The Happy Dyke of Fagradalsfjall

Dalssel at Fagradalsfjall. Image from


The good part about volcanology is that nature will sooner or later test both your theories and your scientific models. In this case what was tested was the original model of the available pent up seismic strain in the currently active area of Reykjanes Peninsula in Iceland.

It turned out that the amount of pent-up strain was higher than the models had given at hand. Science loves when oddballs like this comes around the corner, then you get the opportunity to see if something is wrong or missing with your theories, or if there is something wrong with the models used. I love the smell of science in the morning.

As everyone has noticed by now, well at least everyone interested in Icelandic geology, the seismic activity continued well beyond what could be expected according to the model for the area. Since the start of the seismic crisis an average of 20 000 earthquakes has occurred per week, and the size of the larger earthquakes are comparatively in the same range without any sign of them abating.

During the first week of seismic activity, we mainly saw tectonic earthquakes along the portion where the Mid Atlantic Rift enters the Reykjanes Peninsula in a WSW/ENE-direction spanning most of the peninsula.

After that the activity started to concentrate in an area between Keilid and Fagradalsfjall, with a migrating trend towards the southern part of Fagradalsfjall.

At this point in time this is the largest earthquake swarm recorded instrumentally in Iceland. It is in fact so numerous in earthquakes that it stands for a large portion of all earthquakes ever measured in Iceland. It even dwarfs the famously noisy eruption of Holuhraun in 2014-2015.


Tectonic or Magmatic?

Dykemap, courtesy of the Icelandic Met Office.

One of the most irrelevant questions in Iceland is probably if an intense earthquake swarm is tectonic or volcanic. The reason I am saying this is that sooner or later an initially purely tectonic swarm will turn into a magmatic swarm as intrusions sets in.

Here I am talking about earthquake swarms that are not directly in or around the well known large central volcanoes, they operate a bit differently. I am here keeping to areas like the Reykjanes Peninsula.

The reason for this is simple fluid dynamics. The main tectonic regiment in Iceland is caused by the rifting apart of the Eurasian and North American continents. This is by no means a constant process, instead you get increased strain locally, and that strain is later released as an earthquake swarm (or as a large earthquake, but now we are talking about swarms).

The swarm in turn is in the beginning tectonic as things are pulled apart, and here it is where fluid dynamics comes into action. I will here invoke a rule, nature abhors a vacuum. So, if a void is formed in the crust as it is pulled apart two things will happen.

Scenario number one is the purely tectonic one, and the rift will snap back together giving no dilation of the rift. In other words, the vacuum problem is solved by the sides of the rift coming back together (most often not exactly in the same way).

I am obviously simplifying things quite a bit here, but the principle is correct. In these cases, you might get lateral slip or vertical slip of one edge, but you do not see dilation. This is quite easy to see a few days later on the GPS-network.

The second option happens if the swarm is ongoing for some time and the crust is fractured enough that magma can enter from below. In this scenario you will instead be having fresh magma being sucked into the void, filling it up.

Now we have dilation that will be visible on the GPS system if we are in the happy situation that we have a GPS-station on either side of the rift as the earthquake(s) strike. It is visible since the stations will move apart, and then they stay apart.

The important part here is that it is quite common that this happens in Iceland, it usually happens a couple of times per year, or even more. After all, most intrusions will fail before an eruption occurs since the intruding magma is just enough to fill up that pesky void in the crust.


The Happy Dilating Dyke

Cumulative seismic release over 11 days. Image by Gaz Dale based on data from the Icelandic Met Office.

Now things will become a tad technical, so I need to explain a few expressions as I go. The first one is obviously the word dyke. It is a form of volcanic conduit that is extending from the mantle upwards like a hanging sheet of magma, sometimes spanning a considerable horizontal distance.

It forms like I explained above as the rift is pulled apart by the continental drift, the ensuing void will be filled by magma. If the magma moves far enough upwards it will start to exert force in two ways.

