Seismic activity related to the 2000 eruption of the Hekla volcano, Iceland

Bull Volcanol (2005) 68: 21–36DOI 10.1007/s00445-005-0417-7

RESEARCH ARTICLE

Heidi Soosalu · Pall Einarsson ·Bergflora S. fiorbjarnardottir

Seismic activity related to the 2000 eruption of the Hekla volcano,Iceland

Received: 26 November 2003 / Accepted: 13 December 2004 / Published online: 3 May 2005C© Springer-Verlag 2005

Abstract The 2000 Hekla eruption took place from Febru-ary 26 to March 8. Its seismic expressions were a swarmof numerous small earthquakes related to its onset, andlow-frequency volcanic tremor that continued throughoutthe eruption. A swarm of small earthquakes was observedsome 80 min before the onset of the eruption, and the sizeof the events increased with time. Low-frequency volcanictremor, with a characteristic frequency band of 0.5–1.5 Hzand dominant spectral peak(s) at 0.7–0.9 Hz, became visi-ble at 18:19 GMT on February 26, marking the onset of theeruption. The tremor amplitude rose quickly and was veryhigh in the beginning of the eruption. However, it soon be-gan to decrease after about an hour. In general, the seismicactivity related to the 2000 Hekla eruption was very simi-lar to what was observed in the previous eruption in 1991.Based on knowledge gained from seismicity and strain ob-

Editorial responsibility: J Stix

H. SoosaluNordic Volcanological Center, Institute of Earth Sciences,University of Iceland,Sturlugata 7,IS-101 Reykjavık, Iceland

P. Einarsson (�)Institute of Earth Sciences, University of Iceland,Sturlugata 7,IS-101 Reykjavık, Icelande-mail: [email protected]: +354-552-1347

B. S. fiorbjarnardottirIcelandic Meteorological Office,Busta›avegur 9,IS-150 Reykjavık, Icelande-mail: [email protected]

Present address:H. SoosaluDepartment of Earth Sciences, Bullard Laboratories,Cambridge University,Madingley Road,CB3 0EZ Cambridge, United Kingdome-mail: [email protected]

servations from 1991, this was the first time that a Heklaeruption was predicted.

Keywords Hekla eruption 2000 . Earthquakes . Volcanictremor . Low-frequency earthquakes

Introduction

The volcano Hekla is located at the Mid-Atlantic plateboundary, at the junction of its transform segment, theSouth Iceland seismic zone and its rift segment, the Easternvolcanic zone (Fig. 1). The location is tectonically unstable,which is reflected by the dual nature of Hekla seismicity.The volcano is seismically active almost only during itseruptions (Einarsson 1991). In non-eruptive times the fewsmall earthquakes (magnitude typically ≤ 1) at Hekla fol-low the pattern of the South Iceland seismic zone seismicity(Soosalu and Einarsson 1997). According to observationsfrom June 1990–September 2004 the seismicity is con-centrated in two diffuse N-S lineaments, one crossing theeast-middle part of Hekla and Vatnafjoll volcano south ofit, and the other one along the west margin of these volca-noes (Soosalu and Einarsson submitted). The earthquakehypocentres are typically at 8–13 km in depth, which issimilar to earthquakes in the east end of the South Icelandseismic zone.

The Hekla volcanic system consists of a central volcanoand a fissure swarm with an ENE trend. The volcano it-self is an elongated ridge, which has been built up to aheight of 1,500 m a.s.l. in repeated fissure eruptions. Theproducts of the Hekla system range from basalts throughbasaltic andesites to dacites and rhyolites (Jakobsson 1979).The more silicic eruptives are issued from the centralvolcano itself and the basaltic products from the fissureswarms.

Hekla is one of the most active volcanoes in Iceland andhas erupted at least 18 times during the last 1,100 years. Thelatest eruption took place in 2000 and lasted 12 days, fromFebruary 26 to March 8. The lava was basaltic andesite

22

22˚W 21˚W 20˚W 19˚W

63˚ 40'N

64˚ 00'N

64˚ 40'

0 50

km

64˚ 20'N

64˚ 40'N

HE LJ

BUR

ASB

GYG

HAU

HVO

MID

SAU

SKR

Hekla

V

WVZSISZ EVZ

NVZ

TFZ

Volcanic system:

central volcano

fissure swarm

earthquake faults

glacier

digital seismograph station

analogue seismograph station

Fig. 1 Index map of the Heklaarea. The seismic stationsmentioned in the text arenamed. The strain station BURis shown with a black diamond.The volcanic system featuresare from Einarsson andSæmundsson (1987). V marksthe Vatnafjoll volcano. Thesmall index map shows thelocations of the plate boundaryareas: Western (WVZ), Eastern(EVZ) and Northern (NVZ)volcanic zones, South Icelandseismic zone (SISZ) and Tjornesfracture zone (TFZ)

with an estimated volume of 0.286 km3 (Hoskuldsson et al.in prep). A minor amount of airborne dacite was observed(Olafsdottir et al. 2002). The eruption started with the for-mation of an eruption column reaching up to 12 km inaltitude (Lacasse et al. 2004). The eruption turned effusivesoon after the first explosion. The main eruptive sourcewas a 4.5-km-long fissure crossing the summit (Fig. 5a,b). During the first hours of the eruption, over 15 firefountains were observed along the fissure. Most of thelavas were issued during the first 2 days (Olafsdottir et al.2002).

The 2000 eruption is the fourth in a row of rela-tively small eruptions occurring approximately once adecade, which has been the recent pattern in the activityof Hekla. The previous eruptions were in 1970 (Einars-son and Bjornsson 1976), 1980–1981 (Gronvold et al.1983) and 1991 (Gu›mundsson et al. 1992; Soosalu andEinarsson 2002; Soosalu et al. 2003). These eruptions arerather similar regarding chemical composition and volumeof the products (Table 1). The 1991 and 2000 eruptions alsohave similar characteristics in terms of precursory activityand time progression of the eruption.

The seismicity and strain observations (Linde et al. 1993)of the 1991 eruption showed a distinct pattern before its on-set. Seismometers recorded a swarm of small earthquakes,with sizes increasing in time, half an hour prior to the erup-tion. At the same time a compressive strain signal was ob-served at BUR, a strain station 15 km from Hekla (Fig. 1),interpreted as a result of a feeding dyke propagating atdepth towards the surface. The experience gained in this

eruption led to a successful short-term prediction for the2000 eruption (Agustsson et al. 2000a, 2000b; Stefanssonet al. 2000). Tiny earthquakes related to the eruption began79 min before the onset (Einarsson 2000). In time they grewin size and frequency—showing a pattern that was unprece-dented at Hekla except prior to eruptions. A warning of animpending eruption was given to the civil defence authori-ties about 50 min before the onset. A strain signal of rapidlyincreasing contraction observed at BUR further confirmedthat magma was approaching the surface. A warning wasissued on the public radio about 15 min before the eruptiononset.

Data analysis

This study is principally based on observations of the SILsystem (named after South Iceland Lowland), the digital3-component short-period seismograph network run by theIcelandic Meteorological Office (Fig. 1) (Bo›varsson et al.1999). The SIL system locates and determines focal param-eters of local seismic events automatically and in near-realtime with a delay of a few minutes. To improve the stationcoverage, two local analogue vertical seismometers alsowere used, HE on the flank of Hekla itself and LJ about35 km east of Hekla (Fig. 1). The events were relocatedwith the location program HYPOINVERSE (Klein 1978),using a crustal model consisting of layers with constant ve-locity gradients (Soosalu and Einarsson 1997). The modelis an average model with time delay corrections used for

23

Tabl

e1

Rec

entH

ekla

erup

tions

Dat

esTe

phra

volu

me

(km

3)

Lav

avo

lum

e(k

m3)

Com

posi

tion

Hei

ghto

fer

uptio

npl

ume

(km

)

Prec

urso

ryse

ism

icity

dete

cted

befo

re(m

in)

Max

quak

em

agni

tude

(ML)

Prec

urso

ryst

rain

chan

gede

tect

edbe

fore

(min

)

Ref

eren

ces

7005

05–7

0070

50.

