Stonehenge Landscape Flyover

 Fly over the Stonehenge landscape, and watch the sun rise in 2500 BC

I've put a new tool online: a 3D flyover of the Stonehenge landscape, not just Stonehenge but also Woodhenge and the new Bulford post alignment. It runs in your web browser, there's nothing to install.

https://timdaw37.github.io/stonehenge-landscape-flyover/

The ground is built from Environment Agency lidar, with Ordnance Survey terrain for the distant hills and the rivers. On it sit the monuments from Historic England's mapping. Stonehenge has its stones on their surveyed positions, Woodhenge has its post rings from Maud Cunnington's 1929 plan, and there's the newly found pair of posts at Bulford. A period slider lets you strip the landscape back from the Early Bronze Age to the Mesolithic and watch it fill up again.

You can fly with the mouse, keyboard or the touch pad in the bottom right corner, which also works on phones and tablets. Double-click anywhere to fly there, or press P for a plan view.

The part I find most useful is the alignment check. Choose a site (Stonehenge, Woodhenge or the Bulford post axis), then an event such as midsummer sunrise, midwinter sunset or a lunar standstill, and pick first gleam, half disc or full disc. It works out where and when the sun or moon clears the real skyline, as seen from that spot, for 2500 BC or any year you choose. It then puts you there, looking at it. A Details button gives the full workings: the true azimuth, the time, and the horizon that was used.

Everything behind the site is open on GitHub, so anyone can take it and build on it:

https://github.com/TimDaw37/stonehenge-landscape-flyover

It's shared under a Creative Commons licence (CC BY-SA 4.0). Use it, change it, and add your own sites or ideas, as long as you credit where it came from and share your version in the same way. 

It's still a work in progress. If you spot anything wrong, from a misplaced monument to a sunrise that looks off, please let me know in the comments.

Historic England’s Vanishing Archive

Historic England’s photos didn’t vanish. SmartFrame did. Then they broke their own links.

Not just the aerials. The lot.

Historic England still holds the negatives, the prints and the scans. What the public lost, in early September, is the ability to look at them. The official notice, last updated 25 September 2026, is admirably blunt:

“This is because the third-party service that we use to display Archive images on our website has ceased trading and entered administration.”

The 11 September version named the damage:

“This means that images are not currently available across Archive records and the Aerial Photo Explorer. Images of England images are also not available on the National Heritage List for England.”

Ground shots, architectural stills, catalogue images, aerials, and the old Images of England pictures on list entries. All of it.

The third party, named by Michael Pritchard on British Photographic History, is SmartFrame Technologies Limited. Henry Page and Shaun Walker of Mercer & Hole were appointed administrators on 26 August 2026. BusinessCloud reports assets of about £2.3m, cash of £182k, liabilities of £1.7m, and a firm that “failed to secure further funding.”

SmartFrame did not store the Archive. It streamed it: zoom, captions, right-click protection, screenshot blocking, anti-scrape. A public body wrapped a national collection in a start-up’s DRM. When the start-up died, the pictures went with it.

Confirmed down, because HE said so:

• images on Archive records
• the Aerial Photo Explorer
• Images of England photographs on the National Heritage List

Not named in the photo notice: England’s Places; list descriptions; the Swindon search room; copy orders. The early Aerofilms set is still on Britain from Above.

HE first promised previews “in several weeks.” The 25 September update is cooler:

“We are hopeful that smaller preview images will be available on the website by 2 October… 400 x 400px image thumbnails… a longer-term solution… is likely to take many months to implement.”

And, for listed-building photos on the List:

“Images of England images on the National Heritage List for England are not being replaced at this time.”

Please do not email the Archive team. The page says so. Visit, or order a copy.

In 2022 SmartFrame called Historic England “a great example” of its interactive frames. Critics noted the same wrapper was stopping people saving RAF verticals whose Crown copyright had already expired. HE later said, on X, that it had spent £65,830 net on the licence from 2020/21 to 2023/24, putting over a million images behind it, and that it was “not an income generation tool.” In 2024 there was another £41k tender. Call it a little over £100,000 to stop people saving pictures a public body was already putting online.

The comparison is the London galleries. Douglas McCarthy’s FOI figures show the National Gallery Picture Library paying a £100,000 internal fee to exist, then posting net losses of £37,575 and £47,615 in two of the last three reported years. Andrea Wallace’s 2022 UKRI report was blunter: licensing images of public-domain works “does not appear to be a present or future revenue scheme with potential for growth, and is actually a financial burden on the GLAM sector.” Tate, which also bought SmartFrame in 2022 for Tate Images, at least has living artists to clear. Historic England was spending six figures to wrap photographs the taxpayer had already paid for. Then the wrapper died, and the public cannot see them at all.

