Volcano geodesy and magma dynamics in Iceland - Acri-ST

Journal of Volcanology and Geothermal Research 150 (2006) 14–34www.elsevier.com/locate/jvolgeoresVolcano geodesy and magma dynamics in IcelandErik Sturkell a, *,Páll Einarsson b , Freysteinn Sigmundsson c , Halldór Geirsson a ,Halldór Ólafsson c , Rikke Pedersen c , Elske de Zeeuw-van Dalfsen d , Alan T. Linde e ,Selwyn I. Sacks e , Ragnar Stefánsson aa Icelandic Meteorological Office, Reykjavík, Bústaðavegur 9, 150 Reykjavík, Icelandb Institute of Earth Sciences, University of Iceland, Sturlugata 7, 101 Reykjavík, Icelandc Nordic Volcanological Center, Institute of Earth Sciences, Sturlugata 7, 101 Reykjavík, Icelandd Volcano Dynamics Group, Department of Earth Sciences, The Open University, Walton Hall, Milton Keynes, MK7 6AA, UKe Department of Terrestrial Magnetism, Carnegie Institution of Washington, 5241 Broad Branch Road, N.W. Washington, DC 20015, USAReceived 24 May 2004; received in revised form 19 December 2004Available online 1 September 2005AbstractHere we review the achievements of volcano geodesy in Iceland during the last 15 years. Extensive measurements of crustaldeformation have been conducted using a variety of geodetic techniques, including leveling, electronic distance measurements,campaign and continuous Global Positioning System (GPS) geodesy, and interferometric analysis of synthetic aperture radarimages (InSAR). Results from these measurements provide a comprehensive view of the behavior of Icelandic volcanoes.Between inflation, intrusion, and eruption episodes, volcanoes are likely to deflate or show no sign of seismic activity.Subsidence rates are often in the range of a few millimeters to a few centimeters a year, reducing progressively with time sincethe last eruption or intrusion at the volcano. Subsidence can be caused by cooling and contraction of magma, outflow of magma,it can be related to plate spreading. Volcano subsidence or lack of deformation is often interrupted by episodic magma flowtowards near-surface locations. Such magma recharge has been observed geodetically at Hengill, Hekla, Eyjafjallajökull, Katla,Grímsvötn, and Krafla volcanoes, with inflow inferred to last from a few months up to two decades. In the last 15 years, fivevolcanic eruptions, three intrusive events and two NM6 earthquakes have occurred. In recent years, the Grímsvötn and Katlavolcanoes have exhibited continuous inflation of a few centimeters per year, which at Grímsvötn culminated in an eruption on 1November 2004. Hekla and Torfajökull volcanoes have inflated at rates an order-of-magnitude less. Subsidence is occurringpresently at the Askja and Krafla volcanoes. Within the period of geodetic measurement, signals consistent with no deformationare typical for most of the 35 active volcanoes in Iceland.D 2005 Elsevier B.V. All rights reserved.Keywords: geodesy; volcanoes; Iceland; eruptions* Corresponding author. Tel.: +354 522 6178; fax: +354 522 6000.E-mail address: erik@vedur.is (E. Sturkell).0377-0273/$ - see front matter D 2005 Elsevier B.V. All rights reserved.doi:10.1016/j.jvolgeores.2005.07.010

16E. Sturkell et al. / Journal of Volcanology and Geothermal Research 150 (2006) 14–3424ºW 22ºW 20ºW 18ºW16ºW 14ºW66ºNKrafla65ºN9.7 mm/yrAskjaNorth AmericanPlate64ºN100 kmHHeklaTEGjálpKatlaBGVatnajökullBEGHT9.7 mm/yrEurasian PlateBár arbungaEyjafjallajökullGrímsvötnHengill/HrómundartindurTorfajökullFig. 2. The different volcanoes in Iceland addressed in the text are shown and all of the active volcanic system are outlined in black.jökull, Katla and Hrómundartindur have shown measurabledeformation interpreted as the result of magmamovements at depth. Eysteinn Tryggvason initiatedseveral important time series in the mid-late sixtiesthat are still continued today. The aim of this paper isto provide a summary of recent deformation result.We will address the volcanic systems and eruptionsthat are currently studied by geodetic measurementsof crustal deformation, i.e., Hengill/ Hrómundartindur,Hekla, Eyjafjallajökull, Katla, Torfajökull, Grímsvötn,Askja, Krafla, and Gjálp.2. Geodetic techniques2.1. Early techniquesThe first geodetic project specifically aimed at monitoringof a volcano in Iceland was the establishment ofa leveling line at Askja in 1966 (Tryggvason, 1989).This technique was subsequently applied to severalvolcanoes, including Surtsey, Hekla, Katla, Hengill,and Krafla. A related technique, optical tilt measurements(or bdry tiltQ), has also been widely applied atsome of these volcanoes. Eysteinn Tryggvasoninstalled benchmarks around several lakes in Icelandand utilized lake surfaces as tiltmeters (Tryggvason,1987). By the time of the Hekla 1970 and 1980–81eruptions, electronic distance measuring (EDM) instrumentswere available (e.g., Decker et al., 1971;Kjartansson and Grönvold, 1983). EDM was alsoused extensively during the volcano–tectonic eventsat Krafla in 1975–1984. Electronic tiltmeters weredesigned and applied for continuous monitoring oftilt changes (e.g., Tryggvason, 1986). Extensometerswere installed across fractures to monitor changes intheir width (Hauksson, 1983). Additionally, gravitymeasurements were applied to monitor the extensivevertical movements at Krafla (Johnsen et al., 1980).A network of volumetric strainmeters (Sacks et al.,1971) was installed in the South Iceland SeismicZone in the late seventies for continuous monitoringof strain loading of the seismic zone. These instrumentshave shown strain changes associated with therecent Hekla eruptions and have proven to be veryimportant in the monitoring of Hekla (e.g., Linde etal., 1993).Advances in geodetic methods in the last decadehave brought new techniques that turn out to be veryvaluable in volcano monitoring in Iceland. Theseinclude GPS-geodesy (both campaign and continuousGPS), InSAR, and micro-gravity measurements.Older techniques, such as optical tilt and leveling,are also still widely applied.

E. Sturkell et al. / Journal of Volcanology and Geothermal Research 150 (2006) 14–34 172.2. GPS (campaign)The first GPS-campaign in Iceland was performedin 1986 when a regional network covering most ofIceland was established. The purpose of the networkwas mainly to monitor plate spreading (Foulger et al.,1987). The first GPS survey of an individual volcanoin Iceland was conducted in 1991 around Hekla (Sigmundssonet al., 1992). The GPS network has sincebeen expanded to cover most volcanoes showingsigns of activity.has been gaining popularity since the early 1990s(e.g., Massonnet and Feigl, 1998; Dzurisin, 2003).Interferometric processing of two SAR imagesacquired from about the same orbital position but atseparate times may reveal a phase interference patterncaused by a change in distance from ground to satellitewithin this time span (Fig. 3). A fringe patternemerges with each fringe corresponding to a phaseA2.3. CGPS (Continuous GPS)Crustal deformation in near real-time is currentlyrecorded at 17 CGPS sites in order to monitor platespreading, volcanic deformation and co-seismic displacements.The stations are concentrated around theplate boundary (Fig. 1). The CGPS data are automaticallydownloaded and processed on a daily basisusing predicted satellite orbits and reprocessed withfinal orbits when they become available. The resultsare available on the IMO website for public viewing(http://www.vedur.is/ja). The time series from theCGPS stations are mostly dominated by plate spreading,but deviations are observed at stations close toindividual volcanoes (Geirsson, 2003). The CGPSstations give excellent time resolution of volcanicactivity, but due to high installation and operatingcosts their spatial coverage is necessarily much sparserthan that provided by campaign GPS.5 kmB2.4. InSARApplication of interferometric synthetic apertureradar images (InSAR) in crustal deformation studiesCFig. 3. A) InSAR image showing surface uplift at the Eyjafjallajökullvolcano modified from Pedersen and Sigmundsson (2004). Theinterferogram covers the time period 6 Sep. 1992 to 15 Aug. 1995and shows deformational fringes due to an intrusion event in 1994.One full color fringe corresponds to 2.8 cm of range change.Location of the study area is shown in Fig. 1. The outline of theEyjafjallajökull ice cap is given in black. The interferogram isoverlain on a SAR amplitude image. B) The fringe pattern that ispredicted by a sill with variable opening up to 0.36 m. A green stardenotes the alternative optimal Mogi source. Grey circles shows thebest located earthquakes from the swarm in 1994. C) Residualinterferogram, the difference between (A) and (B).

