexplosions that generated the microearthquake waveforms ranged from 104 to 7 × 107 J.The force component, corresponding to the largest microexplosive events, was estimated ata level from 3.1 × 107 to 3.6 × 108 N. The appearance of microearthquakes before largevolcanic explosions and the sharp increase in their rate of appearance some hours before anexplosion makes them a useful instrument for volcano monitoring.

Key words: Volcano seismology; Explosive volcanism; Eruption mechanism and flowemplacement; Volcano monitoring.

1 I N T RO D U C T I O N

The swarms of microearthquakes (magnitudes about 0 to 2) wererecorded on the different volcanoes of the world (Nunez-Cornuet al. 1994; Orozco Rojas 1994; Neuberg et al. 1998; Carniel et al.2006; Iverson et al. 2006; Macedo et al. 2008; Waite et al. 2008).They are associated with effusive-explosive eruptions. At Volcan deColima, Mexico they were named as ‘hiccups’ or ‘pulgas’, at othervolcanoes ‘drumbeats’, ‘low-frequency (LF)’, ‘long-period (LP)’,‘very-long-period (VLP)’ or ‘hybrid’ earthquakes.

These seismic events were observed at different types of volca-noes: basaltic (Stromboli), andesitic (Volcan de Colima, Ubinas)and dacitic (Mount St Helens).

1.1 Hybrid microearthquakes at Stromboli volcano, Italy

The 924-m high basaltic stratovolcano Stromboli, situated in South-ern Tyrrhenian Sea on the island Stromboli, is an open-conduit

volcano characterized by the permanent emission of a gas plumefrom the craters and by a persistent activity lasting since morethan 1000 yr ago (Rosi et al. 2000). Two recent significant effusiveeruptions interrupted the persistent explosive activity during 2002December–2003 March and beginning from 2007 February 27 toApril 2. During the effusive phases, the visible explosions wereabsent, but with decreasing rate of effusive activity, the large ex-plosions were recorded again. The microearthquakes were recordedbefore, during and after these effusive episodes (Martini et al. 2007;Bonaccorso et al. 2008).

Martini et al. (2007) defined microearthquakes at Stromboli as‘hybrid’ according to their waveforms that begin with a LF pulsefollowed by a high-frequency pulse. The main swarms of theseevents were recorded on 2007 March 6–8 and March 20, before andafter the 2007 March 15 explosion. Their foci, located by using aprobabilistic approach, were clustered within a small volume closeto the summit craters of the volcano at a depth of about 400–500 mbeneath the crater. The study of other sequence of similar eventsrecorded at Stromboli before the 2003 April 5 explosion (Carniel

808 C© 2010 The Authors

Journal compilation C© 2010 RAS

Geophysical Journal International

Downloaded from https://academic.oup.com/gji/article-abstract/182/2/808/571345by gueston 28 March 2018

Page 2: Swarms of microearthquakes associated with the 2005 Vulcanian ...

Swarms of microearthquakes 809

Figure 1. The unfiltered velocity record of the sequence of mi-croearthquakes (‘hybrid’) before the large 2007 February 27 explosion atStromboli volcano, Italy (station STRA). Code is yyyy_mmdd.

et al. 2006) demonstrated a sharp increase in the number of eventssome hours before the explosion.

Fig. 1 shows the sequence of microearthquakes recorded on 2007February 24 at Stromboli volcano, just before the beginning of itseffusive eruption of February 27, by the broad-band seismic sta-tion STRA, installed at a distance of only 300 m from the crater

and equipped by GURALP CMG-40T sensor. Their waveforms andFourier spectra are shown in Fig. 2. It is seen that the waveformsconsist of two signals, the initial LF and small amplitude pulse(between t1 and t2) and the second, characterized by larger am-plitude and higher frequency. The Fourier spectra of these eventsare characterized by two dominant peaks at frequencies of 3.5 and1.5 Hz.

1.2 LP microearthquakes at Ubinas volcano, Peru

The 5672-m high andesitic-to-rhyolitic stratovolcano Ubinas is themost active Peruvian volcano for last 450 yr. Its historical activity,documented since the 16th century, has consisted of intermittentminor-to-moderate explosive eruptions. The recent unrest at Ubi-nas began on 2006 March 25 with small ash explosions. A newlava dome in the crater began to grow beginning in 2006 April.A sequence of large explosions accompanied by the ash columnsand reaching to 7–8 km a.s.l., occurred during 2006 May-October;the ash eruptions and steam emissions of different intensity contin-ued through to at least 2009 January (BGVN, 1969–2009; Macedoet al. 2008). 43 from 134 explosions were preceded by the se-quences of microearthquakes (Macedo et al. 2008). It was shownthat the microearthquakes waveforms are very similar, occur in

Figure 2. Examples of the seismic records (STRA station) of microearthquakes (‘hybrid’) (A, C and E) recorded during the 2006 February explosive sequenceat Stromboli volcano and corresponding Fourier spectra (B, D and F). Arrows with t1 and t2 mark the duration of the first pulse of the seismic signals. Code isyyyy_mmddhhmm.

C© 2010 The Authors, GJI, 182, 808–828

Journal compilation C© 2010 RAS

Downloaded from https://academic.oup.com/gji/article-abstract/182/2/808/571345by gueston 28 March 2018

Page 3: Swarms of microearthquakes associated with the 2005 Vulcanian ...

810 V. M. Zobin et al.

Figure 3. The unfiltered velocity record of the sequence of micro-earthquakes (LP) before the large 2006 May 29 explosion at Ubinas volcano,Peru (station UB1). Code is yyyy_mmdd.

families and characterized by the dominant frequencies at 2.8 and3.6 Hz.

Fig. 3 shows a sequence of microearthquakes (named LP inMacedo et al. 2008) preceding the 2006 May 29 explosion. Thisrecord was obtained at a digital short-period (Kinemetrics SS1,Ts = 1 Hz) seismic station UB1, installed on May 24 at an altitude

of 4840 m, and at a distance of 2.5 km from the crater (Macedo et al.2008). The waveforms and corresponding Fourier spectra of threemicroearthquakes from this sequence are shown in Fig. 4. It is seenthat the waveforms consist of two signals, the initial LF and smallamplitude pulse (between t1 and t2), and the second, characterizedby larger amplitude and higher frequency. The peak frequenciesof their Fourier spectra range between 2.5 and 4 Hz. More detailspectral study of the Ubinas microearthquakes (Macedo et al. 2008)showed that the microearthquakes of the same waveforms with thepeak frequencies of 2.8–3.6 Hz were also recorded during the 2006explosive activity.

1.3 Drumbeat microearthquakes at Mount St Helens,Cascades

The most detailed study of volcanic microearthquake swarmswas carried out during the lava extrusion at Mount St Helens in2004–2005 (Iverson et al. 2006; Moran et al. 2008a; Waite et al.2008). The 2549-m high dacitic stratovolcano Mount St Helensbegan erupting on 2004 October 1 after 18 yr of quiescence. Thenewly extruded dacite formed a series of spines that were repeat-edly formed and disintegrated (Iverson et al. 2006). A sequence of

Figure 4. Examples of the seismic records (UB1 station) of microearthquakes (LP) (A, C and E) recorded during the 2006 May explosive sequence at Ubinasvolcano and corresponding Fourier spectra (B, D and F). Arrows with t1 and t2 mark the duration of the first pulse of the seismic signals. Code is yyyy_mmdd.

C© 2010 The Authors, GJI, 182, 808–828

Journal compilation C© 2010 RAS

Downloaded from https://academic.oup.com/gji/article-abstract/182/2/808/571345by gueston 28 March 2018

Page 4: Swarms of microearthquakes associated with the 2005 Vulcanian ...

Swarms of microearthquakes 811

explosions was recorded during 2004–2005. The largest of themoccurred on 2005 March 9 (Moran et al. 2008a, 2008b).

On October 16, the appearance of regularly spaced, numeroussmall (Md < 2) shallow earthquakes (named as ‘drumbeat’) wasmarked. The drumbeats consisted of repetitive LF (peak frequenciesin the 2–3-Hz range) and hybrid (peak frequencies in the 8–16-Hzrange) earthquakes and accompanied the regular rate of extrusionof several dacite spines. Most of them were occurring in familiesof similar events. The epicentres of these small earthquakes weredistributed around the crater, within an area 1 × 1 km centred onthe vent (Moran et al. 2008a, 2008b).

Fig. 5 shows a sequence of the drumbeat events, preceding the2005 March 9 explosion, recorded by short-period station MIDEthat was installed at a distance of a few hundred metres from thevent. The waveforms and Fourier spectra of three small events fromthis swarm are shown in Fig. 6. It can be seen that they consist oftwo signals, the initial LF and small amplitude pulse (between t1

and t2) and the second, characterized by larger amplitude and higherfrequency. The Fourier spectra of these events are characterized bydominant peaks at frequencies of 6.5 and 2.3 Hz.

The origin of the drumbeat earthquakes was discussed by Iversonet al. (2006) and Waite et al. (2008). Iverson et al. (2006) considerthem as stick-slip motions between the extruding lava and conduitwalls. This shear-faulting process may be modelled as a combinationof a double couple and a near-vertical single force. At the same time,Waite et al. (2008) noted that no stick-slip events were documentedgeodetically; moreover, these earthquakes appeared when the ex-trusion stopped. They consider that nearly identical waveforms andlocations suggest a repetitive, non-destructive source process.

Modelling of the source mechanisms of 68 drumbeat earthquakesas LP events using 23 channels from eight broad-band seismicstations showed that the moment tensor, found for the best fit model,

Figure 5. The unfiltered velocity record of the sequence of micro-earthquakes (drumbeat) before the large 2005 March 9 explosion at MtSt Helens volcano, Cascades (station MIDE). Code is yyyy_mmddhhmm.

is dominated by the three dipole components and a vertical single-force component. The dipole components indicated the mechanismrepresenting a volumetric source. Waite et al. (2008) consider thatthe vertical-force component may be explained by oscillations ofthe growing spine and the entire dome, and the volumetric sourcemay be represented by a horizontal tensile steam-filled crack. Thecentroid depths indicate the position of these cracks directly beneaththe growing lava dome, about 350 m beneath the dome apex.

