Volcanoes of the North: the fire in the ice

A freezer provides storage. And a deep freezer gives deep storage. You can’t go deeper than an original ice-age glacier. In Greenland, those glaciers are still there, being so tall that the top has remained in deep freeze even though the ice age is a distant memory. In this cryogenic environment, that frozen memory is still there. Even the heat of volcanoes was no match against the cold of the ice. The volcanic gas and dust that came down here was captured in snow, frozen in ice, and stored far below the surface. But no one documented the falls. Now we have a record of the past, but without a rosetta stone. We can see the story of the fire in the ice, but reading it requires more information. This post is about that attempt.

In the previous post we looked at how many eruptions we can see in the ice of Greenland and Antarctica, focusing on the centuries between 900 and 1500. The ice shows when something erupted somewhere and can separate the tropical eruptions from the southern and northern ones. Finding out any more, that was hard. It would be nice to know whether eruptions were brief or long lasting. It would be even nicer to know where the volcano was. And the ultimate price is finding out which particular volcano it was. That really does need a rosetta stone.

I will for now focus on the northern hemisphere. This is out of a slight obsession with the Viking world (did you notice?), but also because it should be easier to identify specific volcanoes. At the time, the southern hemisphere was still largely an unknown world, a volcana incognita. In the north, the world was beginning to be documented.

Let’s start reading. One warning: do get some coffee first as this post is quite long. The northern hemisphere turned out to be a big place. (As a test: there will be three millennium eruptions coming up. Can you find them?)

The freezers next door

As helpfully pointed out in the comments on the previous post, a single ice core can miss eruptions, for lots of different reasons. The wind may have blown the wrong way and deposited the ejecta next door. A hit summer may have melted the annual snow and poured the dust off the glacier. It would be better to use more than ice core. The previous post used only the NEEM core, from the northwest of Greenland. But Greenland has been drilled in many places and there are other cores available. Several of these ice cores have been analyzed for volcanic signals and may have seen things that NEEM did not.

Sigl et al published data for three different Greenland ice cores: NEEM, NGRIP and TUNU. The NEEM ice core is from the northwest of Greenland. NGRIP (northern GRIP) is near the centre of Greenland, and a bit closer to Iceland. TUNU is from the northeast of the Greenland icecap, in a location which has airflow from northern Europe.

The sulphur records of the NEEM, NGRIP and TUNU cores are compared below. The two panels show the periods 900-1200 and 1200-1500 . There is clearly much more detail visible in NGRIP. Some peaks are much stronger in NGRIP than in NEEM, and may be absent in the latter. NGRIP shows over 50 separate peaks, or roughly one per decade. Many of these may be from Icelandic eruptions as it erupts at roughly that frequency. The NGRIP core is a bit closer to Iceland and NEEM can be in a rain shadow behind Greenland’s glacier, for winds from the direction of Iceland. For more distant eruptions, the frequency of once per decade suggests that NGRIP will typically VEI 5 eruptions. TUNU shows fewer volcanic peaks and does not appear to add much to NEEM and NGRIP. I will not use TUNU for this post.


A new ice core from east central Greenland has recently been made available by Seyedhamidreza Mojtabavi et al. 2020 https://cp.copernicus.org/articles/16/2359/2020/. The location is a little closer to Iceland so this ice core seemed worth looking at. The available data includes an age-depth profile and electric conductivity measurements done throughout the ice core. The conductivity is due to sulphates, but also to ammonia and salt from the ocean. However, the ice cap is very high and the ocean is far away. I will assume that the conductivity is dominated by volcanic emission.

The data does not give an age for every depth at which conductivity was measured. To calculate the ages, I interpolated the age-depth profile of EGRIP. In the original data tables (accessible from https://www.iceandclimate.nbi.ku.dk/data/), the age-depth profile is given in the so-called GICC05 (Greenland Ice Core Chronology 2005) calibration. GICC05 differs slightly from the modern tree ring record. I have applied the correction from Sigl et al for this. The main effect of this correction is that dates before about 1250 become a bit younger. The difference reaches 5 years for dates before 1000. This correction moves Eldgja from the original date of 934 to 939. The correction is less certain between 1000 and 1250 where I fitted a curve, and here the ages I find may be slightly off.

