Since October 2021 a single date has run through writing on the Vikings in America: 1021. It comes from no saga and no inscription. It comes from tree rings that kept the trace of a cosmic radiation event of 993. This article walks through the study by Margot Kuitems, Michael Dee and their team one step at a time: which wood they measured, how a measurement becomes a calendar year, what the cuts on the pieces can and cannot do – and why "1021" is neither the year of arrival nor a statement about how long anyone stayed.
On 20 October 2021 the journal Nature published online a short paper with a deliberately sober title: "Evidence for European presence in the Americas in AD 1021". It has thirteen authors, with Margot Kuitems and Michael W. Dee of the University of Groningen as corresponding authors, and fills pages 388 to 391 of volume 601 in print. The basic idea fits in a sentence: if a tree laid down an unmistakable stamp in its wood in a known calendar year, you only have to count the rings from that stamp to the tree's last year of growth.
The study claims something very specific and leaves other things open. The site and the sagas are covered in the articles on Vinland and Leif Erikson. This one asks: what does the number 1021 carry, and what not?
There was no shortage of measurements. According to the study, more than 150 radiocarbon dates exist from L'Anse aux Meadows, 55 of them relating to the Norse use of the site. Radiocarbon dating relies on carbon-14 (¹⁴C), a weakly radioactive form of carbon that forms high in the atmosphere, enters plants and then slowly decays. The less ¹⁴C a sample holds, the longer ago it was taken up.
For those 55 samples the outcome was sobering. Their calibrated ranges, the authors note, extend across the whole Viking Age and beyond it. That fixed the broad period, not when work took place. The study gives three reasons. The measurements date from the 1960s and 1970s and are far less precise than modern ones. The technique of that time needed large samples, which restricted the choice. And many samples carried an unknown amount of inbuilt age.
Inbuilt age is the real stumbling block. A radiocarbon measurement dates the moment the carbon entered the wood, not the moment anyone used the wood. The heartwood of an old trunk can be centuries older than the felling. Driftwood may travel a long while before someone picks it off a beach.
The new dating sidesteps the problem by using a sudden rise of ¹⁴C in tree rings. One such jump, in the late eighth century, appears on this site in the article on the dating of Ribe. In 2013 Fusa Miyake, Kimiaki Masuda and Toshio Nakamura reported another in Nature Communications, given in the abstract as a rapid increase from 993 to 994. Later work, including Büntgen et al. and Kuitems et al., places the jump in 993. Jumps of this kind are now called Miyake events.
A broad test followed in 2018: Ulf Büntgen and a large team measured 484 tree rings from the stretches 770–780 and 990–1000 and could place 44 tree-ring chronologies from five continents. So the jump is no local effect but a signal readable worldwide in the same growth year. Only that allows it to serve as a time marker in Newfoundland.
How large is it? Kuitems and her team put the rise of 993 at about 9 per mil, equal to some 70 radiocarbon years. In the measurements, a ring of 993 therefore looks about 70 years younger than its neighbour of 992.
The cause is less certain than popular accounts suggest when they simply say "solar storm". Kuitems et al. write of cosmic radiation events. Miyake and colleagues considered solar activity a plausible cause in 2013, drawing also on matching beryllium-10 peaks in an Antarctic ice core. Büntgen et al. see strong solar proton radiation as the likely explanation. That is the prevailing reading, but the dating needs only a jump in a known year. Even the month is disputed: Büntgen et al. place the onset in the northern spring of 993, while Rakowski and colleagues arrived at late summer 993 from Polish oak in 2022. The accounts differ here, but not on the year.
The material came from Parks Canada, the Canadian national parks agency. According to Margot Kuitems' blog post of 26 October 2021, it came from layers attributed to the Norse phase, and the team looked specifically for wood with clear rings that kept the very last one.
