Our Deep History Our Deep History A museum of unexplained ancient sites and relics
Latvia rooms

Lake Āraiši, Latvia

Āraiši

Felled in AD 835, in the later match (an earlier match gave AD 776 to 780)

The Āraiši lake fortress as rebuilt. The logs are later timber, cut to follow the lowest and best-preserved phase. Window openings and the plank roof are part of that rebuild. The buried originals are not these walls. Egilus, CC BY-SA 4.0. Shown unedited.

One spruce timber from the first phase at Āraiši was felled in 776–780 when its rings were matched to the 2013 calibration, and in AD 835 when the atmospheric spike of 774–775 was found 60 rings in from the bark.

Why this is extraordinary

On an island in Lake Āraiši, in north-eastern Latvia, a timber settlement was built on a platform over the water and fortified. Excavation in 1965–1969 and 1975–1979 separated five construction phases. The lowest, the first, is the one that survived best. Artefacts had already put that phase in the 9th century, probably the 830s. A cross-match of the spruce rings to a chronology from Novgorod once suggested a felling around AD 930 for the platform under the earliest buildings. The same timbers have since been asked a narrower question, and the answers have not stayed still.

A floating sequence of about 93 years of spruce was tied to the radiocarbon calibration by dating ten blocks of wood, each a handful of rings, from one timber. Matched to the 2013 curve, there was a 95 percent chance the last ring fell in 775–784. The last sample also held a one-year jump in atmospheric radiocarbon that the world’s trees record in AD 774–775. On that basis the felling was narrowed to 776–780. That was earlier than the objects liked. A fortress in the late 770s and a set of finds that had been read as the 830s were the dispute.

A later measurement dated the rings one year at a time and found that same atmospheric jump 60 rings in from the outside of the log. The felling year that follows is AD 835, and the building season that follows the felling is the winter or spring of 835–836. The older block dates, matched to the 2020 calibration instead of the 2013 one, also prefer the 830s. The 776–780 result was a real match to the curve then in use. It is not the match the same timber gives once the spike is pinned to a single ring. Both papers are the dates. The objects sit more easily with 835.

What a visitor walks through is not that timber. The originals could not be shown underwater, so the lowest phase was rebuilt in new wood. One house was up in 1981. The lower courses follow the excavated plan. The upper walls are a completion from the pieces that survived and from later building habits that make a roof stand. The photograph is that completion. It is the shape of the first phase as rebuilt, not the spruce that was felled in the 9th century. No azimuth of the island, and no star for the fortress, is published in the dating papers. The sky these studies use is the atmosphere’s carbon, the spike of 774–775, not a sunrise over the lake.

Why it matters

Fifty years is the difference between a fortress that is early for its objects and one that sits where the artefacts had already put it. The buildings in the photograph are a later rebuild in new wood, so the year you are dating is not the year of the wall you can touch.

The photograph is the reconstruction. The two felling years are the argument. The buried logs are not in the frame.

Photo tour

1 photograph, unedited. On a phone, swipe the row. Choose one to look closely.

The Āraiši lake fortress as rebuilt. The logs are later timber, cut to follow the lowest and best-preserved phase. Window openings and the plank roof are part of that rebuild. The buried originals are not these walls. Egilus, CC BY-SA 4.0

What is still unexplained

These sit beside the consensus account, with the same visual weight. None of them is marked solved.

Open The same log, two curves Ten blocks of rings, matched to the 2013 curve, put the last ring in 775–784 and, with the spike in the last block, in 776–780. The same measurements matched to the 2020 curve prefer the 830s. A calibration curve is part of the date. When the curve moves, a wiggle-match can move with it.
Open Sixty rings from the outside Single-year sampling found the 774–775 spike 60 rings in from the outside. The felling that follows is AD 835, and the raising is the winter or spring of 835–836. That agrees with the 9th-century artefacts better than 776 does. It retires the Novgorod cross-match of about 930 only if this timber is the platform’s timber. The papers treat it as the early phase.
Open A roof that was completed The lower walls of the rebuild copy excavated courses. The upper walls were finished so the house would stand, from fragments and from later practice. A photograph of the window and the plank roof is a photograph of that decision. It is not a photograph of AD 835.

