Recent unrest and magma movements at Eyjafjallajökull and Katla volcanoes, Iceland

Abstract

[1] Katla and Eyjafjallajökull volcanoes are situated 25 km apart at the southern tip of the Eastern Volcanic Zone in Iceland. Both have been active in historic time (last 1100 years) and have a history of simultaneous activity. The much more active Katla volcano has erupted at least 20 times, and Eyjafjallajökull's two eruptions were contemporaneous with Katla eruptions. Following a quiet period of several decades, the seismicity beneath Eyjafjallajökull was high in 1994 and again in 1999. The activity culminated in July 1999 when a flash flood occurred from the Mýrdalsjökull ice cap covering Katla, associated with changes in seismicity, bursts of volcanic tremor, and the formation and deepening of ice cauldrons. We report here results of deformation observations of these events, both by GPS geodesy and tilt measurements. The 1999 increase in seismicity at Eyjafjallajökull was associated with significant inflation of the volcano. The deformation data are modeled by a point pressure source at 3.5 km depth beneath the flank of the volcano, about 4 km south of the summit crater. Maximum uplift of the model is 0.35 m. A similar model also explains deformation associated with the 1994 seismic crisis. The deformation field of the Katla volcano is more difficult to ascertain due to the extensive glacier coverage. Movements of points on nunataks on and near the caldera rim indicate inflation and magma movements at shallow level beneath the caldera in connection with the events of July 1999.

1. Introduction

[2] The Eyjafjallajökull and Katla volcanoes belong to a group of volcanoes located south of the rift-transform intersection where the Eastern Volcanic Zone (EVZ) and the South Iceland Seismic Zone (SISZ) meet (Figure 1). Eyjafjallajökull is an ice capped volcano reaching an altitude of 1600 m above sea level. No volcanic activity is known from the time of settlement (874 A.D.) until 1612. Annals for that year mention an eruption in Eyjafjallajökull or Katla, but it is not clear whether one volcano or both were active. A small eruption occurred in 1821–1823 [Larsen, 1999]. Prior to 1991, the Eyjafjallajökull volcano showed insignificant seismic activity for at least three decades (time of seismic coverage). An increase in activity started near the end of 1991 and peaked with an earthquake swarm in 1994. Unrest continued, and a second earthquake swarm occurred in 1999.

Figure 1

Map of Iceland showing the general outline of volcanic systems along the plate boundary [Einarsson and Sæmundsson, 1987]. The neovolcanic zone in Iceland is divided into the northern volcanic zone (NVZ), the western volcanic zone (WVZ), and the eastern volcanic zone (EVZ). The EVZ and the WVZ are connected in the south by the South Iceland Seismic Zone (SISZ). The area including Eyjafjallajökull and Katla is outlined with a box. Stars indicates the epicenter of the M = 6.5 (17 June 2000) and M = 6.6 (21 June 2000) earthquakes.

[3] Much interest has been focused on the very active, neighboring Katla volcano located under the Mýrdalsjökull ice cap (Figure 2). The volcano has a large composite caldera and is known to have explosive eruptions twice a century. The most recent large eruption occurred in 1918. The eruptions of the preceding three centuries have followed a remarkable pattern, occurring around the 20th and the 60th year of each century with only small deviations [Thorarinsson, 1960]. On the basis of this pattern it was commonly assumed that the next large eruption would occur around 1960. It has not happened yet. Instead, a small, confined eruption probably occurred in 1955. This event was accompanied by a sudden jökulhlaup (glacier flood) and several small earthquakes [Tryggvason, 1960], but no eruption column or ash fall. Two subsidence cauldrons were formed in the ice near the eastern rim of the caldera. A similar event occurred on the southern caldera rim in July 1999. A sudden flood from the glacier was preceded by changes in earthquake activity and accompanied by tremor bursts [Sigurðsson et al., 2000; Einarsson, 2000].

Figure 2

Geodetic network around Eyjafjallajökull and Katla (Mýrdalsjökull) showing GPS and tilt stations. All abbreviations are explained in Table 1 except the seismic station Skammadalshóll in Mýrdalur (SKAM in the figure). During the July 2000 GPS survey the network was connected to two points in the Icelandic ISNET geodetic network. At Reynisfjall, the ISNET point REYF is located 74 m from the point REYN, and Botnar (BOTA) is located just outside the map edge to the east. On Mýrdalsjökull glacier, the hachured line is the Katla caldera, while the hachured line on Eyjafjallajökull marks the summit crater.

[4] This paper reports results of crustal deformation studies around Eyjafjallajökull and Katla volcanoes, utilizing GPS measurements and optical leveling (dry tilt) measurements. Tilt measurements were initiated in 1967 at three stations [Tryggvason, 2000], and the first GPS measurements were made in 1986, as a part of a regional survey. The first campaign specifically aimed at Katla took place in 1992. The increased seismic activity at Eyjafjallajökull 1994 led to a remeasurement of the Katla network and installation of additional points around Eyjafjallajökull. Two dry tilt stations were added. The total network now contains 31 GPS points (including three permanent GPS stations) and five dry tilt stations (Figure 2). A compilation of the seismicity and crustal deformation between 1992 and 1998 around Eyjafjallajökull was presented by Blomstrand-Stinessen [1999], who found that the 1994 earthquakes were associated with significant ground deformation, interpreted as a result of shallow intrusion.

