The Nordic Volcanical Institute reported that from late in the evening of 30 September until 13 October a subglacial eruption occurred along part of the East Rift Zone that traverses beneath the NW side of Vatnajökull, Europe's largest continental glacier (Björnsson and Einarsson, 1991; Björnsson and Gudmundsson, 1993). This part of the Rift Zone includes both Bardarbunga and Grímsvötn fissure systems and their respective central volcanoes, each containing a substantial caldera (figure 1).

Figure 1. Area map showing the erupting fissure and recent seismicity along the East Rift Zone in the Grímsvötn-Bardarbunga region. Shaded regions indicate exposed land surface, unshaded regions indicate glaciers; ice-surface contour values are undisclosed. The solid sub-circular lines depict the larger extents of the named central volcanoes; hachured lines indicate the respective caldera topographic margins. Dots show earthquake epicenters for 29 September-2 October. Balloons depict available earthquake fault plane solutions for some events over M 4. Courtesy of the Icelandic Meteorological Office.

The eruption was preceded by an unusual sequence of earthquakes. One, at 1048 on 29 September, was Ms 5.4 and centered near Bardarbunga caldera's N rim (figure 1). Similar earthquakes have occurred beneath Bardarbunga many times during the last 22 years. Unlike this event, however, none of the previous large earthquakes had either significant aftershocks or preceded magmatic activity.

In the two hours following the M 5.4 event there were numerous earthquakes, including five larger than M 3. These were recorded at the two analog seismic stations just NW of Bardarbunga and at the S rim of the Grímsvötn caldera. Shortly after 1300 on 30 September, Science Institute seismologists informed Civil Defense authorities and the scientific community about this unusual seismicity and the possibility of impending eruptive activity.

The seismic swarm continued throughout 30 September, with increasing intensity. Hundreds of earthquakes were recorded each day, including over 10 events larger than M 3. The earthquakes were located in the N part of Bardarbunga and migrated towards Grímsvötn. They were accompanied by high-frequency (>3 Hz) continuous tremor of the same type as was frequently observed during intrusive activity within the Krafla volcanic system during 1975-84.

The Civil Defense Council issued a warning of a possible eruption at 1900 on 30 September. Later that evening earthquake activity near Grímsvötn decreased markedly, while that near Bardarbunga continued. At about 2200 the seismograph at Grímsvötn began recording continuous small-amplitude eruption tremor. The sudden decrease in earthquake activity and the onset of tremor may be taken as evidence that an eruption began between 2200 and 2300 on September 30. Tremor amplitude increased very slowly during the next hours, reaching a maximum at about 0600 on 1 October.

The eruption site was spotted from aircraft in the early morning of 1 October. By that time two elongate, 1-2 km wide and N23E-trending subsidence bowls or cauldrons had developed in the ice surface. These bowls were located to Bardarbunga's SSE, along a fissure on Grímsvötn's N flank (figure 1). The bowls (one of which is shown in figures 2 and 3) appeared in the glacial ice above a 4-6-km-long NNE-trending fissure; ice in this location had been considered 400-600 m thick, though some later estimates put the ice thickness more precisely at 450 m. The eruption was most powerful under the northernmost bowl, causing it to subside 50 m over 4 hours.

Figure 2. A subsidence bowl developed in glacial ice on Grímsvötn's N flank., 1 October 1996. Courtesy of R. Axelsson.

Figure 3. A detail from 1 October showing inward stepping crevasses of the subsidence bowl with a fixed-wing airplane and its shadow for scale. Courtesy of R. Axelsson.

The resulting meltwater drained into Grímsvötn caldera (figure 1) raising the ice shelf above the caldera lake. The lake was covered by 250 m of ice and held in place by an ice dam. Widening and deepening of the bowls during the day added an estimated 0.3 km³ of water to the Grímsvötn lake in less than 24 hours. On 1 October a shallow linear subsidence structure extended from the eruption site to the subglacial Grímsvötn caldera lake, the surface manifestation of the subglacial pathway for water draining into Grímsvötn.

