Laki was one of the defining events of Iceland’s history. Its impact went well beyond the nation, covering much of the northern hemisphere. It seems amazing that something that caused so much suffering was seen by so few people. This it shares with Eldgja, which was even larger and must have devastated Iceland, but of which no eye witness report remains. There are reasons. The largest disasters leave the local population decimated and local areas uninhabitable, and the historical record is lost. The region in which the eruptions occurred, and which Lurking has named the Dead Zone (because of lack of earthquake activity), was unsuitable for habitation due to previous lava floods. So no one saw the rift opening, and the lava was only seen when it came spilling out from the river canyon, some days later.
The best record we have is from the local priest, mentioned by Carl: Jon Steingrimsson. He was quite a remarkable person, well educated and caring, and his observations have stood the test of time. His diary is much more than a scientific record: it is a personal story of happiness, worry, despair, suffering and survival, in a developing disaster that would claim the life of his wife and leave him destitute and starving and decimated his human flock. It is about being human in a time of the inhumane. But it doesn’t answer the major questions that we would like to answer about the Laki eruption. What made it happen? Where did the magma come from? Could it happen again?
Let’s first get the lie of the land. The map shows the main locations in the story of Laki. Klaustur is the place where Jon lived during the events. It is along the river Skafta, which in those days came from a 200-meter deep canyon further west. The canyon has been filled by Laki lava and no longer exists. Nowadays the Skafta is fed by multiple small tributaries. The first lava flows came down the Skafta gorge, after the rift southwest of Mount Laki fissured. Later, the eruption shifted towards the northeast and now the lava flowed through the canyon of the Hverfisfljot river.
Carl has explained the timeline of the Laki eruption. The first explosion happened in the morning of Sunday, 8 June 1783 (Pentecost Sunday), as Jon was getting ready for the church service. The ash turned day to night, but by afternoon the sky had cleared. The ash returned the next day. The Skafta river began to dry up, and by the afternoon had stopped flowing completely. The lava appeared in the lowlands on June 12, having come down the canyon. The following Sunday, some of the local farmers trekked into the highlands, climbing a peak (Kaldbakur, 8km due north of Klaustur – not the mountain of that name in northwest Iceland – still over 10 km from the Laki rift) to get a view of the events. They were the first (and only?) to see the rift in action, reporting twenty fountains of fire. (Following their example, these types of eruptions are now called ‘fires’). The number may well have been exaggerated – how would we know? By mid July, the lava was encroaching on Klaustur, but it stopped on Sunday 20 July, during Jon’s famous fire sermon. The flow ceased because of a combination of reducing lava flow and quenching by heavy rain (and of course the famous sermon of which sadly no transcript survives).
Four day later, the fissures northeast of Mount Laki opened up, with a major explosion on July 29. The new flows came down the Hverfisfljot river. By August 3 this river no longer flowed, and on August 7 lava appeared in the lowland. This was more sedate than the June eruption, but in the end just as devastating. Various eruption episodes continued, each one accompanied by a new lava flow down the river valley. The last of these was late October. After this, there was still fountaining but at rates too low to reach the lowlands. It was stopped by cooling. Lava can flow as long as it is hot enough. When eruption rates get less, the lava flows become thinner and therefore cool faster. Lava tubes can extend the reach by a lot, but for surface flows, the lava reaching less far is a first sign of a tapering off of the eruption. So it was at Laki. The last fire was seen on February 7, 1784.
(In case Jesper is about to ask, the eruption temperature of Laki is estimated at 1140 C.)
Were there any warnings? Was there any sign that it was coming? Here, we only have Jon’s diary. He recalls vivid dreams which he took as a warning of things to come, but of course this carries little scientific weight. Early in the year he noted a blue smoke on the ground, of unknown origin. That probably had nothing to do with Laki. There had been an eruption on Reykjanes which had created a new island (Nyey). It did not last long. The blue smoke may have been small aerosols, blown towards the east from the Reykjanes eruption. Or it may have been a nearby farmer burning something he shouldn’t. Sometime in May, the Skafta river flooded with muddy, bad-smelling water. The cause is guess work: it could have been due to an emptying of a lake that had been caught behind an ice wall, up in the highlands. The bad smell though may be explained as gas escape, signalling encroaching magma. The first obvious indication of trouble came from earthquakes. These started mid-May, and by June became strong enough that people started sleeping outside. (In Iceland!) Knowing what happened during Holuhraun and Leilani, it is not hard to interpret this. A dike was opening up, with magma moving in the rift looking for a weak spot.
