Eldgja: Eruption dating

The previous post described what we think we know about the Eldgja eruption. Our knowledge about one of the largest eruptions in Iceland is somewhat limited, surprisingly so given that Iceland was already well populated. One of the few things which seems secure is the date. Eldgja is believed to date to 934 AD, continuing for up to 7 years. But what is the evidence for this proven fact? Is it correct?



Let’s look at how we were able to get such an accurate date on such a poorly observed eruption. There are three strands: Greenland ice cores, written records from Iceland and elsewhere, and volcanic winters.

An example tephra layer in the NGRIP ice core. This layer is just over 28,000 yr old, long before Eldgja

An example tephra layer in the NGRIP ice core. This layer is just over 28,000 yr old, long before Eldgja

Ice cores Eldgja left a strong signal in the Greenland ice cores. The sulphate deposition was phenomenal, larger even than Laki. The ice cores are dated just like tree rings, by counting annual layers. The most accurate date comes from the GISP2 core, where the sulphate peak is dated to 934+-2 AD. The Greenland GRIP ice core shows it too, and here it has been dated to 938+-4. Ash found within high sulphate layer in the core has been shown to come from Eldgja: evidence doesn’t come much better than this. The GRIP core shows elevated sulphate over an 8 year period, 933-941 and the authors (Zielinski et al) argue for an extended period of eruptions. But this should be taken with care: a large spike may pollute the annual snow layers just below it, and this would show up as several years of slightly elevated sulphate before the main eruption occurred.

Book of Settlement Attempts have also been made to obtain dates from historical records. The problem is that these records don’t give exact dates, and don’t actually mention Eldgja. There is one indirect reference to Eldgja in the Book of Settlement:

Gnup went to Iceland on account of his own and his brother’s manslaughters and settled land between Kuda-fleet and Isles-river, and all Swans Haunts (Alftaver); there was a great standing-water then and swan catches thereon. Molda Gnup sold from his landtake to many men, and it became thickly peopled, until the earth-fire (i.e. lava) flowed adown there; when they fled westward to Head-Brink (Hofda-brekka), and made there tent-dwellings in the place which now is called Tent-field (Tjaldavöllr). But Vermund, the son of Sigmund Kleykir, would not allow them abidance there, so they went to Horse-garth (Hrossgard) and made a house there and sat there over the winter and quarrels and manslaughters befell there among them. But in the following spring Molda-Gnup and his companions went west to Grind-wick and took up their abode there. They had a scanty store of livestock. By then the sons of Mould (Molda) Gnup, Bjorn and Gnup, Thorstein Hrungnir and Thord Leg-wielder, were of ripe age. (part 4 ch. 12).

Book of Settlement. Photo from http://valkyrja.com/260815.html

Book of Settlement. Photo from Valkyrja.com

It gives a fascinating insight in Viking life! It was a world born of violence, especially so during the bleak and idle winters. Molda-Gnup is believed to have been born around 885 AD and his brother Vermund 890 AD, both in Norway. They came to Iceland for fear of revenge for murders they had committed (there seems to be a bit of a pattern here):

There was a man named Hrolf the Hewing, he dwelt at Nordmæri [Norway], at a place named Mould-Town (Molda-tún); his sons were Vermund and Mould-Gnup; they were men great at manslaughters and smiths in iron.

Their move to Iceland would have been around 910-920 AD. Taking time to have children and growing them to a ‘ripe age’ (presumably ripe for marriage) may have taken another 20 year, so this puts the eruption in the region 930-940, in agreement with the ice record. The sequence of events suggests a spring or summer eruption. Molda-Gnup is mentioned several times in the Book of Settlement: he was an ancestor of a wealthy and powerful family. One such text reads:

There was a man named Avang, an Irishman by race, he first settled in Botn (= Bottom). The wood was at that time so abundant there that he built from it a seagoing ship, and put in her cargo at the place which is now called Hladhamar. His son was Thorleif, the father of Thurid, the wedded wife of Thormod, who was the son of Thjoster at Alftanes, and of his wife Idunn, the daughter of Molda-Gnup. (part 1 ch 14).

Avang came to Iceland one generation before Molda-Gnup (perhaps around 880), when the country was still well forested. The statement that Idunn and Thjoster lived at Alftanes is interesting, as this area was overrun by Eldgja lava. Molda-Gnup had time for at least three grown-up children , and for his daughter (the oldest?) to have her own family before the eruption. This adds a few years to the estimates of when Eldgja occured, perhaps around 935 at the earliest.

There is one other oblique reference to the eruption in the Book of Settlement (part 4 ch 12):

Hrafn Haven-Key was a great Viking, he came to Iceland to settled land between Holm’s-river and Isle’s-river and dwelt at Din-Shaws (Dynskogar). He foretold a volcanic eruption, and moved his dwelling to Low-isle (Lágey): his son was Aslak ‘orgodi’ and from him the Lowislanders are descended.

This suggests there were precursors to the eruption. A significant earthquake is most likely, doing enough damage (but also leaving enough time) to make people move away.

Non-Icelandic records Strothers has identified one possible reference to Eldgja in European records, by the 10th century Saxon chronicler Widukind of Corvey:

Indeed before the death of King Henry many prodigies occurred, such as: The brightness of the Sun outdoors in a cloudless sky appeared almost nil, but it streamed indoors, red as blood, through the windows of houses. Likewise for the mountain where the overlord of the states was buried, according to report, because the mountain erupted flames in many places.

King Henry of Saxony died in July 936. The text is difficult to understand but as there are no ‘flaming mountains’ in Saxony, Strothers interprets this as a reference to Iceland. The ‘overlord of the states’ was a title of the leader of the Icelandic Parliament (the Althingi), and the first leader Úlfljótr, disappeared from the records in 934, at the start of his three-year term. He would have traveled to the parliament through the area covered by Eldgja lava. Does Widukind imply that he died in or shortly before this eruption? If so, the eruption can be dated to 934 AD or shortly after, in spring as the Althingi convened in mid-June. But one has to be careful. ‘Flames’ was often used in the old records to describe northern lights, in which case this may have nothing to do with Iceland, or volcanoes. There are some anachronisms: the ‘states’ did not come into existence until 964, and the actual title of the Althingi leader was ‘Law speaker’. Úlfljótr wrote the first law (parts of which survive in the Books) but there is no evidence he was the leader. Histories can be open to wishful interpretations, and here the evidence that Widukind talked about Iceland is not that strong.