CEO Rob Sewell wrote that SmartFrame “had all the hallmarks of a global disruptor” and was “seeking to restore industry value to the photography market.” It signed football clubs and picture agencies. It did not sign a plan B for a statutory archive.

Then they did it again to the words.   

UPDATE 29 Sept 2026 - this problem seems to have been fixed

On 16 September 2026 Historic England switched on a new Heritage Gateway and killed the old one. For twenty years the stable record URL was:

https://www.heritagegateway.org.uk/Gateway/Results_Single.aspx?uid=219528&resourceID=19191

That path is dead. No redirect to the record. heritagegateway.org.uk dumps you on a search box. PastScape hob.aspx?hob_id= went the same way years ago and now has nowhere to land.

Volunteer sites took the hit first: Megalithic Portal, Megalithomania, county walk pages, teaching lists — anything that treated a Gateway UID as a permalink. Annoying. Expected, if you hitch your wagon to someone else’s CMS.

Then look at HE’s own map.

The Aerial Archaeology Mapping Explorer still draws. The banks and ditches are on the Open Data Hub. What died is the click. Every grey monument pop-up was supposed to open the full record. HE baked the old Gateway URLs into the GIS. The published schema says so, in words a child could read:

• HE_URL1 to HE_URL5 — “URL link to the relevant Historic England record in Heritage Gateway”
• HER_LINK_1 to HER_LINK_25 — the same for the local HER
• HER_PREF_REF — “alternative HER UID that is used to create the Heritage Gateway URL”

Thousands of those strings are sitting in a national dataset HE still publishes. The Mapping Explorer landing page still tells you a click “will take you to the Heritage Gateway” and still points at heritagegateway.org.uk/gateway. They relaunched the catalogue and left their own map wired to the corpse. Not a third party this time. Not SmartFrame. Them.

A public body spent six figures locking pictures, then cancelled the only stable identifier the rest of the web — and its own aerial mapping tool — had been told to use. HIAS Principle 5 was “a national overview should continue to be delivered online through the Heritage Gateway.” Principle 6 was that knowledge “should not be at risk of loss, fragmentation… or system obsolescence.” They managed both in a fortnight.

Let the photographs free. Out of copyright: a JPEG. Need a tidy file for print: charge for the file, not for the looking. And if you change a national URL scheme, redirect the old ones. A statutory archive is not a Premier League still, and a mapping explorer with thousands of dead links on the organisation’s own website is not a launch.

Bloodlines in the barrows

Diachronic reuse of Neolithic burial monuments: What the Kinship Test Found

After my posts (https://www.sarsen.org/2026/09/borrowed-tombs-or-ozymandias-in-wessex.html and  https://www.sarsen.org/2026/08/a-stranger-in-family-grave.html) on Vuković et al. 2026, a query arose that the paper's main text doesn't answer. Did the study test whether the people buried at each site were related to each other, especially the three infants buried together in the Fir Tree Field pond barrow? And what exactly were the Monkton four: all siblings, or a mother and three children? The genetics finds two boys among the children, where the bones suggested one. Children can't be reliably sexed from the skeleton, so this is a correction rather than a surprise.

UPDATE from Martin Green: " the adult female at MUW97 was F23c - so they are a mother and 3 children"

The main text only has room for Monkton. The answers are in the supplement. The kinship test found one first-degree family, one third-degree pair, and no link across the gap at the three reused monuments. The triple infant burial turns out to be the most surprising result of all. Figures are from Table 1 and Supplementary Data Files 9 and 10, unless noted.

How the test was run

READv2 was run on pairs, with the Neolithic individuals kept apart from the later ones so the baseline is not mixed. A first-degree pair is a parent and child, or full siblings. The authors specify a pair as siblings only where the overlap is at least 40,000 markers. Lower pairs are printed and not used for that call. The Neolithic pairs are Supplementary Data File 9. The later pairs are Data File 10. There is no published pair that joins an early burial to a later burial at the same site.

Monkton-up-Wimborne

Four people in one oval grave, dates overlapping 3364–3093 BC. The main text treats two arrangements as the most parsimonious: all four full siblings, or a mother of 30–45 with three children of about 5, 9 and 10. The site note says a woman of about 30, two girls of 5–10 and a boy of about 9. Genetic sex does not match that reading of the bones. Table 1 has two males and two females.