18E. Sturkell et al. / Journal of Volcanology and Geothermal Research 150 (2006) 14–34difference of 2k which for the ERS satellites is 2.8cm. In Iceland, utilization of the technique has yieldedgood results (e.g., Vadon and Sigmundsson, 1997;Pedersen et al., 2003; de Zeeuw-van Dalfsen et al.,2004a). Combining three component GPS data withthe spatial coverage of InSAR data is currently themost powerful tool in volcano geodesy.2.5. Micro-gravityMicro-gravity techniques can be used to identifysub-surface mass redistributions related to volcanicactivity (Rymer, 1996). To be able to interpret theresults unambiguously, it is essential that heightchanges are known at least with centimeter precision.The combination of micro-gravity with geodetic dataforms a powerful tool to get better insight into volcanicsystems, rather than with any of the two techniquesalone. Repeated micro-gravity measurementshave been conducted within the Askja volcano since1988 (Rymer and Tryggvason, 1993; de Zeeuw-vanDalfsen et al., 2004b) and at Krafla volcano since1975 (Johnsen et al., 1980; Rymer et al., 1998). Theresulting data are among the most extensive microgravitydata sets in the world.3. Recent magmatic activity and geodeticobservationsSince the early nineties, installation of an extensiveGPS network and introduction of InSAR technology,have improved the ability to measure magma movements.More data has resulted in a more completepicture of volcano deformation in the active volcaniczones. Three volcanoes in Iceland dominate the eruptionstatistics and are the most active: Grímsvötn,Hekla and Katla (Fig. 2). Since 1990 there havebeen measurable movements at nine volcanoes, 5confirmed eruptions, 3 inflation episodes, and severalevents suspected to be small subglacial eruptions.Three volcanoes are currently inflating, two of themat a rate of several centimeters a year.3.1. KraflaThe episode of magmatic and rifting activity withinthe Krafla volcanic system from 1974 to 1989 offereda multitude of opportunities to study crustal deformationassociated with magma movements. Located inthe Northern Volcanic Zone, the Krafla system consistsof a central volcano with a caldera and a transectingfissure swarm. High-temperature geothermalsystems are located within the Krafla caldera, providingenergy for the Krafla power plant. The 1974–1989rifting activity affected a 100 km long section of theplate boundary (see e.g., Björnsson et al., 1977;Einarsson, 1991a; Tryggvason, 1995). During mostof this time magma inflow into a shallow-levelmagma chamber beneath the caldera caused inflationof the caldera region. This inflation was interrupted bysudden deflation events when the walls of the chamberwere breached and dykes were injected into theadjacent fissure swarm (Tryggvason, 1980; Einarssonand Brandsdóttir, 1980; Brandsdóttir and Einarsson,1979). Large scale rifting occurred within the fissureswarm during these events. In nine out of twenty casesthe dyke reached the surface and produced basalticfissure eruptions. Extensive geodetic measurementshave been conducted at Krafla, utilizing most availablemethods.The Krafla power plant was under constructionwhen the activity started in 1974. The foundation ofthe new power station building had been surveyed byprecise leveling in order to monitor its settling duringconstruction. The Krafla station turned out to be ideallylocated for monitoring of tilt variations associatedwith the Krafla magma chamber. A water tube tiltmeterwas installed in the building in August 1976 and tworeadings taken per day during the next 11 years. Thisinstrument was complemented by an electronic pendulumtiltmeter in August 1977 (Tryggvason, 1995).The tilt record in Fig. 4 is a composite of data obtainedby these three methods, i.e., leveling the foundation,water-tube tiltmeter and pendulum tiltmeter. The tiltparameter can be used as a proxy for the pressure in amagma chamber throughout a whole eruptive andintrusive sequence. Leveling and tilt data from Krafla,both during inflation and deflation periods, have beenused by several authors to derive the location anddepth of a magma chamber (e.g., Björnsson et al.,1979; Tryggvason, 1980; Ewart and Voight, 1991).The principal conclusions of these studies are thatthe pressure source is located near the centre of thecaldera at slightly less than 3 km. Seismic studiesreveal a body of S-wave attenuation and low P-wave

E. Sturkell et al. / Journal of Volcanology and Geothermal Research 150 (2006) 14–34 19µrad1000900 20 Dec. 758007006005004003002001000 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 198926 Sep. 7631 Oct. 7620 Jan. 7727 Apr. 778 Sep. 777 Jan. 7810 Jul. 7810 Nov. 7813 May 7916 Mar. 8010 July 8018 Oct. 8030 Jan. 8118 Nov. 81Fig. 4. Record of the one component water-tube tiltmeter at the Krafla power plant from August 1976 to January 1990. Data before August 1976originate from leveling measurements of the power plant foundation. The water-tube tiltmeter was installed along the longer wall of the powerplant building that is orientated N138 E. The figure is modified after Tryggvason (1995). The stars mark eruptions.4 Sep. 84velocity within the Krafla caldera (Einarsson, 1978;Brandsdóttir et al., 1997). Several lines of evidencesuggest a deeper magma chamber at Krafla in additionto the one at 3 km depth. The continuous flow ofmagma towards the shallow chamber for over a decadeduring the Krafla fires suggest magma was transportedfrom a localized reservoir or magma storage area at adeeper level in the crust or mantle. The location of it isuncertain, but geodetic data shows that deformation onthe surface cannot fully be explained by the 3 km deepchamber. The deformation field is wider thanexpected. Tryggvason (1986) interpreted recordsfrom tiltmeters during the September 1984 eruptionin terms of three stacked magma chambers, but locationof the two deeper ones was not well resolved. Therole of two deeper sources, rather than one, in the 1984eruption, is also not certain. Árnadóttir et al. (1998)interpreted EDM and leveling data from the 1984eruption and confirmed that magma was drainedfrom a deeper source in addition to the 3 km deepsource, to form a 1 m wide, 9 km long dyke, extendingfrom the surface to about 7 km depth, and lava flowson the surface. The location of the deeper source wasconstrained to be deeper than 5 km, but narrow apertureof the geodetic network did not allow furtherresolution of its depth. The deformation associatedwith dyke intrusion events during the Krafla fires isfor most events not well constrained. Leveling profilescrossing the fissure swarm in two places (Sigurdsson,1980; Kanngiesser, 1983) show how the flanks of thefissure swarm were uplifted while a central blockabove the dikes was downfaulted.The caldera region inflated after the last eruption(1984) until the pre-eruption level was approached.Then the inflation became intermittent. The last inflationperiod occurred in the summer of 1989. After thatthe caldera region has been slowly subsiding, a fewcentimeters per year, as determined by InSAR (Sigmundssonet al., 1997a) and a combined geodetic andmicro-gravity study (Rymer et al., 1998). In the postriftingperiod, GPS revealed higher than averagespreading rates. The post-rifting deformation is fitby an elastic upper crust overlying a visco-elasticlower crust and even weaker visco-elastic upper mantle(e.g., Foulger et al., 1992; Pollitz and Sacks, 1996).A recent InSAR study (de Zeeuw-van Dalfsen etal., 2004a) has revealed uplift over an about 50-kmwidearea at the Krafla volcanic system (8 cm from1993 to 1999), in addition to deformation caused bysubsidence above the shallow magma chamber, andeffects of plate spreading and post-rifting adjustment(Fig. 5). The favored explanation of de Zeeuw-vanDalfsen et al. (2004a) for the widespread uplift is thatit results from magma accumulation at the crust–mantleboundary at 20 km depth, rather than being causedby post-rifting adjustment.3.2. AskjaThe Askja volcano in the Northern Volcanic Zone(Fig. 1) has three overlapping and nested calderas.The main caldera, formed in the early Holocene, is 8km in diameter. An older caldera is discernable but ithas been filled by later lava flows. The most recentcaldera, Öskjuvatn, is 4.5 km in diameter and is filledby a lake. It was formed after the Plinian eruption of1875. The most recent eruption of Askja took place in1961 when a short E–W fissure opened up near theeastern caldera fault of the main caldera. A shortleveling line was installed in the main caldera in

20E. Sturkell et al. / Journal of Volcanology and Geothermal Research 150 (2006) 14–34AH66NM1GM2KL10 km16WB 93-99curve with a decay constant of 39 years (Sturkell etal., submitted for publication). The geodetic data havebeen modeled and interpreted in terms of a singleMogi-type pressure source (Mogi, 1958) locatedclose to the centre of the main Askja caldera (Tryggvason,1989; Rymer and Tryggvason, 1993; Sturkelland Sigmundsson, 2000). All these authors placed thepoint source at 1.5 to 3.5 km depth. This modelA4291 km66NA406A404DYNG10 kmFig. 5. A) A InSAR amplitude image which includes the outline ofthe Krafla fissure swarm (dotted line) the Krafla central volcano(solid white line), the caldera (white dashed line) and the location oftwo model sources, a shallow magma chamber (M2) and about 20km deep source (M1). The location of the area is given in Fig. 1. B)An interferogram spanning the period from 1993 to 1999 showingseveral features including subsidence above a shallow magmachamber (M2), inflation due to suggested deep magma accumulation(M1) and a N–S linear feature relating to plate spreading.Modified from de Zeeuw-van Dalfsen et al. (2004a).65º10'N65º05'N í1966. The line was extended to 1.7 km length (Fig. 6)in 1968 and measured annually until 1972. Annualmeasurements were resumed in 1983 and the line wasalso extended. Because not all benchmarks are foundevery year due to snow cover, and because the entireline lies within the deformation field of the volcano,benchmark 404 is arbitrarily set to zero. This point hasin recent years been measured annually by GPS.To visualize the changes along this line we haveplotted the height difference between the end points,A429 and A406, as a function of time (Fig. 7). StationA406 is 1 km closer to the caldera centre than A429.The data show that the volcano was inflating during1970–1972. This had turned to deflation when measurementswere resumed in 1983, a trend that hascontinued since. The deflation follows an exponential65º00'Nkm0 5 10Öskjuvatn17º 00’W 16º 50’W 16º 40’W5 cmFig. 6. Horizontal GPS-displacements 1993–1998 in the Askjavolcano. The model with the two Mogi sources gave an estimateof the displacement of the reference station DYNG (white arrow).The horizontal displacements are adjusted for the displacement ofthe reference station. (Modified from Sturkell et al., submitted forpublication). Inset shows the location of leveling benchmarksdescribed in the text.

