Center of Icelandīs Hotspot Experiences Unrest


Páll Einarsson, Bryndís Brandsdóttir, Magnús Tumi Guðmundsson, Helgi Björnsson

Science Institute, University of Iceland, Dunhaga 3, 107 Reykjavík, Iceland


Karl Grönvold, Freysteinn Sigmundsson

Nordic Volcanological Institute, University of Iceland, Grensásvegi 50, 108

Reykjavík, Iceland



A volcanic eruption began beneath the Vatnajökull ice cap in Central Iceland (Fig. 1) in the late evening of September 30, 1996, on a 7 km long fissure located between the volcanoes Bárðarbunga and Grímsvötn. The eruption continued for 13 days and produced on the order of 0.5 km3 of basaltic andesite. Meltwater from the eruption site flowed into the caldera lake of the Grímsvötn volcano, where it accumulated beneath a floating ice shelf. The lakeīs ice dam was lifted off the glacier bed on November 4, and in the next two days more than 3 km3 of water drained out beneath the glacier and flushed down to the south coastīs alluvial plain, causing extensive damage to transportation and communication systems. This major eruption was preceded by several years of unrest, including both earthquakes and small eruptions, which collectively may signal the onset of a period of volcanic activity in this most productive area of Icelandīs hotspot.


The Iceland hotspot is centered on Central Iceland, where it overlaps with the Eastern Volcanic Zone (EVZ, Fig. 1), which represents the mid-Atlantic plate boundary. The plates are separating at an average rate of about 2 cm per year. This area is characterized by large volcanoes, Bárðarbunga, Grímsvötn, Hamarinn, Kverkfjöll and Tungnafellsjökull [Björnsson and Einarsson, 1990], and a large part of it is covered by the Vatnajökull ice cap (Fig. 1). Mapping of the subglacial topography by radio echo sounding has revealed large calderas in the Bárðarbunga, Grímsvötn and Kverkfjöll volcanoes [Björnsson, 1988]. A hint of a yet larger circular structure can

also be seen in the subglacial topography. It extends from the southern flank of Bárðarbunga and encloses Grímsvötn. Hamarinn is situated on its western rim (Fig. 2).


The Bárðarbunga and Grímsvötn volcanic systems are among the most productive systems in Iceland and have fed some of its largest recent fissure eruptions, for example, the Laki eruption in 1783. This eruption produced 12-14 km3 of basalt and was the largest lava eruption ever witnessed by man. It had a pronounced effect on the climate, particularly in Iceland where almost one-fifth of the population perished

in the resulting famine, but also in Europe and North America. The 1996 eruption offers a rare opportunity to study an eruption beneath an ice sheet, but such

eruptions were common during the Pleistocene, that is, prior to 11 000 years B. P.


Premonitory Activity and Warning


The eruption was preceded by an unusual sequence of earthquakes, beginning on September 29 at 10:48 with a magnitude 5.4 (MS) event at the northern rim of the Bárðarbunga caldera. Many similar earthquakes have occurred beneath the Bárðarbunga volcano during the last 22 years, but none had significant aftershocks, nor were they followed by magmatic activity. This time the earthquake was followed by an intense earthquake swarm, including five events with magnitude larger than 3 withinin two hours of the main quake.


Scientists of the University of Iceland notified the Civil Defense authorities as well as the scientific community of this unusual seismic activity and the possibility of eruptive activity. The seismic swarm continued throughout September 29 and 30, with increasing intensity. Hundreds of earthquakes were recorded each day including more than 10 events larger than 3 in magnitude. The earthquakes were initially located below the northwestern rim of the Bárðarbunga caldera (Fig. 2), and then, over 24 hours, migrated 20 km southward toward Grímsvötn. The earthquakes were accompanied by continuous tremor, which indicated that magma from a magma chamber in the Bárðarbunga area was being injected to the south, feeding a dyke.


Following a meeting of the Science Advisory Board of the Civil Defense Council, a public warning of a possible eruption in northwest Vatnajökull was issued on September 30. In the evening the earthquake activity near Grímsvötn decreased markedly, while activity at Bárðarbunga continued. The seismograph at Grímsvötn began recording continuous, low-amplitude eruption tremor. The sudden decrease of the earthquake activity and the onset of the eruption tremor may be taken as evidence that the predicted eruption had begun.


The Eruption


The tremor amplitude increased slowly and reached a maximum on the morning of October 1. The eruption site was discovered early that morning by an observer in an aircraft. By that time two elongate, 1-2 km wide and NNE trending subsidence cauldrons had formed on the ice surface SSE of Bárðarbunga (Fig. 3 a), on the northern flank of the neighbouring Grímsvötn volcano. The cauldron formation showed that the glacier was being melted by an eruption on a 4-km-long fissure at the base of the glacier, which was 400-600 m thick. The northern cauldron deepened some 50 m in 4 hours. The fissure was located within the drainage basin of the Grímsvötn caldera causing the meltwater from the eruption to drain into the caldera lake. A shallow linear subsidence structure was visible at the glacier surface, marking

the subglacial pathway of the meltwater draining into the Grímsvötn caldera.


