October 12, 1918 was just another Saturday. On the fields of France, the Hundred Days Offensive was turning a war which would now soon end, but leave 17 million dead; the devastation would set the scene for an even worse war to come. Far away, Puerto Rico lay in ruins after a magnitude-7 earthquake on the previous day: the damage had been compounded by a 5-meter tall tsunami arriving just minutes after the main shock. In Minnesota on that Saturday, 450 people died in large forest fires which had been set off by sparks from passing trains. And back in Europe, what was called the American flu, also known as Flanders flu (and to be renamed Spanish flu in order not to antagonize the American soldiers), was in its deadly second wave. A children’s rhyme had become popular: I had a little bird\\ its name was Enza \\ I opened the window,\\ and in-flu-enza. In Iceland, everything seemed calm – the flu would not arrive here for another week (on the same day as the independence referendum). The pandemic would soon turn Reykjavik into a ghost town, but that was still weeks away. And October 13 would be a day of darkness in Reyjavik.

Vik is a small, ancient fishing town on the south coast, just west of the ‘land between the mountains’ which we wrote about before. The name means ‘bay’ – because it was the oldest, no further detail was needed, unlike those lesser ones called Grindavik or Reykjavik. Vik has the reputation of being both the warmest and wettest town in Iceland. Its climate is attributed to the influence of the Gulf stream – it is, after all, the southernmost town on the main land. The nearby beach of black sand, Reynisfjara, is famous. You may have seen it in the Rogue One movie. Out at sea, a series of rock stacks show the remnants of an old cliff, now largely removed by the stormy post-ice-age seas. Three of the stacks form the Reynisdrangar, appearing as (and reported to be) three frozen trolls. Invariably, black-sand beaches (and trolls) are a sign of danger. Indeed, 20 kilometer to the north lies Katla: the town was well familiar which its infrequent eruptions (perhaps twice per century) and its somewhat more frequent jokulhlaups, although to some degree protected from both. Katla, of course, is the source of that black, basaltic beach.

Guðrún Larsen (see references) recounts a story of that Saturday, more than a century ago:

Lilja Tómasdóttir (1906–1973), was 12 years old in 1918 and lived with her parents in Vík. On the morning of October 12 she had been washing clothes with her mother (Margrét Eiríksdóttir, 1867–1950) by the small river that runs through the village. At about 11:30 AM she noticed that there were small ripples on the surface of the water in the tub they were using. Her mother then said: “Guð hjálpi okkur, ætli Katla sé að koma?” (God help us, could this be Katla coming?) At 3 PM they had lunch. As the family sits by the table, a man comes and shouts: “Katla er að koma” (Katla is coming). Lilja then ran up the slope above the house and looking east saw the jökulhlaup burst forth carrying large blocks of ice, west of Hjörleifshöfði, 8–10 km away (Larsen’s source was Erlingur Ísleifsson, son of Lilja, 30th of January 2001).

In between, people in Vik had felt a stronger earthquake at 1pm , followed by half an hour of small earthquakes, probably tremor. (Another report from that day gives 1 hour duration for the tremor.) Around this time, people in Mýrdalsandur started hearing the sound of rushing water. And at 3pm, a large cloud quickly rose from the glacier, seen by people in Vik, in Mýrdalur (the ‘boggy valley’ just west of Vik) and at Hjörleifshöfði (the outcrop just north of the label ‘Kötlutangi’ on the map). The cloud was white at first, with the lower part black. One description gives it the shape of an inverted funnel above a large horizontal ring. It was seen as a ‘cloud above the clouds’ from the Vestmannaeyjar islands, 65 km away. From there flashes of light were seen, something we would now identify as volcanic lightning. The photograph of the cloud below is taken from Larsen et al. (2021).

The plume was blown eastwards. By 3:30pm, ash was falling 30 km to the east-southeast of Katla, and by 8pm the ground in this region (Álftaver) was covered in black sand. During this time, as evening fell, the plume became more impressive with a red glow reflected of the lowest cloud. Lightning became incessant, being constant or at most a few seconds apart and lighting up the entire (visible) cloud. The plume reached 15 km height, but the lightning flashed upward to 20-25 km. A drawing of the height measurement done from Reykjavik is shown below, in a figure taken from Larsen et al.

In the evening, the wind became more westerly and the ash was now falling a bit further north, at Skaftártunga. By morning, 1cm of fine black ash covered the ground.