[5] Elevated seismicity from Eyjafjallajökull and Katla in 1999 spurred a remeasurement of crustal deformation networks. All GPS geodetic points and dry tilt stations in the area were measured in 1999 and 2000. In this paper we report on the measured crustal deformation and argue that batches of magma were moved into the crust to about 3.5 km depth beneath Eyjafjallajökull in 1994 and 1999. We furthermore argue that magma movements took place at shallow level beneath Katla in 1999, even though the details of these changes cannot be discerned by the data.

2. Geologic Setting

[6] The Eastern Volcanic Zone is divided in two tectonically different zones, with the dividing line at its intersection with the transform zone, the South Iceland Seismic Zone. North of the transform the EVZ has the characteristics of a divergent plate boundary, i.e., fissure eruptions, normal faults and fissures. The part of the zone south of the transform is characterized by large volcanoes. Rifting structures are inconspicuous. These volcanoes are thus located within the Eurasian plate. In spite of this intraplate setting some of these volcanoes; that is, Katla and Hekla are among the most productive ones in Iceland. It has been suggested that this area represents the tip of a rift propagating into the Eurasian plate, away from the center of the Iceland hot spot. The arguments are mainly geochemical and structural. Fe-Ti volcanism, characteristic for propagating rifts, is found within this area, beginning 2–3 m.y. ago [Johannesson et al., 1990]. Voluminous volcanism has created a plateau of Fe-Ti basalt south of the Torfajökull (Figure 1) area toward the south coast in Mýrdalur. A characteristic feature of the Fe-Ti basalts is their aphyric state, which is ascribed to rapid segregation from a large regional mantle source beneath the (propagating) rift [Sinton and Christie, 1982].

[7] The Katla volcano is a major volcanic massif rising to an elevation of 1500 m. It is partly covered by the Mýrdalsjökull ice cap. The oldest rocks observed are about 100,000 year old (K. Grönvold personal communication, 2000), and the volcano has a 4-km-wide elliptic caldera (Figure 2) [Björnsson et al., 2000]. Since the settlement of Iceland (874 A.D.), Katla has erupted at least 20 times [Larsen, 2000]. A two-dimensional seismic undershooting of Katla revealed a travel time anomaly within the Katla caldera [Gudmundsson et al., 1994], interpreted as a 5-km-wide shallow magma chamber with a base 1.5 km below sea level (3 km below the surface). The top of the magma chamber is unresolved.

[8] Different chemistry and mineralogy distinguish the two volcanoes. Fe-Ti basalt dominates volcanism of the Katla system. The Eyjafjallajökull volcanic system (also referred to as Eyjafjöll), on the other hand, has produced a suite of alkalic rocks that ranges from ankaramites to hawaiite and minor silicic rocks [Jakobsson, 1979].

[9] The Eyjafjallajökull volcano is an elongated, flat cone of about 1600 m height. The Eyjafjallajökull glacier covers the volcano and its elliptical ∼2.5–km-wide summit crater (Figure 2). The outlet glacier Gígjökull originates from the crater and flows toward north through an opening in the crater rim. The most recent eruption (1821–1823) was located within the crater close to its southern rim, and produced intermediate to acid tephra [Thoroddsen, 1925]. This eruption caused a jökulhlaup. The 1821–1823 event is the only confirmed historical event in Eyjafjallajökull, but an event in 1612 is suspected [Þórarinsson, 1975]. A very small tephra layer of intermediate composition occurs at the same level as the Katla 1612 tephra in a few places. It is suggested to originate from Eyjafjallajökull volcano [Larsen, 1978]. The volcano is elongated in an east-west direction, as the 5-km-wide valley north of Eyjafjallajökull (the Markarfljót valley see Figure 3). These dominant E-W topographic features may be the surface expressions of deep-seated structures. Eyjafjallajökull volcano has an alkaline composition, similar to other off-rift volcanoes in Iceland. This type of volcano generates relatively small amounts of material, say about 0.1 km³, during each eruption. Most eruptive fissures and crater rows at Eyjafjallajökull are E-W orientated, but occasional radial fissures are observed around the summit of the volcano. The most conspicuous radial eruptive fissure is Sker (Figure. 2). In the area SSE of the summit crater and NE of Seljavellir farm (Figure 3) Jónsson [1998] reports the presence of highly altered rocks that are cut by numerous dikes and veins. This area is interpreted to be the oldest part of the Eyjafjallajökull volcano, with a suggested age of >0.78 Ma. If this age is accurate, then Eyjafjallajökull is one of the oldest active volcanoes in Iceland. The most pronounced geothermal activity at Eyjafjallajökull is confined to its south flank, in the area around Seljavellir (Figure 3). This area of geothermal activity correlates with the location of our maximum inferred uplift, i.e., the shallowest part of recently formed intrusions.

Figure 3

The M_L > 0.1 seismicity during the period 1 July 1990 to 14 November 2000 at Eyjafjallajökull. The data are extracted from the SIL database (http://www.vedur.is/) at the Icelandic Meteorological Office. Poorly located earthquakes occurring at Goðabunga tend to spread westward to Þórsmörk and southwestward to Skógaheidi (G. Guðmundsson, personal communication, 2000). Because of this, the search for Eyjafjallajökull earthquakes in the SIL database was limited to the indicated box. The star denotes a best fit point source of pressure location that explains geodetic observations for the 1999 episode. The pluses indicate the location of the Þorvaldseyri and Seljavellir farms.