Elsewhere in Iceland To complicate things further, there was another eruption, almost simultaneous with Eldgja, which left tephra over much of north-east Iceland. It is dated to 938+-6. We know little about it apart from the general location. These hardy Vikings just didn’t think large eruptions as worth noting. Such were the pre-VolcanoCafe days.

Date and duration


Based on the ice core dates and these historical records, Eldgja is normally said to have started in 934. The duration would have been at least 2 years, based on 8 episodes lasting 2-4 months each (the latter is derived from the structure of two lava lobes). The strong sulphur peak covers 3 years, which is one year longer than Laki. This gives a duration of 2 years (the extra year is needed for the sulphate to drop out of the atmosphere). However, the fact that the ice cores show two sulphur peaks dated 7 years apart is often taken as evidence that Eldgja lasted for 6-8 years.

There are some holes in these arguments. Can we do better?

Re-dating Eldgja

Let’s return to those ice cores. They show two events, separated by about 7 years, with the first one much stronger than the second. Ash found in the first event clearly identifies it with Katla, and thus with Eldgja. But the dates derived from the GISP2 and the NEEM ice cores for this event differ by 4 or 5 year.

Sulphate deposits in the Iceland ice cores. From Stigl et al.

Sulphate deposits in the Iceland ice cores, showing that the NEEM and NGRIP cores give an earlier date than the GISP2 core. From Stigl et al.

Michael Sigl, in a Nature article last year, has tried to fix this dating problem. In the 7th and 10th century, twice there was a strong jump in 14C in the atmosphere, which left a record in trees that were growing at the time. The tree rings give precise dates for these jumps, which may have been due to major solar flares. Such flares would also have produced 10Be. Sigl managed to identify the beryllium isotope in the ice cores, but the date differed from those given by tree rings by 5 years. He attributes this to an error in the established time line for the ice cores. People had miscounted. Shifting the ice core dates to the ones from the tree records changes the time line for these cores: the date for the large sulphate spike in 934 now moved to 939 AD, for both ice cores.

This is an intriguing proposal. There are some lose ends: they don’t state why or when the error or miscounting in the ice core dating occurred (dates more recent than 1200 seem fine), and in the revised time line the sulphate spike around 1104 AD has moved to 1108 and no longer aligns with the 1104 Hekla eruption it was assigned to. The peak in 77 AD has moved to 88 AD and can no longer be attributed to Vesuvius. But in light of the other evidence above, the case for Eldgja to have started in 939 has certainly strengthened.

Remember the dimming of the Sun described by Widukind? There are two similar records for this. The Spanish historian Juan de Mariana (17th century) states that there was a darkening of the Sun on two days of different months of the same year around this time, 19 July and 15 October. The first is easily identified as the total solar eclipse of 19 July 939, when the path of totality was near Madrid and Lisbon. For the second darkening, 15 Oct 939, the light of the Sun changed to a pale colour. The Irish Annals of Astronomy, 939, mention The colour of blood on the sun from the break of day until the middle of the next day.

Is the revised chronology correct? The authors make a good case (although the evidence for dimming of the Sun is more dubious because of the inconsistent date of Widukind, plus the fact that major peat fires can have similar effects). There are two concerns. First, the GISP2 core which gives the later data has a larger uncertainty of +-4 yr, so it does not disagree with a 934 date. Second, the re-dating is based on tree ring records. If there is an error in these, the rug is pulled from underneath their method. That is not entirely impossible. If a year is so cold that the selected trees (which are normally from the edge of their distribution, to maximise the effect of climate on the growth rings) do not grow at all, that year will be missed and all older dates shift forward. The main risk for this comes from the extreme climate event of 1258, caused by Rinjani. This may be less likely but more evidence is needed.

Volcano weather

John S. Dykes

John S. Dykes

The final piece of evidence is from the weather. The limited records from this period sadly do not include daily weather reports. Only extreme weather events would have been described. The winter of 933/934 was cold: the Black Sea and Bosporus were reportedly frozen (presumably only along the edges, not fully!). But an even colder winter had happened in 928/929 when the Thames was frozen for 13 weeks, so this was not completely exceptional. After that, no extreme weather is reported until 939/940, when a severe winter happened, listed for instance in the Annals of Ulster. This cold weather affected much of Europe and Asia and lasted three years. Famine followed in Germany and France, first in 940 and again in 942. Records of Chinese weather (Fei and Zhou, 2006, Fig. 2) tell us that the winters of the years following AD 934 were not unusually cold. But extreme cold came in mid 939 with snow in July. The following winter was the coldest in the 30 years between 923 and 954, and the next two winters again were cold. To make things worse, this was followed by two years of exceptional drought. In a recent paper the same authors hesitantly suggest that this was caused by Eldgja, although that presumably started 8 years earlier. In the revised dates, the cold and drought follows Eldgja much more closely.

A similar point has been made regarding the Nile. Its water level is sensitive to major eruptions, which tend to affect the rainfall in its source region. Climate models have confirmed there is a relation. The year after Laki was extreme, with the lowest flow rate over 400 year. A lesser low point happened after Katmai. The flow around 934 was normal, bit it was extremely low during and after 939. Luke Oman (2006) has used this to argue already a decade ago that Eldgja may have been misdated.

Nile flow after major northern eruptions. From Oman et al. 2006

Nile flow after major northern eruptions. From Oman et al. 2006

While 934 was cold only in Europe, the period 939-942 was cold across the northern hemisphere. This is similar to what happened after Laki. Laki has been blamed for hastening the French revolution. How about Eldgja? What is the evidence for political upheaval? Actually, surprisingly limited. In China, the Later Jin Dynasty collapsed shortly after and this was clearly infuenced by climate. The Abbasid rulers in Baghdad crumbled and the Buyid Dynasty took over, but this process had been on-going for a decade or more. The Viking wars continued across Europe and the Middle East but did not change much. There may have been a fortuitous element: Eldgja happened at the start of the Medieval Warm Period, when European climate was benign. Laki happened during the Little Ice Age and this may have worsened its effects.