Spot the rift
This was a massive case of rift indigestion. Laki erupted some 15km3 of lava. It is easy to overlook how much that is. The village where I live measures about 4 by 4 km. Laki could have buried our entire community under a kilometer of lava. Who said that mountains couldn’t move? An entire mountain was on the move! Jon’s community was threatened by a molten mountain coming down the rift.
We know which rift system is to blame. The highlands are transected by numerous linear rifts, together forming several rift systems. Each system belongs to a different central volcano. Bardarbunga is most active. It has a well developed rift system, called Veidivotn which runs from Vatnajokul to Torfajokul. It is the cause of most of the rift eruptions in the region. Over the last 9000 years, Bardarbunga produced 12 large rift eruptions. The winner was the massive Thjorsa flood, dated to 8600 years ago, producing 25 km3 of lava.
The second possible culprit is Katla. It is a very different volcano. While Bardarbunga is a typical rift volcano, Katla is an intraplate volcano which does almost all of its eruptions from the caldera. However, it did manage to connect to the spreading rift on one occasion, in 939 AD. What followed was the massive Eldgja eruption. The Eldgja rift runs just southwest of Veidivotn. Eldgja produced typical Katla magma, but at the far end it produced tholeitic lava which clearly came from Vatnajokul. Nowadays we understand this better, because Leilani showed how a rift eruption can push out older magma which had been left behind by an earlier incursion. Eldgja pushed out magma deposited in the same rift but from the other end. It invaded the Bardarbunga rift system.
But Laki did not come from either of these. It erupted on the Grimsvotn rift system, an immature and possibly growing system. Grimsvotn much prefers eruption at its caldera or on short rifts a few kilometers either side. However, on occasion it reaches for the sky and goes for a massive flood basalt. There are two large ones known, and two smaller ones: 4,550BC Laki (Botnahraun: this caused the birth of Mount Laki); 3,550BC Thordharhyrna (Bergvatnsarhraun); 1,950BC Raudholar and Brunuholar; 1,783AD Skaftáhraun (Laki) (names and dates from Carl). (Raudholar has also been associated with Thordarhyrna.) This is the most southerly of the rift systems, and on the map one can see that it has fewer rifts: it is less active.
Sourcing the magma
During the weeks of intensifying earthquakes, the mountain of magma was working its way through the rift. But from where? The obvious answer would be that Laki was fed by the Grimsvotn magma chamber. After all, this is where a third of all Icelandic eruptions happen and Laki was part of the Grimsvotn rift system. Both Veidivotn and Eldgja were clearly fuelled by magma from their central volcano. The erupted lava would nicely fit inside the caldera. In addition, the Eldgja eruption caused a major change in the topography of Katla. Before Eldgja, its jokulhaups would go in all directions – no direction was safe. Afterwards, they only ever went southeast. The mountain had changed.
This of course was also true for Holuhraun and for Leilani: in both cases we can relate the erupted volume with the hole left in the caldera. But not for Laki. It has three problems. First, the caldera of Grimsvotn is too small. We also believe that this crater changed little during the eruption: the crater lake of Grimsvotn causes regular jokulhaups, and these did not change from before to after the eruption. Second, it magma chamber is too shallow. It is only around 3-4 kilometers deep, and this makes the route to the surface too easy. Sufficient pressure to force open the rift would already have opened the summit conduit. Thirdly, its magma chamber is too small. It may be up to 10 km3 of which only a small fraction will erupt at any one time. It can’t hide tens of cubic kilometers of eruptible magma.
But if not Grimsvotn, where did the magma come from? We know it wasn’t the dead zone itself. The reason is that magma and heated rock are less dense than cold rock. Thus, they expand and push up the land. And the dead zone is not particularly high. The big bulge is Vatnajokul, and there is little doubt this is where most magma is created and moves towards the surface. The topography of the land tells you so.
There are in fact two models for feeding these rift eruptions. One model assumes that the entire rift is fed from the mantle below, with magma moving up vertically along the entire rift. The second model assumes a central reservoir, with the magma moving sideways along the rift. The second model has become the main one, since it fits all recent rift eruptions, in Iceland, Hawai’i and Ethiopia.
If you are still in doubt, look at the Laki rift. The same physics that limits how far lava can flow also works for magma. As the flow rates decrease, magma can travel less far. The early eruptions of Laki were the most voluminous. Later ones were smaller. So you would expect that the eruption site would tend to move closer to the origin. And this is exactly what happened. The eruption moved from southwest to northeast, straight as an arrow – pointing at Grismsvotn. Talk about a smoking gun!