3. Earthquake Activity

[10] The Katla volcano has exhibited persistent high seismicity for more than four decades [Einarsson and Brandsdóttir, 2000]. The epicenters of the located earthquakes fall in two distinct clusters: one in the eastern half of the caldera and the second at Goðabunga west of the caldera. Earthquakes under Goðabunga (Figure 3) show a clear seasonal correlation as they tend to occur in the latter half of the year [Einarsson and Brandsdóttir, 2000]. The triggering effects of reduced ice load after summer melting and the elevated pore fluid pressure in the underlying crust have been suggested to explain this autumn activity. Einarsson and Brandsdóttir [2000] consider pore fluid pressure to be the likely trigger mechanism.

[11] Before to the increase in earthquake activity that started in autumn 1991 Eyjafjallajökull was characterized by very low earthquake activity. Since 1973 the Icelandic seismic network had a good coverage around Eyjafjallajökull, and in the period 1973–1991 less than a dozen earthquakes were recorded at Eyjafjallajökull. Four earthquakes were located in 1979–1985 [Einarsson, 1991], and this was a typical rate of seismicity prior to 1991. Earthquakes located at Eyjafjallajökull during the period from 1 July 1990 to 14 November 2000 are presented in Figure 3. The seismic moment release of the earthquakes in Eyjafjallajökull versus time is given in Figure 4a, assuming the moment-magnitude relation logM₀ = 1.5M_w + 9.1, where M₀ is the moment (in N m) and M_w moment magnitude. From autumn 1991, earthquake activity increased steadily and culminated between 29 May and to 22 June 1994. In this swarm, more than 130 microearthquakes were recorded, with magnitudes reaching M_L ≈ 2.3 [Dahm and Brandsdóttir, 1997]. Their depth ranged from 1 to 13 km, with most earthquakes occurring at 5 km depth [Dahm and Brandsdóttir, 1997]. The May–June 1994 earthquakes were located under the northern flank of the Eyjafjallajökull volcano in the region of Steinsholtsjökull (Figures 2 and 3). Dahm and Brandsdóttir [1997] studied their focal mechanisms, and showed that 75% of the events had E-W oriented nodal planes. They interpreted this as an indication of an E-W oriented dike intrusion under the northern flank of the volcano, generating horizontal compression (i.e., thrust faulting) in its neighborhood. After the 1994 swarm, seismicity continued at a higher level in the Eyjafjallajökull area (Figure 4a) than prior to 1991. On 1 April 1999 a group of earthquakes occurred with a maximum M_L of 3.6 (Figure 3). Those earthquakes were located within a limited area north-northeast of the summit crater. Intense earthquake activity started in late July 1999 located south of the summit crater (Figure 3). Two distinct earthquake clusters are apparent in the epicentral distribution of Eyjafjallajökull, one south of the summit crater, and one beneath the northern flank of the volcano in the region of Steinsholtsjökull (Figure 3).

Figure 4

(a) Seismic moment release versus time at Eyjafjallajökull from 1 January 1990 to 14 November 2000 (data from the same area as shown in Figure 3). (b) Observed ground tilt at the optical leveling (dry) tilt station at Fimmvörðuháls (installed in 1992), located between Mýrdalsjökull and Eyjafjallajökull (see Figure 2). The strongest tilt signals are in an eastward direction and occurred 1994 and 1999. GPS data show that the tilt is a result of uplift under the southern slope of Eyjafjallajökull. (c) Time series of the vertical displacements at SELJ (Seljavellir) and STEI (Steinsholt) GPS stations from June 1994 to July 2000. During 1999–2000, the subsidence of STEI seems to lag in time behind uplift at SELJ. The tilt and GPS measurements bracket the time periods of crustal deformation shown as shaded bars. The 1999 episode is well constrained by deformation measurements, while the beginning of the 1994 episode is more uncertain and is tentatively set to correlate with the 1994 earthquake swarm.

4. Tilt Measurements

[12] Tilt has been measured by repeated optical leveling between permanent bench marks (also referred to as “dry tilt”, as opposed to liquid level tiltmeters). Using an L-shaped or T-shaped array of 10 bench marks with a distance of 50 m between markers, tilt can be measured with an accuracy of 1 μrad [Tryggvason, 1994].

[13] Five optical leveling tilt stations are around Katla and Eyjafjallajökull (Figure 2). Three stations were established in 1967, the JOKV (Jökulkvísl), KOTL (Kötlukriki) and HOFD (Höfðabrekkuheiði) stations. They are located more than 10 km away from the Katla caldera. The tilt network was extended in 1992 with a station on FIMM (Fimmvörðuháls) and after the 1994 unrest, the tilt station at DAGM (Dagmálafjall) was installed. The FIMM station is located between Mýrdalsjökull and Eyjafjallajökull, and a potential tilt signal produced by uplift in Goðabunga, might also be produced by deflation in Eyjafjallajökull, and vice versa. The relatively long distance of tilt stations from the seismically identified magma chamber (closest station KOTL at ∼16 km), and the shallowness of the chamber bottom, make potential tilt signals from inflow of magma into this chamber weak.

