This is quite an eruption, with surprises with every new turn of events. It is an unusually tourist-friendly eruption: conveniently located, small, non-damaging (except to the owner of the land) and spectacular. It does appear that not all tourists are as friendly to the virtual tourists, judging from the antics going on in front of the cameras, but that is what it is. Even the local foxes have been photobombing the eruption.
It is very notably a dike eruption. It started out as a dike several kilometers long, forming at a depth of 6-7 km and growing upwards towards the surface. A whole series of cones have formed along a kilometer-long fissure where the dike reached the surface. For a long time multiple cones would be active simultaneous, but the eruption has finally gone conventional with fountaining and effusion only from a single cone. The cones have shown a notable tendency to go double, with two lava exits close together along the line of the rift. In each twin, the cones were acting fairly independently. From the distance between the twins, we can estimate that their lava feeds separated perhaps 50-100 meter below ground. This is not an entirely stable situation, and over time one of the twins will become dominant and finally exterminate its sibling. That indeed happened in all sets of twins. The surviving cone too has extinguished its sibling.
We might have expected the dike to reach the surface at the lowest point along the dike. To some degree this happened as the eruption started in a valley. But strangely, the eruption avoided the lowest ground in the valley and it first erupted on top of the only hill within the valley. When that wasn’t enough, it went for the higher ground next to the valley. It never extended into the much lower ground further north or further south, on either side of Fagradalsfjall. There was a bit of anti-gravity about this. Why was this? Clearly, the road from dike to fissure was not entirely straightforward.
To making of a dike
How does a dike form? Obviously, magma gets in between the rocks underground and slowly pushing them apart. The dike extends both upward and sideways, and eventually becomes a wall which can be kilometers long and tall, but only a meter or so across. It is like a very large sheet of paper, slowly making its way vertically into the crust.
Where does the magma come from? There are two parts to this question: where was the magma to begin with, and how does it get to the dike? As for the origin, the magma can either make its way up from the mantle through a conduit (as seems the case in Reykjanes), or it can come from a magma chamber that is already in the crust (as was the case in Holuhraun). That magma chamber is in most cases (but perhaps not always) located underneath a central volcano. In both cases, the magma feeds the dike from a particular entry point, and the dike grows sideways from there. If the dike is fed from the mantle, then it is also possible that there is no single conduit, and that the dike is fed from below along its entire length, but this has not been observed in practice. As far as we know, a dike starts with a bottleneck.
Why does the magma form a dike, rather than a conduit up to the surface as it does in any normal volcano? This is the all-important question. Dikes form if there is some weakness in the land. Volcanoes grow on strong land. Long-distance dikes feed on weakness, on land divided against itself. They grow best on rifts.
Let’s look at two examples. The first is, of course, Hawai’i. It has impressive, in fact world-leading, volcanoes. ‘World leading’ can sometimes have a rather subdued meaning in the US. For example, the ‘world series’ is not quite what it seems as only teams from the US and Canada are allowed to take part. The ‘world’ is not allowed to play in the ‘world series’. That’s just not cricket. But Hawai’i can and does compete with the rest of the world. Its top-heavy volcanoes are riddled with rifts, and many eruptions come from dikes following those rifts. This happened in the Leilani eruption of 2018, from a rift which was tens of kilometers long. Many of Hawai’i’s rifts follow the line of flank failures on older volcanoes. If you stand on the shoulder of giants, you also inherit their problems. The weakness in Hawai’i is very much self-induced, a consequence of the immense size of its volcanoes.
The second example is Grimsvotn. For such an impressively erupting volcano, it actually looks a bit small. Bardarbunga is far more imposing. Why is this? Grimsvotn is a child of the spreading rift that bisects (possibly quadrusects) Iceland. The rift forms a line of weakness, and it provides the magma with a tempting escape route. Grimsvotn is a young system that recently (geologically speaking) formed on the rift. It does not have a strong a magma chamber, its summit eruptions are small (although very frequent) and most of the magma goes into the rift. Laki stole its magma from Grimsvotn, either directly or by intercepting the supply. ‘My volcano is my castle’ does not work if you built the thing on a rift.
Bardarbunga once also formed on the rift but it is an older system and it has moved with the times. As the spreading rift spread, it took Bardarbunga with it and over time the volcano slowly migrated away from the rift. This made its magma better contained because it is now further from the weakness of the rift. Not perfectly, though: often its magma chamber will spring a leak, the magma will find its way to the rift, turn north or south following the spreading rift, and produce a fissure eruption far from the embarrassed volcano (who, me?). This happened in 2014. But a decent amount of magma remains confined to its magma chamber and this produces summit eruptions, or otherwise just pushes up the volcano. (Much of the growth of a volcano happens underground. Summit eruptions have their role but don’t do all the work.) So Bardarbunga grew large, while Grimsvotn with its much higher volcano-metabolism remained smaller. One day Grimsvotn too will find strength away from the rift and grow to become a giant. And Bardarbunga will have moved too far from its magma supply, and decline.