ID	Skeleton	Date (95.4%)	Coverage	Sex	mtDNA	Y
muw002	F23B	3364–3101	0.34×	XY	K1a+195	I2a2a
muw003	F23C	3363–3101	0.11×	XX	K1a	–
muw001	F23A	3357–3093	0.78×	XY	K1a+195	I2a2a1a1a2
muw004	F23D	3340–3096	0.41×	XX	K1a+195	–

All six pairs are called first degree. Two clear the 40,000-marker line, and both are called full siblings: muw001–muw002 (42,708 markers, kinship 0.24) and muw001–muw004 (49,696, kinship 0.25). The other four are first degree on less data, and the supplement will not say sibling rather than parent for those.

Pair	Markers	Call	Kinship
001–004	49,696	First degree, siblings	0.25
001–002	42,708	First degree, siblings	0.24
002–004	27,645	First degree	0.24
001–003	15,849	First degree	0.26
003–004	10,282	First degree	0.21
002–003	8,788	First degree	0.25

So muw001, muw002 and muw004 are one generation. muw001 and muw002 share I2a2a; muw001 is called to a finer branch. Three of the four are K1a+195. muw003 is called only K1a, on the lowest coverage and the lowest mitochondrial coverage (7.25×, against 37–107× for the others). That is a shallower call, not a second maternal line. The paper's statement is that all four lie within K1a.

muw003 is in every pair below 16,000 markers. She is the person the test cannot place. If she is the adult woman, a mother and three children fits, and the three children are already the sibling set. If the adult is muw004, the test has already called her a full sibling of muw001, so she is an older sister, and muw003 is another sibling rather than the mother. The sample list is F23A = muw001, F23B = muw002, F23C = muw003, F23D = muw004. Which of those four skeletons is the adult is not in the paper. That is the only fact still needed.

The later man is not in that test. muw005, 1519–1321 BC, coverage 0.14×, is male, mitochondrial haplogroup H, Y-haplogroup R1b1a1a2. No READv2 pair links him to the four. The runs were split by period. The paper calls him unrelated on ancestry, not on a kinship coefficient. He was laid on the surface of the partly silted central pit, not in the family's grave, and covered with a small flint cairn. His mitochondrial haplogroup is not theirs, and neither is his Y line.

Fir Tree Field pond barrow

Fourteen centuries later and not far away, on the same stretch of Cranborne Chase, the question is reversed. At Monkton the bones suggested a family and the genetics confirmed it. At Fir Tree Field a triple burial of infants looks like a family, and the genetics says otherwise.

Five individuals, all Early Bronze Age. The site note identifies four of them. F14a, F14b and F14d are the three infant inhumations in F14. F6 is a separate infant, 18 months to 2 years, extended, with a Food Vessel. F27 is in the sample list and not in that note.

ID	Feature	Who	Date (95.4%)	Sex	mtDNA	Y
dfb004	F14d	Infant	2008–1748	XY	U5b2a4	R1b1a1a2a1a2
dfb002	F6	Infant, 18 months–2 years	1955–1781	XY	I1a1	R1b1a1a2a1a2
dfb005	F27	Not described in the site note	1920–1769	XX	U5b2a1a2	–
dfb003	F14b	Infant	1886–1750	XY	T2b2b	R1b1a1a2
dfb001	F14a	Infant	1882–1745	XY	H1b	R1b1a1a2

UPDATE from Martin Green: "F27a was in a separate grave pit on the northern edge (one of only two burialsin that area), the other an unaccompanied cremation in a funnel-shaped pit.
F27a contained an infant between 0-2 months
The grave pit had partly cut one of the Beaker pits (F27) also in that area."

The five mitochondrial haplogroups are five different ones. That excludes a mother and child, and siblings through the mother, among these five. The three in F14 are unrelated to each other: kinship 0.01 (F14a–F14b, 21,445 markers), 0.02 (F14a–F14d, 20,579) and 0.03 (F14b–F14d, 25,940), against about 0.25 for a first-degree pair. A brother or a parent inside the triple would have shown. One was called at Monkton on fewer markers than any of these.

The relationship that was reported is not inside the triple. dfb002 (F6) and dfb004 (F14d) are third degree: 39,862 markers, kinship 0.06. Same Y-line, R1b1a1a2a1a2, and different mitochondrial haplogroups, I1a1 and U5b2a4. The supplement calls it a distant paternal relative. Between two infants, third degree most simply means first cousins, children of two brothers, though the test cannot say which third-degree relationship it is. The authors treat it as one of the two post-Neolithic results that clear their reporting line. The other is not a relative: cfa001 and a previously published Canada Farm individual are the same person sampled twice.