5. GPS Data

5.1. The Data

[14] The first regional GPS campaign in Iceland in 1986 included the installation of two stations near Eyjafjallajökull and Katla (Mýrdalsjökull). Regional measurements were carried out again in 1989 and 1992, and at these times the GPS network was extended. An overview of the 1986, 1989 and 1992 GPS campaigns is given by Sigmundsson et al. [1995]. All different GPS campaigns in the Katla and Eyjafjallajökull area, from 1992 are listed in Table 1. Measurements of subset of stations have been measured in different years. The complete network was measured in one survey for the first time in July 2000. Two points in the Icelandic Land survey network (ISNET) were also included to provide ties to this network.

Table 1. Station Occupation During Different GPS Surveys in the Katla (Mýrdalsjökull) and Eyjafjallajökull Areas^a

Abbreviation	Name	Inscription	Approximate Latitude, Longitude	GPS Campaigns
				1992 Aug.	1992 Sept.	1993 June	1994Aa May	1994Ab June	1994B Sept.	1995 Sept.	1998 July	1999 July	1999A Aug.	1999B Sept.	1999C Nov.	2000A Feb.	2000B July
Mýrdalsjökull
ALFT	Álftagróf	NE9213	63°29′22″, 19°10′38″		X	X			X				X		X		X
AUST	Austmannsbunga	NE9304	63°40′27″, 19°04′50″			X								X			X
BULA	Búland	NE9411	63°48′09″, 18°33′34″														X
EINH	Einhyrningur	OS7385	63°45′13″, 19°27′02″							X							X
ELDH	Eldhraun	OS5847	63°41′05″, 18°21′26″	X		X											X
ENTA	Enta	NE9305	63°42′04″, 19°10′56″			X	X							X			X
FIMM	Fimmvörðuháls	NE9203	63°36′24″, 19°26′15″		X				X		X	X					X
HOFD	Höfðabrekkuheiði	704	63°30′31″, 18°52′13″		X	X			X			X			X		X
HRIS	Hríshóll	NE9202	63°27′38″, 18°52′38″		X	X			X						X		X
JOKV	Jökulkvísl	746	63°40′47″, 18°46′56″			X	X						X		X		X
KJAL	Kjalnatær	NE9306	63°36′58″, 18°29′39″			X	X						X				X
KOTL	Kötlukriki	726	63°37′35″, 18°48′55″	X		X	X					X			X		X
MAEL	Mælifell	OS7416	63°48′06″, 18°57′56″							X							X
REYN	Reynisfjall	OS7377	63°25′06″, 19°01′38″	X	X	X			X	X	X		X		X		X
RJUP	Rjúpnafell	NE9303	63°37′08″, 18°37′57″			X	X			X			X		X		X
SKER	Sker	NE9201	63°33′11″, 18°56′12″		X								X				X
SKOG	Skógaá	OS7486	63°34′35″, 19°26′43″	X	X				X		X	X					X
SNAE	Snæbýli	NE9907	63°44′11″, 18°37′57″												X		X
SOHH	Sólheimaheiði	NE9214	63°32′54″, 19°15′29″		X	X	X					X		X	X		X
SOLH	Sólheimar	NE9215	63°30′26″, 19°18′19″		X	X	X			X			X		X		X
THRA	Þrætutangi	NE9527	63°49′14″, 19°11′53″							X							X
Eyjafjallajökull
DAGM	Dagmálafjall	NE9420	63°37′53″ 19°49′29″								X		X		X		X
HAMR	Hamragarðar	OS7487	63°37′20″ 19°59′08″	X		X		X	X		X	X	X	X	X	X	X
HVAM	Hvammur	VRH7601	63°34′22″ 19°52′38″												X	X	X
MOLN	Moldnúpur	NE9908	63°34′03″ 19°47′33″												X	X	X
MORK	Miðmörk	NE9909	63°39′25″ 19°53′40″												X	X	X
SELJ	Seljavellir	NE9404	63°33′45″ 19°37′57″					X	X		X		X		X	X	X
STEI	Steinsholt	NE9405	63°40′37″ 19°36′31″					X	X		X		X		X		X
Continuous Stations
HVOL	Láguhvolar	NE9906	63°31′35″ 18°50′51″												X		X
SOHO	Sólheimaheiði	NE9905	63°33′09″ 19°14′49″											X	X		X
THEY	Þorvaldseyri	NE0001	63°33′41″ 19°38′36″													X	X
Connecting the Network to ISNET
BOTA	Botnar LMÍ	LM0351	63°39′22″ 18°14′47″														X
REYF	Reynisfjall LMÍ	LM0352	63°25′08″, 19°01′35″

• a A location map with station abbreviations is shown in Figure 2. The August 1992 and September 1995 surveys were subsets of regional surveys of south Iceland [Sigmundsson et al., 1995]. Stations around Katla, stations around Eyjafjallajökull, continuous stations (see http://hraun.vedur.is/ja/gps.html), and ISNET stations in the Iceland Geodetic Survey network are given.