By 1 October the lake's surface had risen 10-15 m (to 1,410 m). During the first week of the eruption meltwater production was thought to be ~5,000 m³/second, but it later slowed. Glacier bursts (jökulhlaups) were thought to be likely, if not imminent. Water from Grímsvötn crater lake was expected to emerge at an outlet at the edge of the glacier ~50 km S. N-directed floods were also expected if the eruptive fissure continued to propagate N.

Helgi Torfason noted that although a previous glacier burst took place last summer (with 3,000 m³/second flow rates), the affected bridges were designed to withstand surges with meltwater fluxes 3x that size. On the other hand, a 1938 eruption, in almost exactly the same place (Gudmundsson and Björnsson, 1991) caused glacier bursts with fluxes ~5 or 6 times as large.

At 0447 on the morning of 2 October a vent on the floor of one bowl broke through the ice and the eruption began a subaerial phase. At 0800 vigorous explosive activity was observed in the crater with the eruption column rising to 4-5 km altitude. One account noted that rhythmic explosions resulted in black ash clouds rising 500 m while the buoyant eruption column rose to 3 km. In the afternoon the opening in the ice was several hundred meters wide. The eruptive fissure apparently extended 3 km farther N, because on the ice surface observers saw a new, elongated, N-trending ice cauldron. Some 2 October reports noted a steam column that rose to ~10 km altitude.

On 3 October the ice bowl over the northernmost part of the fissure had grown ~2 km since the previous day. By this time the glacier had subsided over an area 8-9 km long and 2-3 km wide. Subaerial eruptions pulsated, alternating between quiet periods and explosive activity. Ash mainly dispersed N but also SSW. The opening at the eruption site grew larger. Eruptive intensity began to decline on this day but tremor continued. A TV photographer captured footage of two lightning strikes traveling along the ash cloud that was widely shown on news reports. The water level in the vent was ~50-200 m below the original ice surface. The surface of Grímsvötn lake was at 1,460 m. Ash samples collected on this day had water-soluble fluorine contents of ~130 ppm, ~10% the amount found in Hekla ash, reducing concerns about the immediate danger to grazing animals. Initial electron microprobe analysis of the ash indicated that it was basaltic andesite in composition.

The eruption continued on 4 October. It was noted that the caldera lake was higher than at any point in this century. Poor weather intervened for the next few days, but on 7 and 9 October the eruption continued from the 9-km-long fissure; thin ash covered about half of the 8,100 km² Vatnajökull glacier. On 9 October J-M. Bardintzeff and a visiting French team saw a 4-km-high plume as well as violent phreatic ash emissions between 1230 and 1415.

On 10 October eruptive intensity appeared similar to the low levels seen since 3 October. Occasional eruptions carried black ash clouds to ~3 km and vapor with finer ash to 4 km. Minor ashfall was limited to the Vatnajökull glacier. An 11 October flight confirmed that emissions continued, but lacked rooster-tail-shaped explosions seen previously and may have declined in intensity. The eruptive crater was still water covered. Grímsvötn ice cover had bulged upward but signs of escaping water were absent. The caldera lake's total volume was estimated at >2 km³.

A Canadian Space Agency satellite radar image from 17 October was processed by Troms Satellite Station. In this image they found increased backscatter compared to earlier in the month; they suggested that this may have been due to cooler ice caused by a return to stability around the crater. In accord with this observation, on 18 October NVI announced that the eruption had apparently stopped on 13 October.

The eruption left material piled up to form a subglacial ridge; the highest part of this ridge supported an eruptive crater that reached a few to tens of meters out of meltwater at the eruptive site. Cooling eruptive materials continued to melt significant volumes of ice.

Increased CO₂ and H₂S in N-flowing river water suggested some flow of meltwater from the eruptive site. As of 18 October most of the meltwater was still directed towards the Grímsvötn caldera lake, with no signs of the awaited glacier burst. GPS measurements in October documented the lake's rise on the 12th (1,500 m), 15th (1,504 m), and 17th (1,505 m). Glacier bursts from the crater lake have typically occurred at the much lower lake level of ~1,450 m.