The next two days, the weather was not conducive to eruption watching, at least from a safe distance (which in the case of Katla is far outside its glacier, at least 20 km away). It was cloudy and hazy. The wind was now easterly and a thin layer of ash (less than 0.1mm on average) fell in Reykjavik as the skies there darkened. On Oct 14 there was some ashfall in the north of Iceland, but also in Vik. Ash and tephra was of course only reported from inhabited regions, mainly on the coast and not in the cente of Iceland or near Katla itself! Absence of reports does not mean lack of ash.

On Tuesday (Oct 15) the visibility improved: the cloud was now pulsing, growing higher and lower. The following days the plume was decreasing; it disappeared from view on Friday, Oct 18.

But this was not the end. On Oct 20, two plumes were visible, an eastern one and a weaker western one. It was now dark and clearly an ash cloud, and was lower than before. In the afternoon of Oct 21, the eruption intensified with eruption plumes and lightning similar to Oct 12: there had been another large explosion. Tephra fell again around northeast of Katla, less finely grained than the previous ash. This tephra was damaging: four farms in this region had to be abandoned. On Oct 22, the black ash cloud was higher than it had been before, and seemed to come from a vent further east than previously. This ash plume split into two ash columns on Oct 24, from the eastern and western vents. The cloud was blown towards Vik, which became completely dark: the reports say that windows looked as dark as the walls. Because of the danegrous lightning, the use of electricity in the town was not allowed. The ‘horrible black ash column’ was now stronger from the western vent. The ashfall lasted for 13 hours.

After this, the eruption diminished. The last tephra fell on Nov 1. The black column became ‘black smoke’ on Nov 2 and finally was described as ‘steam’ shortly after. Although some smoke was seen on Nov 10, the eruption was said to be over by Nov 4.

The most famous photos of the Katla eruption were taken towards the end, on Nov 2, when the column was less than 10 km high.These are panel ‘d’ below (Larsen et al. 2021) and the photo below (from Larsen 2010).

When the eruption had ended, the layer of ash was 1 cm thick 20 km from Katla in most directions, a bit more towards the north. Vik had 2 cm. It was thicker on the glacier, mostly 10-50cm. The total volume of tephra was a little over 1 km³.

Jokulhlaup

All Katla eruptions cause jokulhlaups along the river beds. They melt through large amounts of ice, and this has to go somewhere. Nowadays, the water escapes though a gap in the caldera wall towards the east and comes out along Kötlujökull, the eastern glacier. From there it runs off into several rivers, especially the Múlakvísl and Kúðafljót rivers and their tributaries. In the past, it would also om the other side of the caldera, but this has not happened since the Eldgja eruption in 939.

The flood that escaped the glacier around 3pm came down the Múlakvísl river. It far overwhelmed the capacity of the river bed and spread out across the Mýrdalssandur. The flow rate was a staggering 300 000 m³ /s at peak, with a mixture of meltwater, sediment and ice. (The sandur is in fact the deposits of such jokulhlaups which have deeply buried the soil below.) The flow reached the small hill on the sandur, Hjörleifshöfði, where one person, Kjartan Leifur Markússon, found himself surrounded on all sides. He estimated the speed of the flow as 20 km/hr – impossible to outrun.

Other flows came down river beds to the east. Larsen (2021) provides a description from a 1919 report:

About 15 km east of the edge of Kötlujökull a group of farmers from Álftaver was waiting at a sheep pen on the north bank of the Skálm river for their farm hands that were herding sheep to the pen. Between 1 and 1:30 PM the farmers noticed an unusual humming sound coming from the west. The sound gradually grew louder and around 3 PM the youngsters in the group were sent home for safety’s sake. Shortly afterwards it became clear that a jökulhlaup was approaching along the Skálm river. The remaining farmers hastened across the river as a greyish-black wave of flood water, 2– 3 m high, was closing in “about 10 minutes away” from the ford. Although the herding farm hands made a narrow escape, no human lives were lost and all reached safety.

Afterwards, it was discovered that this flood had overtopped the Kötlujökull and had traveled on the surface of the glacier.

A second flood came around 5pm, perhaps from water traveling below the glacier. This seemed even larger, and carried so much ice that Kjartan Markússon described it as ‘snow-covered hills’. The flood was as much as 8 km wide! At this time, the flood in the Múlakvísl had ended, as it turned out because ice had blocked the stream bed. Off-shore, an icepack formed; on-shore, icebergs as tall as 45 meters (one report says 60 meters) were left behind.

The jokulhlaup continued on and off for weeks, even after the eruption had ended. The last flooding occurred on Nov 26. By that time, parts of the sandur had been raised by 5 meters or more.