The EGRIP profile shows over 20 eruptions. This is far fewer than NGRIP: the conductivity measurement is perhaps not as sensitive as the sulphate measurement. If a peak is seen in EGRIP or NEEM, it is also visible in NGRIP. There are some differences in shape. The peak around 1110 is seen in all three ice cores but looks narrower in EGRIP (discussed further below), and this is also seen in the peaks around 1390 and 1460.

Time resolution

Time resolution is also important. For NEEM, we looked at annual data, i.e. the total sulphate over a full layer. But if an eruption is brief, it may only show up during a small part of the year. In that case, the event may stand out much better in higher resolution data, e.g. monthly measurements. In fact such a data set is also available for NEEM. In this post, I will mainly use the monthly NEEM data, rather than the annual data used in the previous post.

Eruptions on ice

So we have melted the ice cores (a little bit of it) and have retrieved the frozen smoke from the ancient volcanic fires. But where was the fire? Can we identify individual eruptions? We have one advantage. The chosen era of 900 to 1500 is when the Vikings had settled Iceland and had begun to live in the most volcanically active region in the world. They left us historical reports for Icelandic eruptions. The accuracy of those records should not be overstated. There are no written documents from early in the settlement: the oral history was written down only centuries later. Eldgja, the worst eruption of the entire post-settlement era, has left us no historical record at all. The first historically documented eruption is from Hekla just after 1100, but in spite of being reported with a precise year, this date appears to be wrong. The historical record becomes more reliable after the late 1100’s, but seems to disappear again after 1400 when the black death reached Iceland.

A brilliant list of reported eruptions (much copied on the interweb) was compiled by Jon Frímann Jónsson and kept at https://icelandgeology.net/?page_id=8046. When reading this list (it is highly recommended), do be aware that the sources on which this is based are themselves not always 100% reliable. Historical descriptions of events also do not directly indicate the size of an eruption.

Iceland is our best case. There are places with better records, but these have few or no volcanoes, whilst other places with troublesome volcanoes lack proper documents. There are no reliable historical records for any tropical or southern eruptions for this period. Nor do we have written records for northern Pacific volcanoes, with one exception: Japan. This makes it very difficult to assign particular ice peaks to specific volcanoes. There is an escape route: if tephra (volcanic dust particles) is found in the ice, it can pinpoint the volcano that is responsible, as the composition of the ejecta is different for each volcano. This has been done for a few eruptions within our time period: tephra has been found for one tropical eruption, one in Korea, one in the US, and two in Iceland. This already gives us an indication where the volcanic culprits will be found.

In the first post, I listed eruptions that occurred in the southern hemisphere, in the tropics, and in the northern hemisphere. The southern ones are absent in the Greenland cores. In many cases it will be almost impossible to pinpoint the southern and tropical volcanoes. As an example, Chile has 90 active volcanoes, has a significant eruption every few decades but has little or information on explosive eruptions during this era. How do you identify which Chilean volcano caused which Antarctic ice core spike? And that is just one country. (Admittedly Chile is not your average volcanic country but a stand-out performer.) Indonesia has 121 active volcanoes – it presents the same problem for tropical eruptions. Eruptions from the northern hemisphere may (perhaps) be easier to identify. Here I will mainly (but not exclusively) look at Iceland: can we relate the smoke in the ice to the fire in the documents?

If you are not fully familiar with Iceland’s magmatic scenery, I am happy to recommend the woolly mammoth guide to Iceland’s volcanoes. Read it, enjoy, but do come back to finish this.


On our journey through Iceland, the first stop should be at the fissure region between Katla and Vatnajokull. It is colloquially known (but only in VC) as the ‘dead zone’ because of the lack of earthquakes. This region seems a funny choice to begin our search: it lacks volcanoes. But the lack of earthquakes tells us there is something going on. Regular magma injections from the volcanoes on either side have kept the upper crust here hot enough to avoid brittle earthquakes. In spite of the absence of deep magma chambers, this is a highly active zone. In fact, the largest eruptions of Iceland have happened here. This is the domain of Eldgja and Laki. And our exploration begins with Iceland’s biggest eruption of modern times: secret Eldgja, unheralded and unrecorded, but 30% larger than Laki.