[15] Two continuously recording GPS receivers were installed in response to the July 1999 unrest at Katla. The two sites are HVOL (Láguhvolar) and SOHO (Sólheimaheiði), both installed in the autumn of 1999 (Figure 2). As elevated seismicity and crustal deformation continued throughout the autumn of 1999 on the south slopes of Eyjafjallajökull, a permanent continuously recording GPS receiver was installed at THEY (Þorvaldseyri) in May 2000 (Figure 2). After this receiver was installed the unrest terminated. The permanent GPS stations are listed in Table 1. On-line results from the continuous GPS sites are available at the Icelandic Meteorological Office home page, http://www.vedur.is/ja/gps.html.

5.2. Processing

[16] GPS data were processed with the Bernese GPS software package [Beutler et al., 2000], versions 3.5, 4.0 and 4.2. The data were collected at 15 s intervals during three 8 hours sessions at each site. Precise orbit information from the Centre for Orbit Determination in Europe (CODE) was used, resulting in coordinates in the International Earth Rotation Service (IERS) Terrestrial Reference Frame (ITRF). We used the Saastamoinen tropospheric refraction model to estimate site tropospheric values, but no meteorological data were collected at the GPS sites. Station coordinates were initially estimated using the ionosphere free linear combination (L3). Next, the software was used to resolve the wide lane (L5) ambiguities, using the previous L3 coordinate solution as a constraint. The fixed wide lane ambiguities and the ionosphere free (L3) linear combination were used to solve for the narrow lane ambiguities. Subsequently, a final coordinate solution was produced for each measurement session. The results (coordinate and covariance files) were then combined into one multisession solution by using a statistically correct combination of the single-session solutions, using the DYNAP (Dynamic Adjustment Program) software from the National Geodetic Survey, USA. In the processing, the HAMR (Hamragarðar) station was used as a reference station (Figure 2).

6. Observed Deformation and Timing of Events

6.1. Katla

[17] Deformation studies from 1967 to 1999 revealed minor deformation around Katla. Tryggvason [2000] gives an overview of results from optical leveling tilt stations between 1967 to 1995 and concludes that ground tilt related to the Katla volcano has not been observed. Before 1999, GPS measurements also showed insignificant crustal deformation attributable to Katla.

[18] On 18 July 1999, a small jökulhlaup occurred in river Jökulsá á Sólheimasandi in connection with the formation of a new ice cauldron in the meltwater catchment of Sólheimajökull (Figure 3). A number of previously existing ice cauldrons, 10–30 m deep and 300–400 m wide, deepened in the followings weeks [Guðmundsson et al., 2000]. The 1999 jökulhlaup was preceded by increased earthquake activity within the Katla caldera and bursts of volcanic tremor were recorded in the hours before the jökulhlaup burst issued from Sólheimajökull [Einarsson, 2000]. A shallow intrusion or a minor subglacial eruption was the probable cause, however no eruption column broke through the ice. As a response to this event tilt measurements were performed at FIMM (Fimmvörðuháls) on 19 July. These measurements indicated an upward tilt of 3.5 μrad in the direction 60° (i.e., approximately the direction toward Katla) since previous measurements in the summer of 1998 (Figure 4b). GPS measurements were also performed (19–23 July) at some of the station around Katla (Table 1). The measured GPS points showed little or no deformation. Comparison of the June 1993 and November 1999 surveys in the area south and southeast of Katla show that all the stations moved almost uniformly relative to the reference station HAMR (Hamragarðar) during this period (Figure 5). There is only small amount of internal deformation between the stations (excluding the reference station) as they have a near uniform horizontal displacement. It is therefore concluded that the relative displacements are caused by a westward displacement of the reference station. Time history of horizontal displacement of the GPS station at REYN (Reynisfjall) relatively to the station at HAMR (Hamragarðar) is displayed in Figure 6.

Figure 5

Horizontal displacements spanning the period from June 1993 to November 1999 relative to the HAMR station. This displacement pattern was “typical” for GPS stations south and east of Katla (Mýrdalsjökull) prior to the June 2000 earthquakes in the South Iceland Seismic Zone (SISZ). Measurements within the 1993 to 1999 period show that displacements increased with time, such that the general direction and the uniform pattern of deformation persisted.

Figure 6

Time series plot of east and north displacement components of GPS station REYN (Reynisfjall) relative to HAMR (Hamragarðar). Step-like eastward extension correlates with the crustal deformation episodes in Eyjafjallajökull. The June 2000 earthquakes dissipated the east-west crustal dilation that had accumulated since 1992.

[19] The two GPS points at the caldera rim (ENTA and AUST, Figure 2) are most sensitive to magmatic movements in Katla. They are not measured regularly because of logistical difficulties. The ENTA and AUST stations were measured in 1993 for the first time, and then remeasured in September 1999 and June 2000. The September measurements indicated a northward displacement of the GPS point at ENTA (Enta), i.e., outward from the caldera. No useful data were collected at AUST (Austmannsbunga) in this survey. Measurements in June 2000 (Figure 7) confirmed that a local source caused considerable part of the displacements of the GPS point at the caldera rim. However, the timing of the inflation in the Katla caldera is not precisely resolved as the measurements span the interval 1993 to 2000.

Figure 7

Horizontal deformation field in the time period from June 1993 to July 2000. The displacement pattern changed after the June 2000 earthquakes in the South Iceland Seismic Zone (see Figure 6 for reference). All displacements are relative to the reference station at HAMR (Hamragarðar). HAMR is located within the deformation field of the June 2000 earthquake, in a compressive quadrant of the north-south right-lateral strike slip that occurred. Southeast motion of HAMR is inferred. The two GPS points on the northern edge of the Katla caldera (under Mýrdalsjökull) show a larger northward displacement than the “regional” pattern of the stations to the east of the caldera. This larger amount may be attributed to some local extension across the Katla caldera.