The first way is through buoyancy, due to being hotter than the surrounding bedrock it will buoyant and will strive to move upwards like a hot balloon. If the magma is shallow an eruption might happen since there is not enough time for the magma to solidify in the dyke.

This is a fairly slow process, and it can be followed since the earthquakes will be slowly migrating upwards.

The other force will kick in if the magma comes close enough to the surface. All magma contains volatile compounds like water and gasses. In Iceland there is comparatively little water in the mixture, but there is quite a bit of gas that can nucleate out of the magma.

Cumulative number of earthquakes over the last 11 days. Do note that the numbers of earthquakes do not correspond to the energy release. The numbers show a decrease, but the energy is increasing. Image by Gaz Dale based on data from the Icelandic Met Office.

The amount of gas becomes higher the closer you come to the mantleplume that is residing near Bárdarbunga below Vatnajökull. In Reykjanes the magma is poorer in gas, so the force is less.

As the magma nears the surface the pressure from the crustal over-burden will decrease, and the gasses that are locked into the magma will start to come out.

Think of it as a warm soda can that has been shaken. If the lid is on (over-burden pressure) the carbon dioxide will stay dormant in the soda, but as soon as you pop the lid you will get covered in sticky sugary soda.

As and when this start to happen the magma will start to expand and that increases the pressure, and this in turn will push apart the dyke even more, and the gasses will push mainly upwards since they are a lot lighter than the magma it is originating from.

By now we know that a dyke has formed at the old (formerly) dormant Fagradalsfjall volcanic system. From GPS measurements and InSAR we know that the dyke is 7km long (horizontal extension) and that it is 5km deep (vertical extension), and that it is 1.2 to 1.5 metres wide (dilation).

Depth of earthquakes over the last 11 days. Image by Gaz Dale from data by the Icelandic Met Office.

We know that in volcanic rift systems dykes prefer to extend horizontally so that they become longer as magma continues to pour into the expanding rift. This is due to the mass of the magma creating a disproportionate pressure lengthwise.

This will continue until the expanding rift encounters harder colder bedrock at the ends. When this happens, the systemic pressure will jump up a notch.

In a system like Fagradalsfjall there is a lot of pent-up strain, up to 22 meters worth of it. This means that the lava can push the sides of the dyke apart until all strain has been accounted for.

What now, 22 meters? Well, that is the maximum figure of the accumulated motion of the MAR at Fagradalsfjall since the last eruption in the region. From this we obviously need to deduct all swarms that have formed dykes since the last eruption in the region.

To this question we do not have an answer. So, the available pent-up strain might be another half a metre, or several meters.

Latitude of earthquake locations for the last 11 days. Image by Gaz Dale over data from the Icelandic Met Office.

And this is the interesting part, because as soon as that is done the pressure will notch up again, and the dyke will start to propagate upwards, plus the buoyancy effect, plus the expansion of the magma as it nucleates out volatiles… By this point there is literally no return and an eruption will occur.

Now we just need two more figures of interest. The first is how much magma is entering the system, and that is roughly 20 cubic metres per second on average. This in turn equates to 50 cubic meters per second of lava in its fully nucleated fizzy state if an eruption would occur. At least if we assume a steady state of magma arrival to eruption rate.

Now we just need to know how much magma has been emplaced at this moment. It is so far a measly 0.05 cubic kilometres and counting.

This equates to what is called a tourist eruption in Iceland, complete with nice lava fountains that may come from one or several vents along the rift. There will obviously be little, or no ash if the eruption does not end up in the ocean. I will soon return to the ocean issue.


The miscalculation

Longitude of the earthquakes during the last 11 days. Image by Gaz Dale using data from the Icelandic Met Office.

This section could also be called: When nature kicked Carl’s model in the teeth. I can though comfort myself with being in good company on this.

My initial modelling was based on historic instrumental data. And this gave at hand that the pent-up strain was significantly less than it was.

This made me assume that the maximum possible seismic release would be akin to the famously noisy eruption at Holuhraun, and that when about as much seismic energy had been expended an eruption would occur. Oh boy was I wrong.