(Was the jokulhlaup a lahar? A lahar is a mudflow, where between 20% and 50% of the volume consists of solid material, not water. More than 50% makes it a debris flow. In this case, sediment accounted for some 10% of the volume. So it was not a lahar.)

The jokulhlaup left a large canyon in the glacier, 1.5 km long, 500 meter wide and 100 meter deep. The flow dropped so much sediment that the coast was pushed out by several kilometers. This formed the Kötlutangi (‘Kötlu spit’) which now stretches south from the cliffs of Hjörleifshöfði.

This caused an unexpected consequence of the jokulhlaup. Before the change to the coast line, the southernmost point of the main land was at Dyrhólaey, near Vik. Now the new spit reached further south. And even though this spit has been steadily eroding ever since, it still extends further south than Dyrhólaey. On the most recent sentinel images, I measured it as 500 meters. At the current rate of erosion, that will last another 50 years before the baton is passed back.

Precursors

Was 3pm on October 12 really the start of the eruption? The eruption had to melt through 400 meters of ice before it could burst into the open. This melt was probably what caused the tremor. The earthquake at 1pm was therefore the likely start of the eruption. Lilja Tómasdóttir’s description of the jokulhlaup shows that this had already broken through the end of the glacier at the time of the eruption cloud. This means there was only 1.5 hours between the first notable sign (ripples in a bath tub) and the major eruption! Even worse, when the cloud appeared, so did the melt water which rushed down the river valley.

So there was very little warning before the eruption, with only the slight trembling 90 minutes before the likely onset of the eruption. Is that common for Katla? Obviously, modern instrumentation might have picked up more indicators. But is the lack of notable earthquakes usual? Both Bardarbunga and Reykjanes gave us plenty of earthquake warning. Hekla did not, at least for recent eruptions.

Katla has erupted typically twice per century. There are usable reports of precursors after the 16^th century. In 1625, Þorsteinn Magnússon at Þykkvabæjarklaustur wrote (text from Páll Einarsson 2019)

. . . Anno Christi 1625 on the second day of September early in the morning, around first light, a few earthquakes were felt here in Ver, small and not large. Shortly thereafter thundering noises and din were heard so strong that the Earth moved under our feet. Because the glacier first now (after the earthquake), together with fire and water, moved and exploded. But about the time milking was finished the river nearest to the farm began flooding. . . .

This is very similar to 1918, with moderate earthquakes shortly before the jokulhaup broke from the glacier. In the next eruption, in 1660, again there were similarities. This report is from Jón Salómonsson near Vik.

Anno 1660 on November 3, in the evening towards sunset, fire was seen erupting north of Höfðabrekka near Mýrdalur in Skaftafellssýsla district, with persistent earthquake for about an hour before the fire erupted, though with interruptions. Later that night, at the end of the working day, when people wanted to rest, a glacier outburst arrived at Höfðabrekka with tremendous and heavy water current, following the course of the Kerlingardalsá river east of Fagridalur.

The eruption in 1721 had stronger precursors:

Anno 1721 on May 11, at 11 h AM an earthquake occurred, so strong that people in Mýrdalur were afraid of staying inside their houses. This earthquake was also felt in Síða, in Eyjafjöll and Fljótshlíð. […] The same day at 1 PM, heavy noises with large cracking sound were heard, followed by terrible fire, column of smoke and fumes, that issued from the glacier north of Höfðabrekka, named Kötlugjá, … At 2 PM a flood of slush with a few icebergs flowed rapidly from the glacier … At 3 PM the real jökulhlaup arrived with great haste …

This report also states that the earthquakes continued for a week, with some activity until summer. While the onset was similar, the continuation was quite different.

In 1755 we have a familiar, reliable reporter: Jón Sigurðsson, from Laki fame, who would give the fire sermon in that eruption. At this time he lived in Holt, southwest of Katla.

…The events began on Friday, October 17, shortly before noon, with strong earthquakes, like houses were rocked, that lasted till the end of the day and the following night, so that many did not dare to stay inside their houses because of the movements of the ground and worries that the houses might collapse. In the evening of the same day at sunset frequent lightning was seen, that continued throughout the night together with tremors and movements of the ground. … In the night following October 17 and the day a flood and jökulhlaup originated in the so called Kötlugjá.

The eruption apparently broke out of the icecap on the Friday evening, as evidenced by the lightning. The earthquakes continued for a number of hours before the eruption; the tremor afterwards may have been the meltwater, attempting the break through the glacier at the depth of the little ice age.