We have written extensively about this eruption. It opened a 70-km fissure, in the same region as Laki but with magma obtained from the other end of the region. The magma was sourced from Katla. There is some suspicion that Eldgja is the reason for the lack of any historical record of the settlement era (870-930) of Iceland. The eruption coincided with the end of this era (not accidentally, I expect), destroyed a coveted area of Iceland which was wooded and fertile at the time, and cut the travel links between east and west. If Laki reduced the population of Iceland by a quarter (through both death and migration), what did Eldgja do? We do not know. The settlers left us no diaries or stories, apart from suggestions that the imagery of Ragnarok was borrowed from Eldgja. Did everyone able to document the events die from it? That is entirely possible. Travel was agonizingly slow, and a sedentary population could easily have been overtaken by events. Farmers like to live near (but not too near) rivers. Lava would have come down from the highlands following those same rivers. Without communications, there would have been no warning of the streaming lava arriving in the dead of night.

Eldgja gives an obvious signature in many Greenland ice cores, in fact it is the largest ice core signal in the entire post-settlement sequence. Small tephra fragments were found in the ice core, and the composition matched those found on the ground around the Eldgja fissure. This proved the association of the sulphate spike with Eldgja. The ice now allowed us to precisely date the eruption.

Click on image for full resolution

All three Greenland ice cores show the extreme spike. It is even (very faintly) seen in Antarctica. That goes against expectation. Eldgja was effusive, with limited explosions. At 20km3 of lava, it put a massive amount of sulphate in the lower and middle atmosphere but little or none in the stratosphere. The wind spread the sulphate across the north and to the tropics. Here, being in the troposphere it passed below the barrier of the stratosphere to make the forbidden crossing into the southern hemisphere. This Icelandic fissure eruption could be smelled over the entire globe. No other eruption could have done this: it took a massive amount of tropospheric sulphate to travel that far. This is perhaps a warning how bad a proper flood basalt would be: such an event could poison the air everywhere in the world.

The high resolution (monthly) NEEM core puts the onset of Eldgja as very sudden in the late spring of 939. The emission peaked within months, declined, showed a second peak towards the end of the year, a phase of constant, slower emission in the first half of the following year, and a slow ending during the remainder of that year. It lasted for 18 months in total, but most of the emission was in the first few months. That is very similar to Laki, which also lasted a year and a half, but spend most of that time petering out. The summer eruption of Eldgja must have devastated Iceland’s farms.

The NGRIP and EGRIP cores show the same pattern as NEEM but with less detail. There is however one clear difference: they have the peak one year earlier. The eruption was in 939 according to NEEM but in 938 in the other two cores. This is not the vagaries of the wind, but a difference in layer counting and age calibration; it is within the uncertainties. (Do note that these dates are in the tree-ring calibration; in the original GICC05 calibration the eruption was in 934. Also note that I adjusted the EGRIP core to the Sigl et al calibration.)

Can we tell which core is right? Here the lack of historical records from Iceland becomes problematic. There are however a few records from Ireland and from Italy which mention a red and faint sun in 939. Tropospheric sulphate, as in a fissure eruption, is different from stratospheric sulphate, as after a large explosive eruption. Tropospheric sulphate moves with the wind and changes from day to day. It doesn’t give months of faint sun like Tambora did. The records talk about days when this happened, not weeks or months. Tree rings show that 939 was a poor year for tree growth, but this lasted only one year. The winter of 939/40 was rather cold in western Europe, with frozen rivers even in Ireland. It was followed by famine which became worse later, especially in France in 942. The details are insufficient to call this a volcanic winter. The cause of the famine is not clear, since the weather was not that poor – drought may have played a role, but it is also very much possible that the clouds of sulphuric acid damaged crops across Europe. The Nile had low water levels in 939. The Nile flood is June – September, and is fed by summer rains in the mountains of Ethiopia. A similar low Nile flood was seen after Laki, and it suggests that sulphate somehow limits the monsoon rains, perhaps by suppressing convection in the atmosphere. It appears that the main influence on the weather was in the summer of 939. The data fits best with Eldgja happening in 939, as in the NEEM core. The case is summarized by Oppenheimer in 2018, https://link.springer.com/content/pdf/10.1007%2Fs10584-018-2171-9.pdf