1.4 Resume and the subject of our study

This short review of volcanic microearthquakes observed at Strom-boli, Ubinas and Mt St Helens shows that the most common char-acteristic of them is a waveform structure: two pulses, initial ofLF and low amplitude, and the second, of high frequency andhigh amplitude. All these events are characterized by significantlylow amplitudes. Their spectral peak frequencies vary from 1.5 to6.5 Hz.

The subject of our study is the andesitic Volcan de Colima,Mexico and its microearthquake swarms. Microearthquake se-quence was first recorded during its 1991 eruption. This earth-quake sequence continued for 1 d only, 28 d after the beginningof lava extrusion. The dominant frequency of their seismic recordswas about 3 Hz (Nunez-Cornu et al. 1994). No microearthquakesequences were recorded before or during the 1994 phreatic ex-plosion (Jimenez et al. 1995). During the 1999–2005 sequence oflarge Vulcanian explosions at Volcan de Colima, the appearanceof microearthquake swarms became regular. The microearthquakessequences were recorded a few hours before the 1999 largest ex-plosions on February 10, May 10 and July 17(Zobin et al. 2002;Vargas-Bracamontes et al. 2009). Most numerous and long-actingmicroearthquake activity at Volcan de Colima was observed dur-ing the 2005 March–September explosion sequence, when 15 ex-plosions with energy ≥1011 J were recorded (Zobin et al. 2006);all these explosions were preceded and accompanied by the mi-croearthquake swarms over several days.

Our paper presents the study of these microearthquake events. Weanalyse their temporal distribution during the eruption process, theirwaveforms and their nature. We show that two volcanic processessupposedly produced the microearthquakes by small rockfalls dur-ing the partial collapse of the lava dome and by microexplosionsgenerated along the crack system of the lava cap.

2 V O L C A N D E C O L I M A A N D I T SE X P L O S I V E A C T I V I T Y D U R I N G 2 0 0 5M A RC H – S E P T E M B E R

The andesitic, 3860-m high, stratovolcano Volcan de Colima is themost active volcano in Mexico. It is located in the western part of theMexican Volcanic Belt, and together with the Pleistocene volcanoNevado de Colima, forms the Colima Volcanic Complex (Fig. 7).Volcan de Colima displays a wide spectrum of eruption styles,including small phreatic explosions, major block-lava effusions andlarge explosive events (Breton Gonzalez et al. 2002).

The most recent stage of eruptive activity at Volcan de Colimabegan on 2004 September 30. The extrusion of andesitic lava, thatoccurred in 2004 September–November, formed two lava flows ofabout 2400 m long and about 300 m wide and of about 600 m longand 200 m wide, on the N and WNW flanks, respectively. A newlava dome was formed in the crater. With the termination of lavaeffusion, the intermittent explosive activity began. Three explosion

C© 2010 The Authors, GJI, 182, 808–828

Journal compilation C© 2010 RAS

Downloaded from https://academic.oup.com/gji/article-abstract/182/2/808/571345by gueston 28 March 2018

Page 5: Swarms of microearthquakes associated with the 2005 Vulcanian ...

812 V. M. Zobin et al.

Figure 6. Examples of the seismic records (MIDE station) of microearthquakes (drumbeat) (A, C and E) recorded before the large 2005 March 9 explosion atMt St Helens volcano, and corresponding Fourier spectra (B, D and F). Arrows with t1 and t2 mark the duration of the first pulse of the seismic signals. Codeis yyyy_mmdd.

sequences were particularly vigorous; they occurred from March10 to 26, from May 10 to July 5 and from September 16 to 27.Some of these explosions issued material that reached an altitude ashigh as 10 km a.s.l., and pyroclastic flow runout distances reachedup to 5.1 km. The 2005 March explosions destroyed the 2004 lavadome. The repetitive building of new small-size domes was thenobserved during the 2005 April to July explosive activity; each ofthem was destroyed by the following explosion (Fig. 8). The total2005 March–September explosive sequence removed the 2004 lavadome and a sequence of small 2005 April–July lava domes andleft a crater of 260 m in diameter and 30 m deep, having emittedvolcanic material with a volume of about 2×106 m3. Energy ofeight explosions of the 2005 sequence exceeded 1012 J; energy ofseven more explosions was between 1011 and 1012 J (Zobin et al.2006). All of them were preceded and accompanied by the swarmsof microearthquakes.

The 2005 eruption activity of Volcan de Colima was monitoredby a seismic network that consisted of four short-period seismicstations and one broad-band seismic station. In our study, we usethe seismic records of the nearest analogue seismic station EZV4

situated at a distance of 1.7 km from the crater on the northernflank of the volcano and equipped with a short-period (Ts = 1.0 s)vertical seismometer, and the digital station EZ5 equipped with abroad-band three-component GURALP CMG-40TD sensor with acorner frequency of 30 s, situated at a distance of 4 km from thecrater on the southern flank of the volcano together with a short-period instrument EZV5 (Fig. 7). The seismograms of the fartherEZV5 station, as filtered by the distance and free of noise, were usedfor the selection of microearthquakes and preliminary waveformanalysis. The waveforms of selected events were processed thenusing the seismograms of the nearest short-period EZV4 stationwhere they were recorded more clearly. The seismic records of thebroad-band EZ5 station were used as auxiliary; the majority ofmicroearthquakes were unreadable on this station records.

We use also the images of two video cameras Sony model CCD-TRV118 that were installed at two sites (Fig. 7): 15 km S (NAR;Naranjal) and 5.5 km N of the crater (NEV; Nevado) for the identifi-cation of explosive events. They were not synchronized exactly withseismic timing and their images were used only for the identificationof the surface manifestation of volcanic activity.

C© 2010 The Authors, GJI, 182, 808–828

Journal compilation C© 2010 RAS

Downloaded from https://academic.oup.com/gji/article-abstract/182/2/808/571345by gueston 28 March 2018

Page 6: Swarms of microearthquakes associated with the 2005 Vulcanian ...

Swarms of microearthquakes 813

Figure 7. The Colima Volcanic Complex (Volcan de Colima, VC; Nevadode Colima, NC) and the system of monitoring. Triangles are the seismicstations EZV4 and EZ5/EZV5; the video stations NAR and NEV are shownas diamonds. The contour lines at 2000, 3000 and 3500 m show the relief ofthe volcanoes.

3 G E N E R A L D E S C R I P T I O NO F T H E S E I S M I C R E C O R D SO F M I C RO E A RT H Q UA K E S

3.1 Waveforms of the seismic signals of microearthquakes

The general description of the seismic records of microearthquakesis based on the seismic records of short-period seismic station EZV4situated at a distance of 1.7 km from the crater (Fig. 7). The mi-croearthquake waveforms may be separated into three groups.

The first group (type I) is represented by the waveforms consistingof two pulses. Two typical waveforms and their spectrograms areshown in Fig. 9. It is clearly seen in the seismogram that the firstpulse (between t1 and t2) is of lower amplitude and lower frequencycompared with the second pulse. The spectrograms, representingthe contour plots of the log of the power spectral density as afunction of time, show that the signal begins at low (about 0.5–1.0 Hz) frequency level and then continues within the frequencyrange between 1.8 and 4 Hz at the time of arrival of the secondpulse. The spectral frequencies (Fig. 10a) are distributed in therange of 1–4 Hz, with three peaks; the highest corresponds to 2.7–3Hz (41 per cent of events). Two other peaks are noted for 1.2–1.5 Hz

Figure 8. Photos of (A) a lava dome (taken by C. Navarro), (B) a charac-teristic explosion and (C) the dome building-destruction during 2005 June2–6, explosive sequence.

(18 per cent of events) and for 1.8–2.1 Hz (22 per cent of events).These waveforms are similar to the waveforms of microearthquakesobserved at other volcanoes and described in the Introduction. Thedurations of the quake records of the first group (Fig. 10b) aredistributed mainly (for 90 per cent of events) between 20 and 40 s.

The second group (type II) of waveforms (two of them are shownin Fig. 11) are characterized by the spindle-shape seismic signalswith a gradual increase–decrease in amplitude, often without a clearbeginning of the signal. The spectrograms show that the signal isrepresented within the range of frequencies between 1 and 4 Hz.The distribution of the spectral peak frequencies (Fig. 10a) showsthat for 68 per cent of the second type events the peak frequenciesare distributed between 2.7 and 3.7 Hz. The durations of the seismicsignals are distributed with two peaks (Fig. 10b); the largest (for69 per cent of events) is between 25 and 45 s and the second peaklies between 55 and 75 s (28 per cent of events).

Generally, it is possible to say that both types of the seismic sig-nals are characterized by similar range of frequencies and signaldurations for the majority of events. The main difference is in theform of the signals: the two-pulse signal of the first type and a

C© 2010 The Authors, GJI, 182, 808–828

Journal compilation C© 2010 RAS

Downloaded from https://academic.oup.com/gji/article-abstract/182/2/808/571345by gueston 28 March 2018

Page 7: Swarms of microearthquakes associated with the 2005 Vulcanian ...

814 V. M. Zobin et al.

Figure 8. (continued.)

spindle-type signal of the second type. The application of the auto-matic Continuous Hidden Markov Model-based recognition system(Cortes et al. 2009) allows us to discriminate between these twotypes of Colima events (defined by Cortes et al. as LP and COL(collapse), respectively) for 80–90 per cent of recorded events basedon the characteristic features of their records. The cross-correlationtest for the 10-s initial part of the seismic records shows that type Iwaveforms, selected with the same duration of the first pulses, havea good intercluster correlation with a good similarity for the sec-ond pulse records (Fig. 12). The intercluster correlation for type IIevents is absent. For this study, the events were manually identifiedconsidering the difference in the shape of records.