Four pieces were measured, with the find numbers 4A 59 E3-1, 4A 68 E2-2, 4A 68 J4-6 and 4A 70 B5-14. The last dropped out: it spans only nine years and lacks the jump. The headline "three pieces of wood" is what remained after that selection. The three come from three different trees and at least two species. Two are fir, probably balsam fir; the third, 4A 68 J4-6, is juniper or thuja, an identification left uncertain because it is compression wood, wood that grew differently under bending stress in the trunk.
As far as I could see, the study says nothing about what the pieces were for; their size is only hinted at by 5 cm scale bars in a figure. Whether they are woodworking waste or parts of an object cannot be read from the paper.
In all there were 127 measurements on 83 individual tree rings, 115 in Groningen and 12 in Mannheim. According to Kuitems' blog post, the rings were separated under a microscope with scalpel and razor blade, cellulose was extracted from each, and it was measured by accelerator mass spectrometry. One complication: all pieces except 4A 68 E2-2 had once been conserved with polyethylene glycol, which itself contains carbon and has to come out before measuring. In Groningen that took 36 hours in hot ultrapure water. The labs differed by 5.1 ± 7.9 radiocarbon years, statistically insignificant.
The waney edge is the outermost layer of wood formed under the bark. It marks the tree's last year of growth. Without it you never know how many rings were hewn away, and a date would only mean "after". With it, you date the end of the tree.
The study reports that the waney edge could be identified in all three cases. That disposes of two objections at once. Driftwood is effectively ruled out in the authors' view, because this outer layer would almost certainly have been stripped off in the water. And the inbuilt age that smeared the old dates vanishes here, since the count explicitly runs to the last ring.
A logical remainder is left. The waney edge shows when the tree stopped growing. That it was also felled at that moment is an inference, not a measurement. Dead wood gathered later would have a last ring as well. Here the study's strongest argument applies: that three randomly collected pieces from three trees should have precisely the same inbuilt age strikes the authors as very improbable. Three trees ending in one year are most simply explained by people felling them in that year. The authors themselves call the convergence notable and unexpected.
Tree rings can date, but they say nothing about who worked a piece. Trees died on Newfoundland in 1021 without any help from Europeans. The link to the Norse comes from the archaeology alone, and according to the paper by two routes: first the position in the Norse context, going back to earlier work by Parks Canada, second the tool marks. The authors describe "characteristically clean, low angle-in cuts" from metal tools. Such implements, they add, were not manufactured by the Indigenous inhabitants of the area at the time.
The study's title rests on this step. Three things are worth holding on to. The paper speaks throughout of metal, not iron. Nor, from what I could see, does it mention an axe, though popular texts, older ones on this site among them, like to paint it that way. An axe or hatchet is an obvious thing to imagine, but that is interpretation.
Second, "not manufactured" is narrower than "not used". A metal tool could in theory have changed hands. As far as I can see the study does not discuss this, and the Norse context speaks against it. Still, the attribution is a well-founded conclusion, not a name tag.
Third, the cuts show only that the wood was worked. Whether a cut was the felling or later trimming does not emerge. Both the "chopping date" of Kuitems' blog post and the study's felling year rest on the last ring, and a cut can only follow that ring.
The team counted rings backwards from the waney edge, treated as ring 0: ring −1 is a year older, ring −2 two years, and so on. Along this ladder they looked for the jump. According to Kuitems' blog post, a first round in the outermost 20 rings found nothing, so they had to go deeper into the wood.
The evaluation took two routes. Wiggle-matching lays a sequence of measurements with known ring spacing against a reference curve, here IntCal20, the international calibration curve, and seeks the best fit. For all three waney edges, the 95 per cent range fell between 1019 and 1024. On that basis the jump of 993 lies somewhere between ring −31 and ring −26.