So the answer lies in the direction of Grimsvotn, but not in its shallow magma chamber. We need to look deeper. For Grimsvotn is fed through dike intrusions from deeper chambers. We know from the composition of the Grimsvotn ejecta that its magma spends time at a depth of 10 to 20 kilometers, where undergoes a bit of crystallization. This should not be viewed as a deep magma chamber. It is more likely a thick region of partly melted material, with pockets, sills, and dikes though which magma slowly percolates up, perhaps collecting in a lens of fully melted material at the top of the region.
Could this region have fed Laki? The similarity between Laki and Grimsvotn lava suggests it may be. For instance, they have identical enrichment in 18O, different from what is found in the mantle. However, there are also differences, such as a slightly different U and Th concentrations. The 18O may also have been affected by water finding its way into the magma.
The composition of Laki lava is remarkably uniform. That suggests it came from a single, well mixed magma chamber, not an extended mush as found underneath Grimsvotn. However, this chamber may have been located within or near the mush region. The depth of the Laki chamber has been estimated at anywhere between 15 km and 7 km. The composition suggests that the magma resided in this chamber for a maximum of 1000 years, possibly much less.
It makes sense that Laki was fed by such a reservoir. Dikes are normally angled upward; if it starts too shallow the dike is likely to break the surface nearby – as in fact Grimsvotn itself tends to do. A dike breaking out from a reservoir more than 7 kilometers deep is more likely to reach as far as the end of the Laki rift, 30 km from Grimsvotn. Such a reservoir, close to but distinct from the Grimsvotn feeding region, would also explain the subtle differences in composition and the fact that Grimsvotn itself was not damaged by the Laki eruption. The mid-crust Laki reservoir cannot itself be a direct feed for the shallow Grimsvotn magma (this is ruled out by the composition) but it can have the same source.
A fire in the making
Icelandic magma reservoir can erupt a maximum of 10% of their volume before the pressure drop ends the eruption. Thus, to feed the Laki eruption requires a magma chamber of at least 150 km3. As magma had been stored for up to 1000 years, one may assume that this chamber has existed throughout that time. But that is not quite true. The magma appears to have been collected at different times at different temperatures and pressures, at different depths and locations. Perhaps only in the last 50 or hundred years or so before the eruption did it coalesce into a single chamber. This is indicated by a variety of crystals found in the erupted lava. Was the coalescing triggered by some event, perhaps a rifting event or a major earthquake? That is not known. There was a major fire eruption in Krafla, from 1724 to 1727, called the Myvatn fires: perhaps there was on-going rifting.
I mentioned that Laki lava was remarkably uniform. That was not entirely true. The lava contains a fraction of crystals, with properties that differ from that of the lava. It is evidence for a crystal mush that was picked up by the Laki magma. The fraction of the mush changed during the eruption: it became larger as the eruption progressed. There is a small group of olivine crystals that acquired an outer rim deposited from the Laki magma. This layer grew over a period of 1 to 2 weeks. These crystals were apparently picked up while the magma was traveling to or through the rift. The time agrees with the duration of the earthquakes preceding the first explosion. But other crystals must have spend months in the Laki magma. This, it is possible that in the months before the eruption, magma was already on the move.
Laki had around 10 separate eruption episodes. Some overlapped but there were also quiescent periods in between that lasted longer than the 1 to 2 weeks that magma took to travel the rift. Thus, it appears that each episode was driven by a separate magma pulse. That may explain another peculiarity. Why are Laki and Eldgja so much larger than other Icelandic fires? Holuhraun produced a bit over 1 km3 and this appears to be typical. But each individual episode of Laki also produced such an amount. Perhaps this is the normal eruptive capacity of an Icelandic magma chamber feeding a rift: once this much has been erupted, the pressure has dropped and the eruption closes. Laki was different because the magma chamber was re-supplied by fresh magma during the eruption, and these pulses re-opened the rift. But the pressure did drop over time, and later magma pulses did not reach as far along the rift.
Grimsvotn itself erupted a number of times during Laki. But these eruptions were during the latter phases of Laki. As the magma pressure in the rift dropped, Grimsvotn came back to life. That may just have been the magma eruption rate. The magma pressure has to keep the dike open. As the pressure reduces, the dike closes a bit and therefore less magma gets through. Now magma is becoming backed up, waiting for its turn and looking for another exit. And the shallow magma chamber underneath Grimsvotn was ready and waiting to act as the valve.