Back when I did my research on the Lakí eruption I modelled that it had by necessity had a very high seismic activity, including several earthquakes above M6. This also seemed to be evidenced by written sources and collapsed houses.

Back then nobody agreed with my estimates, but after seeing this I think that it was correct after seeing this utter melee of earthquakes. After all, the strain potential in the area where Lakí happened is much higher compared to Reykjanes.

I do think that Reykjanes is proving my original point. Alas, I digress…

What I had done was to take the historic instrumental data and just extend it backwards for Reykjanes all the way back to the last eruption.

In hindsight it is easy to see where this model was wrong, and I should really have caught my mistake. After an eruption, the strain build-up will be aseismic. This means that for quite some time there will be no earthquakes releasing the strain, and this creates a bias in that you would assume that there is more released strain compared to the reality of nature.

Second mistake was that I assumed that the crust was more plastic compared to what it turned out to be. This in turn meant that you get more, and larger, earthquakes compared to the size of the dyke since it takes more power to crack the rock.

Therefore, my original estimate of the needed time until an eruption would start was off. Originally my original estimate was 5 to 12 days (starting the clock from the beginning of week two of the swarm).


Current estimate

The Icelandic word for a dyke is Kvikugangur. This is a pretty good graphic of the dyke at Fagradalsfjall. Think of it as a hanging sheet of magma candy… Image from Iceland Review.

I am still convinced that an eruption will occur, and for the same reasons as back when I started to see this as a runup phase.

That being said, pinpointing exactly when and where is something completely different. Let us start with guesstimating the when (well, at least if nature does not throw something new into the works that I have not yet figured out).

Let us start with the length part. I cannot for the life of me see that the dyke will continue much further south, at least not longer than to near the coast. There is just no evidence that Fagradalsfjall has erupted previously that far south.

This means that the crust is becoming ever harder and more resistant the further south the dyke extends. As and when this happens the dyke can extend a bit more to the north, and then the same cold and hard crust will happen again.

After that, the question is how much the dyke can dilate as the magma pushes it apart. This is obviously the big unknown part. We do know that it is less than 22 meters, and significantly less so, but how much? I have a problem seeing more than another 1.5 to 3 meters, but here is where I was very wrong previously.

Let us say that we have another 0-5 days of horizontal extension, and 0 to 20 days of dilation. After that there is less than 2 days of vertical extension until the magma reaches the surface.

Now, is there any sign that the pressure is increasing that we can look for? Well, yes there is.

Currently the magmatic intrusion has been mostly evidenced as horizontal displacement and only locally have upwards motion been detected. As and when the horizontal displacement dies off and the vertical motion picks up pace an eruption could happen anytime.

So, what is the verdict on that? Yes, there is increasing vertical uplift, not by much but enough to take notice of.

So, the guesstimate is that an eruption is probable to occur within 2 to 25 days from now, unless something happens that is stopping things up, and I see that as ever less likely. I also believe that the eruption will occur in the southern part of the rift.


486 thoughts on “The Happy Dyke of Fagradalsfjall

  1. I think theres another dike forming out at the end of the peninsula.

    Now all we need is Brennisteinsfjoll to have its 6.5 quake, that would probably lead into a magmatic swarm and as has been pointed out most of the previous cycles have started here, maybe this is how it all starts, with indirect activation, leading us astray to its neighbors 🙂

    I also wonder if Hekla up to 1104 was activated by the same tectonics. Reykjanes is connected to the SISZ, which connects to Hekla on its other end. 1104 was not an exact date of an eruption on Reykjanes but it is within the active period, a few decades before the Krysuvik Fires.

    It would be interesting to have a Hekla eruption in this mix today, or maybe it would be an eruption at Vatnafjoll, creating a nice little basalt cone row that I will now claim in advance as Chadagigar 🙂
    Grimsvotn is also going to erupt soon too, Iceland is very much making up for its 6 years of nothing after Holuhraun.