E. Sturkell et al. / Journal of Volcanology and Geothermal Research 150 (2006) 14–34 21Cumulative vertical displacements [m]0.100.050.00-0.05-0.10-0.15-0.20(-(year - 1983) / 39.2)y = -49.8 + 45.7e1970 1975 1980 1985 1990 1995 2000YearFig. 7. Time series showing cumulative vertical displacements atAskja between benchmark A406 and A429 since 1968 (modifiedfrom Sturkell et al., submitted for publication). The time seriesprovides a measure of height change at the caldera centre, assuming95% of the deformation (height differences between A406 andA429) is due to the shallow Mogi source (Sturkell et al., submittedfor publication). The decay starting in 1983 is exponential. Tracingthe subsidence curve back and extrapolating the inflation trend, thechange from inflation to deflation occurred in the middle of 1973.The Krafla rifting episode started in December 1975.accounts for most of the observed displacements inthe main caldera and its immediate vicinity. At agreater distance, however, displacements observedwith GPS do not show the same good fit. A moreelaborate model is presented by Sturkell et al. (submittedfor publication) who invoke two Mogi sourcesto account for the far field displacements (Fig. 6). Theshallow source at 3 km depth and the deeper at 16 kmdepth. The shallow source contributes to an estimatedannual subsidence of 4.2 cm and the deeper sourceabout 1.0 cm. This model was also applied to theresults of a micro-gravity study for the period 1988–2003. A sub-surface mass decrease of 1.610 11 kg isderived (de Zeeuw-van Dalfsen et al., 2004b) indicatingthat magma drainage is an important contributor tothe subsurface mass decrease and therefore to thesurface deflation. Interferometric analysis of syntheticaperture radar images from radar satellites (InSAR)have also been conducted (Pagli et al., 2003), andshow clearly the ongoing subsidence at Askja. TheInSAR data are being further analyzed. The geometryof the deeper pressure source has not been resolved.Sturkell et al. (submitted for publication) suggest thatthe onset of deflation at Askja in 1973 is causallyrelated to the beginning of magma flow into the Kraflamagma chamber about 75 km to the north whichpreceded the rifting episode of 1975–1984. We alsopoint out that a series of large earthquakes began atthe Bárðarbunga volcano in 1974 that lasted until1996. The reverse faulting mechanisms of theseevents were interpreted by Einarsson (1991b) as theexpression of deflation of Bárðarbunga. The temporalcorrelation of deflation of Bárðarbunga and Askja andthe beginning of activity at Krafla suggests linksbetween the volcanoes, eventually by pressure connectionthrough the lower ductile crust in Iceland(Einarsson, 1991b).3.3. GrímsvötnThis highly active volcano most recently erupted inDecember 1998 and November 2004. Grímsvötn is asubglacial volcano, and deformation can only be measuredat a campaign GPS station located at the solenunatak on the caldera rim. Fig. 8 shows verticaldisplacement of the GPS point at Grímsvötn(GRIM). GPS measurements showed uplift beforethe 1998 eruption, followed by subsidence as thepressure dropped in the magma chamber. Measurementswere fitted to a Mogi model with the assumptionthat the source was located under the centre of theGrímsvötn caldera complex at a depth of at least 1.6km (Sturkell et al., 2003a). Inflation resumed atGrímsvötn following the 1998 eruption and it wasassumed that the pressure of the Mogi-source wouldhave to surpass a critical level before a new eruptionwas to be expected. The critical level of the verticalcomponent of the GRIM station was suggested to beat least 0.15–0.20 m above the minimum heightobtained immediately after the 1998 eruption (Sturkellet al., 2003a). A measurement in September 2004showed that the inflation had exceeded this level.Seismic activity increased markedly in the latter halfof the year 2003 (Fig. 9) showing that the stress in themagma chamber roof had increased. Further increaseof seismicity in late October and the onset of ajökulhlaup (a flood of melt water) from the Grímsvötncaldera lake on October 27 suggested that an eruptionmight be imminent, triggered by the pressure releaseof the flood. An eruption began in the evening of

22E. Sturkell et al. / Journal of Volcanology and Geothermal Research 150 (2006) 14–340.15vertical displacement [m]0.100.050.00-0.05GjalpGrímsvötnGrímsvötn-0.101995 2000 2005YearFig. 8. The vertical displacement of the GPS statin at Grímsvötn (located at the top of the caldera rim) relative the reference point at Jökulheimar(JOKU in Fig. 1). The vertical displacement shows uplift prior to the 1998 eruption followed by subsidence as the pressure dropped in themagma chamber. Following the 1998-eruption pressure increase is observed, the 1997 inflation level was passed in 2003 and on 1 November2004 the volcano erupted. The timing of the eruptions at Gjálp in 1996 and Grímsvötn in 1998 and 2004 are shown by vertical bars. Modifiedfrom Sturkell et al. (2003a) with recent data.November 1 as indicated by the appearance of lowfrequencyeruption tremor on nearby seismographs.The eruption onset was preceded by an intense swarmof earthquakes. The geodetic and seismic precursorsto the eruption were recognized and formed the basisfor rather detailed eruption warnings issued to thecivil defense authorities, both in the long-term(months and year), intermediate-term (weeks anddays) and short-term (hours) (Vogfjörd et al., 2005).3.4. GjálpIn the beginning of October 1996, a sub-glacialeruption started on a fissure between the Bárðarbunga43start ofthe eruptionM l2102002 2003Year2004 2005Fig. 9. Magnitude of Grímsvötn earthquakes as a function of time, showing increased activity since a seismic station was installed at Grímsvötnin 2002, until the 25 November 2004. The Grímsvötn volcano erupted in the evening on 1 November 2004. All earthquakes larger thanmagnitude 1.0 that have been observed at four or more seismic stations are included.