The third type of waveforms includes the signals recorded at thehigh level of noise or not identified (Fig. 13).

3.2 Possible nature of the seismic signalsof microearthquakes

We had two options: to name our signals as the LP or hybrid events,as was done in the majority of publications about similar signals,or to identify their volcanic nature. Working at Volcan de Colima,we had a great advantage over the volcanological studies at other

volcanoes. We had the simultaneous video images of the processesthat generated the seismic signals similar to our microearthquakerecords. These seismic signals were generated by volcanic explo-sions and rockfalls that were clearly seen on the simultaneous videoimages (Figs 14 and 15).

The comparison of the seismic record of an explosion with theseismic records of the microearthquakes of the first type showsthe similar two-pulse seismic records (Fig. 14). The comparison ofthe seismic record of a rockfall with the seismic records of the mi-croearthquakes of the second type shows the similar spindle-typeseismic records (Fig. 15). Therefore, we consider that the first-type microearthquakes were produced by explosive-type eventswhile the second-type microearthquakes were produced by smallrockfalls.

4 T H E C H A R A C T E R I S T I C S O FM I C RO E A RT H Q UA K E S W I T H R E G A R DT O V O L C A N I C P RO C E S S

As it was noted earlier, all 15 large explosions (E ≥ 1011 J), occur-ring during the 2005 explosive sequence at Volcan de Colima (Zobin

C© 2010 The Authors, GJI, 182, 808–828

Journal compilation C© 2010 RAS

Downloaded from https://academic.oup.com/gji/article-abstract/182/2/808/571345by gueston 28 March 2018

Page 8: Swarms of microearthquakes associated with the 2005 Vulcanian ...

Swarms of microearthquakes 815

Figure 9. The velocity records (top panels) and spectrograms (bottom pan-els) of two 2005 type I microearthquakes (A) March 13, 19:08 and (B)September 16, 15:06. The spectrograms in the palettes represent the contourplots of the log of the power spectral density as a function of time. They wereobtained by the short-time Fourier transform of the seismic signal, computedby fast Fourier transformation of overlapping windowed signal segments.Spectrogram is normalized to the maximum value of energy for every 2.56 swindow, so it shows the relative changes in frequency distribution and dom-inant spectral peaks. Arrows show the duration of the first pulse. Seismicstation EZV4, 1.7 km from the crater. Code is yyyy_mmddhhmm.

et al. 2006), were preceded and accompanied by microearthquakeswarms. These explosions are listed in Table 1. Seven of them hadenergy between 3×1011 and 8.6×1011 J, seven explosions had en-ergy between 1×1012 and 7.1×1012 J and one explosion was withenergy of 1.5×1013 J. Two sequences of explosions occurred withthe intervals of 2–3 d: March 10 to 13 and May 30 to July 6.They represented the most energetic explosions (six out of eightexplosions with energy greater than 1×1012 J). The seismic eventsassociated with these two sequences of the largest explosions wereselected to study the general characteristics of microearthquakes.We selected also the September 16 explosion, which occurred about3 months after the main explosion sequences and may be consideredas independent of previous explosions. The remaining eight explo-sions were rather small, separated in time or their seismic recordswere not completed (as for the May 24 event). Their characteristicswill be used for a general discussion only.

Therefore, three samples of microearthquakes were selected: theseismic records associated with two sequences of large explosions:the 2005 March 10 and 13 explosions (Sequence 1, 193 events),the 2005 May 30 and June 2, 5 and 7 explosions (Sequence 2, 530events) and 243 seismic records associated with the single largeexplosion of 2005 September 16.

Figure 10. The distribution of peak spectral frequencies (A) and the seismicrecord durations (B) for two types of microearthquakes. 1, 2 and 3 mark thepeaks in the frequency distribution for the events of type I. Data were takenfrom the seismic records at seismic station EZV4, 1.7 km from the crater.

4.1 Microearthquakes associated with a single largeexplosion

The large (E = 8.6×1011 J) explosion of 2005 September 16 wasa single event separate from the previous large explosions, whichterminated in 2005 July. It was preceded by a long sequence ofmicroearthquakes. Fig. 16(a) shows a set of seismograms recordedat station EZV5 during 2005 September 14–16. The moment onSeptember 14 when the microearthquakes began can be seen. Intotal, 243 events were recorded. Their rate of appearance increasedabout 17 hr after the beginning of sequence (Fig. 16b). The cu-mulative number continued to grow gradually until the moment ofexplosion that occurred on September 16 at 15:47 a.m., about 60 hrafter the beginning of the microearthquake sequence. The rate of thenumber of microearthquakes strongly decreased after the explosion.

The sequence was composed of the different types of small seis-mic signals; the majority of microearthquakes lasted between 20 and30 s. The microexplosive events were more numerous (154 events)and represented 88 per cent of the recognized events. The rockfallevents (21 events) represent 12 per cent of the events. 68 of 243 mi-croearthquakes (or 28 per cent) were recorded with a backgroundof strong noise, and it was impossible to identify their waveforms.

4.2 Microearthquakes associated with two sequencesof large explosions

The next two samples of microearthquakes were associated with twosequences of large explosions: the 2005 March 10 and 13 explosions(Sequence 1) and the May 30 and (largest in the 2005 sequence,E = 1.5 × 1013 J) 2005 June 2, 5 and 7 explosions (Sequence 2).They were selected to see the behaviour of microearthquakes in theconditions of a sequence of explosions.

C© 2010 The Authors, GJI, 182, 808–828

Journal compilation C© 2010 RAS

Downloaded from https://academic.oup.com/gji/article-abstract/182/2/808/571345by gueston 28 March 2018

Page 9: Swarms of microearthquakes associated with the 2005 Vulcanian ...

816 V. M. Zobin et al.

Figure 11. The velocity records (top panels) and spectrograms (bottompanels) of two 2005 type II microearthquakes (A) May 30, 21:20 and (B)September 15, 18:50. The spectrograms in the palettes represent the contourplots of the log of the power spectral density as a function of time. Theywere obtained by the short-time Fourier transform of the seismic signal, com-puted by fast Fourier transformation of overlapping windowed signal seg-ments. Spectrogram is normalized to the maximum value of energy for every2.56 s window, so it shows the relative changes in frequency distributionand dominant spectral peaks. Seismic station EZV4, 1.7 km from the crater.Code is yyyy_mmddhhmm.

4.2.1 Temporal variations in microearthquake sequences

Figs 17(a) and 18 (a) show the cumulative number of mi-croearthquakes during two sequences. It can be seen that everytime, some hours before the large explosions, the cumulative num-ber of events sharply increased. The durations of these preliminaryintervals varied from 6 to 14 hr without any dependence on energyof forthcoming explosions (Table 1). Between the large explosions,the rate of microearthquakes appearance decreases.

4.2.2 Structure of microearthquakes sequences

The microearthquakes sequences were represented again by ex-plosive and rockfall events described (See Section 3.1). Within theSequence 1, 193 events were recorded; 39 per cent of the signals (75events) were recorded on the background of strong noise and werenot identified. The remaining 118 events represented 105 events (or89 per cent) corresponding to the waveforms of explosions, and 13(or 11 per cent) to the waveforms of rockfalls. Within the secondsequence, 530 events were recorded; 30 per cent of the signals (160events) were recorded on the background of strong noise and werenot identified. The remaining 370 events represented 312 events

Figure 12. Eight seismic records of type I with adjusted relative time (A)and the inter-correlation between them (B) are shown. The coefficients ofintercluster correlation for these eight records, shown in B and marked withthe same numbers as in A, vary between 0.55 and 0.95. The majority ofthe seismic records (5–8) are characterized by the coefficients of correlationbetween 0.85 and 0.95 that indicates a high level of correlation betweenthese seismic records.

Figure 13. The unfiltered, corrected for instrument response, velocityrecords of type III events (unidentified records). Seismic station EZV4,1.7 km from the crater. Code is yyyy_mmddhhmm.

C© 2010 The Authors, GJI, 182, 808–828

Journal compilation C© 2010 RAS

Downloaded from https://academic.oup.com/gji/article-abstract/182/2/808/571345by gueston 28 March 2018

Page 10: Swarms of microearthquakes associated with the 2005 Vulcanian ...

Swarms of microearthquakes 817

Figure 14. The comparison of the seismic records of an explosion (A), identified with a video image (B), with two records of microearthquakes of type I(C and D). Seismic station EZV4, 1.7 km from the crater. Code is yyyy_mmddhhmm. Arrows show the duration of the first pulse. The velocity records areunfiltered and corrected for instrument response.

(or 84 per cent) corresponding to the waveforms of explosions, and58 (or 16 per cent) to the waveforms of rockfalls.

5 Q UA N T I F I C AT I O N O F T H EV O L C A N I C E V E N T S T H AT G E N E R AT E DM I C RO E A RT H Q UA K E S

We can consider the microearthquakes as the products of two typesof volcanic events, microexplosions and small rockfalls. Now wewill try to quantify these events and compare their size with thoseof large explosions and pyroclastic flows.

5.1 Quantification of microexplosions

The majority of the microearthquake waveforms, as it was notedearlier, are similar to the two-pulse seismic records produced byvolcanic explosions and described in Zobin et al. (2006, 2008a,2009) . It allows applying a two-stage conceptual model of a Vulca-nian explosion at andesitic volcano developed in Zobin et al. (2006,2009) for the quantification of this type of microearthquakes.