    • We will see If it will erupt, one of the more sleepy areras of active volcanism in Iceland.
      Any eruption will not be huge, and will be short fissure eruptions. There is no shallow magma chambers in the Reykjanes Penninsula, If that was the case then you woud have collapse craters, and even shallow evolved melts. Reykjanes is almost purely feed by dike intrusions from the deep litosphere I think.
      The peninsula is far from the Hotspot in Iceland, and the magma supply is not high. Magma pockets will take a long time to grow to an eruptible size.
      The central volcanoes on Reykjanes are poorly developed. Perhaps Brennsteinsfjöll may resemble a long lava shield, and Hengil haves a subglacial hydroclastic lava pile construct near thingvallavatn. But mostly eruptions in Reykjanes are short lived fissure eruptions. Some eruptions at Reykjanes Penninsula have made lava shields, but historical ones been fast short fissures.
      Eruptions on Reykjanes have produced alot of pahoehoe lava before, so some eruptions may last months – years. Reykjanes basalts coud be quite hot since they are not cooling in chambers. The pahoehoe near Brennsteinsfjöll are very fluid and thin, reminded me of molten stearine thats been poured. In Google Earth the white mossy Reykjanes flows are the Aa lava, and the green humocky flows are the pahoehoes.

      I do not think that this will erupt, eruptions are uncommon there on a humans lifetime spann. Reykjanes is the very slow spreading Atlantic ridge, and Without much influense from Icelandic Hotspot thats under Vatnajökull area. There is only limited minior plume influense on Sourthen Reykjanes. Reykjanes Penninsula eruption frequency is similar to Saõ Miguel in Azores, perhaps a bit more frequent. But If you live in Reykjavik hoping for an eruption near the city is something that pointless.
      Most magma in Reykjanes Penninsula is also trapped as passive rifting.

    • Hekla 1104 was actually in 1108. It was explosive. Something re-activated the magma, probably by reheating, after it had been going stale. It happens to volcanoes all over the world and is not particularly related to a transform fault. There is no evidence for a flank collapse, as far as I know.

    • Chad, you have a good point here.

      I noticed a minor association between Hekla eruptions and major SISZ earthquakes (one occurring within 1-2 years of the other). This alone is very interesting. Last Hekla eruption occurred within months of M6.5 earthquake in South Iceland. So let’s watch out for Hekla.

      I also find it is interesting that Hekla long dormancy was broken in 1104 during the middle of the active eruptive period at Reykjanes. But this is a single occurance so I can’t say there is a causal link there.

      Researchers have also suggested a link between long term alternating eruptive activity between Reykjanes and Vatnajokull (with one zone calm while the other erupts). But I noticed a different thing. How certain periods in Reykjanes started within years of a major Vatnajokull eruption. Such as the occurrence of Eldgja and Vatnaoldur at the start of the last eruptive cycle in Reykjanes.

      Finally in the last eruptive period in Reykjanes, it’s important to remark that there was also activity at Langjokull (a shield eruption there around 1000 AD) at the early stages of the last cycle in Reykjanes peninsula. But again as I don’t have more data this might be just a coincidence.

      I wouldn’t be surprised if a major 6.5 happens at Brennisteinfjoll, triggers a big swarm there and the eruptive sequence starts there instead of Fragadalsfjoll.

      • It looks like Hekla always had long intervals before 1108, VEI 5 eruptions about every 1000 years or so. There were multiple eruptions in the Vatnafjoll field in the 500 years before Hekla woke up, including one that was in the 1000s so should have been observed. Its not really clear what interaction exists here though, Vatnafjoll I think is an area that is beginning to turn into a proper rift and a distal extension of the mature rift at Veidivotn, beyond the reach of Bardarbunga, and possibly a part of Torfajokull. Hekla I think will absorb Vatnafjoll as it grows and this could already be happening now but its initial formation seems unrelated, its very complicated for sure, but hopefully a little less complicated in 10 years time 🙂

        I am very interested in seeing what effect a large magnitude quake at Brennisteinsfjoll will do to the faults in the area, Hekla is not very predictable long term but has gone quite overdue of its recent trend, 21 years and counting.