6.2. Eyjafjallajökull

[20] The 1994 crustal deformation episode that took place in Eyjafjallajökull is bracketed from tilt measurements at FIMM (Fimmvörðuháls) to have occurred between 29 September 1993 and 19 September 1994 (Figure 4b). Then an upward tilt of 12.4 μrad in the direction 266° was measured (Figure 4b). An earthquake swarm was recorded between 29 May and 22 June. GPS measurements spanning from 1992 to September 1994 covered the whole 1994 episode, but the network was limited and only includes the SKOG (Skógaá) GPS station. The network was expanded during the earthquake activity in June 1994. Using HAMR (Hamragarðar) as reference the displacement at SKOG is 5.6 cm at 113°. Uplift of 2.9 cm and horizontal displacement of 2.4 cm in the direction 220° were recorded from June to September 1994 by GPS measurements at SELJ (Seljavellir). On the other hand, the GPS station at STEI (Steinsholt) during the same period showed insignificant horizontal displacement and a vertical subsidence of 2.2 cm.

[21] The measurements spanning the period September 1994 to summer 1998 yielded insignificant amounts of horizontal displacements around Eyjafjallajökull. The timing of the beginning of the 1999 crustal deformation event is well constrained. A GPS survey 16–21 August 1999 (Table 1) revealed post-1998 uplift at the SELJ (Seljavellir) station of 5.7 cm and a horizontal displacement of 4.4 cm in direction 198°. The tilt station at FIMM (Fimmvörðuháls) was remeasured 16 October 1999 and this point showed an upward tilt in the direction toward Seljavellir (256°). Also the tilt station at DAGM (Dagmálafjall) showed uplift in the direction toward Seljavellir.

[22] In early February 2000 the GPS network around Eyjafjallajökull was partly remeasured. These measurements showed that the deformation rate at SELJ (Figure 4c) had declined slightly since November 1999, and repeated measurements in July 2000 confirmed this pattern. Subsidence occurred at the STEI (Steinsholt) station from November 1999 to July 2000 (Figure 4c), but this station was not measured in February 2000. The bench mark of the GPS station at THEY (Þorvaldseyri) was installed in February 2000 and measured in that campaign, while the continuous measurements started in May. At the time the continuous GPS station at THEY (Þorvaldseyri) became operational in May 2000, crustal deformation ceased. Our measurements show that the 1999–2000 deformation episode at Eyjafjallajökull started sometime between 19 July and 20 August 1999 and ended sometime in the period from February to early May 2000. Figure 8 shows the total amount of vertical and horizontal displacement of the 1999 event, the duration of which was about seven to eight months.

Figure 8

Observed horizontal and vertical displacements during the period July 1998 to July 2000 for the GPS network around Eyjafjallajökull. The black arrows show horizontal displacement, and the open arrows show vertical movements. The shaded arrows from station FIMM and DAGM show the tilt. The star denotes a best fit point source location for the 1999 episode, based on the horizontal displacements. The suggested center of uplift of the 1994 event is indicated with a plus.

7. Modeling

[23] For the 1999–2000 event at Eyjafjallajökull, the GPS stations at SELJ, FIMM, and SKOG show clear displacements (Figure 8). Tilt measurements show that the deformation episode was associated with upward tilt (14.5 μrad) in direction 256° at the FIMM tilt station. The tilt station at DAGM recorded an upward tilt (5.8 μrad) in the direction 145°. The displacement and tilt vectors point toward (or away) from one common area south of the summit crater (Figure 8). This deformation was modeled both with a Mogi point source [Mogi, 1958] and a dike source using the Okada [1985] formulae.

[24] We used forward modeling and searched for the best fitting Mogi point pressure model within a grid of latitude, longitude, depth and maximum uplift, to fit the data shown in Figure 8. GPS data from the STEI station were, however, excluded, as local movements of this bench mark is suspected (see below). In the grid search we used a least squares criterion minimizing the χ² merit function [Bevington, 1969].

where res_i is the residual between observations i and model prediction and σ_i is the standard deviation of observation i. In our case the number of observations, N, is 16 (four three-dimensional GPS displacement vectors, and two two-dimensional tilt vectors). Uncertainties of horizontal GPS displacements were estimated 6 mm, and 8 mm for the vertical. Uncertainties on the tilt components were estimated to be 4 μrad. The forward search resulted in χ_min² equal to 35. For comparison, a null model would give χ² = 1005. The best fitting Mogi model is 19.60° ± 0.02° W and 63.583° ± 0.005° N, at a depth of 3.5 ± 0.6 km. Maximum uplift directly above the Mogi source is 0.36 ± 0.20 m (Figure 9). Reported uncertainties correspond to the parameter space where χ² increases by a factor of two from the minimum value. The χ_min² value depends on the fit of the model and the uncertainty; a larger uncertainty will result in a lower χ² value. The relatively high χ_min² value probably reflects the combination of a too optimistic uncertainty estimate and the fact that the Mogi model does not account for all the observed deformation.