5.1.1 The conceptual model of a Vulcanian explosion at andesiticvolcano and the structure of explosive earthquake waveforms

The conceptual two-stage model of a Vulcanian explosion at anandesitic volcano, proposed by Zobin et al. (2009), describes the

volcanic explosion process as consisting of two stages (the move-ments of the fragmented magma to the surface and a subsequentexplosion). It is based on the field and laboratory experiments thatRipepe et al. (2001) carried out during the study of the explosionsat Stromboli volcano and a laboratory experiment imitating an ex-plosion at a basaltic volcano.

The bubbles in water were formed by air pumped at a constantflow rate from the bottom of a cylindrical tank filled with water.They merged into a pipe imitating a magma conduit. The gas foamaccumulated at the roof of the tank and then collapsed into a largebubble inside the pipe. The bubble began to flow within the pipeand then finally broke at the liquid surface. During the movement ofthe bubble, an acoustic sensor in a tube outside the water recordeda LF signal; when the bubble breaks at the liquid surface, a high-frequency signal was recorded.

The laboratory signal presented strong similarities to the seismicsignal produced by the Stromboli explosion. It allowed Ripepe et al.(2001) to infer that the initial LF seismic signal was generated bythe rapid expansion of gas in the magma conduit, while the high-frequency seismic signal was generated by the explosion at themagma free surface.

The contemporary records of broad-band seismic signal andvideo images of a Vulcanian explosion at Volcan de Colima (Zobinet al 2008b.) showed that the first superficial manifestation of theexplosion (outlet of magma material from the crater) appears onlyabout 15 s after the onset of the seismic record, on the code of thesecond pulse. This indicates that broad-band seismic record wasgenerated by the process occurring within the volcano conduit.

C© 2010 The Authors, GJI, 182, 808–828

Journal compilation C© 2010 RAS

Downloaded from https://academic.oup.com/gji/article-abstract/182/2/808/571345by gueston 28 March 2018

Page 11: Swarms of microearthquakes associated with the 2005 Vulcanian ...

818 V. M. Zobin et al.

Figure 15. The comparison of the seismic records of a rockfall (A), identified with a video image (B), with two records of microearthquakes of type II (C andD). Seismic station EZV4, 1.7 km from the crater. Code is yyyy_mmddhhmm. The velocity records are unfiltered and corrected for instrument response.

Table 1. List of the 2005 large explosions at Volcan de Colima discussed in the paper.

Period of a Number of events recorded atTime sharp increase in seismic station EZV5 during the

Date, GMT, the number of period of a sharp increase in themmdd hhmm Energy, J microearthquakes, hr number of microearthquakes

0310 1409 1.37e + 12 7 390313 2128 1.47e + 12 18 850326 0340 3.15e + 11 11 710420 0156 7.34e + 11 9 440510 1415 7.09e + 11 38 1170516 0201 2.07e + 12 48 1430524 0009 5.44e + 12 No data0530 0826 7.22e + 12 8 460602 0449 3.63e + 12 5 430605 1920 1.46e + 13 7 410607 0404 1.06e + 12 4 170610 0254 3.00e + 11 26 960705 2321 7.92e + 11 45 1620916 1547 8.57e + 11 43 2170927 1008 5.78e + 11 32 69

The similarity between the seismic records of Colima explosionsand the laboratory and field experiments carried out by Ripepe et al.(2001) as well as the result of (Zobin et al. 2008b), showing thatthe seismic records of Colima explosions describe the process ofexplosion generation within the conduit, allowed to use the resultsof (Ripepe et al. 2001) for interpretation of the two-pulse seismicrecords of Colima explosions (Zobin et al. 2006, 2009) with some

corrections for the case of andesitic volcano. The experiments andtheory described in (Ripepe et al. 2001) were applicable to lowviscosity magma with high mobility of gas bubbles. Formation ofgas slugs in more viscous magmas, typical for lava dome formingeruptions, is not possible because relative velocities of bubbles aresmall in comparison with ascent velocity of magma. For an andesiticvolcano may be considered, according to (Eichelberger et al. 1986;

C© 2010 The Authors, GJI, 182, 808–828

Journal compilation C© 2010 RAS

Downloaded from https://academic.oup.com/gji/article-abstract/182/2/808/571345by gueston 28 March 2018

Page 12: Swarms of microearthquakes associated with the 2005 Vulcanian ...

Swarms of microearthquakes 819

Figure 16. The seismic records of 3 d of activity at station EZV5 (A) andthe curve of the cumulative number of microearthquakes (B) before the 2005September 16 explosion. The moment of explosion is shown by an arrow.The seismic record of the explosion is marked by an ellipse.

Melnik & Sparks 1999; Slezin 2003) degassing of silicic magmasoccurs by gas filtration through a system of interconnected bubbles.

5.1.2 The conceptual model of a volcanic microexplosion

The model (Zobin et al. 2006, 2009) was developed for rather largeexplosions. Therefore, it is not directly applicable to microexplo-sions at Colima although seismic records have several similaritiesand its application to the explosive microearthquakes needs someadditional hypothesis. As it was noted, during the 2005 explosivesequence a lava dome building-destruction process was observed,demonstrating a high magma-ascent rate (See Fig. 8). The passageof the gas-and-ash jets of previous small explosions through thelava dome cap (Fig. 19) produces the magma crack system withinthe lava cap and its borders as a result of the brittle failures ofmagma (Zhang 1999; Neuberg et al. 2006; Harrington & Brodsky2007) and forms the cavities under the lava cap. We consider thatthe sources of explosive microearthquakes were situated within thistensile crack system formed at the base of the lava cap and within it(Fig. 19), and propose the following process of microexplosion gen-eration. Magma degasses by permeable flow and gas accumulatesin the sublava-cap cavities. The pressure builds up and cracks at thebottom of the cap start to open. They are filled by the gas mixturereaching high-pressure levels that lead to the crack opening andpropagation to the dome border. As the gas pathway is established,a gas with some ash forms jets and pressure in the cavity decreases.

Figure 17. Temporal variations in the cumulative number of mi-croearthquakes (A), in the durations of the first pulse of the seismic recordsof explosive events (B) and in the peak frequencies of explosive events (C)during the 2005 March explosive sequence. The moments of explosions areshown by the arrows. The characteristics of the microexplosions observedduring the period of sharp increase in the cumulative number of events areshown in B and C by diamantes; the interexplosion events are marked byfilled circles.

It leads to new influx of the gas into the cavity and preparation of anew explosion.

A crack, containing magmatic fluids, is widely accepted as asource of LP seismic vibration (Chouet 1986; Kumagai & Chouet2000; among others). We consider its vibrations during the fillingof it by gas and steam exsolved from ascending magma as a sourceof the first LF pulse in the seismic record. The explosion would bethe result of a piston-like action of the gas-steam jet escaping witha high velocity from the crack. This volumetric process producesthe second high-frequency pulse. These pulses are indicated on theseismogram (Fig. 19) with indexes t1 (arrival of the first pulse), t2

(arrival of the second pulse) and t3 (termination of the explosiveprocess).

The microexplosion activity continues with the process of lavabase brittle fracturing and crushing the base of the lava cap. Theyprepare the passage for the fragmented magma gas-particle disper-sion jets producing large explosions (Fig. 19).

For better understanding of the nature of the first pre-explosionpulse, corresponding to the filling of the tensile magma crack, wedetermine the wave nature of this pulse. Fig. 20 shows the particlemotion for the first pulse of seismic record of microexplosionsrecorded at EZ5 station. It was the only three-component seismicstation; therefore, it was the only seismic station for the study of theparticle motion. To avoid the influence of high-frequency noise, therecords of vertical and radial components were low-pass filtered at1.0 Hz. The record is short (4 s) but it is possible to see that theretrograde motion of particles forms well-expressed ellipses in the

C© 2010 The Authors, GJI, 182, 808–828

Journal compilation C© 2010 RAS

Downloaded from https://academic.oup.com/gji/article-abstract/182/2/808/571345by gueston 28 March 2018

Page 13: Swarms of microearthquakes associated with the 2005 Vulcanian ...

820 V. M. Zobin et al.

Figure 18. Same as Fig. 17 for the 2005 May–June explosive sequence.

vertical plane that is characteristic for Rayleigh waves (Bullen &Bolt 1985).

Therefore, the entrance of gases, or steam, exsolved from as-cending magma, into the pre-existing cracks may generate the first,pre-explosion pulse of seismic signal as a vibration of the crackwalls producing surface Rayleigh wave.

The model is constrained by four source parameters inferredfrom the seismic records (Zobin et al. 2009). Among them are(Fig. 19) a single force F acting in the crack before an explo-sion and governing the gas-particle dispersion movement to thesurface, the time D1 = t2 – t1 of the duration of the first pulse cor-responding to the movement of gas-particle dispersion in the crackbefore an explosion, the duration of the explosion in the conduitD2 = t3 – t2 and energy E of the explosion. So we consider thesource mechanism of a microexplosion as consisting of a singleforce F and a volumetric component characterized by the explosiveenergy E.

We determine the parameters for explosive microearthquakes andcompare them with the parameters obtained for two samples of ex-plosions observed at Volcan de Colima during 2004 and 2005: asample of 236, 2004 September–November small explosions asso-ciated with the 2004 extrusion and a sample of 15 large (E > 1011 J)explosions observed during 2005 March–September (Zobin et al.2006, 2008a). Both samples are complete enough and may serve asthe representative distributions of these groups of explosions.