As an aside, the Icelandic fires affect climate much more than may be expected from there volume. They emit much more sulphur than normal volcanoes, and this has climatic damage beyond their lava size. The distribution is not world-wide because Iceland is so far north. Icelandic sulphate stays mainly at mid-to-high northern latitudes. But this makes things worse because with a much smaller area to affect, it is much less diluted. The concentration may be 5 times or more higher than that of an equivalent tropical eruption. Iceland may be small, but it packs a punch.

From all this, it seems plausible, but not fully proven, that the Eldgja eruption started in spring 939 AD, five years later than generally accepted. It lasted perhaps two years (not 6 or 8 years: this came from attempts to reconcile the different dates from different ice cores), and caused a volcanic winter which affected Europe, China and Africa for three years.

Next time We now know a plausible date. There are some questions that still need to be discussed: what actually caused the fissure eruption? Where did the magma come from? Why did the eruption stop? And when will the next one be? Come back for part 3 of the Eldgja saga!

updated to add the bit about the Nile

Part III

Albert Zijlstra

335 thoughts on “Eldgja: Eruption dating

    • Sorry. I just thought the location (Salton Sea) was very interesting and hoped some experts would comment.

    • A bit about the area.


      In their paper, the oceanographers from Scripps, point out not only the fact that the fault below the Salton Sea is about due to set off an earthquake as large as 7.5 that could very well effect Los Angeles if the shaking moves from south to north following the fault line, but that such a quake could also cause the soil in the local bays and estuaries to liquefy; something that could cause nearby buildings to sink into the earth, such as that which happened in Japan’s latest earthquake.


      Salton Sea Volcano Mystery Solved
      The Salton Buttes, five volcanoes at the Salton lake’s southern tip, last erupted between 940 and 0 B.C., not 30,000 years ago, as previously thought, according to a new study

      Earthquake swarms and a region-wide rotten egg smell recently reminded Southern California residents they live next to an active volcano field, tiny though it may be.
      At the time, scientists said the phenomena did not reflect changes in the magma chamber below the Salton Sea. But now, researchers may need to revise estimates of the potential hazard posed by the Salton Buttes—five volcanoes at the lake’s southern tip. The buttes last erupted between 940 and 0 B.C., not 30,000 years ago, as previously thought, a new study detailed online Oct. 15 in the journal Geology reports. The new age—which makes these some of California’s youngest volcanoes—pushes the volcanic quintuplets into active status

      • It is an interesting area. I will admit to that.
        And in the end it is likely that there will be a new eruption there and if the activity picks up further we will of course return to that volcano in an article.

  1. I have 3 comments.
    1. I loved your reasoning about the “pollution” by sulphates of the following years prolonging a sequence. Important thing to remember.
    2. As a northerner I here have to say something, nobody here would ever date anything in accordance with the northern lights. For people from down south a northern light is a memorable thing. Here it is about as exciting as a rainy day and this was also true for the early Icelanders. One mans wonder is another mans shrugged shoulders. A wonder you see once or twice in your life, the shoulder shrug comes if it happens every third day all of your life 🙂
    3. I have a huge problem with the moving of the dates forward 5 years. That put a lot of very well dated eruptions out of whack. 1104 is non-disputable, the settlement ash is non disputable. And those two pre-date and post-date things. So, someone would then need to find another very large eruption in 1108 and also explain that there is a marked lack of traces for 1104 according to that reasoning. Problem here is that 1104 was on a scale that would leave a trace. The trace would not be as large as the Eldgjá trace since the erupted unwelded rhyodacitic flow field was 1.2km3 and the distal tephra was 2.5km3. (The ignimbrite is a very very light grey and works well as base for mortar and the samples I have stink of sulphur still).
    In all honesty, I feel dubious about this part of the reasoning.
    Also, 77 Vesuvius is also very hard to argue with. You would then need to explain the lack of data for 77 and invent a 88 eruption. It just falls apart in my eyes at this point.

    Thanks Albert for a stunning series!
    But, for the reason stated above I will stay with 934 as the date, and also I would start to search the globe for a 939 eruption since you make a good case for one having existed.

    • I seem to recall having read a paper, on the lack of info on northern lights in nordic mythology, legends, etc. compared to the Sámi who have a lot of myths and legends about the lights(they were also further north)
      The conclusion was basically that the magnetic field of the earth has moved in these last thousand years, so that back in the day most northern lights activity would be over Siberia/Alaska and not the nordics as it is today.
      However I agree with you, there’s no chance dating would be based on northern lights, even though it (perhaps) was a somewhat rarer phenomenon back in the day for us here in the north.

      • What is at play is that the Samí people have a very different mythology and religion. The religion is much more into weather and phenomenon of nature than the forn-Nordic religions were.
        So, for them the northern lights where used to foretell weather and other conditions.
        I think the difference is that the Vikings lived a settled life in farmsteads and that the Samí lived (and do live) much closer to nature (that is far harsher).

    • And now back to happily sit and wait for the conclusion of the series.
      Well written, funny and informative. Loved it Albert!

    • I was doubtful about the re-dating but it does fix a number of problems. The five year difference between the timing of the two (presumed) solar flares between the tree rings and the ice record is a major problem with the old dating, and that goes away if you shift Eldgja to 939. I don’t mind Vesuvius so much: that was locally an important eruption but may not have been big enough to leave a major trace in the ice. A bigger eruption in 88 AD is not a problem, and the new time line does fit older eruptions better. Hekla 1104 is a significant issue, and I have been wondering how accurate that date is.

      Northern lights are commonly (once every few decades) mentioned in german chronicles of this time, normally as flames and swords in the sky. It is so much rarer further south than Sweden! Even in the UK, northern Scotland is the best place for it. But remember that this chronicle is in fact the main evidence for a 934 date, and it is not clear at all it talks about Iceland. Anyway, this is research. It is about questions an making progress in answering them. It is not yet complete.

      • The Hekla event is very well documented in numerous sources from the time, both in Iceland and elsewhere.
        I have a couple of kilos of the ignimbrite at home that I dug up from a dig site. I have to keep it sealed up since it otherwise makes the entire apartment smell like burning rotten eggs.