E. Sturkell et al. / Journal of Volcanology and Geothermal Research 150 (2006) 14–34 23and Grímsvötn volcanoes (Fig. 1). This eruption fissurewas termed Gjálp (Gudmundsson et al., 1997).The eruption began on a 7 km long N–S fissure after36 h of intense seismic activity at Bárðarbunga. Theseismicity propagated towards the eruption site suggestingthat the eruption was fed from Bárðarbunga(Einarsson et al., 1997). The eruptive products werebasaltic andesites, different from recent eruptive productsof both Bárðarbunga and Grímsvötn. The similarityof the isotope characteristics with those ofGrímsvötn has led to suggestions that the eruptionwas fed by Grímsvötn volcano (Sigmarsson et al.,2000). The eruption lasted for thirteen days. In earlyNovember a jökulhlaup of approximately 3.8 km 3 ofmeltwater was released over Skeiðarársandur, shownin Fig. 1 (Gudmundsson et al., 1997).3.5. HeklaHekla volcano has had 18 summit eruptions andfive in its direct surroundings during historic time,the last 1100 years. Precursory events have beendetected prior to recent eruptions. Generally, eruptionrelated earthquakes begin less than two hours beforethe eruption starts (Soosalu et al., in press). Heklawas the first volcano in Iceland to be studied by alocal GPS-network. Measurements before and afterthe 1991 eruption revealed contraction towards thevolcano (Sigmundsson et al., 1992). The data suggesteda magma reservoir at 9 km depth, but uncertaintieson this estimate were large because ofabsence of GPS stations close to the volcano (theclosest being at about 13 km distance). Measurementsprior to the 2000 eruption were conducted in1996. After the eruption measurements have beenconducted annually at selected points. The 1996–2000 displacement reveal clearly deformation associatedwith the feeder dyke for the 2000 eruption(amounting to several tens of cm). Further modellingof these data are needed, in particular to constraineventual inflation/deflation cycle at the volcano. GPSdata combined with all other geodetic data fromHekla, including InSAR and optical leveling tilt hasthe potential to provide improved constraints on themagma plumbing system under Hekla (Sigmundssonet al., 2001).Continuous monitoring of crustal deformation inthe South Iceland seismic zone (including Hekla) isdone by six continuously recording borehole strainmetersoperated by the Icelandic MeteorologicalOffice. The permanent strainmeters are located 15–45 km from Hekla and gave a clear strain signalprior to the eruptions of 1991 and 2000 (Fig. 10).The closest volumetric strain meter (15 km distance,at Búrfell in BUR in Fig. 10) detected compressiondue to dyke injection 30 min before magma reachedthe surface and the eruption started on February 26,2000 (Fig. 10). However, it had long been suspectedthat large strain excursions at BUR were accompaniedby localized recovery, which may bias estimatesof magma source depth. This was confirmed bystrain records for the Mw 6.5 South Iceland earthquakes(the first occurring on 17 June and the secondon 21 June, 2000). The original interpretation ofthe strain data at BUR associated with the 1991 and2000 eruptions of Hekla gave too shallow a sourcedepth. By considering the localized strain recovery,Jónsson et al. (2003) prefer a deeper location thanthe 6.5 km suggested by Linde et al. (1993) for themagma source. The most recent estimate based onre-interpretation of strain data places the magmachamber at 11 km depth (Kristján Ágústsson pers.com. 2004). Despite these complications, the combinationof seismic and volumetric strain monitoringhas given strong evidence of magma propagationtowards the Earth’s surface and was used to issuea warning prior to the 2000 eruption.Seven short leveling arrays (bdry-tiltQ stations)are surveyed repeatedly around Hekla. The beststation, Næfurholt, is located 11 km west of thesummit, allowing for easy monitoring of E–W tilt.The tilt signal from the Næfurholt station is rathersmall but is still significant (Fig. 11). It suggestsinflation beginning directly after the 1991 and 2000eruptions. Eastward upward tilt generally increaseswith time, as the volcano builds up pressure for thenext eruption, when a co-eruptive pressure decreaseoccurs. The most recent measurement (18 Nov.2004) of the tilt station at Næfurholt shows upwardtilt towards Hekla (Fig. 11). The east componentindicates that a pressure increase under the volcanois of the same scale as prior to the 1991 and 2000eruptions. It should be pointed out that one tiltstation gives only an indication of possible magmaaccumulation under a volcano, and it should spursome additional measurements.

24E. Sturkell et al. / Journal of Volcanology and Geothermal Research 150 (2006) 14–34200GEL100SAUSTOHELnanostrain (extension )0-100-200BURSKANANOSTRAIN1000800600400STOSAUGELSKABUR200-3000 17 18 19 20 21 22Time (Days)26 27 28 29 1February 2000March 2000Fig. 10. Five days of data spanning Hekla’s 2000 eruption from borehole strainmeters after removal of the effects of atmospheric pressurevariations and earth tides. Expansion is positive. The inset shows the strainmeter record for the 1991 Hekla eruption, which displays an almostidentical record. The squares show the strain changes calculated from modeling the eruption in two stages. Stage one shows the deflation of adeep reservoir together with the dyke formation and stage two with continued deflation of the deeper source. For location of the strainmeters seeFig. 1. (Modified from Agustsson et al., 2000).East tilt (µrad)50Eruption1985 1990 1995 2000 2005YearFig. 11. Data from the Næfurholt tilt station located 11 km directly west of the Hekla volcano. Its east–west tilt shows the radial tilt componentof deformation of the volcano. The observed tilt pattern repeats it self, suggesting a cyclic pattern, with upward tilt in the direction of the Heklavolcano prior to the eruptions, followed by subsidence as magma drains from beneath the volcano. The most recent measurement (18 Nov 2004)indicates that the east–west tilt is close to the critical value.Eruption

E. Sturkell et al. / Journal of Volcanology and Geothermal Research 150 (2006) 14–34 253.6. KatlaKatla volcano has had 20 eruptions since thesettlement period in 874 AD. (Larsen, 2000). Itscaldera, which is 600–700 m deep and encirclesan area of 100 km 2 (Björnsson et al., 2000) iscovered by the Mýrdalsjökull ice cap. The initialphase of its eruption is phreatomagmatic due to itsoccurrence at the base of the ice cap. Meltwaterfrom the eruption site is released as a jökulhlaup atthe ice margin. Such floods are characterized by arapid rise to maximum discharge within a period ofa few hours (Roberts et al., 2003). The most recentlarge eruption took place in 1918, generating amassive jökulhlaup. In 1955 a smaller jökulhlaupemerged from the eastern side of Mýrdalsjökull andtwo cauldrons were formed in the ice cap (Rist,1967). Guðmundsson et al. (2000) suggested thatthe jökulhlaup was caused by a small sub-glacialeruption that did not break the ice surface. In July1999 a sudden jökulhlaup occurred at the Sólheimajökulloutlet glacier (Fig. 12) on the south side ofthe Mýrdalsjökull ice cap (Sigurðsson et al., 2000).This was preceded by changes in earthquake activityand accompanied by tremor bursts (Einarsson,2000). A minor sub-glacial eruption or a shallowintrusion could have triggered this jökulhlaup(Guðmundsson et al., 2000).The Katla area has shown persistent seismic activitysince the beginning of seismograph observationsin the first half of the last century. Einarsson andBrandsdóttir (2000) analyzed seismic data from theperiod 1978 to 1985, and found that the seismicitywas mostly located in two areas, one within thecaldera and the other in the Goðabunga area to thewest of the caldera (Fig. 12). The seismicity has adistinct seasonal pattern, particularly in the Godabungacluster (Fig. 13), with a large majority ofearthquakes occurring in the second half of theyear. This pattern has persisted for at least fourdecades. The autumn activity in 2001 was intensebut within normal limits. Seismic activity, however,did not stop in the winter of 2002 as it would havein a normal year and seismicity has been high eversince. In a recent study by Soosalu et al. (in press) alocal seismic net was set up on and around the63º45'NHAMRENTAGo a-bungaThe KatlacalderaAUST63º30'N1 cmSólheimajökullSOHOHVOL10 km20ºW 19ºW19º15'WFig. 12. Observed horizontal and vertical displacements at Katla from June 2001 to March 2003, measured by GPS. The black arrows showhorizontal displacement, and the open arrows show vertical movements. All displacements are relative to the reference station at HAMR(Hamragarðar). This uplift and outward displacement from the centre of the caldera suggests magma inflow in a magma chamber, possiblelocated at 3–5 km depth. The vertical error is up to 7 mm.

26E. Sturkell et al. / Journal of Volcanology and Geothermal Research 150 (2006) 14–346050M 0 *10E+13 [Nm]40302010Go abungaKatla caldera01997 1998 1999 2000 2001 2002 2003 2004 2005YearFig. 13. Accumulated seismic moment release from 1997 to 2004 in the Katla volcano. The area is divided into two sub-areas; the Katla calderaand the area west of the caldera, the Goðabunga area. The division is drawn along a north–south line at 198 15V W (see Fig. 12). The moment–magnitude relation log M 0 =M l +10 from Slunga (1991) is used, where M 0 is the moment (in Nm) and M l the magnitude, for the 2775earthquakes with M 1 N1.7 recorded since the beginning of 1997 (2554 eq. in Goðabunga, 221 in the caldera). The earthquakes under Goðabungashow a clear seasonal correlation with a distinct activity during the autumn. This pattern changed in 2002 to more continuous activity. Theearthquakes in the caldera have a swarm character, and the largest earthquakes occur there.glacier, with one station directly above the center ofthe activity in the Goðabunga area. This study gave amuch better depth constraint for the hypocentrallocations and suggests that the Goðabunga clusteris an expression of an ascending, dome-shaped structurecreated by accumulation of viscous magma atshallow depth (cryptodome).The July 1999 event at Katla spurred both tilt andcampaign GPS measurements. Tilt measurementsshowed no detectable displacements at the Katlavolcano, but the stations are located z12 km fromthe center of the Katla caldera. However, inflation ofthe neighboring volcano Eyjafjallajökull wasdetected (see the Eyjafjallajökull section). CampaignGPS measurements on nunataks in the ice cap, theENTA and AUST sites (Fig. 12), have providedimportant results. The first campaign measurementsthat include the nunatak stations were performed in1993, and repeated for the first time in July 2000.The two stations displayed outward displacementfrom the center of the caldera, during this period,indicating that magma accumulation had started(Sturkell et al., 2003b). The nunatak sites havebeen visited annually since year 2000, and recentlytwice a year. The GPS measurements for the period2001 to 2003 is shown in Fig. 12 and display upliftand outward horizontal displacement from the centreof the sub-glacial caldera. These displacements fitwell with a point source at 4.7 km depth within thecaldera and give an uplift rate of about 2 cm/a since2000 at the GPS point AUST (Sturkell et al.,2003c).The Katla GPS-network is complemented by twocontinuous GPS-stations, (SOHO and HVOL), whichwere installed in the autumn of 1999, at the southernflank (Fig. 12). After considering expected platemovements, they display residual outward displacements(southward) from the volcano centre (Fig. 14).This southward displacement has been observed sincethe year 2000. The southward displacement observedfrom the continuous stations is not uniform in timeand shows periods with accelerated southward displacements.This suggests eventual annual component ofdeformation related to variable ice loading. Altogether,the geodetic data suggest progressive calderainflation over time. From 2000, the measured sub-