5.1.3 Modelling of the first seismic pulses and estimationof the force F

The Rayleigh nature of seismic waves representing the first seismicpulses allows us to use the methodology of modelling the firstpulse as a Lamb’s pulse with the estimation of a counter forceof the eruption (Kanamori & Given 1983). To determine a forcecomponent of the source of an explosive event, we calculated the

Figure 19. Illustration of the conceptual model of a microexplosion (A) andits seismic record (B). In the insert (within the conduit, Fig. A), it is shownin general form the process of an explosion as consisted from the movementof fragmented magma to the surface, governing by a single force F, and thefollowing explosion characterizing by a volumetric component with energyE. See other details in the text.

time domain synthetic seismograms excited by a single verticalpulse of a unit force F (1 N) originating at a depth 0.5 km belowthe crater of the volcano and recorded at a distance of 1.7 km.The distance of 1.7 km was the distance to the seismic station EZ4(Fig. 7). The depth of the source of the LF pulse is a problematicvalue. For our case of microexplosions, forming within the magmacracks, it may be the depth of the crack lower point and not thedepth of the following explosion. The depth of 0.5 km was taken asa possible depth of a lava cap bottom in the crater where the systemof cracks may be developed.

For the calculation of synthetics we used codes by Nishimura(1995). These codes were prepared by applying the discretewavenumber method by Bouchon (1979, 1981) and the reflectionand transmission coefficient matrices (Kennett & Kerry 1979). Thesource time function was taken as a triangle with a pulse width τ .The three-layer crust structure used for modelling was taken fromNunez-Cornu et al. (1994) with some simplifications. The modeldoes not take into account the topography of a volcano.

The procedure by Nishimura (1995) does not include any tech-nique of inversion (such as least-squares iterations) for precise com-parison of synthetic and observed records. At the same time, thesimple form of the synthetics allows us to do a visual comparisonof the records with the modelled synthetics (Fig. 21b). We estimatethe values of F and τ selecting an impulse with τ that has the samewidth as the observed impulse and normalizing the average of threefirst peak amplitudes of LF impulse by their synthetic equivalents(Fig. 22).

C© 2010 The Authors, GJI, 182, 808–828

Journal compilation C© 2010 RAS

Downloaded from https://academic.oup.com/gji/article-abstract/182/2/808/571345by gueston 28 March 2018

Page 14: Swarms of microearthquakes associated with the 2005 Vulcanian ...

Swarms of microearthquakes 821

Figure 20. The particle motions of the first pulse of broad-band seismic ve-locity records of type I microearthquake. The seismic records were low-passfiltered at 1.0 Hz. (A) unfiltered, corrected for instrumental response, veloc-ity record (vertical component); (B) the radial and (C) vertical, low-pass fil-tered at 1.0 Hz and corrected for instrumental response; (D) the particle mo-tion. Seismic station EZ5, 4 km from the crater. Code is yyyy_mmddhhmm.

Our assumption about the fixed depth leads to rather small error.Considering that the depth of microearthquake origin can vary from0.1 to 1.0 km, the choice of the depth equal to 0.5 km gives therelative increase of F (+0.11 log unit) compared with the estimationwith the depth of 0.1 km and the relative decrease of F (–0.09log unit) compared with the estimation with the depth of 1.0 km.Additional errors may arise due to some simplifications proposedby the Nishimura’s procedure such as the absence of topographyeffect and the visual comparison of the records. Totally, it does notallow absolute values better than ±0.5 log unit. At the same time,for a sequence of microexplosions, occurring at the same volcanoand recorded at the same seismic station, the topography effect andthe methodology of the seismic records and synthetics fitting wouldbe the same. Therefore, a relative precision of F would be about±0.1–0.2 log unit that includes the error for a fixed depth.

The methodology was originally proposed by Nishimura (1995)for application to LP seismic records; we have used it in ourstudy for short-period records. Considering that the strongest mi-

croearthquakes may radiate more LP vibrations recorded by short-period instruments and could be suitable for our modelling, weestimated the force components for the sources of only 11 ofthe largest microexplosions selected from the events with energyE >107 J. The estimated values of F varied from 3.1×107 to3.6×108 N.

5.1.4 Duration D1 of the first pulses of the seismic signals

The duration of the first pulse, D1, is equal to t2 – t1 (Fig. 21a). Theerrors in picking of these times may vary between 0.1 and 0.5 s fort1 and between 0.5 and 1.0 s for t2. The first pulse of microexplosionrecords is thought to be produced during the upward (or horizontal+ upward) movement of a gas filling the magma cracks duringtime D1 before an explosion. Fig. 24 shows the distribution of thedurations of these pulses for microexplosions. It is seen that themajority of these values vary from 0 to 2.5 s (87 per cent eventsof the March sequence, curve 1; and 94 per cent events from theMay-June sequence, curve 2). Suggesting the constant velocity ofthe gas movement within the cracks, the pulse duration constrainsthe crack length, or the depth of generating of microexplosions.We cannot give any absolute estimation for these depths but wecan estimate the relative position of microexplosions and large andsmall explosions within the volcanic conduit.

D1 values, estimated by (Zobin et al. 2006, 2008a,b) for small ex-plosions, associated with the 2004 September–November extrusion(Fig. 24, curve 3), and the 2005 May–September large explosions(Fig. 23, curve 4), show that large explosions are systematicallycharacterized by D1 ranging from 3.5 to 7.5 s. D1 values, measuredfor small explosions, and vary widely having the peaks character-istic for microexplosions as well as for large explosions. There-fore, the microexplosions are generated within the near-surfacezone of the conduit, above the level of large explosions genera-tion, but coinciding eventually with the level of generation of smallexplosions.

Figs 17(b) and 18(b) show the temporal variations in D1 of explo-sive microearthquakes during two sequences of large 2005 explo-sions. It is difficult to see any regularity in the temporal variations ofD1 during the stages of increase in the number of microearthquakesbefore an explosion or during the stages between the pair explosions.However, it is possible to see the slight tendencies in decreasing ofthe first pulse durations for about 30 hr after the explosions of March10, May 30 and June 2.

5.1.5 Energy of microexplosive events

The second, co-explosion pulse of seismic records of microexplo-sions, as it is shown in Fig. 21(a), is limited by the time of arrival ofthe second pulse t2 and the time of termination of this pulse t3. Thesehigh-frequency pulses are supposed to be generated by the explo-sions of gas at the moment of its outlet from a crack. Therefore,according to the conceptual model, their durations may be associ-ated with the duration of explosions, and their spectral amplitudesare proportional to energy of explosions.

The estimation of the Colima explosions energy in Zobin et al.(2006) was based on the seismic records of the broad-band seismicstation EZ5 situated at a distance of 4 km from the crater. The Fourierspectrum of the second pulse was used to estimate energy of theseismic pulse of explosion recorded at this station, Es. The theoremof Parseval (Weaver 1983) states that it is possible to calculate thetotal amount of energy Ei (for the unit of mass) in a continuous

C© 2010 The Authors, GJI, 182, 808–828

Journal compilation C© 2010 RAS

Downloaded from https://academic.oup.com/gji/article-abstract/182/2/808/571345by gueston 28 March 2018

Page 15: Swarms of microearthquakes associated with the 2005 Vulcanian ...

822 V. M. Zobin et al.

Figure 21. Illustration of the processing of seismic signals. The short-period unfiltered seismic signal (vertical component of velocity) corrected for instrumentresponse is shown in (A). The time indices t1, t2 and t3 indicate the beginning of the first pulse, the beginning of the second pulse and the end of the secondpulse, respectively. The modelling of the first pulse is shown in (B). The heavy dashed line shows the synthetic seismogram; the thin line is the observedseismogram. The Fourier spectrum of the high-frequency velocity impulse is shown in (C). The position of the peak frequency fo is shown by arrow.

signal v (velocity) from the Fourier spectrum of the signal.

Es =∫ 2π

0v2(t)dt . (1)

According to this theorem, we calculate energy Es (in Joules) of theco-explosion pulse as the integral (1) where v is the length of theFourier vector taken from the minimum to maximum frequenciesof the spectrum (Fig. 21b and c). To estimate the total energy ofthe explosion, we multiply energy of the co-explosion pulse, Es,by the coefficient of the seismic portion of the total energy of theexplosion (the ratio between seismic and total energy), p, and thecoefficient of effective attenuation of seismic energy with distanceincluding the effects of geometrical spreading, energy absorption,focal mechanism and station conditions, k. The coefficients p =10−5 and k = 10−13 were approximately estimated for the Colimavolcano explosions recorded by broad-band seismic station at adistance of 4 km from the crater (Zobin et al. 2006). The totalenergy of the explosion Et was calculated as

Et = Es x1018. (2)

Eq. (2) provides the values of E with a precision of about half anorder of magnitude. At the same time, the estimations of E for asequence of explosions at the same volcano can give a resolution ofabout ± 0.1 of log unit.

We selected 11 largest microexplosions, recorded simultaneouslyby EZV4 and EZ5 stations, calculated energy Es of their seismicimpulses and obtained the relation of

EEZ5/EEZV4 = 0.36 ± 0.076. (3)

It allowed modifying the eq. (2) for the seismic records of EZV4station as

Et = Es × 0.36 × 1018. (4)

The estimated energy of explosions producing microexplosionsvaries within about four orders ranging from 1×104 to 7.2×107 J.

Fig. 24(a) shows energy distributions for two groups of mi-croexplosions, occurring in 2005 March (1; 106 events) and 2005May–June (2; 312 events). These distributions are similar with themean values of (3.00 ± 8.83) × 106 and (2.51 ± 6.83) × 106 J,respectively.

The characteristic values for the strength conditions of the rocksmay be obtained as the mean values of these types of distributions(Fellin 2004). Tokarev & Firstov (1967), studying the explosiveearthquakes of Karymsky and Kamchatka, volcano, suggested thatthe existence of the maximum on the frequency–energy curve ob-served for volcanic explosions may indicate the optimal conditionsin the conduit (such as the conduit diameter, the rate of magmadischarge or the physical characteristics of magma) for the occur-rence of an explosion of this characteristic size. Therefore, the mean

C© 2010 The Authors, GJI, 182, 808–828

Journal compilation C© 2010 RAS

Downloaded from https://academic.oup.com/gji/article-abstract/182/2/808/571345by gueston 28 March 2018

Page 16: Swarms of microearthquakes associated with the 2005 Vulcanian ...