070.

20B

asal

tican

desi

te16

254

–fio

rari

nsso

nan

dSi

gval

daso

n19

72;

Ein

arss

onan

dB

jorn

sson

1976

8008

17–8

0082

0an

d81

0408

–810

416

0.06

0.12

Bas

altic

ande

site

1523

3ye

sG

ronv

old

etal

.198

3;St

efan

sson

etal

.19

8191

0117

–910

311

0.02

0.15

Bas

altic

ande

site

1230

2.5

30G

u›m

unds

son

etal

.19

92;S

oosa

luan

dE

inar

sson

2002

;So

osal

uet

al.2

003

0002

26–0

0030

80.

010.

286

Bas

altic

ande

site

1279

2.1

34A

gust

sson

etal

.20

00a,

2000

b;E

inar

sson

2000

;St

efan

sson

etal

.20

00;O

lafs

dotti

ret

al.2

002;

Lac

asse

etal

.200

4;H

osku

ldss

onet

al.i

npr

ep

24

the seismograph stations to improve the location accuracy(Soosalu and Einarsson 1997).

A total of 341 earthquakes were recorded and identi-fied in association with the Hekla 2000 eruption, 340 ofwhich occurred during the initial phase in the evening ofFebruary 26, from 17:00 GMT on. The smallest observedevents were less than 0 in local magnitude and the biggestones of magnitude 2.1. The Icelandic local magnitude issimilar to the classical Richter magnitude, i.e. calculatedusing maximum amplitudes and distances between the sta-tions and earthquakes. The 100% detection threshold wasabout ML 0.5 before the appearance of volcanic tremor andabout ML 1.5 once it started. In describing the amount ofseismic activity we take all the events into account, but forcloser interpretation, e.g. for studying the depth distribu-tion, we only use well-located events. The location criteriafor such events are root mean square time error ≤0.2 s,horizontal error ≤ 1.0 km, vertical error ≤ 2.0 km, andlargest gap between observing stations ≤ 180◦. A total of109 events fulfil these criteria, 79 before the onset of theeruption at 18:19 GMT and 30 in the next few hours.

For volcanic tremor studies, continuous seismic recordsat a few stations are available for the first hours of theeruption. From the stations HAU, about 15 km west ofHekla, and SAU, about 35 km to the west (Fig. 1), dataare available from 17 h until midnight of February 26, andfrom 3–7 h on February 27. SAU is an especially use-ful station regarding the onset of the eruption, becausecontinuous data recorded at SAU also exist for the firsthours of the 1991 eruption, and the two cases can thus becompared (see Soosalu and Einarsson (2002) and Soosaluet al. (2003)). There are continuous records available fromstations GYG 40 km northwest of Hekla, MID 40 km tothe south-southwest, SKR 90 km to the northeast, and ASB115 km to the northwest until 21 h on February 26. Forthe station HVO 65 km southeast of Hekla there are con-tinuous data until 18:30 h. In addition, there are samplesof the tremor recorded by the station HAU throughout theeruption. A more qualitative source for tremor studies isthe paper seismograms of the analogue vertical componentstations HE and LJ.

To study the appearance of the low-frequency volcanictremor, 5-min-average amplitudes for frequency bands of0.5–1.0 Hz, 1.0–2.0 Hz and 2.0–4.0 Hz were calculated andtheir behaviour with time at the beginning of the eruptionwas observed. Spectral analysis was made of the tremordata, and continuous spectrograms were plotted for theexisting data for the first hours of the eruption.

Seismic activity between the 1991 and 2000Hekla eruptions

The 1991 Hekla eruption started on January 17 and endedon March 11. The Hekla-Vatnafjoll area was seismicallyrather quiet between the eruptions of 1991 and 2000(Soosalu and Einarsson 2002). The seismicity in this pe-riod is summarized in Table 2 and Fig. 2. Sporadic smallearthquakes occurred, with the exception of an earthquake

19˚ 50'W 19˚ 40'W 19˚ 30'W

63˚ 56'N

63˚ 58'N

64˚ 00'

64˚ 02'

19˚ 50'W 19˚ 40'W 19˚ 30'W

64˚ 00'N

64˚ 02'N

0 5

km

19˚ 50' 19˚ 40' 19˚ 30'

210

ML

Fig. 2 Epicentral map of the Hekla area in March 12, 1991–February25, 2000. Inverted triangles denote high-frequency earthquakes untilthe end of June 1991 and dots high-frequency events occurring later.Stars are low-frequency earthquakes. Well-located events (rms timeerror ≤ 0.2 s, horizontal error ≤ 1.0 km, vertical error ≤ 2.0 km,largest gap between observing stations ≤ 180◦) are drawn with greysymbols and rather well located events (rms ≤ 0.2 s, erh ≤ 2.0 km,erz ≤ 5.0 km, gap ≤ 230◦) with open symbols. These criteria arealso used in the other epicentral maps. The magnitude scale (localmagnitude, ML) is shown in the inset. The eruptive fissures of the1970 and 1980–1981 eruptions are shown with a thinner line and thefissures of the 1991 eruption with a thicker line

swarm in June 1991. The activity in 1991–2000 can bedivided as follows.

1. The unusual swarm of small earthquakes. On June 1,1991 a swarm of earthquakes occurred suddenly, last-ing about 10 h. The SIL network detected about thirtyevents with magnitudes of 0.6–1.7, and the analoguestation HE at Hekla recorded more than seventy addi-tional small events. The events were concentrated in thenorthern part of the volcano, mainly in the uppermost3 km (see the cluster in Fig. 2). We interpreted this ab-normal seismicity to be an expression of a failed attemptof the volcano to resume the eruptive activity (Soosaluand Einarsson 2002).

2. Low-frequency earthquakes. In our earlier study onearthquakes in a non-eruptive period, July 1991–December 1996, we reported Hekla earthquakes witha peculiar appearance (Fig. 8 in Soosalu and Einarsson1997). In the records of every observing station theycontained only low frequencies but had clear and unat-tenuated S-waves, thus pointing to a tectonic, not vol-canic origin. The first such earthquake was observed inApril 1991, in a period when high-frequency events werealso recorded at Hekla. Between September 1991 andJanuary 1998 all the observed events under the summitof Hekla and north of it had a low-frequency character(stars in Fig. 2). The last low-frequency event prior tothe 2000 eruption was recorded in March 1999.

3. High-frequency earthquakes. All the earthquakes thatwere detected during the 1991 eruption were high-frequency events. After the cessation of the erup-tion high-frequency earthquakes continued to occur atHekla until August 1991. During the following years,no such events were detected at Hekla proper until

25

Table 2 Hekla earthquakes between the 1991 and 2000 eruptions

Date Time Coordinates Depth (km) ML Region Remarks◦N ◦W

910411 024253 64 01.00 19 38.84 5.4 0.3 N Hekla lf, P910416 172747 64 01.39 19 37.55 4.1 1.4 N Hekla hf, W910419 110142 64 00.36 19 39.21 0.0 0.40 Hekla hf, P910419 175703 64 01.01 19 42.90 11.7 −0.1 Hekla hf, P910506 024206 64 00.77 19 37.48 6.1 0.5 Hekla hf, P910601 1535 < 3 N Hekla, swarm of