The cauldrons widened and deepened during the day, and the level of the Grímsvötn lake rose by 10-15 m. About 0.3 km3 of water was added to the lake during the first 15 hours of the eruption. The vigor of the eruption could be monitored in three different ways, that is, by the volume of the depressions in the ice caused by the melting, by the volume of meltwater accumulating in the Grímsvötn caldera, and by the intensity of the volcanic tremor.


The eruption was most powerful during the first 4 days. Most of the activity was hidden below the glacier, but in the morning of October 2 an opening formed in the glacier surface, through which an eruptive column rose to 4-5 km altitude. Later that day the eruptive fissure extended some 3 km farther to the north. Ash dispersed to the north and colored the glacier surface (Fig. 3b). The opening in the glacier grew larger in the following days and the subsidence area grew to 9 km long and 3-4 km wide. An ice canyon melted along the central axis of the depression (Fig. 3 c). Water flowed southward along the canyon toward the Grímsvötn caldera. The volcanic tremor stopped on October 13, indicating that magma transport to the eruption site had ceased.


A schematic section of the subglacial eruption and the path of the meltwater is shown in Fig. 4. The length of the main eruptive fissure was 7 km, but in addition a minor subglacial eruption occurred on the SE-rim of the Bárðarbunga caldera, 6-7 km to the north. Two small depressions formed in the ice surface there.


Vatnajökull Ice Cap and Floods from the Grímsvötn Caldera Lake


Vatnajökull, the largest ice cap in the world outside the Arctic and Antarctic, is a temperate glacier, which means that all but a 15 m thick surface layer is at the melting point during the winter. The ice and bedrock topography of Vatnajökull were mapped by radio echo sounding to trace the path of meltwater at the base of the ice [Björnsson, 1988]. The ice thickness reaches 800 - 900 m in places but the average thickness is 400 m.


The glacier, which has a surface area 8 200 km2, covers some of Icelandīs highest

and most active volcanoes. These volcanoes erupted frequently throughout historic time , but few direct observations have been made.


The Grímsvötn caldera has an active geothermal system which melts ice at the rate of 0.2 - 0.5 km3/yr during normal times [Björnsson and Gudmundsson, 1993; Gudmundsson et al., 1995]. The water is contained by an ice dam that closes an outlet in the eastern caldera wall, so the 250-m-thick ice shelf that floats on the caldera lake rises about 10 - 15 m per year. Eventually, the water penetrates the ice dam, drains through a tunnel in the ice that expands as it melts and flows beneath the ice toward the coast in a catastrophic flood [Björnsson, 1988, 1992]. No observations are available to put constraints on the timescale of this penetration process.


The resulting floods, called jökulhlaups in the geological literature, normally last 2-3 weeks. Jökulhlaups from Grímsvötn have occurred every 4-6 years during the last five decades; each releases 1-2 km3 of water from the caldera lake. In earlier decades the floods were larger and sometimes followed by eruptions in the Grímsvötn caldera, which were presumably triggered by the pressure release when the water load was removed from the caldera floor. The latest event of this type occurred in 1934.


Jökulhlaups can also occur as the result of eruptions. An eruption in 1938 on the northern flank of Grímsvötn formed a subglacial ridge, which coincides with the southern part of the 1996 eruptive fissure [Björnsson, 1988, Gudmundsson and Björnsson, 1991]. The 1938 eruption was almost entirely subglacial, breaching the glacier surface shortly at the very end of the eruption. The meltwater drained into the Grímsvötn caldera, causing a large flood of about 4.7 km3 [Björnsson, 1992, Gudmundsson et al., 1995]. The peak discharge of some 30,000 m3/s was reached in about 3 days, the flood then receded slowly over the next 2 weeks.


The rate of meltwater flow during the first few days of the 1996 eruption was about 5000 m3/s and the ice shelf on the caldera lake rose 15 - 20 m per day. The usual lake level for the triggering of a jökulhlaup was reached in 4 days, but the uplift continued beyond this critical level. The rate of uplift decreased, manly for two reasons: melting slowed as the vigor of the eruption diminished and the surface area of the lake increased. The rate had decreased to one meter per day when the eruption stopped, corresponding to an inflow of 400 - 500 m3/s. By the end of October the lake level was approaching the point where ice closing the outlet of the lake would be lifted off the glacier bed. A jökulhlaup was anticipated and was expected to release more than 3 km3 of water over a few days.