E. Sturkell et al. / Journal of Volcanology and Geothermal Research 150 (2006) 14–34 27EAST (mm)NORTH (mm)UP (mm)1009080706050403020100-10-203020100-10-20-30-40-50-60-70605040302010-10 0-20-30-40-50-60The Hekla eruptionThe June earthquakes2000 2001 2002 2003 2004 20052000 2001 2002 2003 2004 20052000 2001 2002 2003 2004 2005YearFig. 14. Time-series plot of the continuous GPS station SOHO, located at the southern edge of Mýrdalsjökull ice-cap covering the Katlavolcano, relative to REYK (the continuous GPS station in Reykjavík). Times of the Hekla eruption in February 2000 and the June 2000 southIceland seismic zone earthquake sequence are noted with vertical lines. For reference, the predicted movement of SOHO from the NUVEL1-Amodel (DeMets et al., 1994) is shown with solid lines. Outliers have been removed and error bars are at the 2 r level.surface volume increase is ~0.011 km 3 . In comparison,the eruptive volume of the 1918 Katla eruptionwas of the order of 1 km 3 . In summary there areseismic, geodetic and geothermal indications thatKatla has entered an agitated state that most likelybegan in 1999. The most recent GPS measurements(from March 2003 to January 2004) confirm that theunrest at Katla continues with the same rate.

28E. Sturkell et al. / Journal of Volcanology and Geothermal Research 150 (2006) 14–343.7. EyjafjallajökullThe volcano Eyjafjallajökull has an ice cap withthe same name. No volcanic activity is known fromthe time of settlement (874 AD) until 1612. Annals forthat year mention an eruption in Eyjafjallajökull orKatla, but it is not clear whether one volcano or bothwere active. A small eruption occurred in 1821–1823(Larsen, 1999). Prior to 1991, the Eyjafjallajökullvolcano showed insignificant seismic activity for atleast three decades (time of seismic coverage). Anincrease in activity started near the end of 1991 andpeaked with an earthquake swarm in 1994. Unrestcontinued, and a second earthquake swarm occurredin 1999. The seismic activity that started in the earlynineties continues with a higher background level thanbefore 1991 (Fig. 15).The 1994 and 1999 seismic episodes at Eyjafjallajökullwere associated with significant inflation of thevolcano indicating intrusions under its southern flank.The timing of the deformation is bracketed by tilt andGPS measurements (Sturkell et al., 2003b) for boththe intrusive events (Fig. 15). The deformation datafrom 1999 are modelled by a point pressure source at3.5 km depth beneath the flank of the volcano, about 4km south of the summit crater. Maximum uplift of themodel is 0.35 m (Sturkell et al., 2003b). A similarmodel also explains deformation associated with the1994 seismic crisis. However, its location is not thatwell constrained because only two GPS stations andone tilt station recorded the event. The 1994 intrusionhas also been modeled by Pedersen and Sigmundsson(2004) using InSAR images. Their study suggests thata sill with varying thickness accommodates the deformationsignal better than a single point pressuresource (Fig. 3). A similar sill model can be appliedto the 1999 event (Pedersen et al., 2004). The earthquakeactivity associated with the intrusions began afew months before measurable deformation occurred.A large part of it was located below the northern flankof the volcano (Dahm and Brandsdóttir, 1997) and atlarger depth than the inferred intrusions. We suggestthat this activity is related to the feeding channel ofthe intrusions.3.8. Hengill–HrómundartindurIn mid-1994 a very intense earthquake swarmstarted at the Hengill triple junction where the ReykjanesPeninsula oblique rift, The Western VolcanicZone and the South Iceland Seismic Zone meet. Theactivity was concentrated 5 km east of the Hengillvolcanic system, in the less pronounced Hrómundartindurvolcanic system. This enhanced earthquakeactivity continued until 1998. During the 1994–1998period about 85,000 earthquakes were recorded(Jakobsdóttir et al., 2002; Árnadóttir et al., 1999), andstill the Hengill triple junction is one of the mostpersistent source of microearthquakes in Iceland. Theincreased earthquake activity was associated with 243M l2101990 1992 1994 1996 1998 2000 2002 2004YearFig. 15. Magnitude of the earthquakes in Eyjafjallajökull volcano as a function of time, spanning the interval 1 Jan. 1990 to 31 Aug. 2004,including all events above magnitude 1.0. The seismic activity started in the autumn 1991 and it has continued with varying intensity,culminating in a swarm that started in the beginning of 1999. This swarm continued into 2000 and thereafter the earthquake activity has settledat a higher background level. The 1994 and 1999 periods of crustal deformation (intrusion) are shown as shaded bars (Sturkell et al., 2003b).

E. Sturkell et al. / Journal of Volcanology and Geothermal Research 150 (2006) 14–34 29cm/a of uplift and horizontal displacement between1993–1998, interpreted to be caused by magma accumulationat about 7 km depth (Sigmundsson et al.,1997b; Feigl et al., 2000). Hreinsdóttir (1999) measureda maximum uplift of 8 cm and outward horizontaldisplacement from the centre of Hrómundartindur volcanicsystem by GPS-geodesy. The geodetic monitoringof the Hengill triple junction utilized leveling, tilt,GPS and InSAR. An extensive GPS network isinstalled in the area. The unrest in this area (which islocated about 50 km east of the capital, Reykjavík)spurred the installation of 4 continuously recordingGPS stations, that were up and running in the springof 1999, transferring the data daily to the MeteorologicalOffice in Reykjavík. The activity culminated in1998 with a couple of magnitude 5 earthquakes (Cliftonet al., 2002) after which the unrest terminated.3.9. TorfajökullThe Torfajökull volcano is located at the rift–transformintersection where the Eastern Volcanic Zonemeets the South Iceland Seismic Zone. It is uniqueamong Icelandic volcanoes because of its abundanceof acidic lavas (e.g., Gunnarsson et al., 1998). Eruptionsare sometimes triggered by lateral injection ofbasaltic dykes from the rift zone to the NE, theVeiðivötn fissure swarm of the Bárðarbunga volcanicsystem (McGarvie, 1984). This appears to be true forthe last two eruptions, those of 871 and 1477 AD(McGarvie, 1984; Larsen, 1984). Torfajökull is apersistent source of small earthquakes, both highand low frequency. The distribution of the hypocentersof the high-frequency events led Soosalu andEinarsson (1997) to suggest that they were causedby the cooling of a hot body centered at 8 km depthbeneath the 12 km wide Torfajökull caldera. Tiltmeasurements seem to support this as slow deflationis discernible for the period 1991–2002. Our measurementof the tilt stations in 2003, however, shows thatthe subsidence has turned to inflation.4. DiscussionCrustal deformation studies of active volcanoes inIceland were initiated by Eysteinn Tryggvason in1966, when he set out the first part of the levelingline at the Askja volcano. In the following years, heinstalled short leveling lines at the Katla volcano andstarted to use mobile water-tube tilt meters at Hekla.The profile at Askja was extended in 1968 to 30benchmarks. Eysteinn concluded after a few years ofexperiences with mobile water-tube tilt meters thatoptical leveling gave the same precision and the measurementswere much easier to perform (Tryggvason,1994). With the start of the Krafla rifting episode1975–1984, crustal deformation studies became oneof the most important means to monitor the state ofthe volcano, and its shallow magma chamber. Theinstallation of the water-tube tilt meter (Fig. 4) inthe main building of the Krafla power station thengave a time-series of 11 years for the volcano.The installation of leveling lines and tilt stationssince the mid-sixties has made it possible to observelong time series at active volcanoes. Eysteinn Tryggvasoninstalled several long leveling lines and morethan fifty optical tilt stations (dry-tilt) around volcanoesin Iceland. Over the years some leveling linesand tilt stations have proved to be more sensitive tomagma movements under volcanoes than others. Agood example is the tilt station Næfurholt (Fig. 11) 11km due west of Hekla volcano, which shows inflation/deflation events clearly. Some tilt stations are notparticularly sensitive or are unstable with local conditionsmasking volcano signals. With the introductionof GPS and later InSAR techniques the studies ofcrustal deformation on volcanoes have taken a stepforward. Those techniques have much better spatialand temporal resolution and they are less weatherdependent.In the Hengill volcano a persistent swarm ofabout 85,000 earthquakes were recorded in 1994–1998. In May 1999 a permanent GPS station (OLKEin Fig. 1) was installed close to the measured centerof uplift in Hengill as determined by campaign GPS(Hreinsdóttir, 1999). At the time the continuousGPS-station at OLKE became operational in May1999, crustal deformation ceased. Based on tilt andGPS measurements it has been revealed that intrusionevents occurred within the Eyjafjallajökull volcanoin 1994 and 1999 (Sturkell et al., 2003b). Thetwo permanent GPS stations at the southern edge ofMýrdalsjökull that covers the Katla volcano havedetected displacements (Fig. 14) that are attributedto magma inflow under the Katla volcano (Sturkell