        • Hekla is very young and does not have an established pattern. it has done big explosions in the holocene, of which 1108 was one example. It has dormant periods and active periods, rift, etc. 20 year dormancy is nothing unusual: 50 years is probably typical for the past 1000 years. If I would guess, I think the frequent smaller eruptions since 1940 are because the conduit was hot and open. It ha slowly cooled and hardened since. Eruptions are becoming harder to trigger. A large earthquake in the region might act as a trigger. One at Reykjanes would not, I think.

          • Yes, you are right. Its hard to say much about Hekla with certainty, not while it is young and dynamic.

            The magma supply has increased though since the 1947 eruption, that by itself was nearly 1 km3 of magma which is bigger than most eruptions, the only example that was a lot bigger was in 1766 and the volcano slept for nearly 80 years after that. But after 1948 Hekla erupted again after only 22 years, and then again every 10 years until 2000. The volume of lava erupted in the 20th century was twice that of anything previous in the historic period. Hekla has only got a very deep magma chamber so it shows no obvious inflation on the surface but its magma supply is about as high as Etna, so to me it isnt likely to go much longer, its quite a different volcano now than it was 100 years ago.

            I guess we will never know for sure until it erupts though.

  2. Look how drastically the activity temporarily switsched from the assumed dike to the peninsula’s tip:

  3. A question if i may for those that have an understanding of interpreting drum plots.

    When i initially checked the Krisuvik plot, my attention was naturally drawn to the long tornillo like signals around 6 in the morning, but on checking i can see these were the EQs off the tip of the peninsula.

    My question is –

    The drumplot nearest, Reykjanes, looks like this with the sharp signals of a nearby EQ.

    Krisuvik however looks like this with very flat signals tornillo like signals.

    But, Vogsosar which is a bit further away east, and also a little south still does have the sharp signal response, this time separated out in what i assume are the P and s waves.


    So – is the attenuated nature of the signals at Krisuvik a result of the slushy material in the dike?


    And as a follow up question – is it possible to paste images rather than linking, or do you have to log in via wordpress to do that?

    • Or links for that matter? As they have disappeared from my post…..

      • Not an expert. All I can say is that, when we were looking at Holuhraun, there was a similar effect – if there is magma between the earthquake and the station, you get the more elongated signals.

        • That image did not embed, but it is the eqs on the Reykjanes Peninsula with a depth of less than 1.5km within the last 24 hours.

    • The difference in appearance is simply down to the distance. Very close to the quake, P- and S-waves overlap, so it looks sharp. A bit further away they blend into each other to make this smeared appearance. Further away, the P- and S-waves have separated enough so you can differentiate between the two.

      • Correct. The seismograph closest to the event will show the sharpest peak. Further away, the fact that different parts f the wave travel at different speeds means that they will arrive at slightly different times, giving the drawn-out shapes. If the wave of a nearby quake looks fuzzy (or slurpy), it is time to sit up.

    • Pictures in comments is hit and miss. It is not under our control: wordpress deems some as acceptable and others not. The IMO seismographs apparently are the dangerous variety. If it doesn’t work, give the link with the http bit left out and our readers will figure it out. Too many links in one comment is likely to lead to the dungeons, by the way. Our demon likes to intercept those. We recommend not to use more tan 2 links. Also avoid over-sized pictures as you will just be subsidizing the telecoms.

  4. I bet on jelly materials for that drum signals…could be so called wet sounds?!?…anyway another observation: I suppose that African plate is on the push so here we go on iceland and as a matter of fact all north atlantic rim up to svalbard…

  5. I still think there will be an eruption by Sunday night my time (est) but thats just me, and I’m usually wrong. :p

  6. This dyke intrusion…. have beeeetrayyed uuusssssssssss!! There will be no eruption of this one.

    Just give me Spike Pages ( he is the VC volcano in the car ) his eruptinator volcano controll
    And then I will press Flood Basalt

    • There still might be some hope for fagradalsfj-something in the future but I wouldn’t hold my breath. As someone who is still waiting on that perfect mafic-felsic eruption blend from Katla, we can sit down together for some beer and cry together.