The November 1996 Jökulhlaup


On November 4 the sudden appearance of a continuous, high-frequency (> 3 Hz) tremor on the Grímsfjall seismograph indicated that the ice barrier was failing. The amplitude grew steadily for the next several hours as the water front migrated 50 km downstream beneath the glacier. The water emerged at the glacier edge as a flood wave the following morning, 10 1/2 hours after it lifted the barrier at Grímsvötn. The flow rate increased steadily during the day.


The flood began in the easternmost river, Skeiðará, and several hours later the rivers farther west on the Skeiðarársandur alluvial plain were also flooding. In the afternoon the flood had reached all the rivers on Skeiðarársandur. The maximum discharge rate was the highest ever recorded, about 45 000 m3/s) on November 5. The flood ended on the morning of November 7. This is the most rapid course of events ever recorded for a jökulhlaup from Grímsvötn. The total volume of water released from the glacier is estimated at 3.5 km3.


The flood caused $15 million in damage to the transportation and communication systems of southeast Iceland. The road connection was cut when one major bridge was washed away and two others were severely damaged. About 10 km segment of the road disappeared, the main high tension power line was severed and the fiber optics cable of the telephone system broke in the flood.


The Magma and the Eruption Mechanism


Volcanic activity in Iceland is generally confineded to large central volcanoes and their associated fissure swarms. Each central volcano has episodes of unrest separated by longer periods of relative quiescence. During times of unrest magma accumulates in crustal magma chambers, often followed by episodic lateral migration of the magma away from the magma chamber into dikes along the fissure swarm, as observed at Krafla in the northern rift zone from 1975 to 1984 [Einarsson, 1991].


The ash produced in the current eruption is almost aphyric glass with only minute amount of crystals, plagioclase, augite, olivine, and magnetite. Microprobe analysis of the chemical composition shows the glass is basaltic andesite. The composition varies significantly, SiO2 is in the range 51.8 - 54.2 % and MgO in the range 2.5 - 3.7 %. Samples tha were richest in MgO erupted earliest. The lack of suitable samples from previous subglacial eruptions makes it difficult to characterize the different subglacial volcanoes. It is clear, however, that the 1996 eruptive products are different from those of Grímsvötn, as defined by the products of the recent eruptions of 1922, 1934, and 1983. They are also different from typical Bárðarbunga products, that generally contain MgO in the range 7.2 - 8.8 %.


Since the 1996 eruption does not have the chemical characteristics of the main magma systems of Grímsvötn or Bárðarbunga. We suggest it was fed by a subsidiary system producing relatively evolved magma. The migration of seismic activity from Bárðarbunga to the eruption site and the intrusion tremor at the beginning of the present activity strongly suggest, that the eruption was fed by an intrusion from a magma chamber underneath the Bárðarbunga volcano. Most likely this was a small chamber, not the main magma chamber of Bárðarbunga.


Previous Activity of Bárðarbunga


The Bárðarbunga volcanic system erupted from 1697 to 1720, in 1766, 1769, and from 1862 to 1864 [Björnsson and Einarsson, 1990]. The large fissure eruptions of Vatnaöldur in 871 and Veiðivötn in 1480 that occurred up to 100 km SW of Bárðarbunga were fed by the Bárðarbunga volcanic system. The Bárðarbunga caldera and its surrounding are covered by glacial ice that is 900 m thick in places. Melting of this ice during eruptions could cause catastrophic floods much larger than those of historic times. The caldera is thus a likely source of prehistoric catastrophic jökulhlaups at about 7,100 B.P., 4,600 B.P., 3,000 B.P., and before 2,000 B.P., that cut up to 100 m deep glacier-river canyons in northern and northeastern Iceland [Björnsson, 1988]. The eruption in 1938 on the north flank of Grímsvötn was followed by an unusually long, relatively quiet period of the Vatnajökull volcanoes.


The eruption of September - October 1996 marks the end of this quiet period. On a shorter timescale the eruption was preceded by a series of seismic and magmatic events in the Vatnajökull area that possibly began as early as 1974 and escalated in the year before the eruption. The events began in June 1974 with a large earthquake (mb = 5.1) at Bárðarbunga, the first in a series of such events that still continues. An eruption occurred within the caldera lake of the Grímsvötn volcano in May 1983, which may have been followed by a small subglacial eruption in August 1984 in the caldera as well. A jökulhlaup in November 1986 from a subglacial geothermal area east of Hamarinn and NW of Grímsvötn was followed by a distinct tremor episode, presumably a small eruption, triggered by the pressure release [Björnsson and Einarsson, 1990].