Swarms of microearthquakes 823

Figure 22. Illustration of the fitting of the first pulse records with synthetic seismograms. The short-period unfiltered seismic signals (vertical component ofvelocity), recorded by EZV4 station and corrected for instrument response, are shown by thin lines. The heavy dashed lines show the synthetic seismogramscalculated for the source depth of 0.5 km and the different duration of τ . Code is yyyy_mmddhhmm.

Figure 23. Distributions in the durations of the first pulse of the seismicsignals for the microexplosions of the 2005 March sequence (1), the mi-croexplosions of the 2005 May–June sequence (2), the small explosionsof the 2004 November sequence (3) and the large explosions of the 2005March–September sequence (4) at Volcan de Colima.

values of explosive energy, released during the microexplosions atVolcan de Colima, may be considered as the characteristic values ofthe rocks constructing the volcano conduit and associated magmaticactivity.

We show that the groups of microexplosions, occurring during2005 March (1) and 2005 May–June (2), had the distributions be-longing to the same sample with the characteristic mean values and,consequently, occurring in the same conditions of their generation.We applied the Kolmogorov–Smirnov test (Fig. 24b) consideringas a null hypothesis that these two databases, n1 and n2, belong tothe same sample. The maximum difference between two cumula-tive distributions, Dn1, n2, is 0.1644. To reject the null hypothesis atthe significance level α, Dn1, n2 has to be greater than the criticalvalue equal to k1−α

√(n1 + n2)/n1n2, where k1−α is the quantileof the Kolmogorov distribution at the significance level α (Mulleret al. 1979). We checked the significance of D at the level of α =0.01 (k1−α = 1.628) obtaining the critical value of 0.1830, whichis greater than the observed Dn1, n2. It allows us to consider that theboth databases of microexplosions belong to the same sample andmay be considered as having the same nature of generation.

The comparison of energy distributions of microexplosions withenergy distributions obtained for 236 small explosions, associatedwith the 2004 September–November lava extrusion (Fig. 24a, curve3), demonstrates (Fig. 24c) that they, in their turn significantly differ.The maximum difference between the cumulative distributions, D,is 1.00 (Fig. 24c), that is larger than the 0.01-critical value of 0.1402.It is interesting to note, that energy distribution obtained for smallexplosions significantly differs also from energy distribution of the2005 March 15–September large explosions (Fig. 24a, curve 4).The maximum difference between the cumulative distributions, D,is 1.00 (Fig. 24d), that is larger than the 0.01-critical value of 0.4334.

It shows that microexplosions occurred in the conditions thatdiffer from the same of small and large explosions, and may be

C© 2010 The Authors, GJI, 182, 808–828

Journal compilation C© 2010 RAS

Downloaded from https://academic.oup.com/gji/article-abstract/182/2/808/571345by gueston 28 March 2018

Page 17: Swarms of microearthquakes associated with the 2005 Vulcanian ...

824 V. M. Zobin et al.

Figure 24. (A) Distributions in energy of explosions for the microexplo-sions of the 2005 March sequence (1), the microexplosions of the 2005May–June sequence (2), the small explosions of the 2004 November se-quence (3) and the large explosions of the 2005 March–September sequence(4) at Volcan de Colima. (B–D). The Kolmogorov–Smirnov test comparisoncumulative fraction plots for the sequences 1 and 2, 2 and 3 and 3 and 4,respectively. The statistic D is the maximum vertical deviation between twocurves.

generated by the different processes within the volcanic conduit.The separation between the large and small explosions is at a levelof 1011 J and the separation between small and microexplosions isat a level of 107 J (Fig. 25).

5.1.6 Durations of microexplosions

The errors in picking of t3 from the short-period seismic recordsmay vary between 1.0 and 5.0–15 s. It is not so important forenergy estimation, the coda amplitude is low, but it may raise largeerrors in D2. Therefore, we prefer to not discuss this parameter ofmicroexplosions.

5.1.7 Spectral content

Fig. 26 shows the characteristic Fourier velocity spectra for theexplosive events. It is seen that each of them has a few peaks;we use in our analysis the only peak, fo (Fig. 21c) correspondingto the maximum spectral amplitude. These peak frequencies aredifferent from event to event (Fig. 26) but are rather regular withineach of two sequences. The distribution of peak frequencies shows(Fig. 27) that the explosive microearthquakes associated with theMarch explosions were characterized by a dominant peak frequencyof 2.8 Hz. However, for the May–June, three dominant peaks wereobserved, at 1.3, 2.0 and at the same 2.8 Hz as for the March events.

Figure 25. Relation between a counter force and energy of explosion forthree groups of explosive events of Volcan de Colima (large, small andmicroexplosions).

The peak frequency of velocity spectrum is usually a functionof the source dimension (Lay & Wallace 1995). Therefore, theexistence of stable frequency peaks may indicate the repetitive oc-currence of explosive microearthquakes along the same cracks. Itis interesting that the lower frequency microearthquakes with thepeak frequencies of 1.3 and 2.0 Hz appeared within Sequence 2when the largest explosions occurred. The distribution of the peakfrequencies of explosive microearthquakes during the 2005 Septem-ber 14–16 sequence (see Section 4.1) gives the same peak at 2.8 Hzas for the March sequence (Fig. 27, curve 3).

Figs 17(c) and 18(c) show the temporal variations in the peakfrequencies of explosive microearthquakes spectra during two se-quences of large explosions. No regularity is seen in the temporalvariations of the peak frequencies during the stages of increase inthe number of microearthquakes before an explosion or during thestages between each of pair explosions. Only before the first largeexplosion of March 10, the gradual decrease in frequencies wasobserved.

5.2 Quantification of rockfalls

Rockfall signals represent a small part of microearthquake records.These events are believed to occur as a result of the partial collapseof the lava dome during its building destruction. The size of rock-fall deposit may be approximately estimated from the duration ofseismic signal. Zobin et al. (2005) studied the dependence of the du-ration of short-period seismic records at a distance of 4–7 km fromthe crater on the mean volume of deposit of pyroclastic flow rocks.It was shown that the short-period seismic signals can be dividedinto three categories based on their duration: short events with du-rations less than 100 s; intermediate events with durations between100 and 250 s and long events with durations longer than 250 s. Itwas inferred that long events correspond to pyroclastic flows withmean deposit volume ∼2×105 m3, and intermediate events repre-sent pyroclastic flows with mean deposit volume ∼1×103 m3. Fieldobservations suggest that short events correspond to rockfalls witha mean volume of ∼50 m3.

C© 2010 The Authors, GJI, 182, 808–828

Journal compilation C© 2010 RAS

Downloaded from https://academic.oup.com/gji/article-abstract/182/2/808/571345by gueston 28 March 2018

Page 18: Swarms of microearthquakes associated with the 2005 Vulcanian ...

Swarms of microearthquakes 825

Figure 26. Examples of the unfiltered, corrected for instrument response, velocity records of microexplosions (A, C and E) recorded during the 2005 May–Juneexplosive sequence at Volcan de Colima and corresponding Fourier spectra (B, D and F). Arrows with t2 and t3 mark the beginning and the end of the secondpulse. Seismic station EZV4, 1.7 km from the crater. Code is yyyy_mmddhhmm.

Fig. 28 shows that the seismic records of the rockfall-type mi-croearthquakes are characterized by durations less than 100 s, set-tling them within the category of short events which correspond torockfalls with a mean volume of ∼50 m3.

6 R E S U LT S A N D D I S C U S S I O N

Our study allows us to propose the following characteristics ofmicroearthquakes recorded at Volcan de Colima during the 2005explosive sequences.

(1) The swarms of microearthquakes appeared at Volcan de Col-ima before and after its large explosions. Their rate of appearancesharply increased 6 to 40 hr before large explosions allowing analarm to be raised for Civil Protection.

(2) The microearthquake waveforms were identified as thespindle-shape records of rockfalls and two-pulse records of mi-croexplosions. The explosive microearthquakes represent from 84to 89 per cent of the total number of microearthquakes withthe readable waveforms and are similar to the waveforms ofmicroearthquakes (LP, hybrid or drumbeats) observed at othervolcanoes.

(3) The microexplosion waveforms are represented by two pulsessupposedly generated by the source consisting of a single forcecomponent and volumetric component. The force component, cor-

responding to the largest microexplosive events, was estimated at alevel from 3.1×107 to 3.6×108 N. Energy of volumetric component(explosions that generated the microearthquake waveforms) rangesfrom 104 to 7×107 J.

(4) The dominant frequencies of the explosive microearthquakeswere 2.8 Hz for the 2005 March explosion sequence and 1.3, 2.0 and2.8 Hz for the 2005 May–June explosion sequence. The durations ofthe first pulse of the explosive microearthquake waveforms, corre-sponding to the time of the movement of gas, filling the cracks beforean explosion, range between 0 and 2.5 s for all microearthquakes.

The appearance of microearthquakes before large volcanic ex-plosions and lava dome collapse and the sharp increase in theirrate of appearance some days or hours before the explosions makesthem a useful instrument for volcano monitoring. Some alarms offorthcoming volcanic explosions were announced at Colima Vol-cano Observatory during 2005 by G. Reyes for the Civil Protectionof Colima State, once with the publication of this alarm in thelocal newspaper (El Comentario, 2005 September 27: publication‘Dentro de las proximas 72 horas, se espera gran explosion delVolcan’). The explosion with energy 5.8×1011 J occurred onSeptember 27 at 05:08 a.m., local time.