So you have magma, and you have a rift. What happens next? The magma finds it easy to push in the direction of the rift. This pushed the rocks apart in a direction they want to go anway, because it is the direction of the rifting. The magma is pushing at a nearly open door. Going in the spreading direction, perpendicular to the rift, is much harder. That requires pushing the rocks in a direction they do not want to go. Magma, as a liquid, always takes the easiest path. So the magma prefers to move along the rift, not perpendicular to it. That happens both in Hawai’i and in Iceland. In the right circumstances, the magma can travel unopposed for tens of kilometers – but only along one direction.
When a dike begins to grow, it tends to follow the line of the rift. The growth can be followed quite easily, because of its earthquake activity. The tip of the dike has to break the rock apart, and this gives a lot of small shakes – almost explosions. While the dike lengthens, it may also grow thicker, but this is mostly a silent process which does not cause earthquakes. It does not involve new breakages, just widening a crack that already exists. So while earthquakes detect what happens at the ends of the dike, the centre moves quietly and secretly.
So this explains how a dike grows long and wider (‘wide’ is the wrong word though for something that is 10,000 times longer than it is wide). How about moving up? After all, a dike doesn’t do much unless it can reach the surface.
It turns out, up is also a hard direction. Gravity complains, and when gravity has the weight of 5 kilometers of rock behind it (or above it), it is hard to go against it by moving up in the world. It is not as hard as going perpendicular to the rift, because going up pushes the rock in the spreading direction it prefers to go, but it is much harder than going along the rift where the weight of gravity does not work against it. That magma may have some buoyancy which gives a bit of a push upward, but it is still hard. So the magma can move in three directions, but they differ in difficulty. Easiest is horizontal along the rift, harder to move vertical, upward, and impossibly hard to move horizontal but in the direction of spreading. This is why dikes become like sheets of paper, and why this only happens where there is a rift to give a preferred direction.
In the absence of a rift, all directions become hard. Magma does not find it easy to move against an undivided land. That is one reason it collects in magma chamber: it is hard to get out! But physics still acts, and physics says that in such cases it is easier to move sideways than upward. In such a case it forms a sill, a horizontal dike.
As an aside, comparing Reykjanes to Vatnajokull (Grimsvotn and Bardarbunga), one thing immediately stands out. Dikes in Reykjanes are only several kilometers long, perhaps up to 10 kilometers, while Bardarbunga’s dikes can reach over 100 km from the volcano and Grimsvotn, in spite of having quite an immature rift system, also has a long reach. There are two reasons for this difference. One, there is much more magma available in Vatnajokull because it receives more mantle heat. Ten times more heat should give ten times more magma and this should allow for ten times longer dikes, all else being equal. Second, there isn’t a clear spreading rift in Reykjanes. Instead, the rifting is taken up by numerous mini rifts, located along the Reykjanes fault line. As a result, eruptions in Reykjanes are rarely more than a few kilometers from the central fault.
The topography of the surface began to play a role. For there are dungeons and dragons here, or rather, valleys and hills. Where there is a hill, kilometers down there is more weight to carry. This affects the magma dike. To understand how, we need to go back to Holuhraun.
The Holuhraun dike
The Holuhraun eruption was fed by a very impressive dike which did not play according to the rules. It ignored the straight line and followed a roundabout route. First it went east-southeast from Bardarbunga. Then it waited, and changed direction to north-northeast. There were several more changes of direction. Why?
The answer is that the direction a magma dike takes is a combination of the spreading axis and the local topography. The basic principle is an easy one: magma will always take the easiest path. This is, the path that requires the least energy. The energy is used for two things: pushing apart the rock, and pushing up the ground above the dike. The energy needed for the latter should not be underestimated. A 4-kilometer deep dike carries 12,000 tons above every square meter. Pushing this up requires a lot of energy. Pushing up the ground is easier when the ground is lower, so the dike prefers to take the direction where the ground slopes down steepest.
What if this is not the direction of the spreading axis? Then the dike will find a compromise: it will go into a direction which is not quite the spreading axis, and not quite the lowest direction, but is the most efficient. It is called the principle of minimum energy. We can calculate at every point which direction requires the least energy. And indeed, the dike took this direction at every turn. The initial ESE direction was down the slope of Bardarbunga. It reached the rift at the saddle point where it turned north. Now it followed the rift direction in general, but with various deviations. This is described by Sigmundsson et al, Nature, https://www.nature.com/articles/nature14111.pdf
The plot of the rift shows these changes of directions. The coloured circles below indicate at each these location the energy required for evert direction. Red is high energy, blue is low. The dike always took the blue direction. The principle of least energy really works. Magma does not like wasting energy.
