Sven Erik Matzen

Software Architect | Cloud & Security Expert | AI-enabled Solutions

The Year Without a Sun: 536, the Volcanic Winter, and How Ice Cores Redated History

🎧 Listen to this article

History · 2026-09-17

EU label: fully AI-generated content Fully AI-generated article (no prior review).

The Hook: A Sun Without Brightness

In the spring of 536, the Byzantine historian Procopius of Caesarea wrote a sentence that for centuries was treated as a puzzling marginal note. He reported a most dread portent: the sun gave forth its light without brightness, it looked much as it does during an eclipse, and the beams it shed were not clear. This, he wrote, continued throughout the year.

Procopius was no mystic. He was a sober, often acid observer at the court of Emperor Justinian, a man who documented campaigns and liked numbers. And he was not alone. Cassiodorus, a Roman statesman in Ostrogothic Italy, complained in his Variae that the sun's rays were thin and feeble, that the moon shone without brightness, and — a detail no inventor of wonder-tales would have added — that one marveled to see no shadows of one's own body at noon. John of Ephesus, a Syriac church historian, wrote that the sun became dark and that its darkness lasted eighteen months. Michael the Syrian transmits the same report with an oppressive addition: the sun shone for about four hours each day, and even that light was only a feeble shadow; the fruit did not ripen, and the wine tasted of sour grapes.

For centuries, historians treated such passages as what they superficially appeared to be: pious topoi, apocalyptic rhetoric, literary ornament. Late antique authors liked to write of signs in the heavens. The suspicion that something physically real was being described here could neither be confirmed nor refuted with the tools of textual criticism.

Today we know better — and the reason lies not in a newly discovered manuscript. It lies in trees and in ice. Over the past four decades, dendrochronology and ice-core analysis have opened a second archive of the past, one kept entirely independently of human hands. That archive confirms Procopius on one point completely: in 536 a cold phase began that became the strongest and longest-lasting cooling of the Northern Hemisphere in two thousand years. And it corrects science on another point — the ice cores themselves were misdated for decades.

This article tells how two utterly different classes of evidence — written words and layered matter — were brought together, what emerged from that, where the evidence grows thin, and why this case in particular carries a methodological lesson that reaches far beyond late antiquity: the temptation to turn an impressive correlation into a story of cause and effect.

Part 1: The Witnesses on Parchment, and Their Limits

One Event, Three Continents

The striking thing about the written sources for 536 is their geographic spread. Procopius wrote in Constantinople or in the field, Cassiodorus in Italy, John of Ephesus in the Syriac world. From Mesopotamia, the so-called Chronicle of Zuqnin reports a darkened sun. From China, annals record frost in summer for 536 and famines, in some regions even snowfall in months when snow never falls. Irish annals note for the following years a "failure of bread" — a formula that in that genre marks years of hunger.

One should not overstate the precision of these sources. Chronicles were often compiled decades or centuries after the events, they copy from one another, and their dates are sometimes off by years. The Irish annals in particular are notoriously imprecise for the sixth century. A single source proves nothing here.

But the pattern is what matters. Independent traditions in different languages, with no demonstrable common origin, describe the same phenomenon within a narrow window of time: dimmed sunlight over a long period, unusual cold, failed harvests. In source criticism this is called independent multiple attestation, and it is one of the strongest arguments historians have at their disposal.

What Texts Cannot Do

And yet a hard limit remains. Textual sources can say that someone perceived a dimming. They cannot say how strong it was, how far it reached, how long it lasted in physical terms, or what caused it. John of Ephesus's "eighteen months" is not a measurement; it is the recollection of a man who had no instruments.

Above all, texts can say nothing about regions from which no texts survive. For Scandinavia, for large parts of Central Asia, for pre-Columbian America, the written archive is entirely silent on 536. Whoever reads only texts sees the world through the window of literate cultures — and easily mistakes that window for the world.

This is precisely where the second class of evidence begins.