The data on the period of a sharp increase in the number ofmicroearthquakes before a large explosion (duration of the periodand the number of events recorded during the period; Table 1)

C© 2010 The Authors, GJI, 182, 808–828

Journal compilation C© 2010 RAS

Downloaded from https://academic.oup.com/gji/article-abstract/182/2/808/571345by gueston 28 March 2018

Page 19: Swarms of microearthquakes associated with the 2005 Vulcanian ...

826 V. M. Zobin et al.

Figure 27. The distribution of peak frequencies of microexplosions duringthe (1) 2005 March, (2) 2005 May–June and (3) explosive sequences andbefore the 2005 September 16 explosion at Volcan de Colima. Seismicstation EZV4, 1.7 km from the crater.

Figure 28. The distribution of the durations of rockfall microearthquakesand its position comparing with the durations of short-period seismic signalsfor the different volume of deposits of pyroclastic flows and rockfalls. Themean volume of deposits of pyroclastic flows and rockfalls associated withthe seismic records of different durations is shown according to Zobin et al.(2005).

give a mean rate of microearthquakes appearance before the largeexplosions (E ≥ 1012 J) equal to 5.8 ± 1.5 events/hr at stationEZV5 situated at a distance of 4 km from the crater (for the eventswith energy <1012 J the rate is 3.9 ± 1.4 events/hr). At the sametime, the mean rate of appearance of microearthquakes between theexplosions is only 2.0 ± 0.3 events/hr. This significant differencein the microearthquakes rates during different periods of volcanicactivity at Volcan de Colima allows us to carry out a monitoring ofexplosive activity at the volcano in real time and to give an alarmof the forthcoming explosion a few hours before the event.

The 2005 alarms of large explosions at Volcan de Colima werenot realized in the quantitative form, they were based only on the

general intuition of the investigators. Nevertheless, the cumulativecurves of the number of microearthquakes before large explosionsshown in Figs 16–18 are similar to those used by Minakami (1960),Tokarev (1963), Cornelius & Voight (1995), Reyes-Davila & Dela Cruz-Reyna (2002), among others. Considering that the 2005Colima microearthquake swarms were the result of the mechanicaldeformation of the upper (rockfalls) and lower parts of the lava dome(microexplosions) during the building of the new lava domes, theMaterials Failure Forecast Method (Cornelius & Voight 1995) mayserve as a theoretical basis for the quantitative forecasting of thetime of large explosions. The study of the repose intervals betweenthe Vulcanian explosions may give additional information for theexplosion forecasting (Connor et al. 2003).

The microearthquakes were identified as the products of rockfallsand microexplosions. A rockfall as the source of the seismic signalsis not a questionable event. The spindle-shape seismic records ofrockfalls are well known during the dome-building eruptions at an-desitic volcanoes Soufriere Hills (Luckett et al. 2002) and Merapi(Ratdomopurbo & Poupinet 2000). At the same time, the explo-sive nature of other group of microearthquakes needs in additionaldiscussion.

We use in our paper a two-stage conceptual model of volcaniceruption to explain the two-pulse seismic record of explosion ormicroexplosion. The traditional practice in the study of explosiveseismic records is to filter their high-frequency content, obtainingthe LP or VLP signals. After filtering, the seismic records haveno two-pulse waveform but a simple LP pulse to determine thecharacteristics of a certain centroid representing the generalizedsource of the explosion (Chouet et al. 2008; Waite et al. 2008;among many others). Another approach is to consider the first30–60 s of the seismic record for analysis as a signature of anLP event and study the characteristics of this explosive source (Paloet al. 2009). Our model has another ideology. We study the seismicsignature of an explosion without significant filtering that allowsconsidering the observed two pulses as characterizing the differentstages in the source of explosive event.

Our main argument for the explosive nature of type 1 mi-croearthquakes was the similarity between the seismic signals ofvolcanic explosions and this group of microearthquakes. Additionalargument may be taken from the study of the source mechanism ofmicroearthquakes (or LP events) associated with the 2005 July ex-plosions at Mt St Helens (Waite et al. 2008). Waite et al. (2008)obtained that the source mechanism of LP events includes a volu-metric component and a vertical force. A similar mechanism wasobtained by Chouet et al. (2008) for the 1997 Stromboli explo-sions that support our suggestion of the explosive nature of thesemicroearthquakes. The interpretation of the explosion source mech-anism (Chouet et al. 2008) as the result of a piston-like action ofthe liquid associated with the disruption of a gas slug, transitingthrough discontinuities in the conduit direction, is in a good accor-dance with the proposed in our model, a piston-like action of thegas-steam slug transiting through the magma crack.

The dominant frequencies of the Colima explosive mi-croearthquakes are close to those estimated for LP events of Ubinas(2.8 and 3.6 Hz; Macedo et al. 2008) and Mount St Helens (2–3 Hzrange; Waite et al. 2008).

The locations of LP and hybrid events at different volcanoesshowed that they occur within a small area (about 1 km2), centredon the vent, at the depth of about 300–500 m beneath the crater(Moran et al. 2008a; Martini et al. 2007). We can attribute thesecharacteristics to the position of our explosive microearthquakesalso.

C© 2010 The Authors, GJI, 182, 808–828

Journal compilation C© 2010 RAS

Downloaded from https://academic.oup.com/gji/article-abstract/182/2/808/571345by gueston 28 March 2018

Page 20: Swarms of microearthquakes associated with the 2005 Vulcanian ...

Swarms of microearthquakes 827

Having a short-period seismic network (only one broad-bandstation EZ5 was in operation during 2005) we cannot test other,noted earlier, possible mechanisms of the type I seismic events.The construction of the seismic moment tensor solution requires anear-volcano network of broad-band seismic stations; we did nothave them during the 2005 explosive sequence. Now the networkof six broad-band stations is installed around Volcan de Colima.Therefore, the models proposed by Chouet et al. (2008), Waiteet al. (2008), and others, to explain the nature of LP, VLP or hybridseismic signals at volcanoes may be useful for study of the natureof at least a part of type II events during the future explosions atVolcan de Colima.

A C K N OW L E D G M E N T S

The comments of Carlos Navarro and two anonymous reviewershelped to improve the manuscript. We also thank Carlos Navarrofor the image used in Fig. 8(A). We thank Gabriel Reyes andthe personnel of the seismic network RESCO of Colima Uni-versity for providing the seismic records of the Colima earth-quakes. Luca D’Auria and Steve Malone provided us with thedigital seismograms of the Stromboli and Mount St Helens mi-croearthquakes. Raul Arambula helped with the application ofcross-correlation test using the codes GISMO prepared by MikeWest (http://www.giseis.alaska.edu/Seis/EQ/tools/GISMO/). JuliaCrummy improved our English grammar. The processing of thedigital signals was realized using the program DEGTRA providedby Mario Ordaz. This study was partially supported by Fondo Sec-torial de Investigacion para la Educacion SEP-CONACYT, projectNo 79998 (VMZ). OEM acknowledges financial support from theRussian Foundation for Basic Research (grant 08–01-00016).

R E F E R E N C E S

BGVN, 1969–2009. Smithsonian Institution, Washington, DC.Bonaccorso, A., Gambino, S., Guglielmino, F., Mattia, M., Puglisi, G. &

Boschi, E., 2008. Stromboli 2007 eruption: deflation modeling to infershallow-intermediate plumbing system, Geophys. Res. Lett., 35, L06311,doi:10.1029/2007GLO032921.

Bouchon, M., 1979. Discrete wave number representation of elasticwave fields in three space dimensions, J. geophys. Res., 84, 3609–3614.

Bouchon, M., 1981. A simple method to calculate Green’s functions forelastic layered media, Bull. seism. Soc. Am., 71, 959–971.

Breton Gonzalez, M., Ramırez, J.J. & Navarro, C., 2002. Summary of thehistorical eruptive activity of Volcan de Colima, Mexico 1519—2000,J. Volc. Geotherm. Res., 117, 21–46.

Bullen, K.E. & Bolt, B.A., 1985. An Introduction to the Theory of Seismol-ogy, Cambridge University Press, Cambridge, 499 pp.

Carniel, R., Ortiz, R. & Di Cecca, M., 2006. Spectral and dynamical hints onthe time scale of preparation of the 5 April 2003 explosion at Strombolivolcano, Can. J. Earth Sci., 43, 41–55.

Chouet, B., 1986. Dinamics of a fluid-driven crack in three dimensions bythe finite difference method, J. geophys. Res., 91, 13 967–13 992.

Chouet, B., P. Dawson & M. Martini, 2008. Shallow-conduit dynamics atStromboli volcano, Italy, imaged from waveform inversions, in FluidMotions in Volcanic Conduits: A Source of Seismic and Acoustic Signals,pp. 57–84, eds Lane, S. & Gilbert, J.S., Geol. Soc, Spec. Publ. 307,London.

Connor, C.B., Sparks, R.S.J., Mason, R.M., Bonadonna C. & Young, S.R.,2003. Exploring links between physical and probabilistic models of vol-canic eruptions: the Soufriere Hills Volcano, Montserrat, Geophys. Res.Lett., 30, 1701, doi:10.1029/2003GL017384.

Cornelius, R.R., Voight, B., 1995. Graphical and PC-software analysis ofvolcano eruption precursors according to the Materials Failure ForecastMethod (FFM). J. Volc. Geotherm. Res., 64, 295–320.

Cortes, G., Arambula, R., Alvarez, I., Benıtez, C.M., Ibanez, J.M., Lesage, P.,Gonzalez-Amezcua, M. & Reyes-Davila, G., 2009. Analysis of Colima,Popocatepetl and Arenal volcanic seismicity using an automatic Con-tinuous Hidden Markov Model-based recognition system, in The VOL-UME Project, pp. 150–160, eds Bean, C.J., Braiden, A.K., Lockmer, I.,Martini, F. & O’Brien, G.S., University College Dublin, Dublin, ISBN:978-1-905254-39-2.