Jökulhlaups from the same geothermal area in August 1991 and July 1995 were also followed by presumed eruptions. An intense earthquake swarm occurred in February 1996 near the Hamarinn Volcano. A jökulhlaup from a second subglacial geothermal area NW of Grímsvötn in August 1996 was followed by a tremor episode thought to indicate an eruption. This event was followed by increased seismicity during the next few weeks. The historical record of high volcanic activity in Vatnajökull, the unusually long recent quiescence of the area, and the escalating trend of

the present events suggest that we have entered a new period of unrest in

this central area of the Iceland hotspot. The unrest is not limited to one

volcanic system. Both Bárðarbunga and Grímsvötn are involved, as well as

the area between Grímsvötn and Hamarinn. Continued magmatic activity in

these areas in the coming months and years must be considered rather

likely. The activity in this part of the volcanic zones in historical time

is characterized by relatively small but frequent eruptions. Keeping in

mind, however, the potential of the Vatnajökull volcanoes to produce

catastrophic events of global importance, such as the Laki eruption of 1783

from the Grímsvötn volcanic system, there is every reason to worry when

they become restless.

Acknowledgments: This paper benefitted greatly from constructive criticism

by M. Garcia and S. Self. Finnur Pálsson gave invaluable help with figures

and data.




Björnsson, H., Hydrology of ice caps in volcanic regions. Soc. Sci.

Islandica, 45, Reykjavík, 139 pp and 21 maps, 1988.


Björnsson, H., Jökulhlaups in Iceland: prediction, characteristics and

simulation. Annals of Glaciology 16, 95-106, 1992.


Björnsson, H., and P. Einarsson, Volcanoes beneath Vatnajökull, Iceland:

Evidence from radio-echo sounding, earthquakes and jökulhlaups. Jökull, 40,

147-168, 1990.


Björnsson, H., and M.T. Gudmundsson, Variations in the thermal output of

the subglacial Grímsvötn caldera, Iceland. Geophys. Res. Lett., 20,

2127-2130, 1993.


Einarsson, P., Earthquakes and present-day tectonism in Iceland.

Tectonophys., 189, 261--279, 1991.





Gudmundsson, M.T., H. Björnsson, Eruptions in Grímsvötn, Vatnajökull,

Iceland, 1934 - 1991. Jökull, 41, 21- 45, 1991.


Gudmundsson, M. T., H. Björnsson, and F. Pálsson, Changes in jökulhlaup

sizes in Grímsvötn, Vatnajökull, Iceland, 1934-91, deduced from in-situ

measurements of subglacial lake volume. J. Glaciology, 41, 263-272, 1995.





Fig. 1. Tectonic map of Iceland showing fissure swarms along the plate

boundaries, also volcanoes and calderas in the Eastern Volcanic Zone (EVZ).

Letters mark the volcanoes Krafla (K), Tungnafellsjökull (T), Bárðarbunga

(B), Kverkfjöll (Kv), Hamarinn (H), Grímsvötn (G), and the Skeiðarársandur

alluvial plain (S). The central area of the Iceland hotspot is defined by

the volcanoes Bárðarbunga, Grímsvötn, Kverkfjöll and Tungnafellsjökull.

Glaciers are shown with grey shading. The box shows the area of Fig. 2.


Fig. 2. Map of the NW-Vatnajökull area, showing bedrock topography mapped

by radio echo sounding, epicenters of earthquakes September 29 - 30 (dots),

and the eruption fissure of 1996 (bent line). The calderas of the

Bárðarbunga and Grímsvötn volcanoes are clearly seen in the topography, as

well as a hint of a larger circular structure (marked with a dashed line).

The seismic stations of Vonarskarð and Grímsfjall, shown with stars,

provided invaluable data on the seismicity accompanying the eruption and

the ensuing flood. The thin line in the NW corner is the glacier margin.


Fig. 3. Photographs of the eruption. a) Subsidence cauldrons forming above

the subglacial erupting fissure 14 hours after the start of the eruption.

The diameter of the cauldrons is about 2 km, and the depth about 100 m. b)

Aerial view from the north showing the subsidence cauldrons above the

eruptive fissure on the third day of the eruption. Most of the 7 km long

fissure is erupting under the ice, but one crater near its center is

erupting through the ice. The Grímsvötn caldera is seen in the background,

the Öraefajökull intraplate volcano (about 2000 m a.s.l.) in the distance.

The glacier surface in the foreground is colored by light ash fall. c) The

ice canyon on the 12th day, with crater rim emerging from the meltwater



Fig. 4. Vertical section along the eruptive fissure and through the

Grímsvötn caldera, showing conditions at the beginning of the eruption and

at its end. The ridge formed in 1938 is shown and a schematic

representation of the new volcanic construct. The meltwater flows at the

bottom of the glacier and crosses a ridge on its way to the Grímsvötn

caldera lake. It is driven by a gradient in a potential, which is mainly

determined by the surface topography of the ice. The bedrock topography

only has a small effect.