There is a bonus eruption in the ice. Did you notice the small peak at 946? Tephra in the ice core has identified this as the Paektu eruption, also known (not entirely accurate) as the millennium eruption. It is clearly seen in the NGRIP core (note that this should be shifted up by one year). Not all Greenland volcanics are from Iceland! But the Paektu signal, in all its high-VEI-6 strength, is dwarfed by Eldgja.


There is another dead-zone rift eruption hidden in the ice. This rift eruption is Veidovotn, much less well known than Laki or Eldgja. It took place in the same general area of as Eldgja, a bit to the northwest along a 40-km rift between Bardarbunga and Torfajokull. While Eldgja was fed from Katla, Veidivotn counts as a Bardarbunga eruption.

This was an unusual eruption: even though it was a fissure, it erupted mainly explosively, with 3.5 km3 of tephra but less than 0.5 km3 of lava, all basaltic. The lava has been dated to 1480+-11 years. And just as for Eldgja, there is not a single written record of the eruption. It is as if Iceland lives in denial of its rifts. I blame the Vikings.

The eruption is very obvious in the ice signal. It is strong in NGRIP but weaker in NEEM. Unlike Eldgja, it is not present in the Antarctica (blue) core. The signal begins in the autumn of 1476, peaks early 1477, perhaps February, and ends in spring, perhaps April. The eruption was during the winter, and peaked at the coldest time of the year. This may explain the lack of a written record. It was not the season to be outside in the harsh landscape of the dead zone, the lava did not get far, and the eruption had limited impact on farming.

click on image for full resolution

Does the time of the year explain the explosive nature? This is an area currently known for its lakes. In winter, the lakes would have been frozen and the eruption needing to break through the ice.

Let’s move on from the dead zone to the massive glacier of Vatnajokull, source of the majority of Iceland’s eruptions. We can be rather quick, since eruptions here are generally not documented. It is too far from any settlement, and eruptions here were largely ignored. Often only the jokulhaups were mentioned, the floods coming down the rivers which often -but not always- have a volcanic origin. There are likely to have been a large number of eruptions at Grimsvotn during this time. some of which may have reached VEI 4 or 5, but we know nothing about these. There is however one exceptional Vatnajokull eruption.


In June 1362, a large explosive eruption of Oraefajokull turned the remaining farmable region on the south coast into a wasteland. The output is estimated at 2.3 km3 tephra and 0.1 km3 pyroclastics, and the eruption column may have reached 30 km. This was the largest rhyolitic explosion in Iceland after the settlement era. A stratospheric VEI-5 just 130 km from the ice should give a nice, strong signal.

click on image for full resolution

It didn’t give any signal. This large, nearby, perfect eruption is invisible in all three ice cores. We know that the ash reached Greenland, because tephra has been found there. But sulphates show nothing. It is a mystery.

There are several possible reasons for the missing signal. Wind direction may also play a role. Ash is distributed lower in the troposphere and sulphate higher: if the wind direction was different at different altitudes, ash and sulphates may have taken different paths. The answer may be in the rocks. Rhyolitic explosions blow out lava which has been in the crust for some time, perhaps centuries. Over time, the sulphur escapes from the lava, especially it it is within a few kilometers from the surface. This affects fissure eruptions much less because those erupt fresh magma. Rhyolitic eruptions get much of their sulphur not from the magma (which is degassed) but from the rock which they pulverize. The sulphur budget of such as eruption depends on the composition of the mountain. In the case of Oraefajokull, the rocks perhaps contained rather little sulphur.