Figure 9

Observed horizontal displacements (open arrows) for the period July 1998 to July 2000 and calculated displacement for the best fitting model (solid arrows). The vectors point outward from an area on the south slopes of Eyjafjallajökull, indicating inflation. The best fitting Mogi source gave an estimate for the displacement of the reference station HAMR, resulting in the vector pattern. The “best fitting” Mogi model gives a displacement vector of 1.1 cm at 282° for the reference station, with the parameters, h₀ = 0.35 m (the vertical displacement h₀ over the point source) and depth is 3.5 km. Star denotes best fit point source location for the 1999 episode.

[25] The vertical vector at the Steinsholt point contradicts the inflation trend, as it signifies subsidence. Including the STEI station in the grid search would lead to a shallower depth (3.2 km) for the inflation center but at the same horizontal location. In the search for the best fitting parameter, data from STEI station were ignored because it seems influenced by a different deformation field. The tilt and the GPS data are independent of each other and both methods show uplift located in the same area. We therefore consider it likely that the deviation at STEI is of a very local origin eventually reflecting instability of the site or due to local faulting, as the bulk of the seismicity occurs close to the station. Upward tilt in the direction toward the south edge of Eyjafjallajökull is observed at tilt stations FIMM and at DAGM. The two tilt vectors suggest a horizontal location about 700 m from that obtained from all data. The tilt data alone suggest a deeper source, at 4.0 ± 0.5 km, and maximum uplift of 0.6 ± 0.2 m. Considering uncertainties, this solution is, however, not significantly different from the optimal one based on both GPS and tilt.

[26] When fitting a dike model to our observation, we did not conduct a grid search because of the high number of unknown variables and our few data points. Rather, we relied on field evidence of eroded dikes in the area, and tested forward models against the observed data. Loughlin [1995] mapped 118 dikes on the south side of Eyjafjallajökull. The area with the highest density of eroded dikes is in the valleys and gorges above Seljavellir and Þorvaldseyri farms (Figure 3). The vast majority of the observed dikes are 0.1–0.7 m wide, and about 50% of them have a dip of 90°. The distribution of dike strike is bimodal with the bulk of the observations in a bell-shaped distribution with the center value of 45°, and the second peak (small and narrow) at 180°. Using a selection of these parameters we tested several dike geometries and concluded that we did not obtain any fit close to the fit obtain by the Mogi point pressure source model.

[27] Dahm and Brandsdóttir [1997] suggested on basis of focal mechanism solutions a vertical intrusion beneath the northern flank of the Eyjafjallajökull volcano during the 1994 unrest. They used a numerical simulation of a stress field generated by a vertical, 10-km-long east-west striking dike, in order to explain the earthquake focal mechanisms and distribution. The average thickness of the dike is 5 m, and it is 10 km high [Dahm and Brandsdóttir, 1997, Figure 11]. Using these parameters in a dike model gives a deformation field that bears little resemblance to the measured displacements. The model displacements are four to five times larger than those observed in 1994 and 1999.

8. Discussion

[28] Crustal deformation was observed at both Eyjafjallajökull and Katla volcanoes during the 1999–2000 period of unrest. In the Katla network most of the GPS points and all tilt stations are located at a distance of 9 km or more from the caldera center. Two GPS points (AUST and ENTA in Figure 2) are situated on or near the caldera rim of Katla. The displacement vectors at those two GPS points over the period 1993 to the summer 2000 strongly indicate magma emplacement at shallow depth under Katla.

[29] Our set of GPS measurements makes it possible to estimate a 1992–2000 time series of the relative displacement between the GPS stations at HAMR and REYN (Figure 6). These two stations span the 55-km-wide southward continuation of the EVZ into the Eurasian plate. The horizontal deformation pattern between the two GPS stations is dominated by an east west extension that significantly exceeds the uncertainty (Figure 6). The two steps in the east west displacement correlate with the 1994 and 1999 events in Eyjafjallajökull. Our modeling of the intrusion activity shows that the Hamragarðar station is within the deformation field of the inferred intrusion under Eyjafjallajökull and the steps can partly be explained as a consequence of small westward displacement of the reference station in response to the intrusive activity.

[30] The regional pattern as exemplified in Figure 5 remained until two Ms 6.6 earthquakes occurred in the south Iceland seismic zone in June 2000 (Figure 1). The earthquakes changed the stress field in the region after a right-lateral slip on two parallel N-S planes. Relative to a reference point in REYK (Reykjavík) the horizontal displacement of HAMR (Hamragarðar) GPS point is almost double the size the of displacement of the more eastern stations SOHO (Sólheimaheiði) and HVOL (Láguhvolar) spanning the time period of the June 2000 earthquakes. The reference station at HAMR (Hamragarðar) is affected by three transient signals, the two intrusion events and the June 2000 earthquakes.

[31] The amount of horizontal displacement (the SKOG point) and tilt (the FIMM point) of the 1994 episode seems to be slightly smaller than the 1999 episode and pointing to a slightly different location for the center of uplift. Using the SKOG point and FIMM point and also including the SELJ point the vectors radiate from an area 2–3 km north-northeast of the “best” fitting location of the 1999 uplift event (Figure 8).