Part 2: The Archives of Nature

Trees as Thermometers

A tree growing at the cold limits of its range — high in the Alps, far north in Siberia, at the treeline of the Californian White Mountains — grows in thickness essentially in proportion to summer warmth. If the summer is cool, the annual ring stays narrow. That makes such trees thermometers with year-level resolution reaching back thousands of years, provided living trees, construction timber, and subfossil trunks are joined by overlap into a continuous chronology.

In extreme cases the wood shows not merely narrow rings but so-called frost rings: cell layers damaged and deformed by frost during the growing season. A frost ring is not a statistical signal but a physical scar — proof that it froze in the middle of summer.

The Northern Irish dendrochronologist Michael Baillie pointed out as early as the late 1980s that around 536 an exceptionally narrow ring appears in chronologies from Ireland to Scandinavia, followed by a multi-year growth depression. North American bristlecone pines show frost rings around the same time. Later, chronologies from the Russian Altai were added, which had been missing and which closed the gap between Europe and East Asia.

Ice as a Chemistry Lab

The second archive lies in Greenland and Antarctica. Every snowfall deposits a thin layer that never melts again; beneath, the layers are compressed by the weight of younger ones but not mixed. Drill a core and you read the atmosphere of the past year by year.

For volcanism, the decisive marker is sulfur. A large explosive eruption hurls sulfur dioxide into the stratosphere, where it reacts to form sulfuric acid aerosols. These particles scatter sunlight back into space — they cool the Earth's surface — and after one to three years are washed out, among other places as sulfate in polar snow. A sulfate peak in the ice is the chemical signature of a stratospheric eruption. If the same peak is found simultaneously in Greenland and in Antarctica, that argues for a tropical volcano whose aerosol cloud spread across both hemispheres; if it appears only in Greenland, the source was in all likelihood northern.

Sometimes the ice additionally contains tephra: microscopic glass particles of volcanic ash. Their chemical composition is something like a fingerprint of the magma source — in the ideal case, it allows the volcano itself to be identified.

Two Clocks That Did Not Agree

Here lay the problem for decades. Tree rings are absolutely dated: one ring is one year, and counting errors show up when many trees are cross-compared. Ice cores, by contrast, are dated by counting annual layers — a procedure that grows less certain with depth, because layers can become thin, indistinct, or disturbed by wind redistribution. Small counting errors accumulate.

The result was a stubborn contradiction. The tree rings said 536, unambiguously. The established Greenland chronology GICC05 placed the corresponding sulfate peaks at roughly 529 and 533/534. Seven years of difference — too much to ignore, too little to resolve easily. Some researchers concluded that the tree rings were wrong. Others held that the ice was displaced. The debate was unpleasantly deadlocked, because there was no independent referee.

The referee was provided by the sun.

The Cosmic Anchor

In 2012 the Japanese researcher Fusa Miyake demonstrated in Japanese cedars a sharp, single-year rise in radiocarbon 14C for 774/775 CE — an event now named after him and most likely caused by an extreme solar particle event. Such events simultaneously produce other cosmogenic nuclides, among them 10-beryllium, which is deposited in polar ice.

That gave researchers a global time marker visible in both archives at once: in trees as a 14C spike, in ice as a 10Be spike. Where both clocks show the same event, they must be set to the same time.

That is exactly what Michael Sigl and colleagues did in a paper published in Nature in 2015, which today serves as the reference chronology for volcanism over the past 2,500 years. The result: the Greenland ice chronology was about seven years too old in the first millennium. After the correction, the sulfate peaks fell exactly on the years the trees had been indicating for decades — 536 and 540.

It is worth pausing on this finding. Two independent measurement systems contradicted each other. A third, physically quite unrelated phenomenon — a particle shower from the sun — settled the dispute, because it had marked both archives at once. This is science at its most persuasive: not consensus by authority, but by a new piece of evidence independent of the question in dispute.