30E. Sturkell et al. / Journal of Volcanology and Geothermal Research 150 (2006) 14–34et al., 2003c). The Icelandic Meteorological Officeprovides online access to real-time geophysical data,(see http://www.vedur.is/ja). Epicenters and straindata are displayed on this homepage and areuploaded automatically.Of the 35 active volcanic systems in Iceland, deformationdue to magma movements has been measuredat eight of them in the last 15 years. Using a combinationof the different geodetic techniques we have beenable to locate magma chambers with a better precision.The Askja, Krafla, Grímsvötn and Katla volcanoesappear to have a shallow magma chamber atabout 3 km depth. Recent observations suggest thatat least the Askja and Krafla volcanoes have deepermagma reservoirs, suggested to be at 16 and 20 kmdepth, respectively (Sturkell et al., submitted for publication;de Zeeuw-van Dalfsen et al., 2004a,b). Thelocation of the magma chamber under the Heklavolcano has been debated and depth estimates rangefrom 6.5 to 11 km depth. Petrological observationssuggest a magma reservoir under Hekla, as magmaundergoes differentiation with time. However, a studyof the attenuation of seismic waves beneath Heklasuggests that a significant volume of magma doesnot exist above 14 km depth (Soosalu and Einarsson,2004). The discrepancy between these data is thesubject of ongoing research.Comparison of seismicity and deformation duringvolcanic unrest demonstrates great variability. In theHengill area 85,000 earthquakes were recorded duringinflation, but the total uplift was 8 cm during the sametime period. The uplift in the Katla caldera since 1999has been 12 cm but only 221 earthquakes MwN 1.7were recorded within the caldera (Fig. 13). In theneighboring volcano Eyjafjallajökull, which experiencedintrusion events in 1994 and 1999, the earthquakeactivity began before the intrusion events. Thispattern could be observed for both intrusions (Fig.15). These observations suggest volcanoes behaveindividually, e.g., depending on prevailing stressfields.Over this period (the last 15 years), the volcanoesHengill, Hekla, Katla, Eyjafjallajökull, Torfajökulland Grímsvötn had inflow of magma from depthinto shallow-level magma chambers (Fig. 2). TheGrímsvötn volcano was the fastest inflating volcanowith a rate of 3 cm/a at the closest GPS point (Fig. 8)until it erupted 1 November 2004. It is likely thevolcano will start to re-inflate after the eruption in asimilar way as happened in 1998. The Grímsvötncaldera hosts a lake that can contain up to 2–3 km 3of water and can be drained in one flood (Björnsson,2002). Removing such amount of mass directly abovean inflating magma chamber may serve as an eruptiontrigger. The eruption of Grímsvötn on November 1,2004 was preceded by a jökulhlaup that began onOctober 27, suggesting that a pressure drop equivalentto 15 m of water was sufficient to trigger an eruptionfrom the inflated magma chamber. The center of theKatla volcano inflates with a steady rate of 3 cm/a.since year 2000. In addition to the uplift in the centerof the Katla caldera, uplift is observed in the Goðabungaarea (Fig. 12) with 2.6 cm in six months. Thisis interpreted to be in connection with an ascendingdome-shaped structure (Soosalu et al., submitted forpublication).Subsidence takes place at present time at theKrafla and Askja volcanoes. Cooling and consequentcrystallization of magma cause a volume reductionof 10–11%. This process is probably the most dominatingfor the 3 km deep magma camber at Krafla,but also magma draining may be a contributingfactor. The Krafla volcano began to subside in1989, initially at a rate of a few cm/a. Since around2000 the subsidence rate has decreased to a few mm/a. Currently the largest subsidence signal is causedby the extensive exploitation of the geothermal field.The extraction and increased water circulation maycontribute to the cooling of the shallow magmabody. Volcanic geodesy allows observations of thissubsidence and shows its decay rate. Combined withmicro-gravity, it gives an estimate of the amount ofcrystallization and the amount of draining from theshallow magma chamber (de Zeeuw-van Dalfsen etal., in press). The Askja volcano has subsided forover twenty years, and extrapolation of the verticalchange into the 1972–1983 period suggests a totalsubsidence from 1973 to 2003 of 1.9 m at the Askjacenter (Sturkell et al., submitted for publication). Acombination of magma drainage and cooling andcontraction of the shallow magma reservoir at 3km and the deeper one at 16 km depth is ourfavoured model, and is consistent with the integratedmicro-gravity and deformation observations (deZeeuw-van Dalfsen et al., 2004b; Sturkell et al.,submitted for publication). We suggest that exten-

E. Sturkell et al. / Journal of Volcanology and Geothermal Research 150 (2006) 14–34 31sional tectonic forces generate a pressure drop in thelower and ductile part of the crust to accommodateongoing magma drainage from the shallow magmachamber.A combination of different monitoring methods isessential to obtain the best possible information onmagma sources and movements within volcanoes.Today, the real-time network in Iceland consists of40 seismic stations, 17 permanent GPS stations andsix volumetric strainmeters. Five analogue seismicstations maintained by the Science Institute at theUniversity of Iceland complement these networks.Both networks play an important role in eruptionprediction. Data from volumetric strain meters haveenabled detection and monitoring of dyke propagationpreceding eruptions of Hekla. The initial sign of aforthcoming eruption of Hekla in 2000 was the onsetof small earthquakes, known from previous eruptionto be associated with opening of a dike under thevolcano. The BUR strain station then recorded contraction(Figs. 1 and 10) beginning 30 min before theeruption, and was at that time recognized as resultingfrom a dike propagating to the surface (Agustsson etal., 2000). It was used along with the seismic data in asuccessful prediction of the eruption that was issued tothe civil defense authorities and broadcast as well tothe public.5. ConclusionsCrustal deformation measurements in Iceland havesignificantly advanced our understanding of volcanicprocesses, with new geodetic techniques allowingincreasingly detailed views of magma dynamics. Seismicmethods complemented with various geodeticmethods, including leveling, GPS, volumetric strainand InSAR, have provided the most important data.The longest time series extend from the mid-sixties tothe present time. Forecasting volcanic activity is challenging,but geodetic measurements are providinguseful constraints (location, amount and timing ofmagma movements), which can be interpreted alongwith other data. The most successful monitoring programsare based on a combination of seismic anddeformation methods, as the recent eruptions inGrímsvötn demonstrated. In recent years Katla andGrímsvötn have shown the highest inflation rates withca. 3 cm/a. and tilt data from Hekla suggest thatmagma accumulation is taking place. At Katla, continueduplift coincides with elevated seismicity. Ifuplift and internal pressure build-up continues atthese volcanoes, eruptive activity is a likely consequencewithin several years.AcknowledgmentsThe paper is dedicated to Eysteinn Tryggvason, thepioneer of deformation studies at Icelandic volcanoes,who turned 80 on July 19, 2004. We like to thank allthe people who participated in collection of geodeticdata. Guðmundur Gudfinnsson commented on earlierversions of this paper. The colleagues at the IcelandicMeteorological Office are thanked for their help, especiallyMatthew J. Roberts. Comments from the twoanonymous reviewers helped us to significantlyimprove the paper. Also the suggestions of the guesteditor Mike Poland have been most helpful. Thefigures were prepared with the public domain GMTsoftware. This work was supported by a grant fromthe Icelandic Research Council RANNÍS, and by theRetina project (EVG1-CT-00046).ReferencesAgustsson, K., Stefansson, R., Linde, A.T., Einarsson, P., Sacks,I.S., 2000. Successful prediction and warning of the 2000 eruptionof Hekla based on seismicity and strain changes. Eos Trans.AGU, Fall Meet Suppl., Abstr., V11B-30.Árnadóttir, Th., Sigmundsson, F., Delaney, P.T., 1998. Sources ofcrustal deformation associated with the Krafla, Iceland, eruptionof September 1984. Geophys. Res. Lett. 25, 1043–1046.Árnadóttir, Th., Rögnvaldsson, S., Ágústsson, K., Stefánsson, R.,Hreinsdóttir, S., Vogfjörð, K., Thorbergsson, G., 1999. Seismicswarms and surface deformation in the Hengill area, SW Iceland.Seismol. Res. Lett. 70, 269.Björnsson, H., 2002. Subglacial lakes and jökulhlaups in Iceland.Glob. Planet. Change 35, 255–271.Björnsson, A., Sæmundsson, K., Einarsson, P., Tryggvason, E.,Grönvold, K., 1977. Current rifting episode in north Iceland.Nature 266, 318–323.Björnsson, A., Johnsen, G., Sigurðsson, S., Thorbergsson, G.,Tryggvason, E., 1979. Rifting of the plate boundary in northIceland 1975–1978. J. Geophys. Res. 84, 3029–3038.Björnsson, H., Pálsson, F., Guðmundsson, M.T., 2000. Surface andbedrock topography of the Mýrdalsjökull ice cap. Jökull 49,29–46.