          • Yes, in sub-atomic print the smell is labeled 15 October 1104 +/-45 days.
            The 45 days issue has to do with discrepancies in the calendars used and in some cases travelers gave the news and it was written down and dated in ports and so on.
            But, for the time it is about as well dated as if we used an atomic clock at Holuhraun to decide the exact time of eruption.

    • She is really persistent this year.
      I am still waiting for one of the swarms to pick of into a prolonged swarm. These have been to short.

  2. The more I read about Icelandic volcanoes the more I realise how complicated they are to understand! 😀

    • And then you have Hekla, that one is enough to crack your brain completely 🙂

      • I still have a theory about a sort of domino effect that happens in Iceland. of course I have no scientific reasoning behind this really but I just imagine there is a network of similar to transform faults radiating out from the spreading area to form a lacy network, some interconnected. Each time the MAR gives a bit of a stretch then that triggers who knows what each side…. and the plume just adds fuel to the fires to the East and may account for the peculiarities of the Hekla modi operandi. Not Modus as Hekla has a few different modi. It may also explain the odd magma mixes.
        ..and another mystery… why when I now try to type quotations marks I get @? It was Ok until Windows updated itself this morning…Grrrrrr!

        • That theory works quite well for Grimsvötn, Bárdarbunga and Katla, but not for Hekla. No, she has to do something quite different just to drive me nuts 🙂

          • Lol
            It must be a Male Mountain. Just awkward and stubborn and not too bothered about fashion and etiquette.

          • No, it is short tempered and you never know where the next blow will come. Definitely female as the name indicates 😉

        • Normally when this happens (the quotation marks becoming @) the language settings have reset to a discern keyboard setup, easy fix.

          I’ll try and catchn you tomorrow and get this sorted 🙂

  3. These earthquakes today and an earthquake swarm of 45 recent earthquakes in California are being discussed on another website .

    Do they need keeping a watchful eye on ?

    2016-09-26 14:33:44.5 33.35 N 115.71 W 15 km 3.0 SOUTHERN CALIFORNIA
    2016-09-26 14:31:08.0 33.30 N 115.71 W 2 km 4.3 SOUTHERN CALIFORNIA
    2016-09-26 13:14:19.4 33.29 N 115.71 W 4 km 3.3 SOUTHERN CALIFORNIA


    • Yes and no, it is in an interesting volcano, but since it has not erupted in a long time it would take quite a lot of activity before an eruption would start.
      I would though be happy if the Salton Buttes erupted, they are an intriguing form a volcanism.

      • Just to make things clear, this would not be a large eruption. It is likely that a dome extrusion would occur and that it would be a “tourist eruption”. Well worth to go and have a look at if an eruption occurs.
        Very small risk for life and limb.

        • Sayeth he, just like the shaman famously told the chief of the Oranui tribe when the latter wondered about the earth tremors at Hatepe some eighteen centuries ago… :mrgreen:

          • Difference is that Salton Buttes have never done anything more than producing new Buttes.
            Oranui had produced previous large eruptions.

          • Oh yeah of little history.
            New Zealand was inhabited about 500 years later, so no shaman said anything about anything there 😉

    • I am here going back to the Salton Buttes.
      In a report issued in 2014 there was 5 new steaming vents found at Red Island. There has been previous earthquake swarms in the region.
      Earthquake swarms there may be originated in the sinking of the Salton Sink caused by a regional spreading, or by intruding magma as volcanism in the Salton Buttes Volcanic Field may resume, or by geothermal energy extraction.
      The new steaming vents point to the possibility of volcanism as the cause, but it may well be more than one factor at play.

        • Are these earthquakes happening all along the San Andreas Fault line ?

          • No. This is activity that comes the day after a swarm ran directly above Salton Buttes that was more clearly related to the volcano. That swarm stopped and a day later this happened just a short distance away.
            This is not directly related to the San Andreas.

      • Is there any significance to the quakes occurring in a depth range of 1 – 14 km?

        • If you are on our Facebook page I just posted the USGS specialist remark there with a comment.
          But I can answer your question. This was a strike/slip tectonic main shock with after-shocks. This means that the earthquakes will be fairly well constrained in the same area.

          • Thanks. I was hoping volcanic as the prospect of your tourist friendly volcano was quite appealing. But not yet it seems.

          • If you live in California it seems that you have to go south for some friendly tourist volcanism. I recommend Fuego, sure to deliver and you can watch it sitting in a roof top bar in Antigua drink in hand and observe the eruption.

    • There have been multiple swarms in this area for at least the last six years or so. The Borrego Springs 5.1 jolt + aftershocks in 2014 and a 5.2 in June of this year comes to mind…as well as a potent swarm slightly SE of the Salton Sea in 2010? Here’s a quick summary of near-term activity near Borrego Springs over the last year). http://earthquaketrack.com/us-ca-borrego-springs/recent
      55 earthquakes today
      65 earthquakes in the past 7 days
      108 earthquakes in the past month
      860 earthquakes in the past year

  4. Interesting article; thank you, Albert 🙂

    I can see reasons why the sulphates could peak later than the onset of the eruption. The eruption may have occurred in phases, some being larger than others. Also the sulphates may take a while to reach Greenland (if they went the long way round). OTH how accurate is ice-core dating? Ice isn’t deposited evenly – it is affected by precipitation, wind and temperature. An eruption as large as Eldja might well have affected local weather patterns in the North Atlantic. Ice core samples represent the position they were taken from.

    • It would take the sulphate a few weeks to months to get around to Greenland. We know from the tephra distribution that the wind was mainly northwesterly during the explosive parts of the eruption. But the earth isn’t that big that close to the pole, so stuff does get around quite quickly. The ice core dating is assumed to be accurate to one year if you can count the annual layers well enough. It is the winter snow and the summer melt that gives distinct layers. In the past, we would have given a uncertainty of 5-10 years this far back. The counting would have used certain ‘events’ to calibrate, so ice cores of the 10th century would have been counted from the known date of Eldgja (934). Without this, now that the three ice cores shown above do seem to give a 5 year difference in dates. Sigl used an automated technique to count the layers which gets better accuracy and does away with the need for this calibration. That technique gives consistent dates between the different cores, and that is an improvement.