∼ 100hf

−0602 −0114 events, max ML

1.71910603 193141 64 02.49 19 37.15 1.4 1.5 N Hekla hf, W910606 121815 64 00.91 19 36.56 3.2 0.8 Hekla hf, P910606 121831 64 01.06 19 36.09 2.4 0.8 Hekla hf, P910825 204639 63 59.22 19 40.67 4.8 – Hekla hf, P930214 044046 63 58.62 19 47.98 14.5 0.8 W Hekla lf, R931015 115348 64 02.13 19 40.67 1.6 0.8 N of Hekla, near lf, P940328 204604 64 03.31 19 32.93 6.4 0.8 fissure swarm NE

of Heklalf, R

940505 134931 64 02.30 19 37.87 26.3 0.9 N of Hekla, near lf, W940804 105433 63 58.62 19 42.19 13.4 0.3 middle Hekla lf, W940922 125458 63 57.77 19 41.23 12.2 1.0 S middle Hekla hf, W941014 152113 63 58.58 19 38.35 7.6 0.3 Hekla lf, R941113 133516 64 00.68 19 42.66 6.9 0.7 N edge of Hekla lf, P941124 024736 64 00.48 19 48.34 5.3 0.6 W of Hekla lf, P950304 193336 64 02.52 19 45.36 12.9 1.0 N of Hekla hf, R951013 184014 64 00.94 19 42.72 10.7 1.0 N of Hekla, near lf, W960427 232808 63 58.83 19 43.52 12.9 0.7 Hekla lf, R980203 123116 64 00.54 19 36.65 8.0 0.9 NE Hekla hf, W980203 123125 64 01.63 19 38.50 7.7 – N of Hekla, near hf, W980203 123218 64 01.64 19 44.23 6.5 0.6 N of Hekla hf, W980929 074512 63 58.21 19 36.13 9.5 0.6 E Hekla hf, W990327 005149 64 01.29 19 38.89 16.0 1.2 N edge of Hekla lf, W990706 215309 63 59.46 19 42.51 6.0 0.8 middle Hekla hf, W

Remarks: hf high-frequency earthquake, lf low-frequency earthquake, W well-located, R rather well located, P poor location accuracy.W fulfils the criteria: rms time error ≤ 0.2 s, horizontal error ≤ 1.0 km, vertical error ≤ 2.0 km, largest gap between observing stations ≤180◦, and R the criteria rms ≤ 0.2 s, erh ≤ 2.0 km, erz ≤ 5.0 km, gap ≤ 230◦. In the 910601– 0602 swarm there were events with all thethree sorts of location accuracies

February 1998, when three high-frequency events wererecorded. An additional event occurred in July 1999.After that time, no earthquakes were observed at Heklabefore the onset of the 2000 eruption.

Initial earthquake swarm

Hekla is a notorious volcano because of the short warn-ing time before its eruptions. The seismicity related to theeruptions has started gradually and reached the detectionthreshold very shortly before the onset (Table 1), approx-imately 25 min in the eruptions in 1970 (Einarsson andBjornsson 1976) and in 1980 (Gronvold et al. 1983), andhalf an hour in 1991 (Gu›mundsson et al. 1992; Soosaluand Einarsson 2002).

No long-term precursory seismicity was observed priorto the 2000 eruption, either, and the related earthquakesstarted only 79 min before its onset (Fig. 3) (Einarsson2000; Stefansson et al. 2000). The first tiny events wererecorded by the analogue seismograph station HE onthe flank of Hekla (Fig. 1) and the SIL network from17:00 GMT on. A representative set of records of initialearthquakes is shown in Fig. 4a. All the first events werevery small (Fig. 5a) and those that could be located wereshallow, at 0–4 km depth (Fig. 6). After 17:36 the mainactivity jumped to 4–9 km depth. The deepest eventswere located down to 14 km, and only a few eventswere observed to be shallower than 4 km. Towards theonset of the eruption the earthquake sizes generallyincreased from an initial local magnitude of −0.5 up toML 2.1.

26

4

6

8

10

12

14

16

ln o

f int

ensi

ty

17 18 19 20 21 22 23 00

HAU 2000

SAU 2000

SAU-Z 1991

-1

0

1

2

3

ML

mag

nitu

de

26 Feb 2000BUR contr.

starts(17:45)

eruptionstarts

(18:19)

BUR contr.stops (19:20)

-1

0

1

2

3

ML

mag

nitu

de

17 18 19 20 21 22 23 0017 Jan 1991

(16:32)BUR straincontraction starts

eruption starts(17:02) (18:40)

BUR straincontraction stops

Fig. 3 The magnitude ofearthquakes as a function oftime at the onsets of the 2000and 1991 eruptions, MLmagnitude scale is shown to theright. The graphs are adjustedthus that the onset times of theeruptions are aligned. Changesin strain are also marked. Aboveis shown the intensity ofvolcanic tremor at allcomponents of stations HAUand SAU in 2000, with thenatural logarithm scale to theleft. Vertical components areshown with a bold line, east, i.e.“radial” components with a thinline and north, i.e. “transverse”components with a thick greyline. The vertical component ofSAU in 1991 is shown forcomparison (dashed line). Thefrequency band used is0.5–3.0 Hz. Note the similaritiesin both the patterns ofseismicity in the two cases

The initial earthquakes occurred at short intervals withas many as several events per minute. A total of 208 earth-quakes were observed before the start of the eruption at18:19 GMT. At that time low-frequency volcanic tremorappeared on the seismograms (Fig. 4b). A total of 132earthquakes were observed after the onset of the eruption(Fig. 5b). The earthquakes of greatest magnitudes culmi-nated around the onset, with two ML 2.1 events at 18:17and 18:26 (Fig. 3). After that time, the earthquakes hadmagnitudes up to 1.9 with a slightly declining trend. Thedetection threshold of the earthquakes was about ML 1.5because of the high-amplitude tremor masking them. Afterthe start of the eruption, earthquakes were observed at alldepths from the surface down to 14 km, mainly at 2–12 km.

Seismic energy released in the detected initial earth-quakes was small. In total it corresponded to a magnitudeML 3.2 earthquake. The energy released in earthquakes be-fore 17:45 when the contraction strain signal at BUR wasobserved represents a single magnitude 1.8 event, and allthe earthquakes before the onset of the eruptive activityone magnitude 2.6 event. The earthquakes after the onsetcorrespond to an event of magnitude 3.1. The earthquake

swarm started fading soon after the onset of the eruption,and the events became infrequent after 19 h (Fig. 3). To-wards midnight the events became sporadic with the lastevent observed at 23:50. During later phases of the eruptiononly one earthquake of ML 0.8 was detected on March 1.It does not have a well-constrained location, due to largegaps between observing stations. It is not likely that laterearthquakes had occurred unnoticed. After the first hours ofthe eruption, the amplitude of tremor decreased drastically(see intensity graph in Fig. 11) and the earthquake detec-tion threshold likely reached soon the pre-eruption level ofML 0.5.

As in the 1991 eruption (Soosalu and Einarsson 2002), theobserved earthquakes during the 2000 eruption were high-frequency earthquakes, as far as could be discerned in themidst of the volcanic tremor. No low-frequency volcanicearthquakes were detected.