The earth didn’t shake
The main sign of the impending catastrophe was seismic. That was how Jon knew something was coming. There were several weeks of intense earthquakes as the dike forced its way through the crust. But what about the years prior to Laki? Did the accumulating magma body not produce earthquakes? We know that Grimsvotn announces its intentions in this way, with slowly increasing earthquake activity over several years. Bardarbunga also had some earthquake swarms in the decade before Holuhraun. But these events are small, in fact surprisingly small in view of how much magma accumulates. To put it in context: Bardarbunga had months of M5 quakes as the caldera collapsed by 60 meters. Why did the preceding inflation not generate the same quakes? And why was the observed inflation before the eruption so much smaller than the collapse afterwards? It is as if much more magma was extracted than had been inserted. Was that also the case for Laki? Wouldn’t the intrusion of a magma body the size of Laki have caused significant earthquakes long before hell broke loose?
The lack of strong precursor quakes tells us two things: accumulation takes time, and it happens below the brittle-ductile transition, i.e. at depths where the crust is hot enough to become a bit elastic. New models for Bardarbunga’s Holuhraun eruption were published this week which suggest that the behaviour can be explained by a buoyant magma body within a visco-elastic crust. Visco-elastic means that if you deform something, it slowly creeps back to its original shape. You all know materials like that. Honey is the liquid equivalent, and memory foam is a solid form. How quickly it returns to its original shape depends on the ratio between the elasticity of the material and the viscosity. For the hotter mid-crust of Iceland, it is something like 25 to 50 years. The idea is that if you drip-feed magma over this time, everything adjusts fine without stress – no large earthquakes. And because the pressure hardly increases, it can spend a long time very close to failure before breaking. This allows much more magma to accumulate.
In fact this may explain another aspect of Icelandic fire eruptions: why they can erupt so much. How can 50 centimeters of inflation leads to 50 meters of collapse? Shouldn’t the eruption stop when the 50 centimeters has been lost, as this clearly was a stable situation? A visco-elastic magma chamber can do this. As magma is expelled, the chamber does not contract quickly – it takes its time. This leaves the chamber underpressured, and makes it possible for buoyant magma from below to replace the erupted magma. (It also allows the caldera above to begin collapse.) It is a neat model which can explain the typical size of Icelandic eruptions (although I haven’t seen those numbers plugged in to the models), as well as the multiple eruption pulses of Laki and Eldgja. It requires a large magma chamber at or below 10 km depth. But scarily, it also predicts that there are few signs of the coming storm. And there is no way to tell whether it will be a one-trick pony like Holuhraun, Leilani, or Mayotte, or a long-term disaster like Laki – or Pu’u’O’o.
Visco-elastic crust can not cope with a magma influx that is too fast. It works as long as the chamber inflates slowly, over decades to centuries (or millennia). If the magma moves in too quickly (for the size of the magma chamber), the behaviour becomes more brittle-like and pressure rises: the magma will now be pushed out. This may be why Grimsvotn and Katla have so many summit eruptions but so few fires: their magma supply rates are too high. (Katla, being an intraplate volcano, will have less ductile crust. Its supply limit will be lower than for Grimsvotn.) Bardarbunga is a bit more sedate, and this allows deeper magma to slowly accumulate for the next fire. If this is correct, it would indicate that the magma reservoir that fed Laki is indeed separate from the one that feeds Grimsvotn. But Laki and Grimsvotn have similar lava production rates (in km3 per year). To reconcile this, we would need to pose a nearby reservoir which is much larger than the one feeding Grimsvotn, but with a similar magma input rate. The race is on to find this reservoir.
The story of Laki is still being uncovered. Research is now mainly focussing on two things: quantifying the impact of Laki on the environment, and studying how those crystals in the Laki lava can trace what happened underground. But the final question is whether this could happen again. Laki devastated Iceland and affected every nation around the North Atlantic. A re-occurrence would have predictable but unforeseen consequences: life stock suffering fluorine poisoning, acid rain causing injuries, people having to stay indoors for months on end. Any warning would help. (Yes, I know about how much good that did for the current epidemic.) What would those warnings be? It appears that we could see something brewing, by inflation and small seismic swarms, followed by unmissable shaking in the weeks leading up to the eruption, perhaps with increasing gas emissions. But these might not tell you the size of what was to come.
Jon took his warnings from his dreams. Science can’t do that – we need measureables. We are getting to the stage where we can detect signals of coming eruptions – even if many of these fail to progress to actual eruptions. But to distinguish the coming storms from the hurricanes, that may still be beyond us.
Albert, May 2020