      You have to be careful with the ancient dates. People tend to accept dates that agree with the authorative one, but question when they find a different answer. There is a bias towards ‘known’ dates. The evidence that Eldgja occured in precisely 934 does seem a bit shaky. If only we had just one clear historical record of it! But we don’t. Icelandic history is patchy for this time and for instance the time line for the law speakers at this time gives the impression some retro-fitting has been done.

      The later date fits the weather patterns (which are well documented) much better. Look at things such the flow of the Nile. After Laki, it almost dried up, the worst for 400 year, something that climate models have related to the eruption. A similar low extreme happened in 939 – but not in 934. I’ll add this figure to the post. Bailey finds that in the period 500 – 700 AD, the NEEM ice core is consistently 7 year out of date with the frost rings seen in tree records – and that includes the very well known event in 540.

      But it is up for discussion: the case has been made, and now it is up to other scientists to proof it wrong. At this point, I wouldn’t dismiss the later date.

      • I do not really argue the 934 moving to 939, I just have a big argument with Hekla and the settlement ash. Those are well dated, but I am quite happy with moving the 934 and admit that most of my stubborness spins out of it being “the figure”.
        To put it in another phrasing, I find most of Alberts reasoning plausible, but I have a problem with the general moving of dates since it causes abundant problems with more well defined dates.

        What I found most interesting is that Albert aptly points to a big problem in Icelandic, and for that matter across the globe, and that is that dates are uncertain for most eruptions quite close to our time.
        Let us take a look at Brennisteinsfjöll.
        It is close to areas settled since man came to Iceland.
        The 875 eruption is so badly constrained that it has a +/- value of 50.
        The 910 eruption is even worse. That has a +/- value of 75. So, for all points of purposes this could be two hrauns from one eruption.
        The 950 eruption is so badly constrained that it does not even have a +/- value. So here is a third eruption that could be the same event as the previous two.
        Then the Icelandic people get their shit together and do an almost perfect dating with the June 25 eruption of 1000. That one is +/-4 days. Calender issues are the reasons for the uncertainty here.
        Now you would think that they would keep it up? No…
        The 1200 eruption is so badly constained that it does not even have an uncertainty value.
        The 1341 eruptions is +/-1 year so slightly better.
        All of this is of course due to the well placed date being attached to a momentous moment in Icelandic history.
        This kind of pinpoint Alberts point I think.

        • The Hekla 1104 eruption is the hot potato in the argument. The group working on the tree records finds that the error in the ice core dates should have happened somewhere between 1000 and 1100, keeping Hekla as a fixed point. The ice core group arguing for the revised dates shifts the dates more smoothly so that Hekla changes. This will not be resolved until they can show what caused the counting error, and when did it happen. Without that piece of evidence, the date change is a possibility, not yet a fact. In the scientific literature, people seem taken with the new dates, but you have to be careful. They could still be wrong.

          • I am a simplistic man, if we know one date to be certain. And that date has to mysteriously slide away, then something is wrong.
            I did though find something interesting in a paper that I mislaid after writing about Aniakchak/Thera. And that was an outcry from one of the guys countin ice core layers pointing out that one ring sometimes could constitue several years that had melted together as one.
            So, one “year” could contain several years worth of ash and sulphates.
            Apparantly this problem affected the warm period around the ice age the most.
            It was hidden in a side paragraph, but it still stuck in my head since it would make the ice core dating more problematic than normally stated.
            Obviously this is a known phenomenon and something that people try to correct by using known stratigraphic markers to correct things. But, this makes things in between slightly more fluid than one would think between these strong stratigraphic markers.
            So, in theory the stratigraphic marker of 1104 is correct, but the area from there back to the Settlement Ash is fluid. That would leave a period from 870 to 1104, or 234 years that may not be entirely correct for at least two reasons, one being the reason Albert gives, and the melding of years by melt.
            As Albert said, this is what makes science so fun 🙂

  5. I propose that rather moving dates 4-5 years, that there might have been actually two large volcanic eruptions (probably two episodes of Katla-Eldgjá), one circa 933-934 and another circa 939-940. This would make them separated by 7 years and fit most of the data.

    We can´t really move the dates of Vesuvius, Hekla 1104 and Vatnaoldum Settlement ash, those seem to be pretty well defined.

    • I have an idea in that regard.
      The southern parts are unquestionably Katla. Now imagine a smaller Thordharhyrna rift eruption on the northern end that mixes with left over magma to the north in 939. Speculative yes.

      • It’s sort of like how we once thought the Huckleberry Ridge Tuff was the result of a single eruption, but it turns out some of the material was from a second smaller blast, i.e. a very large eruptive unit may be several smaller units combined into one.

          • I read the posts on that, it’s a shame we don’t seem to have a fully clear picture of the events. Even the VOGRIPA database, which is supposed to be very accurate, still lists the Saksunarvatn as a single 10km3 eruption. As it turns out, even that database can’t seem to make its mind up half the time!

            I’ve been compiling my own figures and databases for loads of eruptions lately and have discovered just how much discrepancy there is in published figures (where there are any at all). Having to learn to triple-check everything and sometimes end up having to make fairly informed guesses, if only things were a lot simpler! Can someone hurry up and find the holy grail of volcano research and facts already!! 😀

          • Saksunarvatn must be more than 10km3.
            Oraefajokull, Hekla3 and Veidivotn all reach near the 10km3 mark. And Saksunarvatn is a much bigger event.

          • Now we have to wait until the rest of the academic world catches up and then we can start to pinpoint everything exactly! You know, this volcanoholic bug seems to be quite a pest! 😀

  6. And while we are at it.
    There is another badly constrained very large eruption that could have caused the 939 track. Baekdus 940 eruption is so badly constrained that it has a +/-60 value. It would of course have caused a rather big spike.

  7. I am also curious about which other eruption in Iceland circa 938 you talk about, Albert. Was that Katla itself?

    Consider other events, like the 536 famine/plague/climate chaos event. It has plenty of historic records. But the culprit is not well known yet. Nevertheless you can check the ice cores for that one too.

    There were other big eruptions in Iceland before Hekla 1104 and after Katla 934/939.