The other two boys in F14 are called only to R1b1a1a2, the clade above that longer type. That can be as far as their data go. It is not a second reported relationship. F27 is unrelated to the other four on the pairs that could be read. She is the one individual in the barrow without a described context: female, dated with the others, with ordinary Bronze Age ancestry, but not in the site note, so her age, grave and position are unknown here. Her mitochondrial haplogroup, U5b2a1a2, sits on the same U5b2 branch as the F14d infant's U5b2a4, but on a different twig. That shared maternal ancestry lies far too deep to count as family.

Across the gap

Which leaves the question the paper was built around. At the three monuments where a later burial sits in an older one, the later person does not share a mitochondrial haplogroup with anyone earlier sampled there.

Site	Earlier	Later
Monkton-up-Wimborne	Four, all within K1a, 3364–3093 BC	muw005, H, 1519–1321 BC
Sisters Long Barrow	K1a4a1b, J1c3, H4a1a1, 3888–3527 BC	slb016, H1ak1, 1534–1436 BC
Sale's Lot	lot005, K1b1a1, 3946–3654 BC	lot001, H6a1b, 2622–2467 BC

No READv2 pair was published between the early and late individuals at these sites. A first-degree link across those gaps was not a plausible thing to find, and was not tested in the published runs. A maternal line from the particular people sampled is ruled out by the haplogroups. A paternal line is ruled out only at Monkton, where both the early boys and the later man have Y data and they differ (I2a2a against R1b). At Sisters and Sale's Lot the later burials are women, so there is no Y line to compare. Descent from Neolithic Britain in general is a question for the ancestry models, and for my earlier post.

Within the whole set, the only kinship READv2 reports between different people is the Monkton cluster and the one third-degree pair at Fir Tree Field. Everywhere else the within-site pairs are unrelated, or the overlap was left unused.

What it adds up to

So, to answer the query: they did look. The kinship test was run on everyone with enough data, and it simply found very little. At Monkton it found the family everyone expected, and it is one fact away from saying which family. If the adult woman is F23C, she is the mother of three full siblings. If she is F23D, all four are siblings. Someone with the skeletons can settle that.

At Fir Tree Field it found something nobody expected. The three infants laid together in F14 were not brothers, and not close kin at all. Whatever put them in one grave, whether a shared death, a shared season or a shared rite, it was not close family. The one family thread in the barrow runs between graves instead: the infant in F6 and one of the F14 three were cousins, or something like it, through their fathers. The pond barrow held a lineage, loosely, but not a household.

And across the gap, nothing, which is what anyone should have expected. Kinship tests find parents, siblings and cousins. They cannot bridge twelve centuries, let alone twenty-two. The strangers in the old graves were not the builders' grandchildren. Whether they were, in some remote and untraceable way, their descendants is a question kinship tests cannot answer.

---

Vuković, N. et al. (2026). Diachronic reuse of Neolithic burial monuments by Bronze Age newcomers in Southwestern Britain. Scientific Reports 16:26819. doi:10.1038/s41598-026-66094-z. Table 1; p. 6; site notes for Monkton-up-Wimborne and Fir Tree Field pond barrow; Supplementary Data Files 9 and 10. The position of muw005 on the surface of the infill is from the excavator's account.

Borrowed tombs, or Ozymandias in Wessex?

A paper in Scientific Reports gives the genomes of thirty people from twelve burial places in Gloucestershire and Dorset, buried between about 3800 and 1400 BC. Those laid down before 3000 BC belong with the Neolithic farmers who raised the long mounds. Those laid down from about 2550 BC are a different people, with the ancestry that comes into Britain among the Beaker dead. At three of the old monuments the later dead were put into earth their own ancestors had not built. I looked at the science: https://www.sarsen.org/2026/08/a-stranger-in-family-grave.html but there is a another puzzle.

Continued use is not the puzzle. The puzzle is what the use was for. It may be a tradition, taught on, even after the bloodline has changed. Or it may be Shelley's traveller, who comes on "two vast and trunkless legs of stone." He reads the boast on the pedestal. He does not inherit the kingdom. 

The graves look like that second kind of use, and the years are the reason. At Sale's Lot a woman was laid in a mound twelve centuries after the people who built it. At Monkton-up-Wimborne a man was set down in an old hollow seventeen centuries after the family buried there. At Sisters Long Barrow a young woman went into a mound twenty-two centuries after the men who lay in it. Not a rite kept up year by year. Two moments, far apart: one around 2500 BC, as the Beaker dead were beginning to arrive, and one around 1500 BC, when axes were being carved on the Stonehenge sarsens. The later dead were mostly not the builders' descendants. They wanted the places. Wanting a place is what the traveller does.