A strainmeter network of six stations operating in southIceland recorded strain changes related to a propagatingdyke in both the 1991 and 2000 eruptions (Linde et al.1993; Agustsson et al. 2000a, 2000b). In both cases thestrainmeter station BUR (Fig. 1) at 15 km distance from

27

-0.50

0.5 17:50-17:52

17:52-17:54

17:54-17:56

17:56-17:58

17:58-18:00 P Sam

plitu

de, x

10-5

m/s

18:00-18:02

18:02-18:04

18:04-18:06

18:06-18:08

18:08-18:10

18:10-18:12

0 20 40 60 80 100 120seconds

18:12-18:14

a

-101 18:20-18:22

18:50-18:52

19:20-19:22

19:50-19:52

20:20-20:02

ampl

itude

, x 1

0-4 m

/s

20:50-20:52

21:20-21:22

21:50-21:52

22:20-22:22

22:50-22:52

23:20-23:22

0 20 40 60 80 100 120seconds

23:50-23:52

b

Fig. 4 a Vertical component seismograms at station HAU in the halfhour before the onset of the 2000 Hekla eruption. At first occasionalsmall earthquakes occur (one event is circled as an example, P andS arrivals are shown). Towards the start of the eruption they growin size, become more frequent and finally merge into a continuous-looking signal. The data are high-pass filtered at 2 Hz. b Vertical

component seismogram samples with 0.5 h intervals at the stationHAU at the onset of the eruption. The eruption started at 18:19 GMT.The low-frequency tremor amplitude increases quickly. It stays highfor 1.5–2 h, and then clearly starts to decline. N.B. The scale in b is18 times smaller than in the seismograms before the onset in a. Nofiltering is done

Hekla showed a start of a contraction signal caused by theintrusion half an hour before magma reached the surface.After the swarm of small earthquakes started at Heklaat 17:00 on February 26, 2000, the strain records wereclosely monitored. At 17:45 BUR recorded the onset ofthe intrusion-related contraction signal while the moredistant stations showed expansion, caused by emptyingof the magma reservoir (Agustsson 2003, personalcommunication). The strain observations of the network

are interpreted to indicate that a conduit was openingto the surface from 17:45 to 18:17. Later the conduitcontinued to expand much slower until 19:20. After19:20 the conduit was fully developed, and expansioncaused by magma flowing from the magma chamberwas observed by all the strain stations, including BUR(http://hraun.vedur.is/ja/englishweb/heklanews.html#strain).

A magnitude 1.2 earthquake, located at 6 km depth co-incided with the onset of the contraction signal at BUR

28

20˚ 00'W 19˚ 50'W 19˚ 40'W 19˚ 30'W

63˚ 56'N

64˚ 00'N

64˚ 04'N0 5 10

km210

ML

a

20˚ 00'W 19˚ 50'W 19˚ 40'W 19˚ 30'W

63˚ 56'N

64˚ 00'N

64˚ 04'N0 5 10

km210

ML

b

Fig. 5 a Earthquakes prior to the onset of the Hekla 2000 eruption,grey dots are well-located events and open circles rather well locatedones. The black dot is the magnitude-2.1 earthquake at 18:17. b Theearthquakes during the first hours of the eruption. The fissure of the2000 eruption is shown with a thick line. The thick bars at about 19◦58′ W show the surface expressions of the easternmost fault of theSouth Iceland seismic zone

at 17:45. The cessation of contraction signal at BUR at19:20 coincided with a small time-gap in seismicity, simi-lar to observations in 1991 (Fig. 3) (Soosalu and Einarsson2002). The strain rate observations at BUR show that thecontraction rate increased rapidly until 18:17 (Agustsson2003, personal communication), 2 min before the appear-ance of volcanic tremor. The time 18:17 was interpreted asthe onset of the eruption (Agustsson et al. 2000a). At thattime one of the largest earthquakes, ML 2.1, occurred at3 km depth. After 18:17 the strain rate started to decrease.

Volcanic tremor

At Hekla, volcanic tremor has been observed to occur onlyduring its eruptions. Tremor starts together with the erup-tive activity, thus its appearance in seismograms is takenas the seismic expression of the onset of the eruption. The

0

2

4

6

8

10

12

14

dept

h (k

m)

210

ML

Feb 26

18:00 19:00 20:00 21:00

Straincontr. starts

Eruptiononset

Straincontraction stops

Fig. 6 Depth distribution of the well-located events in time. Verticalerror bars are shown. Times for the strain changes at the station BURand the onset of the eruption are marked

tremor continues throughout the Hekla eruptions and ceasestogether with the eruptive activity.

On February 26, 2000, the low-frequency volcanic tremormade its first appearance in the seismic records at 18:19GMT. This marks the beginning of the eruption, whichwas verified by an eyewitness account over the radio. Bylistening to the news account and simultaneously viewingthe seismograph at LJ it was observed that the eruption andthe low-frequency tremor began within a few seconds ofeach other. The appearance of the tremor is illustrated inrecordings of the stations HAU and SAU in Figs. 7a–c and8a–c, respectively, where the amplitudes in the frequencybands of 0.5–1.0 Hz and 1.0–2.0 Hz increase rapidly. Amagnitude 2.1 earthquake had occurred two minutes earlier.Around the onset time of the tremor there occurred severalearthquakes per minute, with magnitudes from 1.2 to 1.5.The tremor increased quickly in amplitude and was mostvigorous during the first couple of hours of the eruption.Later it had a continuous generally declining trend towardsthe end of the eruption.

As observed at the onset of the 1991 eruption (Soosaluet al. 2003), the tremor at the beginning of the 2000 eruptionwas distinctly restricted in the frequency band of about0.5–1.5 Hz. The vertical component spectrograms for thestations HAU (Fig. 9a) and SAU (Fig. 9b) are shown asexamples. All the seismic stations available regardless oftheir distance from the eruptive source or azimuth showedconsistently the same frequency range. However, there arealso local features, e.g. at the station SAU the spectrumextends to slightly higher frequencies than at the otherstations. The amplitude of the tremor attenuates rapidlywith distance. Stations at distances over 65 km recordedthe tremor only vaguely or not at all due to the backgroundnoise.

Generally, it is observed that volcanic tremor has apeaked spectrum, with typically one dominant and a few

29

0

5

10

15

ampl

itude

, x 1

0-6m

/s

17 18 19 20 21 22 23 00

HAU - Z

0

5

10

18:10 18:20 18:30

0.5-1 Hz1-2 Hz2-4 Hz

0

5

10

15

ampl

itude

, x 1

0-6m

/s

17 18 19 20 21 22 23 00

HAU - "R"

0

5

10

18:10 18:20 18:30

0

5

10

15

ampl

itude

, x 1

0-6m

/s

17 18 19 20 21 22 23 00

February 16, 2000

HAU - "T"

0

5

10

18:10 18:20 18:30

a

b

c

Fig. 7 Tremor amplitude timeseries with different frequencybands, 0.5–1.0 Hz (solid line),1.0–2.0 Hz (dashed line), and2.0–4.0 Hz (dotted line) forstation HAU: a The vertical,b The approximate radial(E–W), c The approximatelytransverse (N-S) component.The values are 5-min amplitudeaverages. The insets show thebehaviour of the frequencybands around the very onset ofthe eruption, starting at 18:00

subdominant frequency peaks (e.g. Aki et al. 1977; Chouet1992; Seidl et al. 1981; Ferrick et al. 1982; Konstantinouand Schlindwein 2002). The Hekla tremor also had thispattern. Most of the time a single major peak existed withinthe frequency band 0.7–0.9 Hz (Fig. 10). Occasionallythere were two or three approximately equal peaks. Ahandful of subdominant peaks sporadically appearedwithin the 0.5–1.5 Hz band. The frequency of the maxi-mum peak was remarkably constant. Throughout the firsthours, for which we have continuous data, it was mainlywithin this narrow band. Although the general pattern issimilar at several stations, there are some local differencesin the spectra, apparently due to path and site effects. Thestation MID (Fig. 1) had a prominent peak at 0.55–0.6 Hzin the first hours of the eruption, whereas at other stationsthe dominant peak occurred at this frequency mainly inthe very beginning, if at all. In these cases, a peak at 0.7–0.75 Hz often was subdominant. Increasing distance maypartly cause the lower-frequency spectral peaks to be pre-served better than the higher-frequency ones. However, thestation GYG is as far from Hekla as MID (40 km), havingdominant peaks at higher frequencies than MID. Also, at

about 20:00–21:00 the station SKR 90 km from Hekla hasmain peaks at the same or higher frequencies than MID.1

A study of the spectral content of the tremor from thesecond eruption day on was made using 2-min seismogramsamples available for the station HAU, located 15 km awayfrom the volcano. A spectrum was calculated over 1-minsamples every hour until 11 a.m. on March 8. The generalpattern with one dominant peak, and occasional subdom-inant peaks, continues after the first day of the eruption.The peak is also at similar frequencies as before, within theband of 0.6–1.0 Hz, in most of the cases within 0.7–0.9 Hz.No apparent shift of the frequency of the dominant peakin time is evident. The signature of the tremor in the spec-tra becomes more vague towards the end of the eruption.Occasionally the signal is clearer, but in general it starts todisappear into the background noise around March 3.