32E. Sturkell et al. / Journal of Volcanology and Geothermal Research 150 (2006) 14–34Brandsdóttir, B., Einarsson, P., 1979. Seismic activity associatedwith the September 1977 deflation of the Krafla central volcanoin NE-Iceland. J. Volcanol. Geotherm. Res. 6, 197–212.Brandsdóttir, B., Menke, W., Einarsson, P., White, R.S.,1997. Færoe–Iceland Ridge Experiment: 2. Crustal structureof the Krafla central volcano. J. Geophys. Res. 102,7867–7886.Clifton, A.E., Sigmundsson, F., Feigl, K.L., Guðmundsson, G.,Árnadóttir, Th., 2002. Surface effects of faulting and deformationresulting from magma accumulation at the Hengill triplejunction, SW Iceland, 1994–1998. J. Volcanol. Geotherm. Res.115, 233–255.Dahm, T., Brandsdóttir, B., 1997. Moment tensors of microearthquakesfrom the Eyjafjallajökull volcano in South Iceland.Geophys. J. Int. 130, 183–192.Darbyshire, F.A., White, R.S., Priestley, K.F., 2000. Structure ofthe crust and uppermost mantle of Iceland from a combinedseismic and gravity study. Earth Planet. Sci. Lett. 181,409–428.Decker, R.W., Einarsson, P., Mohr, P.A., 1971. Rifting in Iceland:new geodetic data. Science 173, 530–533.DeMets, C., Gordon, R.G., Argus, D.F., Stein, S., 1994. Effect ofrecent revision to the geomagnetic reversal time scale onestimates of current plate motions. Geophys. Res. Lett. 21,2191–2194.de Zeeuw-van Dalfsen, E., Pedersen, R., Sigmundsson, F., Pagli, C.,2004a. Satellite Radar Interferometry 1993–1999 suggests deepaccumulation of magma near the crust–mantle boundary at theKrafla volcanic system, Iceland. Geophys. Res. Lett. 31 (13),L13611. doi:10.1029/2004GL020059.de Zeeuw-van Dalfsen, E., Rymer, H., Sigmundsson, F., Sturkell,E., 2004b. Net gravity decrease at Askja volcano, Iceland:constraints on processes responsible for continuous calderadeflation, 1988–2003. J. Volcanol. Geotherm. Res. 139 (3–4),227–239. doi:10.1016/j.jvolgeores.2004.08.008.de Zeeuw-van Dalfsen, E., Rymer, H., Williams-Jones, G., Sturkell,E., Sigmundsson, F., in press. Integration of micro-gravity andgeodetic data to constrain shallow system mass changes atKrafla Volcano, N Iceland. Bull. Volc.Dzurisin, D., 2003. A comprehensive approach to monitoring volcanodeformation as a window on the eruption cycle. Rev.Geophys. 41, 1001. doi:10.1029/2001RG000107.Einarsson, P., 1978. S-wave shadows in the Krafla caldera in NE-Iceland, evidence for a magma chamber in the crust. Bull.Volcanol. 41, 1–9.Einarsson, P., 1991a. Umbrotin við Kröflu 1975–1989 (The activityof Krafla 1975–1989, in Icelandic). In: Einarsson, Á., Garðarsson,A. (Eds.), Náttúra Mývatns. Hið Íslenska Náttúrufræðifélag,pp. 97–139.Einarsson, P., 1991b. Earthquakes and present-day tectonism inIceland. Tectonophysics 189, 261–279.Einarsson, P., 2000. Course of events connected with the flood inJökulsá áSólheimasandi in July 1999 (In Icelandic). GeoscienceSociety of Iceland, February Meeting 2000, 14.Einarsson, P., Brandsdóttir, B., 1980. Seismological evidence forlateral magma intrusion during the July 1978 deflation of theKrafla volcano in NE-Iceland. J. Geophys. 47, 160–165.Einarsson, P., Brandsdóttir, B., 2000. Earthquakes in the Mýrdalsjökullarea, Iceland, 1978–1985: seasonal correlation and relationto volcanoes. Jökull 49, 59–73.Einarsson, P., Sæmundsson, K., 1987. Earthquake epicenters 1982–1985 and volcanic systems in Iceland, map. In: Sigfússon, T.,(Ed.), Í hlutarins eðli: Festschrift for Thorbjörn Sigurgerirsson,Menningarsjóður, Reykjavík.Einarsson, P.B., Brandsdóttir, B., Guðmundsson, M.T., Björnsson,H., Grönvold, K., Sigmundsson, F., 1997. Center of theIceland Hotspot experiences volcanic unrest. Eos 78, 369–375(Sept. 2).Ewart, A., Voight, B., Björnsson, A., 1991. Elastic deformationmodels of Krafla Volcano, Iceland, for the decade 1975 trough1985. Bull. Volcanol. 53, 436–459.Feigl, K., Gasperi, J., Sigmundsson, F., Rigo, A., 2000. Crustaldeformation near Hengill volcano, Iceland 1993–1998: couplingbetween magmatic activity and faulting inferred from elasticmodelling of satellite radar interferograms. J. Geophys. Res.,25655–25670.Foulger, G., Bilham, R., Morgan, W.J., Einarsson, P., 1987. TheIceland GPS geodetic field campaign 1986. Eos, Transact., Am.Geophys. Union 68, 1809.Foulger, G.R., Jahn, C.-H., Seeber, G., Einarsson, P., Julian, B.R.,Heki, K., 1992. Post-rifting stress relaxation at the divergentplate boundary in northeast Iceland. Nature 358, 488–490.Geirsson, H., 2003. Continuous GPS measurements in Iceland1999–2002. Icelandic Meteorological Office, Reykjavík, Rep.03014. 94 pp.Gudmundsson, M.T., Sigmundsson, F., Björnsson, H., 1997. Ice–volcano interaction of the 1996 Gjalp subglacial eruption, Vatnajokull,Iceland. Nature 389, 954–957.Gunnarsson, B., Marsh, B.D., Taylor Jr., H.P., 1998. Generation ofIcelandic rhyolites: silicic lavas from the Torfajökull centralvolcano. J. Volcanol. Geotherm. Res. 83, 1 – 45.Guðmundsson, M.T., Högnadóttir, Th., Björnsson, H., Pálsson, F.,Guðmundsson, G., 2000. Geothermal activity at Mýrdalsjökulland events in the summer 1999 (in Icelandic). GeoscienceSociety of Iceland, February Meeting 2000, 13.Hauksson, E., 1983. Episodic rifting and volcanism at Krafla inNorth Iceland, growth of large ground fissures along the plateboundary. J. Geophys. Res. 88, 625–636.Hreinsdóttir, S., 1999. GPS geodetic measurements on the ReykjanesPeninsula, SW Iceland: crustal deformation from 1993to 1998. M.Sc. thesis, Univ. of Iceland, Reykjavík, Iceland.114 pp.Jakobsdóttir, S.S., Guðmundsson, G.B., Stefánsson, R., 2002. Seismicityin Iceland 1991–2000 monitored by the SIL seismicsystem. Jökull 51, 87–94.Johnsen, G.V., Bjornsson, A., Sigurdsson, S., 1980. Gravity andelevation changes caused by magma movement beneath Kraflacaldera, northeast Iceland. J. Geophys. 47, 132–140.Jónsson, S., Segall, P., Ágústsson, K., Agnew, D., 2003. Localfluid flow and borehole strain in the south Iceland SeismicZone. Eos Trans. AGU, Fall Meet Suppl., Abstr., vol. G31B,pp. F500.Kanngiesser, E., 1983. Vertical component of ground deformation innorth Iceland. Ann. Geophys. 1, 197–206.