While growing in length, the magma also tried to rise, driven by its buoyancy. It was a slow process, but once an opening had been made, the dike did not progress further but instead used the easy way, to the exit. The last stretch where the dike begins to feed the surface eruption is called the feeder dike.
So this is how the Geldingadalir dike (the name has a ring to it) formed. The magma quickly moved from the mantle through the lower crust, either following or triggering the series of earthquakes along the main fault. In the upper crust, 7 km deep, the magma began to move sideways, horizontally along the spreading axis. Unusually, in Reykjanes the spreading axis does not run along the main fault but goes off at angle. So did the dike. The earthquakes traced the tips of the dike, which as expected moved towards the south coast in one direction and towards Keilir in the other direction. The earthquakes gave the impression that the two tips of the dike were primed for eruption. But that was misleading. The tips of the dike remained at a depth of 4-7 km. Instead, most of the magma that entered the dike was used to bulk out the central region. This bulking up was aseismic, although not invisible: the growing dike was seen on the interferograms which showed local inflation and extension, and those indicated where approximately the actual eruption would take place. Of course, hindsight is always 20/20.
The Geldingadalir dike feeling its way towards Keilir was small, but it followed the same physics as the Holuhraun dike did. The map shows the line of earthquakes which traced the dike. I took this from a plot showing the locations of earthquakes between Feb 24 and March 12: https://twitter.com/krjonsdottir/status/1370280535050878980
The dike is shown by the orange line. The location is not exact but is taken as the approximate midline of the earthquake activity While moving north and south, the dike made a small change in direction, and at one place going south there was a bifurcation and the dike (as traced by the earthquakes) backtracked and went into a slightly different direction. The point of bifurcation is uncertain: it might be a bit further to the south.
One would expect the dike to follow the spreading axis. The spreading axis in Iceland runs nearly north-south (about 10 degrees east of north). However, the land itself is also moving. If you take the direction relative to the (moving) plate boundary, the axis becomes more eastward, and follows the line of the obvious volcanic ridges on the Reykjanes peninsula. This direction is shown by the blue dashed line. The northern section of the dike indeed closely follows this direction. The land here has only a small slope towards the northern coast. The cost of gravity varies very little with direction and the dike is completely dominated by the spreading axis. But this is not the case underneath the mountains of Fagradalsfjall, and here the dike chose a different direction. It avoided the peaks of Langholl, Kistufell and Storihrutur (where the cost of gravity would be high) and followed the lower ground in between. The peaks here are far smaller than the mountains of Vatnajokull and the effect is less than seen in Holuhraun. But topography still matters.
Sometimes a dike takes a wrong turn. One of the two southern branches ran into a high ridge, found no way out and terminated. There was not enough energy to continue in any direction. This caused a bifurcation: the alternate channel found a way into the deep valley of Natthagi, and ended there. Towards the north, the dike ran into the large cone of Keilir. This also was an unsurmountable obstacle, and the dike could neither go through or go around. It terminated here.
Underneath the mountains, the direction is always a compromise between the spreading axis and the topography. We can largely understand why the dike took the directions it did. The only surprise is that it crossed a small ridge but this ridge is quite narrow.
The best route also depends on where the dike first formed. If it had formed along the southernmost segment, the preferred route would have been directly into Natthagi and Meradalir. If it had formed north of Kistufell, it would have gone only to the north. The most likely place for the first magma intrusion is therefore along a line between Kistufell and Langholl, to a point left of Meradalir. This is indicated by the blue oval.
The dike and the fissure
How does this compare to the fissure? An excellent Landsat shot of the series of cones (and lava flows) has just appeared, showing all 6 cones.
The series of cones follows the direction of the main dike quite well. However, the fissure is offset to the west by about 200 meters. I measured this several times but could not get the mid-line locations to match. Between the deep dike and the surface rupture, there is a slight tilt. This seems problematic: it is a principle of geology that dikes are vertical and do not tilt. Perhaps the dike is not quite at the central line of the earthquakes, the dike has a few more small but unrecognized changes of directions that bring it in line with the fissure, or the fissure is not actually the top of the dike but there is a tilted conduit between them. Most likely, of course, is that I did not get the measurements quite right.