The Hekla tremor attenuates faster with distance than theearthquakes, and was evident at remote digital seismographstations only at the beginning of the eruption. This points toan origin of the tremor that is shallower than the earthquakesin the first hours of the eruption. A curious detail is thatthe Hekla earthquakes were remarkably easily observable

30

0

1

2

3

4

ampl

itude

, x 1

0-6 m

/s

17 18 19 20 21 22 23 00

SAU - Z

0

1

2

18:10 18:20 18:30

0.5-1 Hz1-2 Hz2-4 Hz

0

1

2

3

4

ampl

itude

, x 1

0-6 m

/s

17 18 19 20 21 22 23 00

SAU - "R"

0

1

2

18:10 18:20 18:30

0

1

2

3

4

ampl

itude

, x 1

0-6 m

/s

17 18 19 20 21 22 23 00

February 16, 2000

SAU - "T"

0

1

2

18:10 18:20 18:30

a

b

c

Fig. 8 Tremor amplitude timeseries with different frequencybands, 0.5–1.0 Hz (solid line),1.0–2.0 Hz (dashed line), and2.0–4.0 Hz (dotted line) forstation SAU: a The vertical,b The “radial” (E–W), c The“transverse” (N–S) component.The values are 5-min amplitudeaverages. The insets show theblowups of the onset of theeruption, starting at 18:00

at a distant station, ASB, 115 km away. Tremor was barelyvisible that far, even initially. The tremor appears to beclosely related to degassing, as it first appears when theconduit is open and is most intense in the beginning whenthe eruption was violent and explosive in nature. When theeruption calmed down to flow of lava, the tremor amplitudealso declined. Occasional tremor bursts had duration of tensof seconds to minutes, similar to observed bursts of gas.Similar observations were also made for the 1991 eruption(Soosalu et al. 2003).

To follow the development of the vigour of the tremorin a simple manner we calculated its intensity using theprocedure of fiorbjarnardottir et al. (1997). Data of all thethree components were used and band-pass filtered between0.5 and 3.0 Hz to eliminate the effect of the low-frequencymicroseism of oceanic origin and the high-frequency earth-quakes. The intensity is calculated as averaged energy over60-s intervals, and its natural logarithm plotted as a func-tion of time (Figs. 3 and 11). This procedure was donefor the two closest stations, HAU and SAU. Because boththese stations are located west of the volcano, no rotation ofcomponents was performed but the east components weretaken to “radial” and the north components as “transverse”.

For the first hours, the intensity is shown for all threecomponents in Fig. 3, together with the temporal evolutionof the earthquake activity. Because continuous observationsexist at the station SAU for the onset of the 1991 eruption,the tremor intensity graph of the 1991 eruption is added forcomparison and adjusted so that the start times of the twoeruptions match. For clarity, only the vertical componentdiagram is shown. The horizontal components followedthe same pattern, but had somewhat higher values. Intensityvalues of the data samples at the stations HAU and SAU areshown in Fig. 11. As was observed with the spectra, aroundMarch 3–5 the tremor amplitude at HAU had declined tothe level of the background noise (in of intensity ∼8), andit is difficult to distinguish tremor from noise.

The analogue station HE on the flank of the volcanowas the last one to record Hekla tremor for certain, untilthe morning of March 8. During the last days tremor wasvisible on the seismograms of HE as occasional burstsrising above the detection threshold. The recordings at thestation HE give a qualitative account of the tremor duringthe eruption. In the beginning, the tremor had such a highamplitude that the records became saturated and the gainwas lowered by 30 dB. After the first phases, the tremor

31

18

19

20

21

22

23

00

01

02

03

04

05

06

07

26-2

7 F

eb 2

000,

tim

e [h

]

0 1 2 3 4

frequency [Hz]

0 1 2 3 4

frequency [Hz]

HAU - Z< 9.104

9.104 -3.105

3.105 -5.105

5.105 -9.105

9.105 -3.106

> 3.106

18

19

20

21

22

23

00

01

02

03

04

05

06

07

26-2

7 F

eb 2

000,

tim

e [h

]

0 1 2 3 4

frequency [Hz]

0 1 2 3 4

frequency [Hz]

SAU - Z< 1.104

1.104 -4.104

4.104 -6.104

6.104 -1.105

1.105 -4.105

> 4.105

a bFig. 9 a Spectrogram of theHAU vertical component data atthe beginning of the eruption,February 26–27. Thespectrogram is constructed ofnon-overlapping amplitudespectra of 1-min-long datasamples. b Spectrogram of theSAU vertical component data atthe beginning of the eruption,February 26–27. The amplitudeunits are arbitrary

0.0

0.5

1.0

Hz

0.0

0.5

1.0

Hz

19 20 21 22 23 00 01 02 03 04 05 06 07

HAU - 15 km to W

0.0

0.5

1.0

Hz

0.0

0.5

1.0

Hz

SAU - 35 km to W

0.0

0.5

1.0

Hz

0.0

0.5

1.0

Hz

GYG - 40 km to NW

0.0

0.5

1.0

Hz

0.0

0.5

1.0

Hz

MID - 40 km to SSW

0.0

0.5

1.0

Hz

0.0

0.5

1.0

Hz

19 20 21 22 23 00 01 02 03 04 05 06 07

26-27 Feb 2000, time [h]

SKR - 90 km to NE

Fig. 10 Location of the dominant peak (occasionally two or threepeaks) of the Hekla tremor at various seismic stations at the onset ofthe eruption; the vertical component and available continuous dataare used. Distance to each station is given. The plots are constructedby stacking five consecutive 1-min spectra

started to decrease and from February 28, about 9 a.m.the gain was increased gradually to keep the tremor abovethe detection threshold. The general trend of the tremorwas declining, but occasionally it increased, as is visiblein the intensity graph in Fig. 11. A characteristic featureof the tremor was that it was fluctuating, i.e. the amplitudeincreased and decreased in a spindle-like manner. Tremorbursts lasted from a few seconds to a few tens of seconds.Figure 4b illustrates this feature, see e.g. the sample at19:20–19:22.

5

6

7

8

9

10

11

12

13

14

15

16

17

ln o

f int

ensi

ty

27 28 29 01 02 03 04 05 06 07 08

Feb - Mar, 2000

HAU - Z

SAU - Z

Fig. 11 Intensity of the tremor at HAU (black), constructed usingsamples available throughout the eruption. SAU data (grey) onlyexist for the first hours of the eruption. The frequency band used is0.5–3.0 Hz

32

-0.5

0.0

0.5

Z, v

ertic

al

-0.5 0.0 0.5E-W, "radial"

18:27

vv

-0.5

0.0

0.5

N-S

, "tr

ansv

erse

"-0.5 0.0 0.5

18:27

v

v

v

v

-0.5 0.0 0.5E-W, "radial"

18:35

v

v

v

-0.5 0.0 0.5

18:35

v

vv

-0.5 0.0 0.5E-W, "radial"

18:41

v

v

v

v

-0.5 0.0 0.5

18:41

v

v

v

v

-0.5

0.0

0.5

Z, v

ertic

al

-0.5 0.0 0.5E-W, "radial"

19:19v

vv v

-0.5

0.0

0.5

N-S

, "tr

ansv

erse

"

-0.5 0.0 0.5

19:19v

v

v

v

-0.5 0.0 0.5E-W, "radial"

19:39

v

v

v

-0.5 0.0 0.5

19:39

vv

v

v

-0.5 0.0 0.5E-W, "radial"