[32] In 1994 the crustal deformation measurements indicate intrusion under the south slope of Eyjafjallajökull, but the earthquake swarm during this year was located beneath the northern flank of Eyjafjallajökull. Similar discrepancy was observed for the 1999 event. Then the largest earthquakes occurred before any observed crustal deformation. Figure 3 shows the location of the center of uplift 1999 (star) together with the earthquake activity. The earthquake swarm is located off the area of maximum uplift.

[33] The GPS station at STEI shows subsidence while our model suggests uplift at the station. This pattern is observed during both 1994 and 1999 inferred intrusion events. In 1994 and especially in 1999 the subsidence at STEI seems to lag behind the events at SELJ (Figure 4c). The subsidence at STEI is the largest deviation from our Mogi model. It is unclear if it is due to a local instability effect, or signifies that our deformation source model is incomplete.

[34] The crustal deformation in the Eyjafjallajökull volcano is centered outside the summit crater both in 1994 and 1999. The 1999 uplift at the south slope is confirmed by two independent methods (tilt and GPS). These two methods point to a horizontal location separated by about 700 m for the 1999 event. Our observations for the 1999 event can be fitted to a Mogi type point source. A dike model has not found to fit better than the Mogi model. The 1999 episode generated an integrated surface inflation volume of 0.027 km³. This volume is given by ΔV_e = 2πh₀d² (h₀ is the vertical displacement over the point source and d is the source depth), with a maximum uplift of h₀ = 0.35 m and a depth of d = 3.5 km (assuming a Poisson's ratio of 0.25). The corresponding subsurface volume change of the Mogi source is 2/3 of the integrated surface inflation volume, or about 0.018 km³.

[35] The event that trigged the jökulhlaup from the Mýrdalsjökull ice cap in July 1999 was most likely a minor eruption, which did not break the ice surface. An alternative explanation is intrusion to very shallow levels. It is thus shown that contemporaneous shallow magmatic movements took place at both of the neighboring Katla and Eyjafjallajökull volcanoes, culminating in 1999.

[36] The 1994 and 1999 intrusive rocks in Eyjafjallajökull are most likely of the alkalic rock suite, the “Eyjafjallajökull” type, or the regional Fe-Ti basalt type. Recent eruptions of Eyjafjallajökull released evolved alkalic magma of volumes up to about 0.1 km³, as calculated from the map by Jónsson [1998]. This may be compared with eruptions of Fe-Ti basalts, which tend to be of large volumes, up to several cubic kilometers (e.g., Skógaheidi that is about 4 km³). The modeling gives a small volume of magma and earthquake distribution is limited to the Eyjafjallajökull volcano. These circumstances point to a small magma volume, probably of the alkalic local type. The center of uplift is on the south slope of the volcano. If an eruption takes place at this location, regardless of its magnitude, farms at the base of the southern flank of Eyjafjallajökull are within the hazard zone.

9. Conclusions

[37] Intrusion events occurred in 1994 and 1999 under the southern slopes of Eyjafjallajökull. GPS and tilt data constrain the location of the 1999 center of uplift to be ∼4 km outside the summit crater edge. The 1994 center is uncertain but was probably about 3 km northeast of the center of the 1999 episode (Figure 8). Deformation associated with the two events is very similar. The modeling suggests a maximum uplift of 0.35 m in 1999.

[38] The two intrusion events were associated with elevated earthquake activity. The earthquakes are not spatially constrained to the center of uplift on the southern slope but are also to a large extent located below the northern slope of Eyjafjallajökull. Almost all earthquakes in 1994 occurred in the north; this pattern changed slightly in the 1999 episode, with some earthquakes occurring in the south.

[39] The two GPS points ENTA and AUST at the caldera rim of Katla show an outward (from the caldera center) displacement in the period 1993 to 2000. This extension across the caldera signifies magma accumulation. The GPS and tilt network at greater distance shows no coherent nor distinct pattern of deformation attributed to inflation in the center of Katla. The cause of the inflation in the Katla caldera is located at shallow depth. A burst of seismic tremor preceded a jökulhlaup from the glacier covering the Katla volcano on 17–18 July 1999. The tremor was observed at seismic stations more than 100 km from Katla, and is also taken as evidence for magmatic movements beneath Katla, even though the details of these changes cannot be discerned by the data.

Acknowledgments

[40] The authors express their gratitude to Eysteinn Tryggvason and Halldór Ólafsson for all their work in the area. The fieldwork would almost have been impossible without Halldór Ólafsson, with his great knowledge of the land. Sigrún Hreinsdóttir and Malou Blomstrand-Stinessen analyzed some of the older GPS data. We also like to thank all the many people that have participated in the collection of geodetic data during all the different campaigns, Anna Engström, Einar Pálsson, Guðmundur H. Guðfinnsson, Iikka Ylander, Peter Momme, Rikke Pedersen, Sigrún Hreinsdóttir, Sigurjón Jónsson, and Vala Hjörleifsdóttir. For fruitful discussions and information we would like to express our gratitude to Halldór Geirsson and Níels Óskarsson. Matthew J. Roberts commented on earlier versions of this paper. Comments from the reviewers, Akira Takada and Mark Simons, helped us to significantly improve the paper. We used the public domain GMT software [Wessel and Smith, 1991] to make the figures. This work was supported by a special grant from the Icelandic government and by the University of Iceland research fund, and the Icelandic Research Council RANNÍS.