Beacon Hill, south of Newbury, is the same act with the paperwork kept. The hill was already an Iron Age fort, and the new grave was cut into it some two thousand years after the rampart. There were older barrows on it, it had carried a beacon, and skeletons had been dug up there long before. In April 1923 the fifth Earl of Carnarvon, who had stood at the opening of Tutankhamun's tomb and then died in Cairo, asked to be buried there in a mound of earth, with trees and stones, and not to have more than fifty pounds spent. He was laid in a shaft in the corner of the fort. We know what the grave was for, because he wrote it down. Nobody that day was keeping the faith of the people who dug the ditch. They were taking a high old hill and giving it a meaning of their own.

Twelve centuries, or seventeen, or twenty-two, is long enough for a place to stay standing and lose its name. It does not show that anyone could say who lay inside it. Stonehenge, which might have been the exception, is not much stronger. The Avenue, dug across the change of people, does carry the solstice line out to the river, and that is the best case for a plan remembered. But the later pits outside the sarsens are irregular, and nobody agrees what they were for. What is plain is a long habit of adding a mark: axe-heads cut in the stones, then names, one said to be Wren's, and cremated ashes still placed there, unofficially, in our own time. A claim, repeated. Not a plan obeyed. Shelley's man could read every word that had been cut. The sands were level all the same.

The genomes settle only the blood. It largely changed. A memory could still have been carried across that break: strangers can be taught, and children can forget. Nothing in these graves shows the teaching. The nearest we can come to watching a handing-on is watching one being made. At Stonehenge now, the druids keep a faith they have given the place, one the builders could not have handed them. The handing-on had already stopped, more than once, before anyone in the eighteenth century thought to begin it again. That is not a tradition out of the Neolithic. It is the traveller, returned, finding a family for the ruin. Carnarvon may have his mound. The druids may have their dawn. Neither are keeping an older faith. Both are giving a high old place a meaning of their own.

Vuković, N. et al. (2026). Diachronic reuse of Neolithic burial monuments by Bronze Age newcomers in Southwestern Britain. Scientific Reports 16:26819. 

Carnarvon’s instruction is the codicil to his will, April 1923. 

Shelley, P. B. (1818). Ozymandias.

Woodhenge: the line and the rings

Abstract. The claimed solstitial alignment of Woodhenge is not supported by the layout of its posts. It rests on a line drawn on Cunnington's 1929 plan at the midsummer sunrise bearing (49.4° true, the first gleam in 2500 BC). When the 156 marker posts are fitted ring by ring, every long axis lies north of that bearing. The outer rings A and B are 10–12° north of it, and their 95% ranges exclude the sunrise. The inner rings C–F are 4–8° north. They are too loosely defined to rule the sunrise out, but they do not demonstrate it. Along Cunnington's line from the centre, the view passes through ordinary gaps between posts, and the gaps themselves are centred a few degrees further north.

The June post on Woodhenge accepted the textbook position: the long axis of the timber rings points, broadly, at midsummer sunrise. That position rests on one line, the dashed axis on Maud Cunnington's 1929 plan, labelled "line of midsummer sunrise, elevation of horizon 30′". The new Woodhenge 3D model allows that line to be tested against the post positions themselves. The result is that the line meets the sunrise, but the rings do not follow the line.

The model and its checks

The model places the 156 ring posts (A 60, B 32, C 16, D 18, E 18, F 12) on the concrete markers set out by the Cunningtons, in Ordnance Survey coordinates, on lidar-derived terrain. The twelve supernumerary holes are not included. Two independent plans were used to check the geometry:

• Cunnington's 1929 plan, with its postholes detected automatically and matched to the model by a scale-and-rotation fit: median residual 0.24 m (90% within 0.48 m), rotation −0.1°, no measurable stretch.
• The 2008 GPS survey of the markers by the Jenks brothers: median residual 0.06 m, rotation 0.0°. Three ring-A markers in the east-south-east are missing from that sheet but present on Cunnington's plan.

The centre of the post pattern lies within 1 m of the published position of the monument. The markers are an interpretation of the holes rather than the holes themselves, and they do not precisely preserve the excavated positions (The Past, 2025), but the fit to Cunnington's own plan shows that any such error is well below a metre.