20:15

vv

v

-0.5 0.0 0.5

20:15v

v v

a

b d f

h j l

ig k

c eFig. 12 Six 5-s samples of theparticle motion of the Heklatremor, observed by the stationHAU. The data were band-passfiltered at frequencies0.5–1.0 Hz, the trend and themean were removed and a0.5-s-long cosine taper wasused. For each sample, theradial-vertical plane is shownabove and the radial-transverseplane below. The start times are:a, b 18:27, c, d 18:35, e, f18:41, g, h 19:19, i, j 19:39 andk, l 20:15. Arrows show thedirection of motion

From the afternoon of March 6 the gain at HE was normalmost of the time, i.e. the same as it was around the onsetof the eruption. Had there been earthquakes at Hekla, atthat gain HE could have detected them, down to ML 0. Theburst-like behaviour of the tremor continued, sometimeswith considerable amplitudes, but otherwise the tremor dis-appeared below the detection threshold. Towards the endof the eruption the bursts became sporadic, and their maxi-mum amplitudes became smaller as well. In the early hoursof March 8 there were lengthy periods, tens of minutes orlonger, with no visible tremor on HE records. Two clearbursts of tremor, lasting about a minute, were seen before10:00 GMT. Very small amplitude bursts continued until16:00. In the evening of March 8 and during the first 2–3 hof March 9 tiny signals were recorded by HE, but it is hardto judge if they were eruption tremor or small earthquakes.

To study the particle motion in the Hekla tremor we tookvarious 5-s samples of the data at the station HAU duringthe first hours of the eruption. Because Hekla is located di-rectly east of this station, the east-west component is takento be approximately radial to the source and the north–

south component as approximately transverse to it. A setof samples is shown in Fig. 12. Although the particle mo-tion looks somewhat diffuse in the plots, some consistentfeatures exist in most of them. In the radial-vertical dia-grams, retrograde elliptical movement can often be seen,pointing to Rayleigh waves, and a shallow source of thetremor. In the radial-transverse plots one would expect tosee motion principally in the east-west direction, but in-stead a clear component of NW-SE movement is visiblein most of the plots. The reason for this direction may behorizontal refraction of the surface waves. One explanationis that the mountain edifice has a focusing effect on thesurface waves in the direction of the Hekla ridge.

Seismicity at Hekla after the 2000 eruption

After the 2000 eruption, Hekla has again become seismi-cally quiet. Eighteen events with ML magnitudes between0.3 and 1.6 have been detected since the cessation of vol-canic activity until September 2004 (Table 3 and Fig. 13).

33

Table 3 Hekla earthquakes after the 2000 eruption until September 2004

Date Time Coordinates Depth (km) ML Region Remarks◦N ◦W

000511 015137 63 59.64 19 43.02 2.1 0.6 N-middle Hekla hf, W000528 065714 64 00.96 19 44.36 13.0 0.2 N of Hekla lf, P010709 191634 63 58.14 19 39.25 9.0 1.0 S Hekla hf, W010803 164618 63 59.85 19 44.97 11.2 0.8 N of Hekla lf, W010909 175842 64 00.63 19 29.71 8.3 1.0 NW of Hekla hf, W011121 071943 64 00.32 19 31.41 10.2 1.6 NW of Hekla hf, W011128 155618 64 01.34 19 37.85 18.3 0.8 NE Hekla lf, W020422 124728 63 58.16 19 40.09 12.0 0.8 S Hekla lf, W020505 021836 64 00.44 19 45.36 10.5 0.6 N of Hekla lf, W021205 190020 63 58.45 19 45.33 8.4 1.2 SW-middle Hekla hf, W030119 221746 63 58.87 19 43.42 11.1 0.8 Middle Hekla lf, W030520 030826 63 59.17 19 43.38 10.9 0.3 Middle Hekla lf, W031126 070344 64 01.09 19 29.23 9.5 0.5 NW of Hekla hf, P031228 011252 63 58.56 19 42.24 12.6 0.6 Middle Hekla lf, W031231 221132 64 00.86 19 44.84 12.1 0.3 N of Hekla lf, P040606 024339 63 56.41 19 47.13 12.6 0.8 SW edge of Hekla hf, W040828 183948 64 02.53 19 37.06 1.6 0.9 N of Hekla hf?, P040929 011733 63 59.81 19 42.39 4.9 0.9 N-middle Hekla hf, W

Remarks: hf high-frequency earthquake, lf low-frequency earthquake, W well-located, R rather well located, P poor location accuracy.W fulfils the criteria: rms time error ≤ 0.2 s, horizontal error ≤ 1.0 km, vertical error ≤ 2.0 km, largest gap between observing stations ≤180◦, and R the criteria rms ≤ 0.2 s, erh ≤ 2.0 km, erz ≤ 5.0 km, gap ≤ 230◦. In the 910601– 0602 swarm there were events with all thethree sorts of location accuracies

The events that have occurred in the middle and north partof Hekla before the fall 2004 appear similar to those thatwe observed with the Hekla earthquakes after the 1991eruption (Soosalu and Einarsson 1997). The events appearto be tectonic in the sense that all the seismograph stationsobserved clear P and S arrivals, but the records consist ofonly low frequencies at every station. An example of a typ-ical Hekla low-frequency earthquake with its spectrum isshown in Fig. 14a. To contrast these with high-frequencyevents Fig. 14b illustrates a high-frequency Hekla earth-quake that occurred during the initial swarm of the 2000eruption, at 17:45, before tremor appeared on the seismo-grams. A well-located ML 0.9 high-frequency earthquakeoccurred at 5 km depth slightly north of the main Hekla fis-sure in late September 2004, being the first high-frequencyHekla event since May 2000 (Table 3).

Most of the Hekla earthquakes from the post-eruptiontime period had hypocentral depths of 8–12 km, whichis typical for non-eruptive times (Soosalu and Einarsson1997). One high-frequency event that occurred in earlyMay 2000 was shallow, at 2-km depth. One low-frequencyevent that occurred in the northeast part of Hekla had aconsiderable depth of 18 km. Interestingly, its location isnear that of another “deep” event that occurred in May 1994at the depth of 26 km.

Discussion and conclusions

The seismicity related to the 2000 Hekla eruption was re-markably similar to what was observed during the 1991eruption. It is thus possible to draw some general con-

19˚ 50'W 19˚ 40'W 19˚ 30'W

63˚ 56'N

63˚ 58'N

64˚ 00'N

64˚ 02'N

63˚ 56'N

63˚ 58'N

64˚ 00'N

64˚ 02'N

0 5

km

210

Fig. 13 The earthquakes at Hekla and in its immediate vicinity afterthe 2000 eruption until mid-November 2003. All the plotted eventsare well located. Grey stars stand for low-frequency events and greydots for high-frequency events. The fissures of the 1970, 1980–1981and 1991 eruptions are shown with a thinner line and the fissures ofthe 2000 eruption with a thicker line

clusions on the behaviour of Hekla. Neither eruption hadobservable long-term seismic precursors. The eruption-related earthquakes were detected only shortly before theonset of the eruption, which was also the pattern in 1970and 1980. The earthquakes were first small in size but grewin time, culminating around the onset of the eruptions. Theyoccurred with very short intervals, often less than a minute.After the start of the eruptions, the events continued fora few hours, but during later phases there was very littleearthquake activity. The total seismic energy released in theentire swarm of the initial earthquakes was rather modest

34

-2

0

2

0 2 4 6 8 10 12 14

P Z

-2

0

2

ampl

itude

, x 1

0-6m

/s

0 2 4 6 8 10 12 14

R

-2

0

2

0 2 4 6 8 10 12 14seconds

TS

-5

0

5

0 2 4 6 8 10 12 14

P Z

-5

0

5

ampl

itude

, x 1

0-6m

/s

0 2 4 6 8 10 12 14

R

-5

0

5

0 2 4 6 8 10 12 14seconds

S T

0.0

0.5

1.0

norm

aliz

ed a

mpl

itude

2 4 6 8 10 12 14Hz

Z

0.0

0.5

1.0

norm

aliz

ed a

mpl

itude

2 4 6 8 10 12 14Hz

Z

a

b

Fig. 14 An example ofa a low-frequency Heklaearthquake with its spectrum(March 27, 1999, ML 1.2). b Forcomparison a high-frequencyHekla earthquake is also shown(February 26, 2000, 17:45, ML1.2). Z vertical, R radial, Ttransverse component. The dataare 0.5-Hz high-pass filteredbecause of microseismic noise.The spectra (verticalcomponent) of both events aretaken over the first 10 s. Theplots are clipped in the low endat 0.5 Hz

in both these cases, corresponding to a single magnitude3.4 event in 1991 and a magnitude 3.2 event in 2000. Appar-ently, little seismic energy is associated with the openingof an eruptive conduit at Hekla.