    At least one confirmed eruption of the shield volcano in Langjokull, which was circa 10km3. And perhaps an eruption of Snaefellsnes, not sure about it. These might have left sulfate and ash traces in Greenland records, respectively.

    • There is a tephra layer in northeast Iceland dated to around this time (with an uncertainty of +- 6 years). It is not Eldgja (or Katla) but will have been a local volcano. As far as I know, it does not overlap with Eldgja tephra so we don’t know whether it occured before or after. It is just one of those things.

      • Askja? Hofskokull? Bardarbunga? Kverfjoll? Grimsvotn? Langjokull?

        Can´t be many culprits, towards the northeast.

        • Northwest of the area that Albert states has many interesting possibilities.
          There is a VEI-5 lurking there around 1200 that has no known volcano. It is believed to be from Háagöngur (not the same Hagöngur) near Eldgigur.
          You also have Hamarinn, Eldgigur, Geirvörtur and so on and so forth.
          There could also have been a Maar event.
          You also have an undated eruption at Urdarhals that may be responsible (polygenetic shield volcano).
          But if we look at the usual suspects we get a 940? eruption at Bárdarbunga that is badly constrained. 910 and 960 badly constrained from Grimsvötn. Nothing from Kverkfjöll. Same for Askja.
          So, once more we are pretty much stuck with it being the usual suspects that are suspects. Unless of course it is one of the unusual suspects I started with.

  8. So I guess you could now say the date of the eruption is 937+/-4? (alternatively “something happened in some place at some point, a long time ago”) 😀

    • Hmmm… perhaps Eldgja was from 934-937, and Baekdu in 939, leaving 938 as year when there were lingering effects (like how Tambora affected 1816 as well as 1815) before a whole new disturbance?

      • Or how about … Grimsvotn 934, high VEI 5 or low 6, then Eldgja 939, mid 6 with very high sulfur? Explains some traces and references to “fires on a mountain” in 934 and the northeastern tephra fall, *and* the major Laki-level climate disruptions in 939-941.

        • I can with a fair amount of certainty say that Grimsvötn did not produce anything on that scale in that timeframe. The big tephras out of Grimsvötn have all been accounted for and after such large eruptions Grimsvötn tends to calm down for an extended period, something that we have not seen.
          Whatever is causing the dual spike it is either a secondary eruption at Eldgjá/Katla or something from outside of Iceland.

  9. On Eldgjá/Katla, where might a new eruption be? There is a spot in the south edge of Katla ice cap, which has seen an intrusion consistently over the past few years. Not a risk of a Eldgjá-like event there, but a small fissure eruption could erupt there, and that´s very near populated areas.

    Besides this let´s remember the frequency of Laki and Edlgjá regional events. There are only two or three of each in the Holocene, giving them a frequency of one event per ~5000 years. So – unless there is an underestimation of these events – I wouldn´t worry about a future Eldgjá or Laki event! Next one could be thousand years in the future.

    While for Veidivotn-located events, there seem to have been at least around 10 in the Holocene, giving them a frequency near once every ~1000 years. And they have been pretty regular throughout the Holocene. From dating records, the average frequency of the past 3 Veidivotn-like events was around every ~700 years. Which makes us “due” to another Veidivotn event, sometime in the next 200-300 years.

    Thus I would worry more about the risk of a new Veidivotn than a new Eldgjá or a new Laki.

    Another interest of mine is the relative abundance of Holocene shield volcano eruptions around Langjokull. There are around 10 of them. This makes a frequency of around one event every ~1000 years. Nevertheless this frequency is not linear, as more shield eruptions happened in early Holocene than recently. The last shield eruption was around year 1000, so we could perhaps be “due” to another of them.

    North of Vatnajokull, there are about 5 Holocene shield eruptions. It has been long since one of them. So this area is also at risk for one such eruption. Holuhraun was “almost” the start of one. And in Iceland, such eruptions nearly always start in new locations rather than in pre-existent shield volcanoes.

    In conclusion, apparently, there is a bigger risk of a shield volcano eruption in Iceland or a Veidivotn event, than a Laki or Edlgjá repetition!

    • I believe that the most intense quake activity has generally been focused more on the northeast region of the Caldera, which is also implied as a loci of uplift and inflation over the past 20 or so years. Obviously, other areas are also creaking, but this is a region that has not erupted in the last 4-5 eruptions, so it’s tough to really say what would occur.

      • I agree with you fully on that. It will be interesting though to follow when it does occur.

    • There was a paper just a couple of weeks ago about that ‘hot spot’ outside the caldera:


      Shallow LP seismicity is observed on Katla volcano’s south flank since 2011.

      Repeating seismic events occur with regular time intervals and seasonal modulation.

      The seismicity is most likely volcano-related but glacial origin cannot be ruled out.

      The seismicity began in coincidence with an unrest episode within the caldera.

      A concealed hydrothermal system is suggested as source of the seismicity.

      • Very interesting, thanks for the link.

        The location for the quakes they discuss – Gvendarfell – is or is near the location of some earthquakes just off the edge of Myrdalsjokull that have occurred in recent months, and which have been mentioned on here, iirc.

        The paper is also interesting in relation to the discussion on here some weeks back about earthquakes with glacial origins.

  10. All this complexity is giving me PTIVD: post-traumatic Iceland volcano disorder!

    • “If we knew what we were doing, we wouldn’t call it research”

      • Good quote! Who is it from?(the quotation marks suggest it isn’t yours) It sounds like Richard Feynman, it’s about his style

        • I think it is from “Surely you are joking Mister Feynman”.

          • OMG! I have something in common with Einstein! Next time I stand researching my thoughts in the washouse/Freezer room .”Now why did I come in here. What was I doing to make me perambulate towards the freezer”? I shall call it “An Einstein Moment”

    • Maybe tomorrow Katla starts a series of M4s and runs up to an eruption in late October…. and that makes your PTIVD worse!

      Today at 13:31 Katla had a M3.9
      This was 7 minutes after a M2.4 quake.
      Then 3 minutes after the large quake, it seems that Katla was continuously shaking for more 9 minutes !!!
      Several aftershocks, every minute, between 13:34 and 13:41

      Everything was shallow. Between 4km and 0km deep.

  11. On a total sidenote here.
    I have just discovered that it is far more fun to debate and comment when you are debating and commenting someone else’s article compared to when you do field comments on your own article.