E. Sturkell et al. / Journal of Volcanology and Geothermal Research 150 (2006) 14–34 33Kjartansson, E., Grönvold, K., 1983. Location of a magma reservoirbeneath Hekla volcano, Iceland. Nature 301, 139–141.Larsen, G., 1984. Recent volcanic history of the Veiðivötn fissureswarm, southern Iceland: an approach to volcanic risk assessment.J. Volcanol. Geotherm. Res. 22, 33–58.Larsen, G., 1999. Gosið í Eyjafjallajökli 1821–23. Rep. RH-28-99,Science Institute, University of Iceland. 13 pp.Larsen, G., 2000. Holocene eruptions within the Katla volcanicsystem, south Iceland: characteristics and environmental impact.Jökull 49, 1 –28.Linde, A.T., Agustsson, K., Sacks, I.S., Stefansson, R., 1993.Mechanism of the 1991 eruption of Hekla from continuousborehole strain monitoring. Nature 365, 737–740.Massonnet, D., Feigl, K.L., 1998. Radar Interferometry and itsapplication to changes in the Earth’s surface. Rev. Geophys.36, 441–500.McGarvie, D.W., 1984. Torfajökull: a volcano dominated bymagma mixing. Geology 12, 685–688.Mogi, K., 1958. Relations between the eruptions of various volcanoesand the deformation of the ground surface around them.Bull. Earthquake Res. Inst. Univ. Tokyo 36, 99–134.Pagli, C., Sigmundsson, F., Árnadóttir, Th., Einarsson, P., 2003.Subsidence of Askja volcano, north Iceland: InSAR observationsand different modeling approaches. Eos Trans AGU FallMeet Suppl. Abstr., vol. V22A-0561, pp. F1520.Pedersen, P., Sigmundsson, F., 2004. InSAR based sill model linksspatially offset areas of deformation and seismicity for the 1994unrest episode at Eyjafjallajökull volcano. Geophys. Res. Lett.31, L14610. doi:10.1029/2004GL020368.Pedersen, R., Jónsson, S., Árnadóttir, Th., Sigmundsson, F., Feigl,K.L., 2003. Fault slip distribution of two June 2000 Mw 6.5 earthquakesin South Iceland estimated from joint inversion of InSAR andGPS measurements. Earth Planet. Sci. Lett. 213, 487–502.Pollitz, F.F., Sacks, I.S., 1996. Viscosity structure beneath northeastIceland. J. Geophys. Res 101, 17771–17793.Rist, S., 1967. Jökulhlaups from the ice cover of Mýrdalsjökull onJune 25, 1955 and January 20, 1956. Jökull 17, 243–248.Roberts, M.J., Tweed, F.S., Russell, A.J., Knudsen, Ó., Harris, T.D.,2003. Hydrological and geomorphic effects of temporary icedammedlake formation during jökulhlaups. Earth Surf. ProcessesLandf. 28, 723–737.Rymer, H., 1996. Microgravity monitoring. Monitoring andMitigation of Volcano Hazards. Springer-Verslag, Berlin,pp. 169–198.Rymer, H., Tryggvason, E., 1993. Gravity and elevation changes atAskja, Iceland. Bull. Volcanol. 55, 362–371.Rymer, H., Cassidy, J., Locke, C.A., Sigmundsson, F., 1998. Posteruptivegravity changes from 1990 to 1996 at Krafla volcano,Iceland. J. Volcanol. Geotherm. Res. 87, 141–149.Sacks, I.S., Suyehiro, S., Everton, D.W., Yamagishi, Y., 1971.Sacks–Evertson strainmeter, its installation in Japan and somepreliminary results concerning strain steps. Pap. Meteorol. Geophys.22, 195–207.Sigmarsson, O., Karlsson, H.R., Larsen, G., 2000. The 1996 and1998 subglacial eruptions beneath the Vatnajökull ice sheet inIceland: contrasting geochemical and geophysical inferences onmagma migration. Bull. Volcanol. 61, 468–476.Sigmundsson, F., Einarsson, P., Bilham, R., 1992. Magma chamberdeflation recorded by the Global Positioning System; the Hekla1991 eruption. Geophys. Res. Lett. 19, 1483–1486.Sigmundsson, F., Vadon, H., Massonet, D., 1997. Readjustmentof the Krafla spreading segment to crustal rifting measuredby Satellite Radar Interferometry. Geophys. Res. Lett. 24,1843–1846.Sigmundsson, F., Einarsson, P., Rögnvaldsson, S., Foulger, G.R.,Hodgkinson, K.M., 1997. The 1994–1995 seismicity and deformationat the Hengill triple junction, Iceland: triggering ofearthquakes by minor magma injection in a zone of horizontalshear stress. J. Geophys. Res. 102, 15151–15161.Sigmundsson, F., Pedersen, R., Feigl, K., Sturkell, E., Einarsson,P., Jonsson, S., Agustsson, K., Linde, A., Bretar, F., 2001.Joint interpretation of geodetic data for volcano studies: evaluationof pre- and co-eruptive deformation for the Hekla 2000eruption from InSAR, tilt, GPS and strain data. Eos. Trans.AGU, 82, Fall. Meet. Suppl., Abstr., vol. V21E-E07.Sigurdsson, O., 1980. Surface deformation of the Krafla fissureswarm in two rifting events. J. Geophys. 47, 154–159.Sigurðsson, O., Zóphoníasson, S., Ísleifsson, E., 2000. The jökulhlaupfrom Sólheimajökull July 18th 1999 (Jökulhlaup úr Sólheimajökli18. júlí 1999, in Icelandic with English summary).Jökull 49, 75–80.Slunga, R., 1991. Earthquake source mechanism determination byuse of body-wave amplitudes— an application to Swedish earthquakes.Bull. Seismol. Soc. Am. 71, 25–35.Soosalu, H., Einarsson, P., 1997. Seismicity around the Hekla andTorfajökull volcanoes, Iceland, during a volcanically quiet period,1991–1995. J. Volcanol. 59, 36–48.Soosalu, H., Einarsson, P., 2004. Seismic constraints on magmachambers at Hekla and Torfajökull volcanoes, Iceland. Bull.Volcanol. 66, 276–286.Soosalu, H., Einarsson, P., þorbjarnardóttir. B., in press. Seismicactivity associated with the 2000 eruption of Hekla, Iceland.Bull. Volc. (Published on line ahead of print, doi:10.1007/s00445-005-0417-7).Soosalu, H., Jónsdóttir, K., Einarsson, P. submitted for publication.Seismicity crisis at the Katla volcano. Iceland — signs of acryptodome? submitted.Sturkell, E., Sigmundsson, F., 2000. Continuous deflation of theAskja caldera Iceland, during the 1983–1998 non-eruptiveperiod. J. Geophys. Res. 105, 25671–25684.Sturkell, E., Einarsson, P., Sigmundsson, P., Hreinsdottir, F., Geirsson,S., 2003a. Deformation of Grímsvötn volcano, Iceland :1998eruption and subsequent inflation. Geophys. Res. Lett. 30, 1182.doi:10.129/2002GL016460.Sturkell, E., Sigmundsson, F., Einarsson, P., 2003b. Recent unrestand magma movements at Eyjafjallajökull and Katla volcanoes,Iceland. J. Geophys. Res. 108 (B8), 2369. doi:10.1029/2001JB000917.Sturkell, E., Sigmundsson, F., Einarsson, P., Geirsson, H., Roberts,M.J., 2003c. Magma inflow into Katla, one of Iceland’s mosthazardous volcanoes. Eos Trans. AGU, Fall Meet Suppl., Abstr.,vol. V12G-07, pp. F1503.Sturkell, E., Sigmundsson, F., Slunga, R., submitted for publication.1983–2003 decaying rate of deflation at Askja caldera: pressure

34E. Sturkell et al. / Journal of Volcanology and Geothermal Research 150 (2006) 14–34decrease in an extensive magma plumbing system at a spreadingplate boundary. Bull. Volcanol.Tryggvason, E., 1980. Subsidence events in the Krafla area, NorthIceland,1975–1979. J. Geophys. 47, 141–153.Tryggvason, E., 1986. Multiple magma reservoirs in a rift zonevolcano: ground deformation and magma transport during theSeptember 1984 eruption of Krafla, Iceland. J. Volcanol.Geotherm. Res. 28, 1 –44.Tryggvason, E., 1987. Myvatn lake level observations 1984–1986and ground deformation during a Krafla eruption. J. Volcanol.Geotherm. Res. 31, 131–138.Tryggvason, E., 1989. Ground deformation in Askja, Iceland: itssource and possible relation to flow of the mantle plume.J. Volcanol. Geotherm. Res. 39, 61–71.Tryggvason, E., 1994. Observed ground deformation at Hekla, Icelandprior to and during the eruptions of 1970, 1980–1981 and1991. J. Volcanol. Geotherm. Res. 61, 281–291.Tryggvason, E., 1995. Optical levelling tilt stations in thevicinity of Krafla and the Krafla fissure swarm. Observations1976 to 1994, vol. 9505. Nordic Volcanol. Inst, Reykjavík,Rep, pp. 218.Vadon, H., Sigmundsson, F., 1997. Crustal deformation from 1992to 1995 at the mid-Atlantic Ridge, Southwest Iceland, Mappedby Satellite Radar Interferometry. Science 275, 193–197.Vogfjörd, K.S., Jakobsdóttir, S.S., Gudmundsson, G.B., Roberts,M.J., Ágústsson, K., Arason, T., Geirsson, H., Karlsdóttir, S.Hjaltadóttir, S., Ólafsdóttir, U., Thorbjarnardóttir, B., Skaftadóttir,T., Sturkell, E., Jónasdóttir, E.B., Hafsteinsson, G.,Sveinbjórnsson, H., Stefánsson, R., Jónsson T.V., 2005. Forecastingand Monitoring a Subglacial Eruption in Iceland, EOS86, 245, 248.Wolfe, C.J., Bjarnason, I.Th., VanDecar, J.C., Solomon, S.C., 1997.Seismic structure of the Iceland mantle plume. Nature 385,245–247.