Because IntCal20 is smoothed and weakens the jump, the team also compared the values with the Büntgen et al. dataset, where it stands out sharply, using a chi-squared test, a standard measure of how well a series and a reference agree. This gave 1021 as the best solution for all three pieces. But the year does not name itself. At 95 per cent, 1022 also passes for two pieces, and for 4A 68 J4-6 so do 1019 and 1020. What decides it is the position of the drop: in each case it fits ideally when ring −29 corresponds to 992. Then ring −28 is 993, and ring 0 is 1021.
So anyone reading that the trees are dated "to the exact year" may take it this way: 1021 is the reasoned preferred solution, supported by three independent pieces turning at the same spot. The statistics alone leave a margin of a few years.
A tree ring grows in two phases, first the pale earlywood, then the denser latewood. How far the last ring got reveals the season of death. For 4A 68 J4-6 the wood anatomy indicated spring, for 4A 68 E2-2 summer or autumn. For 4A 59 E3-1 the earlier conservation prevents any determination.
Taken at face value, wood was therefore cut at different times in one and the same year. As far as I could see, the study does not interpret this further. Was someone already there in spring because they had wintered on the spot? Possibly, but it does not have to follow, since two fellings could also go back to two groups. If both pieces really belong in the same year, the work at any rate did not happen on a single day. Whether anyone overwintered there cannot be established with certainty from these three pieces.
Before the limits, the strength of the result deserves its due. The authors call 1021 the only secure calendar date for the presence of Europeans across the Atlantic before the voyages of Columbus.
Some reports of autumn 2021 went further. Broadcasters and portals wrote that researchers had fixed when the Vikings arrived in Canada, or that they settled there "as early as 1021". Both turn the result around. A felling year is a moment when people were already on the spot. Logically it is an upper limit: they were there by 1021 at the latest. It does not say when the first person set foot on that shore, and "as early as 1021" is exactly backwards, since they may well have come earlier.
The study itself makes no such claim. That the Norse were the first Europeans to cross the Atlantic is background it takes from earlier work, not something it measures. Nor does it tie the date to individuals: by my check Leif Erikson appears in the main text only in the title of a cited article, Karlsefni and Freydis not at all; only an extended-data overview lists the saga voyages and their leaders. Even the famous phrase, that 1021 is the earliest known year by which human migration had encircled the planet, is carefully put: the earliest known year "by which" this had happened, not the year in which it did. The abstract is looser here: the first known point at which humans encircled the globe.
A second point concerns the distance from tradition. According to Kuitems' blog post, the founding had been placed around the turn of the millennium, and according to a Legion Magazine report (3 November 2021), the archaeologist and co-author Birgitta Wallace called the date slightly later than expected. It does not follow that the sagas are wrong. A felling in 1021 and a first visit some twenty years before do not exclude each other. Nor does it follow that the traditional date is confirmed. The wood simply does not reach back to the year 1000.
"We have one date, but we don't know how long before and after they were there," Legion Magazine quotes Wallace as saying.
The study works with an estimate from the sagas, and periods of time turn up only there. In an extended-data figure caption the authors count seven winters in Vinland for the Saga of the Greenlanders and estimate about thirteen years between first arrival and final return, because a voyage out and back would not fit one sailing season. For the Saga of Erik the Red they set about three years as a minimum. These are extrapolations from narratives, not measurements.
The 2021 authors write that the archaeological and written records are consistent with a very brief occupation, and put the wide spread of the old radiocarbon dates down to the limits of the technique of the time. According to Legion Magazine, Wallace herself calls her view of the time spent there a hunch, based on thin deposits between phases of use. The 1021 pieces, the report says, came from the top layer, and 1021 could well have been about the last year wood was cut there. That is a professional judgement, not a conclusion from the tree rings.
Against this stands a 2019 paper. Paul Ledger, Linus Girdland-Flink and Véronique Forbes set out in PNAS to take peat samples for an environmental study, came upon an unexpected horizon with non-native traces, and inferred from a Bayesian re-analysis of older radiocarbon dates that Norse activity may have lasted around a century (0 to 195 years in the model), explicitly not continuously. They date the horizon itself from the mid-12th to the late 13th century. They leave open who made it, though they say the dates point rather to Indigenous activity. Wallace called so late a Norse return "highly unlikely" in Live Science, because matching structures are missing.