Location and dating

Location
An island in Lake Āraiši, 57 degrees 15 minutes 7 seconds north, 25 degrees 16 minutes 58 seconds east. The pin is the island. The museum buildings stand on it now.
Period label
Current consensus — open to revision Five phases, late 1st millennium AD on the artefacts. The first phase’s felling is AD 835 in the single-year match, and 776–780 in the earlier wiggle-match.
How this dating was done
Tree rings, tied to radiocarbon one decade-block at a time and then one year at a time, using the atmospheric spike of AD 774–775. The artefacts are the typology that 835 fits and 776 strained.
What could change it
A second timber from the same platform, with its own spike 60 rings in, would show whether 835 is the fortress or one log. A bark edge that is missing would move the count of those 60 rings.
Published ranges, drawn as bars. A bar is a claim about a span, not a point.
Earlier wiggle-matchContested
Single-year felling
776 CE800 CE836 CE
  • Earlier wiggle-match. 95 percent for the last ring was 775–784 on the 2013 curve, narrowed to 776–780 because the 774–775 spike fell in the last sample.
  • Single-year felling. The spike sits 60 rings in from the outside. Felling AD 835, construction in the winter or spring of 835–836. The same blocks on the 2020 curve also prefer the 830s.

What is actually there

Open questions

Question Status
Which felling year is the first phase? The single-year spike points to AD 835, and the artefacts agree. The 776–780 match is what the 2013 curve did with the same timber.
Is the roof you see the 9th-century roof? The lower courses follow the dig. The upper walls were completed so the building would stand. The photograph is the completion.

The current narrative

Where publication stands today. It is the best-supported reading, and it is open to revision. It is not a verdict.

The current narrative — open to revision

Raised in 835–836, on a platform felled that year

The first phase was built in the winter or spring of AD 835–836, after a spruce was felled in 835, and the late-770s date was an earlier calibration’s reading of the same rings.

What supports this

  • The 774–775 atmospheric spike is 60 rings in from the outside, which fixes the count in a way a smooth curve does not.
  • The 2020 calibration prefers the 830s for the older block dates as well.
  • Ninth-century artefacts had already pointed at the 830s, and 835 sits with them.

What pushes back

  • The 2013 match really did give 776–780. A curve can move again.
  • One timber stands in for the platform. A second log could differ if the platform mixed fellings.
  • The Novgorod cross-match of about 930 shows how far a ring comparison can wander without the spike.

What would settle it

It is the centre. A spike in a known ring is stronger than a wiggle on a curve that has since been redrawn. The rebuild in the photograph is not evidence for either year.

Emerging narratives

Other published readings. Each may be right. None is adopted here unless the evidence carries it.

An emerging narrative

A 770s fortress, with later objects

The first felling was 776–780, and the artefacts that seemed to want the 830s are heirlooms, replacements, or a later phase sitting on an older platform.

What supports this

  • The 2013 wiggle-match was a 95 percent result, not a sketch, and the spike was already inside its last sample.
  • Five phases leave room for an early platform and a later house.
  • A single-year count depends on the outside ring being the true bark year. A trimmed log would move 835.

What pushes back

  • The same blocks on the newer curve leave the 770s.
  • The artefact reading and the spike reading now land in the same decade.
  • A trimmed log is possible and not shown. The paper presents an outside edge.

What would settle it

The 770s were the best match for a while. They are the history of the date, not the date the spike now gives. They stay in the room because the curve did that work in print.

Should the current view be reevaluated?

Open to revision

A five-phase lake fortress is firm, and AD 835 is the felling the single-year spike supports. The 776–780 match is published and superseded by a newer curve and a finer sample, which is a reason to show it, not to build the roof on it. The roof in the photograph is a reconstruction. No sky alignment of the island is published.

What drives this

  • A 93-year spruce sequence and a spike in AD 774–775.
  • 776–780 on the 2013 curve, AD 835 when the spike is 60 rings in.
  • A rebuild in new wood, lower courses copied, upper walls completed.

This is the museum’s own weighing of the evidence. It is not a visitor vote, and it is separate from the slider.

Weigh it yourself

Anonymous. Nothing here is stored except on this browser. Your lean does not change the assessment beside it.

Raised in 835–836, on a platform felled that year Unsure A 770s fortress, with later objects

If the contested reading is right

What would follow if a reading that is not the consensus turned out to be true, and what could settle it. This explains the stakes. It does not adopt the reading.

Every early timber was felled in 835

The first phase would be one winter’s work, and the 770s would be only the old curve.

Why it is still worth a test

The spike on a second log from the platform, at the same ring count.

The visible roofs are the ancient ones

A visitor would be looking at the 9th century. The upper walls were finished in the rebuild so the houses would stand.

Why it is still worth a test

A joint or a tool mark on the standing timber that cannot be modern. The dating papers do not offer one, because the standing timber is new.

Sources

  1. Single-year radiocarbon dating of the lake-fortress at Āraiši Radiocarbon
  2. A lake fortress, a floating chronology, and an atmospheric anomaly Geochronometria

Image credits

← Latvia World map