The volcanic tremor during these two eruptions was alsovery similar. In both cases the low-frequency tremor startedsimultaneously with the eruption itself and had a stablefrequency band at about 0.5–1.5 Hz throughout the firsthours. The maximum peaks were around 0.7–0.9 Hz. Thesame pattern appears to continue during the later stages ofthe eruption. The characteristic frequencies of the Heklatremor are near the lower end of the frequency band gener-ally observed at active volcanoes around the world, about0.1–8 Hz (Konstantinou and Schlindwein 2002).

The tremor had a lower intensity at the onset of the 1991eruption than in 2000 (see Fig. 3: the vertical componentof the station SAU), although the initial earthquake activ-ity accompanying the 1991 eruption was larger. In 1991the intensity diminished rather rapidly after the first hour,followed by a kink in the graph with the decline slow-ing. This was not seen in 2000, but the tremor intensityreached its highest values soon after the onset and had amore gradual exponentially declining trend during the firsthours.

The very first earthquakes were shallow, with well-located events during the first 50 min before the observablestrain signal at station BUR at 17:45 mainly located inthe uppermost 3 km. No propagating front of seismicity

35

from depth towards the surface could be detected. Wehave studied the location of the magma chamber of Hekla(Soosalu and Einarsson 2004) and concluded that noseismic indicators of a prominent magma chamber atHekla can be found in the uppermost 14 km.

The initial earthquake swarm is an expression of changesin stress and strain. Curiously, the seismicity started 50 minbefore the strain signal was observed. The first earthquakeshad magnitudes from –0.4 to 1.2, with total released seismicenergy corresponding to one ML 1.8 event. Possibly theinitial deformation was too small or diffuse to be observedwith the strainmeter network.

Some 10 min before the strain signal was seen earth-quakes at greater depths started to occur (Fig. 6). The firstearthquake with a hypocentre at 6 km depth occurred at17:36 (ML 0.7), and after the start of the strain signal at17:45 earthquakes down to 14 km were observed. Earth-quakes thus started to occur at more variable depths, andactivity was no longer concentrated close to the surface.Both in 1991 and 2000 there was a pause in earthquakeactivity at about the time when contraction ceased at sta-tion BUR (Fig. 3), showing a temporary break in the stressrelease.

In the 1991 eruption we examined attenuation differ-ences between the tremor and the earthquakes (Soosaluand Einarsson 2002; Soosalu et al. 2003). We concludedthat the tremor must have a shallow origin, and the earth-quakes must originate deeper, because of the large attenu-ation of tremor with distance compared to the earthquakes.The tremor waves travel in loose and faulted surface layers,and the earthquake waves in deeper solid layers with lowerattenuation. This feature is also seen in the 2000 eruption.The tremor falls almost to the same amplitude as the back-ground noise during the first hours at the distant stations,located 90–115 km from the volcano.

Another factor favouring a shallow origin of the Heklatremor is that degassing apparently plays a major role. Bothin 1991 and in 2000 the tremor was initiated when the con-duit to the surface was first opened. Also, no tremor hasbeen observed at Hekla except during eruptions. Particlemotion observations point to a large amount of surfacewaves in the tremor signals. On paper seismograms of ana-logue station HE on the flank of the volcano occasionalbursts of tremor with a typical duration from some tens ofseconds to a minute or two can be seen. This behaviouris more pronounced during the latter days of the eruptionwhen the general amplitude of the tremor was already low,sometimes barely detectable. During a field trip to Heklaon March 1, the fourth day of the eruption, a blowing soundfrom the volcano was audible at times. It resembled exactlythe sound produced when a balloon is inflated with occa-sional gusts of hot gas to keep it aloft, i.e. is a sound of de-gassing. In comparison, acoustic observations of degassingcorrelating with tremor observations have been made alsoat other volcanoes, e.g. Arenal (Benoit and McNutt 1997),Karymsky and Sangay (Johnson and Lees 2000).

Some models for tremor, e.g. Seidl et al. (1981) andFerrick et al. (1982), suggest that the frequency content ofthe tremor is dependent on the size of the magma conduit. In

the case of Hekla the characteristic frequencies at the onsetof both the 1991 and the 2000 eruptions have been low,around 1 Hz. If there is such a connection, this may meanthat the magma channel is very large, i.e. the conduit wouldextend to a considerable depth and thus the magma chamberwould be at a deep level. Although the degassing-relatedorigin of the tremor is shallow, the resulting vibration canoccur in a long channel and produce the characteristic lowfrequencies. Lack of evidence for a considerable Heklamagma chamber at the uppermost 14 km (Soosalu andEinarsson 2004) may give an estimate for the length of theeruptive conduit of Hekla.

Both in the 1991 (see Soosalu et al. 2003, Fig. 10) and2000 eruptions (Fig. 11) the semicontinuous plots of tremorintensity at the station HAU show a similar pattern. In thevery beginning the tremor is vigorous but after some hoursdeclines and continues to decrease throughout the eruptionwith some fluctuation. The pattern is the same for the twoeruptions, although the 1991 eruption was longer in dura-tion (52 days) than the 2000 eruption (12 days). Similarly,the amplitude of the tremor in the paper seismograms ofthe analogue station HE showed a general decreasing trendin both cases. A difference is seen in the last days, withtremor in 1991 appearing more continuous than in 2000.In 2000 it was more typical that the signal was below thedetection threshold between tremor bursts.

Low-frequency volcanic earthquakes, common at manyvolcanoes in the world (e.g. Chouet 1996; Kumagai andChouet 1999), have almost never been observed at Hekla.The only definitive exception so far has been in the be-ginning of June 1991, when a swarm of more than onehundred small earthquakes occurred at Hekla (Soosaluand Einarsson 2002). We interpreted this swarm to be anexpression of a failed attempt to resume the 1991 erup-tion. The swarm consisted mainly of high-frequency earth-quakes. During the swarm, HE, the analogue station onthe flank of the volcano, recorded three volcanic-lookingevents, which were similar to hybrid events (Chouet 1996),with a high-frequency onset followed by a low-frequencywavetrain.

The low-frequency nature of the inter-eruption Heklaearthquakes was first detected after the 1991 eruption; be-fore then there is no quantitative knowledge on the exact ap-pearance or spectral content of these inter-eruption events.It is curious that the high-frequency earthquakes at Heklaproper are observed almost exclusively during its erup-tions. Both types of events in the Hekla case are tectonic inorigin, as argued above, representing brittle failure. High-frequency earthquakes are produced during times whenstrain is high at Hekla, i.e., during eruptions and duringoccurrences such as the earthquake swarm in the beginningof June 1991, a failed attempt to resume the eruption. Weconclude that earthquakes in the inter-eruption times havea low-frequency nature because of a low stress drop.