Eichelberger, J.C., Carrigan, C.R., Westrich, H.R. & Price, R.H.,1986. Non-explosive silicic volcanism, Nature, 323, 598–602,doi:10.1038/323598a0.

Fellin, W., 2004. Assessment of characteristic shear strength parameters ofsoil and its implication in geotechnical design, in Analyzing Uncertaintyin Civil Engineering, pp. 1–15, eds Fellin, W., Lessmann, H., Oberguggen-berger, M. & Vieider, R., Springer, Berlin.

Harrington, R.M. & Brodsky, E.E., 2007. Volcanic hybrid earthquakesthat are brittle-failure events, Geophys. Res. Lett., 34, L06308,doi:10.1029/2006GL028714.

Iverson, R.M. et al. 2006. Dynamics of seismogenic volcanic extrusion atMount St Helens in 2004–05, Nature, 444, 439–443.

Jimenez, Z., Reyes, G. & Espındola, J.M., 1995. The July 1994 episode ofseismic activity at Colima volcano, Mexico. J. Volc. Geotherm. Res., 64,321–326.

Kanamori, H. & Given, J.W., 1983. Lamb pulse observed in nature, Geophys.Res. Lett., 10, 373–376.

Kennett, B.L. & Kerry, N.J., 1979. Seismic waves in a stratified half space,Geophys. J. R. astr. Soc., 57, 557–583.

Kumagai, H. & Chouet, B., 2000. Acoustic properties of a crack containingmagmatic or hydrothermal fluids, J. geophys. Res., 105, 25 493–25 512.

Lay, T. & Wallace, T.C., 1995. Modern Global Seismology, Academic Press,San Diego, 521 pp.

Luckett, R., Baptie, B. & Neuberg, J. 2002, The Relationship Between De-gassing and Rockfall Signals at Soufriere Hills Volcvano, Montserrat,the Eruption of Soufriere Hills Volcano, Montserrat, From 1995 to 1999,pp. 595–603, eds Druitt, T.H. & Kokelaar, B.P., Geol. Soc., Memoirs 21,London.

Macedo O., Metaxian, J-Ph., Taipe, E. & Ramos, D., 2008. Actividad sismo-volcanica asociada a la erupcion del volcan Ubinas en 2006–2008. XIIICongreso Latinoamericano de Geologia, Lima, Peru, 29 setiembre – 3octubre, 2008.

Martini, M. et al. 2007. Seismological monitoring of the February 2007effusive eruption of the Stromboli volcano, Ann. Geophys., 50, 775–788.

Melnik, O. & Sparks, R.S.J., 1999. Nonlinear dynamics of lava dome extru-sion, Nature, 402, 37– 41.

Minakami, T., 1960. Fundamental research for predicting volcanic eruptions,Part I, Bull. Earthq. Res. Inst., 38, 497–544.

Moran, S.C., Malone, S.D., Qamar, A.I., Thelen, W., Wright, A.K. &Caplan-Auerback, J., 2008a. Seismicity associated with the reneweddome-building eruption of Mount St. Helens, 2004–2005, in A VolcanoRekindled: The First Year of Renewed Eruptions at Mount St. Helens,2004–2006, U.S. Geolog. Surv. Prof. Paper, pp. 27–60, eds Sherrod, D.R.,Scott, W.E. & Stauffer, P.H., U.S. Geological Survey, Washington DC.

Moran, S.C, McChesney, P.J. & Lockhart, A.B., 2008b. Seismicity and infra-sound associated with explosions at Mount St. Helens, 2004–2005, in AVolcano Rekindled: The First Year of Renewed Eruptions at Mount St. He-lens, 2004–2006, U.S. Geolog. Surv. Prof. Paper, pp. 11–127, eds Sherrod,D.R., Scott, W.E. & Stauffer, P.H., U.S. Geological Survey, WashingtonDC.

Muller, P., Neumann, P. & Storm, R., 1979. Tafeln der MathematischenStatistik, VEB Fachbuchverlag, Leipzig, 272 pp.

Neuberg, J., Baptie, B., Luckett, R. & Stewart, R., 1998. Results fromthe broadband seismic network on Montserrat, Geophys. Res. Lett., 25,3661–3664.

Neuberg, J., Tuffen, H., Collier, L., Green, D. & Powell, T., 2006. Thetrigger mechanism of low-frequency earthquakes on Montserrat, J. Volc.Geotherm. Res., 153, 37–50.

C© 2010 The Authors, GJI, 182, 808–828

Journal compilation C© 2010 RAS

Downloaded from https://academic.oup.com/gji/article-abstract/182/2/808/571345by gueston 28 March 2018

Page 21: Swarms of microearthquakes associated with the 2005 Vulcanian ...

828 V. M. Zobin et al.

Nishimura T., 1995. Source parameters of the volcanic eruption earth-quakes at Mount Tokachi, Hokkaido, Japan, and magma ascending model,J. geophys. Res., 100, 12 456–12 473.

Nunez-Cornu, F., Nava, F.A., De la Cruz Reyna, S., Jimenez, Z., Valencia, C.& Garcıa-Arthur, R., 1994. Seismic activity related to the 1991 eruptionof Colima volcano, Mexico, Bull. Volcanol., 56, 228–237.

Orozco Rojas, J., 1994. Analysis of the peculiar swarm of low-frequencyearthquakes detected at Colima volcano, Mexico, Individ. Studies Par-ticip. Intern. Inst. Seismol. Earthq. Engin., Tokyo, 30, 95–109.

Palo, M., Ibanez, J.M., Cisneros, M., Breton, M., Del Pezzo, E., Ocana,E., Orozco-Rojas, J. & Posadas, A.M., 2009. Analysis of the seismicwavefield properties of volcanic explosions at Volcan de Colima, Mexico:insights into the source mechanism, Geophys. J. Int., 177, 1383–1398.

Ratdomopurbo, A. & Poupinet, G., 2000. An overview of the seismicityof Merapi volcano (Java, Indonesia), J. Volc. Geotherm. Res., 100, 193–214.

Reyes-Davila, G.A., De la Cruz-Reyna, S., 2002. Experience in the short-term eruption forecasting at Volcan de Colima, Mexico, and public re-sponse to forecasts. J. Volc. Geotherm. Res., 117, 121–127.

Ripepe, M., Ciliberto, S. & Della Schiava, M., 2001. Time constraints formodelling source dynamics of volcanic explosions at Stromboli, J. geo-phys. Res., 106, 8713–8727.

Rosi M., Bertagnini, A. & Landi, P., 2000. Onset of the persistent activity atStromboli Volcano (Italy). Bull. Volcanol., 62, 294–300.

Slezin, Yu. B., 2003. The mechanism of volcanic eruptions (a steady stateapproach), J. Volc. Geotherm. Res., 122, 7–50.

Tokarev, P.I., 1963. On a possibility of forecasting of Bezymianny volcanoeruptions according to seismic data, Bull. Volcanol., 26, 379–386.

Tokarev, P.I. & Firstov, P.P., 1967. Seismological investigations at Karymskyvolcano. Bull. Volcanol. Stat., 43, 9–22 (In Russian).

Vargas-Bracamontes, D.M., Nava, A. & Reyes-Davila, G.A., 2009. Time-scale wavelet patterns related to the 1998–1999 eruptions of the Colima

volcano, and their possible implications for eruption forecasting, J. Volc.Geotherm. Res., 184, 271–284.

Waite, G.P., Chouet, B.A. & Dawson, P.B. 2008. Eruption dynamics at MountSt. Helens imaged from broadband seismic waveforms: interaction ofthe shallow magmatic and hydrothermal systems, J. geophys. Res., 113,B02305, doi:10.1029/2007JB005259.

Weaver, H. J. 1983. Applications of Discrete and Continuous Fourier Anal-ysis, John Wiley & Sons, New York.

Zhang, Y., 1999. A criterion for the fragmentation of bubbly magma basedon brittle failure theory, Nature, 402, 648–650.

Zobin, V. et al. 2002. Overview of the 1997–2000 activity of Volcan deColima, Mexico, J. Volc. Geotherm. Res., 117, No 1–2, 1–19.

Zobin, V.M., Orozco-Rojas, J., Reyes-Davila, G.A. & Navarro, C., 2005.Seismicity of an andesite volcano during block-lava effusion: volcande Colima, Mexico, November 1998-January 1999, Bull. Volcanol., 67,No 7, 679–688.

Zobin, V.M., Navarro, C., Reyes-Davila, G., Orozco, J., Breton, M., Tellez,A., Reyes-Alfaro, G. & Vazquez, H., 2006. The methodology of quan-tification of volcanic explosions from broadband seismic signals and itsapplication to the 2004–2005 explosions at Volcan de Colima, Mexico,Geophys. J. Int., 167, 467–478. doi:10.1111/j.1365-246X.2006.03108.x

Zobin, V.M., Varley, N.R., Gonzalez, M., Orozco, J., Reyes, G.A., Navarro,C., Breton, M., 2008a. Monitoring the 2004 andesitic block-lava extrusionat Volcan de Colima, Mexico from seismic activity and SO2 emission,J. Volc. Geotherm. Res., 177, 367–377.

Zobin, V.M., Reyes, G. A., Guevara, E., & Breton, M., 2008b. Seis-mological constraints on the position of the fragmentation surfacesin the volcano conduit, Earth planet. Sci. Lett., 275, 337–341.doi:10.1016/j.epsl.2008.08.034.

Zobin, V.M., Reyes, G. A., Guevara, E. & Breton, M., 2009. Scaling relation-ship for Vulcanian explosions derived from broadband seismic signals,J. geophys. Res., 114, B03203, doi:10.1029/2008JB005983.

C© 2010 The Authors, GJI, 182, 808–828

Journal compilation C© 2010 RAS

Downloaded from https://academic.oup.com/gji/article-abstract/182/2/808/571345by gueston 28 March 2018