  12. So far 7 earthquakes since midnight in Southern California most above 3.5 the strongest a 4.3 all around similar co-ordinates. Three are showing depths of only 2 km .

    Magnitude Mw 4.3
    Date time 2016-09-27 03:23:58.2 UTC
    Location 33.30 N ; 115.71 W
    Depth 5 km


    • Lots of discussion on the internet this part quoted by Dr Jennifer Andrews SCSN :

      26 Sep 2016 07:31:08 PDT, (33.298, -115.713), depth 2.4km, 6km SSE of Bombay Beach, California
      This event is the largest so far in a swarm that started on 26 Sep 2016, 04:03AM PDT, and is occurring in the Brawley Seismic Zone near the southern terminus of the San Andreas Fault.
      The swarm includes 37 events so far (26 Sep 2016, 09:10AM PDT) in the magnitude range M1.4 to M4.3. The M4.3 exhibited strike-slip motion with one nodal plane sub-parallel to the strike of the San Andreas Fault. Relocations of these events show that the are located in the depth range 4 to 9 km.
      This swarm is noteworthy because it is happening near the southern terminus of the San Andreas Fault. This is the same are as the two previous swarms in this region, which occurred in 2009 and 2001. No swarms prior to 2001 with a M4.0 have been recorded in the area since 1933.
      The 2009 swarm had two large strike-slip events: M4.8 and M4.0.
      The 2001 swarm had two large strike-slip events: M4.1 and M3.4.

      Full report can be found on :


    • If you read the replies posted above you will see that your question has already been answered.

      • I might also add that the southern end of the Salton sink is the beginning of the rubble pile that is Southern California. Quite a few fault blocks and uneven stress systems. The Yuha-Wells fault Is the major fault connecting the Elsinore, San Jacinto and San Andreas faults. You can’t rule out that this swarm set is not an artifact of the large Baja quake of a few years ago. While the Elsinore an San Jacinto showed a lot of follow on aftershocks, the San Andreas segment was oddly silent then. It could be tgat it’s finally adjusting to the new stress regime after that event.

  13. And in the regards of Icecream trucks and parking spots in Iceland.
    There was just a M0.3 to M0.8 earthquake at the northern corner of Hekla.

  14. Did the Anglo-Saxons in England also mention the nasty phenomena of this eruption like haze, a sulphurous smell, ashfall in their books?

    • There is one possible german record from 939: ‘the sun did not have any strength brightness or heat . The colour of the sky changed, as if flushed. Others said the sun was seen as if halved’. I don;t have a precise date other than the year: the quote is from Stigl. The similar Irish record is mentioned in the post.

      • Thanks for interesting excursion through viking records ash deposits and tree-rings. Excited to read the third part!

  15. Tuesday
    27.09.2016 13:54:52 63.623 -19.120 0.1 km 3.1 99.0 4.9 km NNW of Hábunga

  16. Katla continues her show off and keeping me awake with a new shallow single M3.1 very close to the yesterday swarm and 3.9 quake..

  17. Thank you Albert! Such a comprehensive post. Lost of evidence, lots of criticism en doubts mixed with alot of research. sounds plaussible :c

  18. Bromo raised to Orange alert, Rinjani raised to Yellow, looks like a small eruption


  19. And around volcanoes what you see is not what you get.

    Momotombo has been a steamy staple for me the last week or so. During that time it has shown every volcanic signal known to mankind and at the same time the amount of steam produced has steadily increased.
    All of a sudden today the steam disappeared while the volcanic signals remained at a high level with interesting “throbbing” signals. So, the activity remains high, but the steam is gone. Either the rising magma has boiled of the aquifers inside the mountain, or there is a blockage. Either of it is a sign that problems may be on the way.

      • There are some puffs, but quite a lot less than the big continuous billow that took place in the last week.

        • Returning to the glow we see at night. Is is not an evidence, of the “lid” not being entirely shut? What besides magma would cause glow?

          • Fumaroles hot enough to heat the surrounding rock to incandescence might do it; Momotombo has always had such fumaroles, even before the present ‘active’ period began. Magma is still ultimately responsible (it provides the heat) but indirectly

  20. And when you clog something up that is under pressure something will break, sometimes sooner than later.

  21. Could someone please explain if these charts mean anything regarding the latest earthquakes early this morning in Southern California are they LP events.


      • On the graph I posted these look like LP earthquakes to me and I think this means Magma is on the move so therefore it must be volcanic related.

        I may be wrong could maybe someone else confirm if this is the case.

    • No, those are still tectonic events, they are a bit clipped that makes them look a bit off.
      The length is caused by reverberation against something, I do not know what the signal is reverberating against though.
      The current activity is happening NNW of the magma reservoir under Salton Buttes, so it is not magmatic as such.

  22. Janet, I’m not sure but I think it looks like LP events, but I’d rather you got that from one of the others who are more experienced than I am.

    • It says ‘without warning’. Should have listened to mjf above.

    • Quite a valid point. Especially since many had started their trek from the monitoring station.

    • Sadly I have been waiting for something like this to happen for a long time.
      It is not like there was not a warning in place. Still you have tour-operators taking people up volcanoes that are pretty darn unsafe. And tourists who wishes to see an active volcano.
      Yes, I do climb volcanoes, but I know what I am doing (fairly at least). I am still aware that there is a risk involved. And, there are volcanoes I do not climb since I prefer to live.
      I hope this was the end of idiotic irresponsible volcano tourism. I do not mind sensible safe volcano tourism though.

      @Albert: Rinjani was on the warning list complete with an exclusion zone. Still the tour guides had taken them there.

      • We’ll have to wait and see on this. It all depends on what is meant by ‘climbing Rinjani’ If they mean the main cone, Rinjani proper, there wouldn’t necessarily be a problem. Cl;imbing Barujari however would be insane. Rather like the difference between climbing Santa Maria and climbing Santiaguito

    • And here it is straight from our contact at Badan Geologi, Professor Igan S. Sutawijaya:
      “Of course Carl, we have a hazard zone map, and we are monitoring the volcanic activity in 24 hours a day, but the eruption appears to be a phreatic eruption that occurs suddenly because of a shallow hydrothermal pressure conditions.”