  7. Just a M3.1 earthquake at 0.1 km depth (verified) at 1.2 km NE of Fagradalsfjall, so very shallow. Interesting.

    • I went and downloaded all quakes from the table (888 in all) around 15:50 UTC 19-March-2021. Of those 888 quakes I selected 573 with quality factor 99 and used gnuplot splot to plot them. 25 have a depth of 100 meters, 30 are 200m to 900m depth, and 54 are depth 1km to 2km. The splot clearly shows the dyke and the recent activity off the Reykjanes peninsula. Assuming that 99% quality means that they have been checked, we are seeing a score of quakes in the Fagradallsfjall area around 100m depth. This might be significant.

    • Yes, the depth of the quakes in this part has turned more shallow. Jesper, you should go over there and try to smell the magma😁

      • Have Jasper use the One Ring to summon the magma, make us a Mount Doom!

      • Basaltic eruptions are Soooo colourful … ssssoooo beautyfulll …….

        its ouuuuuurrr..Prreeecccuuuuiiiiiouuussssssssssssssssssss……………………
        Sssssshhhhheeeeeäääää…. heeeeeeeeeee hääää gollum! gollum!!! cough

  8. The webcam on Borgarfjall is no more. There is a message stating that it stopped sending a signal yesterday. The days before that, it was obscured by fog anyway.
    The Keilr webcam is scanning the ridgeline of Fagradalsfjall.

  9. It would be cool to get a microphone on the ground near the expected eruption site, to hear precursor rumblings from quakes and magma moving up.

  10. What was with the idea from the meeting in Iceland that the current activity is purely tectonic? Is there enough evidence of dike intrusion? Be a bit anti-climactic to get all excited over tectonic adjustment with no fireworks.

  11. There are so many possible explanations. Most probably it’s just imagination.

  12. That has to be an eruption surely (on the Mila Reykjanes cam) – underlit clouds from a bright point on the horizon

  13. No eruption plume yet .. magma haves lots of water in it and bubble out as thick steam

    I think this is city and traffic ligths

  14. I’m guessing that bright spot is a few miles East of Reykjanes…. Fagradalsfjall ?

      • “An eruption seems to have started at Fargradalsfjall right in this. Kristín Jónsdóttir, group leader of nature conservation at the Icelandic Meteorological Office, confirmed this correctly in this to a RÚV reporter.” (GGl transl. from RUV webpage)

  15. There is a glow showing on the Keilir cam (non thermal) but they can’t decide where to point it yet.

  16. Possibly confirmed on RUV:
    “A volcanic eruption appears to have begun on Mount Fargradals just now. Kristín Jónsdóttir, Group Director of Natural Resources Monitoring at the Icelandic Meteorological Office, confirmed this correctly to a reporter for RÚV”.

  17. Just like at Kilauea, the run-up was obvious but when or if it would lead to anything we were not sure, and then suddenly there is a glow with no warning 🙂

    The Keilir Fires have now begun 🙂

  18. Should have known the eruption started, this comment section lit up like a Christmas tree! I want more 150 km3 of red hot ultramafic magma

  19. Preeeeeeeeecccuuiiiiouuuusssssssssss!!!!!!!!!!!!!! preeeeeeeeeciouuuuuusssssssssssssss!!!

  20. The world press have not seen it yet.
    Well done Carl and everybody here, you got it right on.

  21. Yeeeessssss yeeeeeeeeesssssssssssss PREEECCIIIOUUSSSSS
    Preeeeeeeeecccuuiiiiouuuusssssssssss!!!!!!!!!!!!!! preeeeeeeeeciouuuuuusssssssssssssss!!!😍🌋🌋🌋

  22. I hear plenty of noise from helicopters. I suppose they are off to see what’s up.

Comments are closed.