The sun

Sunrise and sunset are computed with the astronomy-engine library (apparent positions, standard refraction, solar semi-diameter 0.27°) against a skyline measured from the lidar terrain at an eye height of 1.6 m above the centre. In the solstice direction that skyline stands at 0.45–0.51°, in good agreement with Cunnington's 30′. For 2500 BC the most northerly midsummer sunrise falls at the following true azimuths:

• first gleam: 49.4°
• half disc: 49.8°
• full disc on the horizon: 50.2°

The modern equivalent (first gleam) is 50.1°. Midwinter sunset in 2500 BC falls at 229.7–230.6°, over a skyline of 0.28°. On this terrain the south-western horizon is lower than the north-eastern one; any obstruction of the midwinter sunset would have to come from something not in a bare-earth model, such as vegetation.

Cunnington's line

Registered to the markers, Cunnington's dashed line lies at 49.4° true and passes within 0.3 m of the centre of the post pattern. It therefore points at the first gleam of the 2500 BC midsummer sunrise to within a tenth of a degree. On the plan the line is marked 50¾° from her north arrow. When the plan is registered, that arrow sits about 1.3° west of true north, which accounts for the difference between the label and the registered bearing.

Figure 1. Cunnington's 1929 plan with the model's marker posts (blue circles) superimposed. Her dashed line runs south-west to north-east at 49.4° true. The orange and green lines are the long axes fitted to rings A–B and C–F. Plan: Cunnington 1929.

The line is drawn at a sunrise bearing through the centre. Nothing on the plan shows it being derived from the posts, and the question is whether the posts agree with it.

Where the rings point

The long axis of each ring was found by fitting an ellipse to its post positions (least-squares conic fit). Uncertainty was estimated by bootstrap: 3,000 resamples of each ring's posts, each post also displaced by a random error of 0.3 m, the fit repeated each time. As a check, the axis of best mirror symmetry of each ring was found independently. It agrees with the ellipse axis to within 3.5° for rings A–E. For ring F the fit is unstable.

Ring	Posts	Size (m)	Long axis	95% range	Short of 49.4°
A	60	44.8 × 40.7	39.7°	35.3–44.1°	9.7°
B	32	39.0 × 34.6	37.5°	32.0–43.2°	11.9°
C	16	29.9 × 25.3	41.0°	32.7–49.6°	8.4°
D	18	23.1 × 19.2	44.4°	34.0–53.3°	5.0°
E	18	18.1 × 14.0	45.2°	35.7–54.6°	4.2°
F	12	11.8 × 8.6	42.7°	26.6–59.7°	6.7°

Azimuths are true (OSGB grid bearing + 0.17°).

Figure 2. The fitted long axis of each ring, with its 95% bootstrap range, against the 2500 BC midsummer sunrise (shaded) and Cunnington's line (dashed).

Every best estimate lies north of the sunrise. None of the rings can point at any sunrise, since 49.4° is as far north as the sun ever rises here. The rings fall into two groups:

• Rings A and B lie about 10–12° north of the sunrise, and their 95% ranges exclude it.
• Rings C–F lie 4–8° north. Their ranges include the sunrise, or in the case of C only just reach it, because rings of 12–18 posts this close to circular do not fix an axis well. If only positional error is allowed for, without resampling the posts, the range for C (35.9–46.0°) also excludes the sunrise, while D (38.6–50.3°) and E (39.4–50.8°) still just include it.

This division matches the suggestion by Chadburn (2010) and Chadburn and Ruggles (2017) that only rings C–F follow the astronomical axis. The fits support it, and add that rings A and B are measurably off it. In Ruggles and Chadburn's grades of precision, C–F sit at the "broadly solstitial" level of about 5°. A–B fall outside it.

The sightline

The Cunningtons watched the midsummer sunrise from the site in 1927 and believed that a gap in the posts could have been used to observe it (The Past, 2025). Along Cunnington's line from the centre, the nearest post centres are as follows:

Ring	Nearest post, NE (m)	Gap centre, NE	Nearest post, SW (m)	Gap centre, SW
F	0.94	46.9°	0.18	238.5°
E	0.30	42.6°	1.07	227.9°
D	1.16	46.5°	0.18	220.8°
C	1.79	45.4°	1.31	223.6°
B	1.04	45.9°	0.37	223.8°
A	0.43	47.4°	0.42	231.0°

Distances are from the post centre to the line. The gap centre is the bearing of the midpoint of the gap the line passes through.