In 1823, the earthquakes were weaker:

About 6 pm an earthquake was felt here in Vík (…), a few sharp and short shocks, but not hard enough to rock houses significantly, followed by weaker but persistent movements, later mixed with harder shocks similar to the first one, until 9h PM, then all became quiet for a quarter of an hour. Then an eruption column ascended at high speed … Near midnight … a water flood began, mixed with ice slush and glacier clay, east along the Víkurfjall mountain …

The 1860 eruption had a longer time between the earthquakes and the jokulhlaup:

On Tuesday May 8 at 6 to 7h AM earthquakes occurred large enough to shake houses and make cracking noises. At 5 1/2 h PM a jökulhlaup came down the Múlakvísl river with great speed …

These reports indicate that the sequence of events in 1918 is indeed typical for Katla, with earthquakes in the hours before the eruption, and jokulhlaups following after a few to 12 hours. There are a few hours of warning, which should be used well since when the water come it is too late for running.

Why is there no more warning? Many eruptions are preceded by weeks of earthquakes, but not here. Magma clearly does not have to do much rock breaking in Katla. There are a few options. First, all these were summit eruptions. Katla does do rift eruptions (Eldgja comes to mind) as well but has not done so recently. Unlike the other big beast volcanoes in Iceland, it only has one long rift, stretching for 70 km northeast. There is also a short rift going south, but this has not erupted since the ice age. Summit eruptions do not require long dikes, which removes one reason for long earthquake swarms. (The 1721 eruption may have activated a rift, from the continuation of the earthquakes, but the eruption remained confined to the summit.) Second, all these eruptions are basaltic. Katla has done silicic eruptions as well, with the large Vedde ash eruption as a good example, but even though these account for 5-10% of its eruptions, they have not happened in the last 1500 years. It is not clear, though, why this would be a reason for less earthquake activity. Finally, the magma may already reside at shallow depth, with the upper ground layers above it softened by the water percolating down from the ice cap – after all, this is among the wettest areas in Iceland. Katla is among the largest sources of volcanic CO2, constantly degassing even though it is not erupting: this also suggests that magma resides at relatively shallow levels. The large effusive Eldgja eruption occurred only 20 years after a normal Katla eruption: the deep large magma source which fed Eldja is apparently not affected by or directly involved with the summit eruptions.

The 1918 eruption was therefore in many ways typical. But the long time of quiescence is not. Katla eruptions tend to be spaced 40-60 years apart (apart from the minor phreatic explosions under the ice which happen on occasion), whilst it has now been over a century since the last one. There is no obvious reason for this delay. Global warming has affected Iceland, but it is not clear how that could affect Katla! (In the UK one would blame a lack of funding.) The most similar period was perhaps from 1500 onwards. Around 1500 Katla had a large eruption – it is not documented but there is a significant tephra layer from around this time. After that, there were small eruptions in 1580 and 1612, whilst the next major eruption was in 1625. Thus, the current interregnum could be similar to that period if the weak eruptions this time were too small to break through the ice.

Is Katla overdue? No volcano is overdue – eruptions do not occur on schedule. Neither is there much evidence for build-up of magma, apart from Katla’s tendency to emit quite a lot of CO2. The next eruption will remain unpredictable and will happen when we least expect it. It could still be 50 years away – but one day, someone in Vik will again see a ripple in the bathtub. A few hours later, the plume will rise and the waters will come. One day.

Albert, June 2024

References

The explosive basaltic Katla eruption in 1918, south Iceland I: Course of events, tephra fall and flood routes: Guðrún Larsen, Maria H. Janebo and Magnús Tumi Gudmundsson.

Jökull, No. 71, p. 1-20, 2021. https://jokull.jorfi.is/wp-content/uploads/2021/12/J71-1-20.pdf

Haukur Tómasson: The jökulhlaup from Katla in 1918. Annals of Glaciology, Volume 22, p. 249 – 254, 1996. https://www.cambridge.org/core/journals/annals-of-glaciology/article/jokulhlaup-from-katla-in-1918/6209392C96BF5A987B1E2E697B044C3E

Guðrún Larsen: Katla: Tephrochronology and Eruption History. Developments in Quaternary Sciences Vol. 13, 2010, p. 23-49, 2010

Páll Einarsson: Historical accounts of pre-eruption seismicity of Katla, Hekla, Öræfajökull and other volcanoes in Iceland. Jökull, No. 65, p. 35-52, 2019.