      And this put the finger straight onto what makes volcanoes so dangerous, they can all of a sudden do the unexpected, especially if there is a hydrothermal field around that can suffer from phreatic detonations.

  23. The Neighbourhood has gone to poop.
    I am seriously interested in buying a ticket to Mars.
    How about it Albert, you wanna come so we can argue about volcanoes all the way?

  24. Mila’s Katla webcam looks odd right now. Is that steam or just a weather effect?

    • I believe it is clouds forming from the pressure drop, when the wind sweeps down from the glacier.

    • Trust me on this, the large Icelandic volcanoes does not steam prior to erupting.
      Instead you get days if very intense earthquake swarm, and then hell opens up and you will see something looking like very large nuclear bombs going off. But one thing is sure, there will not be fluffy small clouds around 🙂

    • Hmmm i do see what you mean but i suspect PederP is correct in that these are probably orographic/katabatic clouds. Has anyone got the coordinates for this web cam? From the shadows i assume it is looking in a southerly direction towards katla?

      • From the topography, vegetation, and the shadows, I’d say the webcam is located SE of Katla, looking towards the eastern side of Myrdalsjokull, *possibly* in the vicinity of Vantsá.

        • While Ian was giving his precise reference, I was taking longer doing a dead-reckoning comparison of the image and the map. Allow me to congratulate myself on being spot on!

          • Spot on, but not with my spelling – Vatnsá. Presumably the camera is on Háfell.

  25. And eruption may have started at Momotombo, or there is phreatic explosions. It is hard to say so far. Both cams are down for the moment.
    The seismometers are indicating strong activity.

    • This is an intense M5+ earthquake swarm, no eruption yet. But these are majestic precursor earthquakes. Feels like it is the Loch Ness Monster crawling up that chute.

      • M 5.1 – 16km ENE of La Paz Centro, Nicaragua
        2016-09-28 16:48:55 (UTC)
        12.423°N 86.548°W
        6.7 km

      • That’s quite an intense quake. Does anybody know quake intensity prior to the 2015 eruption that occurred here? Would be an interesting comparison.

        • This is more intense. Especially since there are at least 5 M5s now.

    • Nothing on the webcam 11:43 local time. But apparently a 5.5 mag eq according to local Tweets.

  26. Momotombo webcam 3 image from 2016 09 28 11:14:

    Admin Edit!
    No ned for tags. It will just put the comment in the pending folder.

  27. …and one more from 11:43 for comparison (and to help with the traffic):

    • Still “just” puffing.. and me trying to insert picture here. Sorry if it fails.

      • Almost seems like the quiet before the storm here.. the puffing almost died/dried out?

  28. According to INETER the volcanic observatory was damaged and had to be evacuated.

    • Initial earthquake is now manually corrected by USGS at M5.9 at 6km. 4 aftershocks are at M5.

      • Wow, well this is worth keeping an eye on. Gigantic blob of magma working its way up from very deep?

      • Reports are saying quakes are not related to the volcano

        CRA. Rosario: not are of origin volcanic, not have relationship with none of the two volcanoes (pit and #Momotombo)

        • INETER says it was right under Momotombo and the signal is volcanic.
          Very odd comment that.

          • The build-up started with an M5 a week or two ago, close to the volcano, so not it is part of a sequence. It doesn’t matter if it is directly volcanic: a tectonic quake of this size can also create magma conduits and so indirectly affect the volcano. An eruption may not be imminent but the risks will have increased.

        • There was significant harmonic tremor prior to the current event, right? If so, that’s a dead ringer for magma overpressurizing a lower region, only for it to pop, and then start pushing towards the upper conduits (creating the m5 quakes).

          • Yes there was, and there has also been many HT and VT events in the last couple of days leading up to this.

      • That is caused by the facing side of the steam plume being in shade. No smoke yet.

      • Are we quite sure that is natural? Only that the POV of that camera is from the geothermal station, and geothermal plants do vent steam. Carl?

        • It is a view of the geothermal field yes, but that is from a hydrothermal well that is natural.

    • Thank you Spike!
      This makes leafing through the world of volcano webcams so much easier in the morning.

    • also added the webcams onto the top of the sidebar, meaning the cameras are easily accessible from the top of each page and post 😀

    • wow, thank you so much for that resource Spike! That just makes life so much easier.

    • help! Non of the links are opening now. Windows updated last night! I have Flash Player that is working. What happened?

  29. There is a strange signal which appears very widespread across central to NE Iceland that appears for the last couple of hours?

    I do not see any significant quakes somewhere else that could be the source and it does not show up as you go towards the east.

    Seems to be low energy, not enough to be so widespread.


    Here is one plot, also shows as a peak on tremor plots


    • There is a series of deep earthquakes north of Kolbeinsey causing this.

      • Thanks Carl, I see that now.

        The signals are showing up almost as strong on the plots SW of BB as they are on Grimsey, that’s what fooled me

  30. USGS now showing only a single shock, 5.5.? Moment tensor indicates a slightly oblique lateral strike slip (transform) fault mechanism. One thing’s for sure though, the old beer bottle got a good shaking.

  31. Thank you Spike, I have something to look at while my wife is doing genealogy on her computer…

    • I know how you feel, my problem now is that with 40 new trailer parks on the peninsula,
      I need to start evacuating 2 weeks early to be able to get out.

      • Trailers (Caravans) in the UK are a motorists nightmare on summer coastal roads or on major roadways leading to such areas. I can see where you are coming from or trying to! Are you on the same side of Florida as Lurking?

        • Probably not likely. I’m far enough west to be concidered lower Alabama.

        • Culturally, we are the same. I’m from MS and this area is sociologically no different.

    • It really shows the ‘curtain of fire’ type of fissure eruption.

      BTW is there any way to edit comments?

      • Only admins and editors can edit comments for security reasons.

  32. …okay…

    “Why did the chicken cross the road?”


    • It did so to get the Burmesian slow-cooker to boil the breakfast eggs.
      A Burmesian slow-cooker is the latest must have item for all chickens according to the Norwegian Chicken Style Review.

  33. Admin/
    I removed to comments that contained image links that was dead. I do not know why the first one did not work.

Comments are closed.