Part 3: The Double Event and the Late Antique Little Ice Age

Two Eruptions, One Decade

After the redating, a clear picture emerges. In 536 a large eruption occurred whose sulfate appears mainly in Greenland — hence most likely situated in the Northern Hemisphere. In 540 a second followed, whose signal is bipolar, pointing to a tropical volcano and distributing the stratospheric aerosol load globally. In 547 a third event was added.

The decisive quantity is not the single eruption but the cumulative effect. The radiative forcing of this volcanic cluster was, over a decade, larger than that of any other period in the past two thousand years. For comparison: the eruption of Tambora in 1815, which produced the notorious "year without a summer" of 1816 in Europe and North America, was a single event. The years 536 to 547 were a series — and the system had no time to recover between blows.

LALIA

In 2016 Ulf Büntgen and colleagues published in Nature Geoscience a reconstruction of summer temperatures over the past two millennia based on new tree-ring measurements from the Russian Altai combined with Alpine chronologies. They showed a sudden, spatially synchronized and unusually prolonged cooling beginning in 536 and lasting until around 660 CE. For this phase they coined the term Late Antique Little Ice Age, or LALIA.

The label deliberately echoes the "Little Ice Age" of the late medieval and early modern periods, and it has not gone uncontested in the field. In the same journal, a comment appeared in 2017 under the title Limited Late Antique cooling, judging the extent and duration of the cooling far more cautiously; Büntgen and colleagues replied. A 2022 modeling study by Evelien van Dijk and colleagues asked explicitly in its title whether there had been a volcanically induced long-lasting cooling over the Northern Hemisphere in the mid-sixth to seventh century at all.

I am of the opinion that this controversy is the most interesting part of the story — not despite its unfinished state but because of it. What is undisputed is the sharp cold snap of the years immediately following 536 and 540; it is consistent across tree rings, ice, and models. What is disputed is how long and how widely the cooling persisted afterward, how volcanism must be weighted against ocean circulation and internal climate variability, and whether "little ice age" is the right term for a phase with a cooling on the order of one to two degrees in the summer mean. Whoever assigns a label shapes how others think about the matter — that too is a methodological lesson.

Part 4: Where Did the Volcano Stand?

One would assume that the question of the source is the easiest of all. It is the hardest.

In 2018 Christopher Loveluck and colleagues published in Antiquity an analysis of a 72-meter ice core from the Colle Gnifetti glacier in the Swiss Alps, measured at extremely high resolution using a laser-based technique. They found tephra in the 536 layer and attributed it to an Icelandic source — an interpretation that went through the press and to this day appears in popular accounts as settled.

The scholarly literature is more skeptical. In the Greenland core NEEM-2011-S1, the 536 layer contains a mixed tephra made up of several chemically distinct glass populations. In 2022 Gill Plunkett and colleagues concluded that the 536 sulfur signal is probably a composite of three simultaneous eruptions of uncertain origin. Individual glass shards from this layer match, geochemically, a tephra from 88 CE whose composition points most strongly to the Aleutian arc in the North Pacific; the authors therefore consider it possible that the same volcano contributed to the 536 signal. That does not identify a specific volcano — and an Icelandic source is not established.

So we face a peculiar situation: the event is dated to the year, its climatic effect is quantified, its imprint documented on three continents — and nobody can say which mountain exploded. I am of the opinion that this open point is the most honest part of the whole body of research, because it shows how differently robust individual claims within the same topic can be. "536 was a volcanic winter" is well supported. "The volcano stood in Iceland" is not.

Part 5: What Cold Means for People

From Degrees Celsius to Crop Failure

A temperature anomaly is not yet a catastrophe. The path from the physical signal to human experience runs through agronomy, and there it becomes concrete.

An extensive study by Evelien van Dijk and colleagues, published in Climate of the Past in 2023, modeled that path for Scandinavia. The team combined climate model runs of the double event with a growing-degree-day model — a standard agronomic method that computes the accumulated heat sum a crop needs to ripen — and compared the result with archaeological and pollen-analytical findings from southern Norway.