There is no indication of any effect on the weather in Europe from Oraefajokull but that is not unexpected as the eruption was a bit too small for that. In 1362, the storm of the century devastated Ireland, England and Western Europe, with reports that trees were sucked out of the ground complete with their roots, but this happened in January, 5 months before the eruption, and can’t be blamed on it. (As discussed earlier in this blog, in Germany a flood disaster has been named after this storm, the Marcellus flood. That flood, however, did not happen. There were major floods at other times in the 14th century but not in 1362.)

In our Icelandic journey, we turn to the southern volcanic area. Here, the usual suspects are Katla and Hekla, two very different volcanoes.


Katla is a fairly regular eruptor. Since 1918 she has taken a break, and is doing a Sleeping Beauty impression while waiting for the kiss of a passing volcanologist, but don’t let this fool you. Before falling asleep she had an explosive temper, and you would be well-advised to check compatibility before marrying the beauty even to a volcanologist. Katla has produced 20 known eruptions since the settlement, not including Eldgja. 11 of those occurred in the period 900-1500: 6 small, 2 moderate and 2 large. The ‘official’ list is tabulated below.

From T. Thordarson, G. Larsen, Journal of Geodynamics 43 (2007) 118–152

For the 920 eruption, there is a peak in the ice core in 919 but it is also seen in Antarctica and so was not Icelandic. The date of this Katla eruption is rather uncertain. The same is true for its 1179, 1357 and 1440 eruptions, where the dates are too uncertain to allow for a search of the ice record.

The 1245 eruption coincides with a feature in NEEM. The 1262 eruption may also be detected: the NGRIP core shows a clear peak in this year. (There is a larger one in 1261 and the eruption of the millennium -which is not the millennium eruption – was around 1259: this is not an easy time to identify eruptions!) There is a weak spike in 1415 in NGRIP which may or may not be the 1416 eruption. That leaves us with three possible detections for Katla of which 1245 and 1261 seem likely identifications; the second one coincides with a large Katla eruption which is encouraging. Katla does not often leave its sign in the Greenland ice, probably because its eruptions are not energetic enough. Perhaps we should let it sleep and convince prince (m/f) volcanologist to move any romantic attentions elsewhere.


This is Iceland’s dark horse. For such a small mountain it is surprisingly active, but it is also utterly unpredictable. Its magma differs from the usual Icelandic compositions. Hekla does both explosive and effusive eruptions, and sometimes both together. It gives no warning and is happy to either sleep for centuries or erupt twice within a few years. Even in the Middle Ages, Hekla was the one that terrified the hell out of people, to the degree that Hekla became synonymous with hell.

Hekla started out as its own sleeping beauty. For the first 200 years of the Viking presence, it slept. It was woken early in the 12th century. This was unwise, in hindsight. This beauty turned out not so much a dreamy princess as an evil queen. Sometimes it really is best to let sleeping beauties be. Ever since, Hekla has left its marks both on Iceland and on Greenland.

Hekla had a major eruption during its rude awakening in 1104 (or thereabouts) when it explosively erupted 2.5 km3. Strangely, the 1104 eruption is not apparent in the ice cores, neither sulphates nor tephra. The Hekla tephra was deposited mainly to the north and east, and so the wind direction may have kept it away from Greenland. However, the date is not fully certain. The eruption must have happened before 1118 but the dating to 1104 comes from one document, written later. There is an ice core peak eruption around 1110 and it makes sense to consider this as a candidate for the Hekla eruption. This peak is most obvious in the NEEM core. This ice core feature is a complex one which we’ll come back to in a minute. There is a second candidate peak in 1115.

After this explosive awakening, Hekla had eruptions in 1158, 1206, 1222, 1300 (0.5 km3), 1341, and 1389. The first three were small, the 1300 event was large and explosive, that of 1341 long lasting and locally destructive, and 1389 was both explosive and effusive where the lava flow produced the 0.25 km3 Norduhraun.

The 1158 and 1206 eruptions are not visible in the ice cores. The 1222 and 1300 events are seen only in NGRIP. The other two eruptions have produced clear signals. 1341 produced a strong peak in NEEM and NGRIP, and 1389 is very clear in all three cores. The pattern from this is that small Hekla eruptions (1158, 1206) do not reach the ice of Greenland. 1158 does not even show tephra in Iceland itself, and 1206 was also small but did produce some local tephra.