There is one further surprise. The dike will find it easiest to erupt at the lowest point, where there is the least amount of rock to melt through. The eruption happened close to the point where the magma came up from the mantle. Meradalir would have made a good eruption point, being a deep valley, but the dike bypassed this. The next lowest point was the valley of Geldingadalir where indeed it erupted. But in this valley, there is a small hill. Strangely, it erupted near the top of this hill. The later cones similarly avoided the deepest valley and went for the higher ground to the north. This remains unexplained. For some reason, the hill was a weak spot which allowed easier access to the surface for the feeder dike. We will come back to this.
The Reykjanes fault
The Reykjanes fault was crucial to the development of the dike. This was the fault that initiated the immense swarm of earthquakes which announced the eruptive intent. It gave way in several M5 earthquakes. This fault is mostly a transform fault. The south side is moving east relative to the north. The interferogram shows the movement beautifully. Red (eastward motion) delineates the fault perfectly.
Let’s put this on the map. Here is the map of the dike and fissure, on top of the digital elevation model. The black line shows the Reykjanes fault, as traced by the area where the surface moved to the east. There are slight offsets along the line but I have ignored those. The white lines shows the region of earthquake activity along the fault in the weeks leading up to the eruption. These were largely aftershocks of the main earthquakes where the sides of the fault made small adjustments with minor failures. The dashed white line shows the location of an earlier series of earthquake which happened in 2017. The earthquakes were seen in an area north of the surface motion.
The fissure is closely related to the fault and the fault earthquake zone. The fissure terminates on the fault itself and did not cross the fault. This eruption belongs to the North American plate and did not manage to cross into the European plate. Interestingly the dike had no problem with this and happily crossed over. It seems that at depth (hotter rock) the plate boundary is easier to cross.
The second point to note is that the length of the fissure largely coincides with the width of the fault earthquake zone. This eruption occurred where the dike crossed the fault crumple zone. It made use of the weakness of the rocks in this area.
This immediately show the importance of the Reykjanes fault. The fissure similarly began exactly at the fault line, and extended northwards from there.
When the eruption began, the first fissure followed the ridge of a small hill in the Geldingadalir. The origin of this hill was not clear. It was not a result of an earlier eruption. It may have been glacial.
However, the location close to the Reykjanes fault suggests another possibility. The Reykjanes fault may have step-overs, where the fault consists of segments which are slightly offset. This is very common for faults that are young and still in the process of organizing themselves. An offset in a transform fault can cause conflicting movements. The step-over can cause either cause compression or extension in the area in between. In the first case, a pressure ridge can form a small hill form in the area of the step-over. I speculate that the hill of Geldingadalir is (was) a pressure ridge: it is the location of the transverse fault where at one time a small step-over occurred.
Pushing this speculation a little further, the transverse crumple zone may have given the magma the easier pathway to the surface for the feeder dike. This would explain why the fissure ran along, an in fact started on, the ridge of the Geldingadalir hill.
Where pressure ridges occur, you can also find sag pools, or small basins, where the fault pulls the crust apart. Was Geldingadalir itself such a depression? This is becoming even more speculative.
The Reykjanes fault is more than just a transform fault. There is also a bit of extension here, a little bit of spreading. The sides of the fault slowly pull apart. If I am allowed to speculate even further (we are getting on very thin ice here), the notably deep valley of Meradalir could be a result of this extension, as a small pull-apart basin approximately along the line of the fault.
Can I see any evidence of the movement of the fault? That is notoriously difficult. It means finding features in the landscape on either side of the fault, and matching them up. But landscapes evolve and the two sides may not look alike any more.
I can attempt this for the ridge that separates Meradalir from Geldingadalir. It continues across the fault but the appearance is quite different on either side of the Reykjanes fault. Perhaps this isn’t the same ridge? There is another ridge a bit to the southeast which looks more like it. If I shift this ridge according to the spreading direction and transform movement, the two ridges line up fairly well. The offset is some 700 meters. Assuming an average movement of 2 cm per year, this makes the ridge 35,000 years old. The offset nicely delineates Meradalir which becomes a small pull-apart basin. The ridge would have formed during the depth of the ice age.
I can push this further. On the south side of the fault, there is an eroded plateau which lies just south east of the ridge. There is no such plateau on the north side of the fault. There is however one on the west side of the area: the high plateau of Fagradalsfjall. Could that be the same plateau? In that case the offset is around 2 kilometers. The age of the plateau would be 100,000 years. Was this a shield eruption? The lack of ridges would imply there was no ice when it formed. And indeed, the time is close to the last interglacial when Iceland was ice-free.
This is deep speculation, in severe need of data, even if those data may well show it to be in error. Lots of handwaving involved! But the ages make sense. The Geldingadalir eruption came in an old landscape which had seen little or no volcanic activity for tens of thousands of years. It is a unique event. We are very lucky to have seen geology in such action in our life times. The last time this happened, we were still into cave paintings.
Albert, May 2021