At every ring the line crosses an ordinary inter-post gap. Only ring B's gap is wider than its normal spacing (13.0° against 10.9°), and no corridor is left open along the line. Towards the north-east, the view clears the posts as long as the E posts were less than about 0.6 m across and the A posts less than about 0.85 m. The A posts are estimated at 0.3 m (Marshall et al. 2024). The diameters of the D–F posts are not known. Towards the south-west the line passes 0.18 m from a D post and from an F post, which would block it unless those posts were under about 0.35 m across. The gaps themselves are centred on 42.6–47.4° in the north-east, averaging about 46°, and 220.8–238.5° in the south-west. The open corridor through the rings therefore runs a few degrees north of Cunnington's line, not along it.

Figure 3. The 3D model at first gleam, midsummer 2500 BC, in plan view. The sun ray (yellow) runs along Cunnington's line (white). The fitted C–F axis (green) and A–B axis (orange) diverge from it. Posts are drawn at a uniform 0.64 m diameter.

What this changes

The solstitial axis at Woodhenge is a property of Cunnington's line, not of the timber rings. The line was drawn at the sunrise bearing through the centre of the rings, and the alignment as usually stated derives from it. Measured against the posts:

• rings A and B are oriented about 10–12° north of the 2500 BC midsummer sunrise, and their 95% ranges exclude it;
• rings C–F are oriented 4–8° north of it, which is consistent with a broad solstitial intent but does not establish one;
• the clear sightline along the line depends on post diameters that are known only for rings A–C, and the free corridor through the rings is centred a few degrees further north.

Because the axes lie north of any possible sunrise, they cannot be solar on any date. Lunar and stellar targets are available. At the major standstill of about 2498 BC, the most northerly moonrise (upper limb, over the same skyline) falls at 39.7° and the most southerly moonset at 218.0°. These coincide with the ring A axis (39.7°) and the A–B mean (38.6°/218.6°). Capella in 2600 BC, first visible 1–2° above the horizon, rises at about 43–45°, on the C–F axis (43.3°). Neither coincidence carries much weight. Solar and lunar extremes together offer about a dozen target directions, and the rising and setting points of the first-magnitude stars fall within 3° of roughly a third of all azimuths. The ring axes span 37.5–45.2°, each with a 95% range of 9° or more, so choosing a target ring by ring enlarges the chance further. Nothing independent points to the moon or to Capella at Woodhenge, and no lunar claim for the rings has been published. The coincidences are recorded here so that "north of the sun" is not read as "aligned on something else".

The June post concluded that the long axis of the rings "does point, broadly, the right way". That holds only for the inner rings, and only at the broad end of the precision scale. It does not hold for rings A and B.

Limits

All of the geometry uses marker positions checked against Cunnington's drawn plan, which is itself her record of the holes. An error common to both would not be detected. An ellipse is one model of the ring shape. Thom's egg-shaped constructions define the axis differently and would give different figures. The bootstrap treats posts as interchangeable, which is crude for rings of 12–18 posts. The results would be most improved by the D–F post diameters and an independent survey of the excavated holes.

Sources

• Chadburn, A. 2010. "Case study 2.1: Stonehenge World Heritage Site, United Kingdom." In C. Ruggles and M. Cotte (eds), Heritage Sites of Astronomy and Archaeoastronomy in the Context of the UNESCO World Heritage Convention. Paris: ICOMOS/IAU, 36–40.
• Chadburn, A. and C. L. N. Ruggles. 2017. "Stonehenge World Heritage Property, United Kingdom." In C. L. N. Ruggles (ed.), Heritage Sites of Astronomy and Archaeoastronomy in the Context of the UNESCO World Heritage Convention: Thematic Study no. 2. Paris: ICOMOS, 41–62.
• Cunnington, M. E. 1929. Woodhenge: A Description of the Site as Revealed by Excavations. Devizes.
• Marshall, P., A. Chadburn, I. Hajdas, M. Dee and J. Pollard. 2024. Woodhenge, Durrington, Wiltshire: Radiocarbon Dating and Chronological Modelling. Historic England Research Report Series 94/2024.
• Ruggles, C. L. N. 2006. "Interpreting Solstitial Alignments in Late Neolithic Wessex." Archaeoastronomy 20: 1–27.
• Ruggles, C. L. N. and A. Chadburn. 2024. Stonehenge: Sighting the Sun. Liverpool University Press / Historic England.
• The Past. 2025. "100 years of Woodhenge: tracing an archaeological icon, from discovery to new dating evidence."
• Jenks brothers. 2008. GPS survey of the Woodhenge marker posts (unpublished plan).
• Cross, D. Astronomy Engine (software library).
• Model and data: github.com/TimDaw37/stonehenge-block-3d.