The numbers: in the ensemble mean, the Northern Hemisphere growing season cooled by roughly 2 °C, Scandinavia by about 1.5 °C, and regionally in southern Norway by up to 3.5 °C. Under an assumed cooling scenario of 3 °C, the growing-degree-day model indicates that crop failures were likely in the regions studied. Added to this was a significant reduction in precipitation; the 5 °C isoline of the growing season shifted roughly six degrees of latitude southward in the first two years after the eruptions.

The crucial quantity underneath is internal variability. For the Northern Hemisphere extratropics the model puts it at 0.4 °C; for Scandinavia at 1.2 °C. That is, the regional climate there fluctuates strongly in any case, and the volcanic signal can be separated cleanly from the noise only in the first years after the eruptions. Individual ensemble members even show wetting rather than drying. Anyone who builds a local narrative out of a model mean exceeds the resolution of their data.

The Fimbulwinter

In Scandinavia there is an archaeological pattern that stands out for the period around 536: abandoned farmsteads, declining cereal pollen and returning forest, in places substantial abandonment of settlement. And there is a literary reflex. Old Norse tradition knows the Fimbulwinter — three consecutive winters with no summer in between, harbinger of the end of the world.

Several Scandinavian researchers, among them Bo Gräslund and Neil Price, have proposed reading the Fimbulwinter myth as a poetic rendition of a real sixth-century crisis. The conspicuous clustering of gold hoards from this period points in the same direction; they are often interpreted as votive offerings made in a situation experienced as existential.

The interpretation is attractive, and it is unprovable. The Eddic texts were written down in the thirteenth century — seven hundred years after the event. Oral transmission across such a distance is possible, but it cannot be verified, and the myth of a world-winter is a motif that can arise without any concrete trigger. I am of the opinion that one should distinguish cleanly here: the archaeological abandonment of settlement is a finding. The Fimbulwinter as a memory of 536 is a hypothesis, no more.

Part 6: The Plague, and the Temptation of the Grand Narrative

541: The Second Blow

Five years after the dimming, in 541, an epidemic broke out in Egyptian Pelusium, reached Constantinople in 542, and spread across the entire Mediterranean world. Procopius described it at length and wrote of tens of thousands of deaths per day in the capital. It bears the name of the reigning emperor — the Justinianic Plague — and is regarded as the opening of the First Pandemic, which recurred in waves until about 750.

That it really was plague was long a diagnosis inferred backward from descriptions of symptoms — hence uncertain. That question has been settled molecularly since the 2010s. From early medieval cemeteries in Upper Bavaria, first Aschheim and later Altenerding, Yersinia pestis DNA was recovered; in 2016 Michal Feldman and colleagues reconstructed a high-coverage genome of the pathogen from an individual at Altenerding. In 2019 Marcel Keller and colleagues presented genomes from across Western Europe in PNAS and demonstrated an early diversification of the pathogen during the First Pandemic.

The identification of the pathogen is thus robust. It says only that plague occurred and where — not how many people died.

The Dispute Over the Numbers

Here one of the most instructive debates in historical scholarship has been raging for years. The traditional, "maximalist" picture attributes to the Justinianic Plague deaths in the tens of millions and a structural upheaval of the Mediterranean world — often linked with the volcanic winter into a causal chain: climate shock, crop failure, weakened population, pandemic, collapse.

In 2019 Lee Mordechai, Merle Eisenberg, Timothy Newfield, Adam Izdebski, Janet Kay, and Hendrik Poinar published in PNAS a paper with the provocative title The Justinianic Plague: An inconsequential pandemic? They systematically examined several mutually independent datasets: written sources, legislation, coinage, papyri, inscriptions, pollen profiles, ancient DNA, and mortuary archaeology. Their finding: individually and taken together, these data do not support the maximalist picture. Instead of a rupture, coinage, papyrological administration, inscription practice, and land use show broad continuity across the plague period. The high mortality estimates, they argue, rest on a small subset of narrative texts — above all Procopius — which tend toward dramatization in any case.