1222 was very similar to 1206 but it is very clear in the ice. This eruption is described as causing a ‘red sun’. The eruption of July 1300 was large but is weak in the ice. 1341 and 1389 left strong signals. May 1341 was a major eruption leaving ankle-deep ash and turning day to night. 1389 started with a large explosion which was heard across Iceland, but evolved into a fissure eruption which produced an 8-km long lava flow and lasted into 1390. The ice cores indeed show that the sulphate peak lasted a year.

Short eruptions may be quite dependent on wind direction. The 1222 eruption was smaller than that of 1300 but left a bigger impression in the ice. The wind may have been more favourable. however, the tephra of 1300 was carried northward and that of 1222 eastward, so that doesn’t fit. Perhaps the 1222 eruption reached higher in the atmosphere, as suggested by the ‘red sun’, allowing sulphate to spread further.

That leaves the 1104 eruption. There is nothing visible at that year. However, knowing that it could have been any time before 1118 gives us two strong candidates, one at 1115 and the other around 1110. The 1110 event lasted from 1109 to 1112. It consists of two patterns. One starts suddenly in 1109 and slowly tapers off towards the end. This pattern is strong in Greenland but is in part also visible in Antarctica. That points at an eruption at lower latitudes. The sudden start and slow decline is typical for an effusive or rift eruption. Superposed on this is a strong sharp peak in 1110, seen mainly in NEEM. It needs high time resolution to separate it from the long-lasting eruption. There may be a counterpart in Antarctica but that is not clear. Confusingly EGRIP agrees with the long-lasting eruption but shows the sharp peak earlier. That leaves us with two potential identifications for the first Hekla eruption: late 1110, or late 1114. We don’t have enough data to decide between these.

Volcanic winter?


This needs more discussion. And we have a bonus eruption, for what was the long-lasting eruption of 1109-1112? There is one candidate for this: the Japanese Asama volcano which erupted in 1108 in its largest holocene eruption (VEI 5). A contemporary writer documents: “October 13: According to a report from the province of Kōzuke, there is a high mountain in the middle of the province, Mount Asama. In the years 1065–1069, a slight smoke rose above the volcano but later became imperceptible. On August 29, there was a fire at the top of the volcano, a thick layer of ash in the governor’s garden, everywhere the fields and the rice fields are rendered unfit for cultivation. We never saw that in the country. It is a very strange and rare thing.” Guillet et al in 2020 https://www.nature.com/articles/s41598-020-63339-3/ give this story. They also mention that there was a very dark total lunar eclipse in 1110, which is often an indicator for volcanic dust in the atmosphere, and that 1109 was an exceptionally cold summer as indicated by tree rings. They attribute the events to a combination of three volcanoes, a tropical one and two northern hemisphere ones, one of which is Asama. This leaves some open questions. A VEI 5 eruption is rather small to cause the weather events they mention, and in fact even three VEI 5’s would not do. What happened, and which volcanoes were to blame?

Click on image for full resolution

Let’s look at this in more detail. Asama may indeed have contributed to the sulphate peak. The plot shows that the onset of the sulphate was at the precise time of this eruption, in October 1108. (Japanese records can be impressively detailed compared to the Icelandic ones of this era!) It is only seen in Greenland. The second peak is around October of 1109, and is seen in EGRIP and NEEM, shortly afterwards followed by a peak Antarctica. This would seem closer to being a tropical eruption, but still with the sulphate reaching Greenland first. The cold summer of 1109, as indicated by the tree rings, cannot be due to this eruption, nor can it come from Asama unless that was very much larger than appears from the reports. English reports mention quite a poor summer in 1109, but there was worse weather in 1110 with frost damage in May and too much rain which might be related to the October 1109 eruption.