Hekla is notorious for giving no long-term warning ofits eruptions, but it is possible that the appearance ofthe rare inter-eruption earthquakes under the Hekla ed-ifice and slightly north of it can help in long-term an-ticipation of eruptions by giving hints of strain build-up.

36

After the summer of 1991, Hekla was seismically quiet un-til February 1993, when the first low-frequency event wasdetected. From then the sporadic Hekla earthquakes allwere of low frequency, until three successive small high-frequency Hekla events were detected in February 1998.The last low-frequency earthquake before the 2000 erup-tion was observed in March 1999. On July 6, 1999, a clearlyhigh-frequency earthquake occurred right in the middle ofHekla, at 6 km depth (Table 2). No further seismicity wasrecorded in Hekla before February 2000. The reappearanceof high-frequency events, although few in number, mayshow that stress is starting to build up at Hekla and a neweruption is in preparation. The signal is vague, however,due to the small number of the events. However, monitor-ing the frequency content of the earthquakes in the areawhere low-frequency inter-eruption events occur may beuseful. Following the 2000 eruption the Hekla events haveagain had a low-frequency character. A change in this pat-tern was observed in late September 2004 when a clearlyhigh-frequency Hekla earthquake was detected at 5 kmdepth, close to its main fissure. The future will show if it isa sign that strain has started again to accumulate beneaththe volcano and if a new eruption is under preparation.

Acknowledgements The Icelandic Meteorological Office providedthe digital SIL data. The National Power Company of Iceland fundsthe analogue seismograph network. All the figures were made usingthe GMT public domain software (Wessel and Smith 1998). Con-structive criticism of Jeff Johnson, Stephen McNutt and the editorJohn Stix improved the manuscript.

References

Agustsson K, Linde AT, Stefansson R, Sacks IS (2000a) Boreholestrain observations for the February 2000 eruption of Hekla,south Iceland (Abstract). AGU 2000 Spring Meeting, p 442

Agustsson K, Stefansson R, Linde AT, Einarsson P, Sacks IS,Gu›mundsson GB, fiorbjarnardottir B (2000b) Successfulprediction and warning of the 2000 eruption of Hekla basedon seismicity and strain changes (Abstract). AGU 2000 Fallmeeting, p 1337

Aki K, Fehler M, Das S (1977) Source mechanism of volcanictremor: fluid-driven crack models and their application to the1963 Kilauea eruption. J Volcanol Geotherm Res 2:259–287

Benoit JP, McNutt SR (1997) New constraints on source processes ofvolcanic tremor at Arenal volcano, Costa Rica, using broadbandseismic data. Geophys Res Lett 24:449–452

Bo›varsson R, Rognvaldsson STh, Slunga R, Kjartansson E (1999)The SIL data acquisition system—at present and beyond year2000. Phys Earth Planet Int 113:89–101

Chouet B (1992) A seismic model for the source of long-periodevents and harmonic tremor. In: Gasparini P, Scarpa R, Aki K(eds) Volcanic seismology. Springer, Berlin Heidelberg NewYork, pp 133–156

Chouet BA (1996) Long-period volcano seismicity: its source anduse in eruption forecasting. Nature 380:309–316

Einarsson P (1991) Earthquakes and present-day tectonism inIceland. Tectonophysics 189:261–279

Einarsson P (2000) The seismograph station on Litla-Hekla andthe 2000 eruption (abstract, In Icelandic). Geosci Soc IcelandSpring Meeting 2000, p 41

Einarsson P, Bjornsson S (1976) Seismic activity associated withthe 1970 eruption of volcano Hekla in Iceland. Jokull 26:8–19

Einarsson P, Sæmundsson K (1987) Earthquake epicenters 1982–1985 and volcanic systems in Iceland. In: Sigfusson fiI(ed) I hlutarins e›li, Festschrift for fiorbjorn Sigurgeirsson.Menningarsjour, Reykjavık (map)

Ferrick MG, Qamar A, St. Lawrence WF (1982) Source mechanismof volcanic tremor. J Geophys Res 87:8675–8683

Gronvold K, Larsen G, Einarsson P, fiorarinsson S, SæmundssonK (1983) The Hekla eruption 1980–1981. Bull Volcanol46:349–363

Gu›mundsson A, Oskarsson N, Gronvold K, Sæmundsson K,Sigur›sson O, Stefansson R, Gıslason SR, Einarsson P,Brandsdottir B, Larsen G, Johannesson H, fior›arson fi (1992)The 1991 eruption of Hekla, Iceland. Bull Volcanol 54:238–246

Hoskuldsson A, Oskarsson N, Vogfjor› K, Gronvold K, Pedersen R(in prep) The millennium eruption of Hekla in February 2000

Jakobsson SP (1979) Petrology of recent basalts of the EasternVolcanic Zone, Iceland. Acta Naturalia Islandica 26:1–103

Johnson JB, Lees JM (2000) Plugs and chugs—seismic and acousticobservations of degassing explosions at Karymsky, Russia andSangay, Ecuador. J Volcanol Geotherm Res 101:67–82

Klein FW (1978) Hypocenter location program HYPOINVERSE.US Geol Surv Open-file Rep 78–694, 113 pp

Konstantinou KI, Schlindwein V (2002) Nature, wavefield propertiesand source mechanism of volcanic tremor: a review. J VolcanolGeotherm Res 119:161–187

Kumagai H, Chouet B (1999) The complex frequencies of long-period seismic events as probes of fluid composition beneathvolcanoes. Geophys J Int 138:F7–F12

Lacasse C, Karlsdottir S, Larsen G, Soosalu H, Rose WI, Ernst GGJ(2004) Weather radar observations of the Hekla 2000 eruptioncloud, Iceland. Bull Volcanol 66:457–473

Linde AT, Agustsson K, Sacks IS, Stefansson R (1993) Mechanismof the 1991 eruption of Hekla from continuous borehole strainmonitoring. Nature 365:737–740

Olafsdottir R, Hoskuldsson A, Gronvold K (2002) The evolu-tion of the lava flow from Hekla eruption 2000. The 25thNordic Geological Winter Meeting, Reykjavık, 6–9 Jan 2002,p. 149

Seidl D, Schick R, Riuscetti M (1981) Volcanic tremors at Etna: amodel for hydraulic origin. Bull Volcanol 44:43–56

Soosalu H, Einarsson P (1997) Seismicity around the Hekla andTorfajokull volcanoes, Iceland, during a volcanically quietperiod, 1991–1995. Bull Volcanol 59:36–48

Soosalu H, Einarsson P (2002) Earthquake activity related to the1991 eruption of the Hekla volcano, Iceland. Bull Volcanol63:536–544

Soosalu H, Einarsson P (2004) Seismic constraints on magmachambers at Hekla and Torfajokull volcanoes, Iceland. BullVolcanol 66:276–286

Soosalu H, Einarsson P (submitted) Seismic character of the Heklavolcano, Iceland. Submitted to Jokull

Soosalu H, Einarsson P, Jakobsdottir S (2003) Volcanic tremorrelated to the 1991 eruption of Hekla volcano, Iceland. BullVolcanol 65:562–577

Stefansson R, Sacks IS, Linde AT (1981) The Hekla eruption of1980 – the mechanism of a ridge volcano. Carnegie Inst Yearb80:511–514

Stefansson R, Agustsson K, Gu›mundsson GB, fiorbjarnardottir BS,Einarsson P (2000) A successful prediction and warning of aneruption in the Hekla volcano, Iceland (Abstract). AGU 2000Spring Meeting, p 442

fiorarinsson S, Sigvaldason GE (1972) The Hekla eruption of 1970.Bull Volcanol 36:269–288

fiorbjarnardottir BS, Bjarnason Ifi, Einarsson P (1997) Seismictremor in the Vatnajokull Region in 1995–1996. ScienceInstitute, University of Iceland, Report RH-03-97, 37 pp

Wessel P, Smith WHF (1998) New, improved version of GenericMapping Tools released. Eos 79:579