The paper has provoked contradiction; the debate is not settled. Critics point out that absence of signal in coarse proxies does not prove absence of mortality, and that pollen profiles and coin series have poor temporal resolution for an event lasting a few years. That is a legitimate objection. But the burden of proof has shifted, and that is the point: anyone writing today of "millions dead" must say which piece of evidence that number comes from.

Why the Causal Chain Is So Seductive

One can watch the narrative assemble itself. There is a spectacularly well-dated natural event. There is a pandemic five years later. There is, in the same century, the East Roman withdrawal from Italy, the crisis of the Sasanian empire, the rise of the Avars, and shortly afterward the Arab expansion. The building blocks lie so close together that the causal arc seems almost to draw itself.

But: the sixth century would have been a century of upheaval without any volcano. Justinian's wars of reconquest in Italy and North Africa were political decisions, not climate consequences. The Persian wars had causes of their own. And a climate signal amounting to 1.5 °C in the Scandinavian growing season, nearly submerged in an internal variability of 1.2 °C in the same region, cannot carry such a chain on its own.

The serious position is more uncomfortable, and runs roughly as follows. The volcanic winter of 536/540 is well supported and was, with high probability, a severe burden on agrarian societies in marginal environments — in parts of Scandinavia probably existential. The Justinianic Plague is confirmed as plague, but its demographic effect is open. Whether and how strongly the two were connected is an open research question, not settled knowledge. And "536, the worst year to be alive" — a popular formulation from a science magazine — is a headline, not a research result.

Part 7: The Methodological Lesson

Four lessons seem to me transferable, far beyond late antiquity.

First: independent archives beat deep single sources. Neither Procopius alone, nor the tree rings alone, nor the ice alone would have produced the picture. Only the convergence of three methodologically unrelated approaches made the case solid. Where several systems with different failure modes show the same thing, the shared statement is more reliable than any of them individually — the same principle by which one would rather investigate a production incident with logs, metrics, and traces together than with any one of them.

Second: a contradiction between two measurement systems is not resolved by argument but by an independent anchor. Seven years of difference between ice and wood were the subject of opinion for two decades. The Miyake event of 774/775 ended the discussion because it was a marker neither side had constructed for its own position. Whoever calibrates should always ask whether their reference point is genuinely independent of the question in dispute.

Third: the resolution of the data limits the resolution of the claim. An ensemble mean over decades permits no statement about a single village. A pollen profile with decadal resolution cannot capture a three-year catastrophe. Many faulty inferences in this field arise not from wrong data but from claims finer than the data they rest on.

Fourth: well-supported individual building blocks do not add up to a well-supported chain. The dating is excellent. The climatic effect is well quantified. The pathogen identification is robust. And yet the narrative "volcano caused famine, famine caused plague, plague caused the fall of the ancient world" is not evidenced but assembled. Causality is not a property that emerges from stringing together established facts. It has to be shown in its own right.

The Central Takeaway

The year 536 is one of the few pre-modern events where we know the event more precisely than its consequences. We know the year, we know the mechanism, we know the radiative forcing — and we do not know which volcano it was, and we argue about how many people died in the pandemic that began five years later.

That asymmetry is the real lesson, and it transfers directly to one's own work. For every claim you make or adopt, check separately: what is measured, what is modeled, what is interpolated, and what is narrated? These four categories blend together in every good story, and precisely because the story is good, the blending goes unnoticed.

The concrete call to action: this week, take a claim treated as self-evident in your working environment — an attribution of cause for an outage, a justification for an architectural decision, a number in a report — and decompose it into exactly these four layers. For each layer, ask: what independent evidence supports it? Does a second archive exist that would have to show the same thing if the claim were true? In most cases you will find that the core is solid and the edges consist of convention. It is precisely those edges where the errors later surface.

Reflection Question

Which conviction in your work rests on a single, especially impressive source — and which independent "second archive" could you consult that would be free to contradict you if you were wrong?

Cross-References in the Vault

Sources

← All articles