A third eruption occurred late in 1110. This is seen mainly in NEEM – there may be a small counterpart in Antarctica at the same time but this could be noise in the data. The fact that the 1109 and 1110 peaks are not seen in all three Greenland cores suggests that these were fast and local events, something you may expect from a VEI-5 eruption. The winter of 1110/11 was severe but the summer of 1111 (how is that for a year number!) was rather dry with good crops. Overall, the weather was not that unusual and there is not enough here to call it a volcanic winter. Neither was there a record that the sun appeared faint. (A comet was seen above the northern horizon between 9 and 30 June 1110, and so at that time the sky must have been reasonably clear.) The case that Guillet et al make for a major volcanic weather event is not fully convincing. The data does suggest three VEI-5 eruptions but that is insufficient to impact the weather.

The one open question is the long tail towards 1113. That looks like a fissure-type eruption and Iceland is the most likely culprit, but no such eruption is reported. There is a possible candidate in Iceland: the 35 km long Frambruni lava flow due west of Askja, part of the Bardarbunga fissure system. The field contains around 4 km3 of lava, several times larger than Holuhraun. It formed before 1362 but the date remains uncertain.

This leaves the identification of the Hekla eruption open. It could be late 1110 or late 1114.

Other Icelandic eruptions

Hallmundarhraun ~950

This is a 8 km3 lava field northwest of Langjokull. It is thought to have formed in the 10th century but a precise date is not available. The lava flow extends over 50 km in length, is riddled with lava tubes and seems to result from a good attempt at building a shield. Such a large eruption should have left an ice record, and as it was effusive a significant sulphur output is likely. The lava tubes became caves, which were occupied by the early Vikings and later by a variety of outlaws including a group of priests(!?). Carbon dating of animal remains has indicated that the caves were occupied by 930. The eruption must have occurred earlier than this. The ice core has a peak around 908 and this is a potential, but far from certain, candidate. For more information, see Samir Patel, The blackeners cave, https://www.archaeology.org/issues/255-1705/features/5468-iceland-surtshellir-viking-cave.


The series of eruptions at Reykjanes (see Hector’s post) from the 10th to the 13th century has left little record in the ice core. That is understandable, seeing that the current Fagradalsfjall eruption is minor, is only affecting a small area and is also somewhat sulphur-poor. There are exceptions though. The ice core show a large number of small peaks in the period 1200-1250, especially around 1210, 1226 and 1240. These roughly coincide with a phase of eruptions off the coast which formed several islands. Subsea explosions can be very ashy and indeed these eruptions are recorded as having caused darkness and sand winters. The sulphates of these eruptions may have reached Greenland. However, no Reykjanes eruption is detected with certainty in the Greenland ice cores.

An eruption happened off the coast around Reykjanes in the Eldey volcanic system in 1422, which over a few years created a short-lived island. There is also no obvious peak in the ice core signal from this eruption.

The case against Iceland

At the moment, with the on-going and well-documented Fagradalsfjall eruption which seems to last forever while always doing something new, it becomes easy to forget that there are other volcanic areas in the world! The Greenland icecap contains more than just Iceland. However, although it is sometimes possible to associate the Icelandic records with the ice, to do the same with other volcanoes is much harder, for lack of accurately recorded eruptions. Who knows what Alaska did in the 12th century? How about the Azores which erupted perhaps 4 times per century, but lacked a trained observer with pen and paper?


There are European volcanoes with historical record. We have some reports of eruptions from Vesuvius, at 991, 999, 1006, 1037 and 1139. Of these, only 999 coincides with a peak in the NEEM ice core, and 1139 coincides with a peak in both NGRIP and NEEM. The 1139 peak is worth investigating, but it is not possible to confirm an association because the sulphate peak was 6 months before the eruption. Vesuvius is more distant from Greenland, and its eruptions may not have been large enough to spread sulphates that far.


This volcano is surprisingly poorly documented in this period. There was a damaging flank eruption around 1130. The ice record shows an extended but weak peak in 1129, and this is a potential candidate. But to confirm this would require more accurate history.


This volcano has been mentioned already. The 946 caldera-forming eruption is seen in the ice, although unexpectedly weak for such a large (high VEI 6) event. Tephra has been found: it is perhaps surprising that tephra traveled such a distance but sulphate was limited. The eruption may have been quite sulphur-poor. There are no European reports of poor weather around this time.