The recent eruption was a continuation of geophysical events in the Vatnajökull area that began in 1995 and possibly earlier. In July 1995 and August 1996 there were glacial floods from subglacial geothermal areas NW of Grímsvötn. In both cases, after the water reservoir drained, distinct tremor episodes occurred. Presumably, these pressure releases triggered small eruptions. In February 1996 there was an intense, week-long earthquake swarm centered on Hamarinn volcano (figure 1).

Besides the prospect of glacier bursts, the eruption was watched closely because the 1783-84 Laki (Skaftár Fires) and 1783-85 Grímsvötn eruptions vented on the Rift Zone within ~70 km of the current eruption. The 27-km-long Laki fissures active in 1783-84 start ~40 km SW of Grímsvötn's center. The Laki eruption produced 14.7 ± 0.1 km³ of basaltic lavas (Thordarson and Self, 1993) making it the largest known lava eruption in history. Sulfur and other gases released produced an acid haze (aerosol) that perturbed the weather in Western Eurasia, the North Atlantic, and the Arctic. An estimated 9,350 Icelanders died in the "haze famine" from 1783-86, an interval that included two severe winters, crop failures, livestock and fish deaths, and various illnesses, including fluorine poisoning (Stothers, 1996).

References. Björnsson, H., and Gudmundsson, M.T., 1993, Variations in the thermal output of the subglacial Grímsvötn caldera, Iceland: Geophysical Research Letters, v. 20, p. 2127-2130.

Björnsson, H., and Einarsson, P., 1991, Volcanoes beneath Vatnajökull, Iceland: evidence from radio-echo sounding, earthquakes and jökulhlaups: Jökull, v. 40, p. 147-168.

Gudmundsson, M.T., and Björnsson, H., 1991, Eruptions in Grímsvötn, Vatnajökull, Iceland, 1934-1991: Jökull, v. 41, p. 21-45.

Stothers, R.B., 1996, The great dry fog of 1783: Climatic Change, Kluwer Academic Publishers, v. 32, p.79-89.

Thordarson, T., and Self, S., 1993, The Laki (Skaftár Fires) and Grímsvötn eruptions in 1783-1785: Bulletin of Volcanology, Springer-Verlag, v. 55, p. 233-263.

Further Reference. Worsley, P., 1997, The 1996 volcanically induced glacial mega-flood in Iceland - cause and consequence: Geology Today, Blackwell Science, Ltd., v. 13., no. 6, p. 222-227.

Geological Summary. Grímsvötn, Iceland's most frequently active volcano in recent history, lies largely beneath the vast Vatnajökull icecap. The caldera lake is covered by a 200-m-thick ice shelf, and only the southern rim of the 6 x 8 km caldera is exposed. The geothermal area in the caldera causes frequent jökulhlaups (glacier outburst floods) when melting raises the water level high enough to lift its ice dam. Long NE-SW-trending fissure systems extend from the central volcano. The most prominent of these is the noted Laki (Skaftar) fissure, which extends to the SW and produced the world's largest known historical lava flow in 1783. The 15 km3 basaltic Laki lavas were erupted over 7 months from a 27-km-long fissure system. Extensive crop damage and livestock losses caused a severe famine that resulted in the loss of one-fifth of the population of Iceland.

Information Contacts: Nordic Volcanological Institute (NVI), Grensásvegur 50, 108 Reykjavík, Iceland (URL: http://nordvulk.hi.is/); Páll Einarsson, Bryndís Brandsdóttir, Magnús Tumi Gudmundsson, and Helgi Björnsson, Science Institute, Dunhagi 3, 107 Reykjavík, Iceland (URL: https://www.hi.is/); Icelandic Meteorological Office, Geophysics Department, Reykjavík, Iceland (URL: http://en.vedur.is/); J-M. Bardintzeff, Lab. Petrographi-Volcanologie, bat 504, Universite Paris-Sud, 91305 Orsay, France; Helgi Torfason, National Energy Authority, Grensasvegur 9, 108 Reykjavík, Iceland; Tromsø Satellite Station, N-9005, Tromsø, Norway; R. Axelsson, Morgunbladid News (photographer), Reykjavík, Iceland.