Notably, Ledger and Forbes are also co-authors of the 2021 study. The 2019 paper is cited in the new study's abstract but, as far as I can see, not discussed in the main text. I found no explicit resolution of the contradiction there, and no follow-up study that settles it, though my search was limited. The number 1021 holds under both scenarios; it does not decide between them.
It is tempting to link the wood with the tradition, since in the Saga of the Greenlanders timber explicitly becomes the ship's cargo. In its story, Leif gives his men an order in the morning:
"[…] pick grapes or cut vines and fell the forest, so that it makes a cargo for my ship." […] "Now a cargo was cut for the ship."Grænlendinga saga, ch. 3, my own translation from the Icelandic text in modern spelling ("lesa vínber eða höggva vínvið og fella mörkina svo að það verði farmur til skips míns"; "Nú var hogginn farmur á skipið"), read through a retrieval tool on the private site wrdingham.co.uk, edition not named there; chapter numbering varies by version
The saga survives only in Flateyjarbók, a manuscript of the late 14th century, and its date of composition is disputed; English Wikipedia gives around 1200 (Jón Jóhannesson) and the 14th century (Helgi Þorláksson). It tells of a Vinland whose location it does not give in our sense and counts in winters, not calendar years. The study does not tie its wood to this passage at all, and nothing marks the three pieces as part of a ship's cargo. Whether the felling of 1021 goes back to any voyage in the stories cannot be established with certainty. Anyone saying the wood "confirms" the saga really means: it does not contradict it.
With 1021 research holds a point independent of the tradition, which is why it should not be built straight back into it. The authors themselves leave the task to the future: how 1021 relates to overall transatlantic activity is a matter for further research.
Why not just count the rings and compare them with a tree-ring calendar? The study does not take that route but uses the worldwide simultaneous ¹⁴C jump as its marker. All it needs is a short ring sequence from one piece reaching back to 993.
How firm is 1021 as against 1020 or 1022? For all three pieces 1021 is the best solution, but 1022 also passes the statistical test for all three, and 1019 and 1020 for one piece. The team prefers 1021 because the drop then lies at the same fitting spot in every piece.
How do we know the cuts were not made by Indigenous people? The authors argue that people of the area did not manufacture metal tools at the time. The attribution to the Norse phase also rests on the find position, following earlier work by Parks Canada.
How long were the Norse at L'Anse aux Meadows? The date does not say. From the sagas the study roughly estimates three to thirteen years, and the authors consider a very brief occupation consistent with the evidence. A 2019 paper weighed considerably more. The accounts differ here.
Does 1021 hold for the Vinland voyages as a whole? No, it holds for three pieces of wood at one site. Whether L'Anse aux Meadows is the Vinland of the sagas, and how the year relates to that, is a separate question.
Vinland: sources, site and disputes · Leif Erikson · Thorfinn Karlsefni · Freydís · Vikings in the North Atlantic · Ribe and its coins · The Kensington Runestone · Vinland Saga fact check
L'Anse aux Meadows: what eight turf buildings in Newfoundland prove, and what they don't · Vinland · Leif Erikson · The Vinland Map at Yale: old parchment, modern ink and a real voyage that never needed it · Bjarni Herjólfsson.
All History & Finds articles →
Yes, a maternal DNA lineage belonging to haplogroup C1 – a group found today mainly among Indigenous peoples of the …
Read more →Vinland is the name Norse sailors from Greenland gave to a stretch of North American coast around the year 1000 – and …
Read more →The Vocontii lived in the Alpine foothills between the Rhône and the Durance, and they do not fit the story of a Gaul …
Read more →Some legendary figures live on through their deeds, a few through their works – and very few through both. Völundr …
Read more →