We do have historical records from Japan, and two eruptions are of interest.


This 1108 eruption is already mentioned above.There is very good agreement between the historical date and the ice core record.


This volcano had a large eruption (2 km3 DRE with extensive ignimbrite) carbon-dated to 1250 BP. It has been identified with an eruption dated to 915, Aug 17. It is large enough (high VEI 5) to potentially be visible in the ice but none of the ice cores have a peak in 915. The date comes from a single old document. The lack of an ice core signal may not be a problem (other VEI 5’s have been missed) but it suggests that the date may need to be reconsidered.

Ceboruco (tropical)

This is a case where the carbon dating is unusually precise. The large (plinian) eruption (VEI 5-6) of this Mexican volcano is dated to 1005+-15. It is called the Jala eruption. The date is based on multiple carbon fragments. The best one by itself gives a date (calibrated calendar) of 945-1022. The volcano is at 21 degrees north and is therefore tropical. The peak is indeed seen in both Greenland and Antarctica. However, in this time period, the relative calibration of the ice core dates is not as precise. As can be seen, none of the ice cores line up perfectly. This caused the algorithm used in the previous post to miss the association, and wrongly classify this eruption as ‘north’.

Within the time range, the most likely peak is in late 1001. This one is in fact the true millennium eruption! However, as it is tropical it does not belong in the list here, and is mentioned only because of the misidentification as ‘northern’ in the previous post.

And finally, one further eruption has been identified through ice core tephra. It is American.

Mount St Helens

St Helens is a frequent exploder (and rebuilder), so it may not be a surprise to find it in the ice core record. Its eruption of 1479 caused widespread tephra in the region and was powerful enough to deposit glass in Greenland – although to be fair, only a single micro-shard has been found so you should not envisage a hail storm of volcanic glass. I can’t say any more about this detection because it was published in a journal that no university I am aware of subscribes too. It is behind the most insurmountable paywall I have ever seen. (One does wonder what is the point of publishing important work where it can’t be read. But I’ll follow the preference of the authors and won’t cite the work.) The eruption has been accurately dated though the local tree ring record; it is classified as a VEI 5. There was a second eruption 2 years later, in 1482, also classified as VEI 5 but it was smaller. Two ice cores indeed show a very clear peak at the end of 1479 or early 1480 – it is missing only from the less sensitive EGRIP data. It looks small compared to the Veidivotn peak two years earlier, but is clearly detected. The second eruption may be seen in the NGRIP core, in the summer of 1481, but the other two ice cores do not show it. Note that there is also a bit of a feature in Antarctica, but this is not strong enough to be a certain detection. A counterpart in Antarctica would be unexpected.

That brings our list of possible identifications for northern-hemisphere eruptions to a close. The remaining parts of the north lack sufficiently accurately dated eruptions to allow a detailed comparison with the ice. For instance, Augustine has two reported eruptions within our time window but they are rather small and have dates accurate to +-50 years. That is too much uncertainty for a small eruption to be identifiable. But if you know of any more identifiable eruptions, please let us know!

Here is the plot with the identifications that we have been able to make. This shows the NGRIP core with the tropical and southern eruptions removed: it only shows eruptions from the northern hemisphere. The labels indicate the eruptions. Oreafajokull is in because of its tephra in the ice.

Northern hemisphere eruptions in the NGRIP ice core. Tropical and southern eruptions were removed. Identifications are listed. H is Hekla, K Katla, O is Oraefajokull (detected only in tephra)

Clearly, we have had some success. But many eruptions remain unclaimed, including some large events: VEI 5, perhaps an occasional VEI 6. Iceland will be responsible for some; it may even include some fissure eruptions we had not recognized or dated correctly. The Azores may be there as well. The northern powerhouse from the Kurils to Alaska will be responsible for some, not just St Helens, and this region can do large and undocumented eruptions. There is a lot of research to be done.

And we still need to look at the tropical and southern volcanoes.

Albert, August 2021

319 thoughts on “Volcanoes of the North: the fire in the ice

Comments are closed.