Woodhenge - The 3d Model

 

An interactive 3D model of Woodhenge, accurately georeferenced with the sun, sky and moon modelled over the age of the monument: https://timdaw37.github.io/stonehenge-block-3d/woodhenge/

Stonehenge - The Complete 3D model

 

Autumn Equinox Sunrise

Stonehenge, rebuilt stone by stone and set moving: timdaw37.github.io/stonehenge-block-3d is a free, interactive 3D model of the monument, built from survey-locked stone positions on real LiDAR-derived ground — open it in a browser and walk around all 93 stones, or watch the sky itself move over them.

The sun and moon travel across the horizon in real time, not as a fixed marker: a time slider and "Sun's arc" / "Moon's arc" controls animate the rise, with Gleam, Half orb, and Full orb settings for exactly when a disc counts as up.

An Epoch control swings the whole sky between Modern and 2500 BC, or any year typed in. Go back to 2500 BC and the pole star marker moves too — from Polaris to Thuban — the slow wobble of the Earth's axis, not a rounding error.

A Newham mode draws the station stones' own sightlines — long sides for the moon's furthest rising and setting points, short sides for the solstice sun — as lines running through the stones themselves, with Most Northerly / Most Southerly toggles for the moon's extremes, and a readout comparing the stone-pair bearing against the live sky azimuth side by side.

A Complete view shows the monument with a theoretical complete sarsen build: every lintel in place, the fallen stones ghosted rather than erased. Terrain, orbit and top-down views, and hover-for-coordinates in both OSGB and WGS84 are built in throughout.

Source and data are open on GitHub.

Ramson Cliff erratic: a second thin section still points Cornubian, not glacial

A second thin section, same answer: still more Cornubian than glacial.

Prof Slack has kindly loaned a second thin section cut from the original hand specimen of the Ramson Cliff (Croyde) epidiorite — the ~700 kg block at ~80 m OD on Baggy Point. Discussion with Dr Mik Markham, who knows Cornubian greenstones and the axe-head groups as well as anyone, prompted a fresh transmitted-light description. It confirms the account in Daw, Ixer & Madgett (Quaternary Newsletter 167, 2026) with only small corrections.

The rock is a highly altered ophitic microgabbro with relict clinopyroxene, no olivine and no quartz. Secondary minerals are dominated by brown-green hornblende and colourless-to-green actinolite, two generations, with actinolite the later, and by minor chlorite, epidote and clinozoisite. Biotite and white mica were not confirmed. That assemblage matches the outer aureole of the Dartmoor Granite as described in the BGS Okehampton memoir, and is consistent with Cornubian contact metamorphism more broadly. It does not match non-aureole Devon or Cornish greenstones, which reach actinolite grade only and carry albitised, sericitised feldspar; nor does it match the Welsh Ordovician metadolerites on the Irish Sea route. A Scottish Dalradian source cannot be excluded on the amphibole evidence alone, but remains less likely on transport grounds; further tests on the relict pyroxene chemistry and plagioclase composition would settle it.

What has not changed is the emplacement argument. The block has been used as evidence that Irish Sea ice reached ~80 m OD on this coast. That reading was already thin: first recorded standing upright in pasture in 1969, absent from earlier maps and photographs, angular and unabraded, and the only claimed erratic on the south Bristol Channel shore above ~30 m OD. A Cornubian source does not prove human movement, but it does make a lone high-level glacial drop even less likely. A second slide from the same specimen does not turn an isolated, poorly documented boulder into proof of high-level ice, nor into a stepping-stone for glacial bluestones on Salisbury Plain.

Full description below.

---

Rob Ixer's Analysis

A second thin section from the original hand specimen was generously loaned by Prof Slack and prompted discussions with Dr Mik Markham, an authority on Cornubian greenstone axes and on the petrography of Cornubian altered greenstones.

A petrographical examination of it confirmed the original petrographical description (Daw et al. 2026) albeit with some minor corrections. The un-named green amphibole is identified as actinolite. Trace amounts of white mica/muscovite and mixed muscovite-chlorite were not confirmed (the latter was probably misidentified mixed chlorite–limonite-stained chlorite) but trace amounts of epidote with high interference colours were. The suggestion that the rock could be Cornubian in origin is maintained.

The rock is a highly altered microgabbro, hence locally plagioclase–pyroxene have their characteristic ophitic relationship. It comprises relict primary pyroxenes showing both high and low interference colours, unaltered to highly altered plagioclase feldspar, and skeletal and equant-shaped opaques including probable titanomagnetite. The degree of alteration varies on a small scale and secondary alteration minerals are dominated by brown-green amphibole (‘hornblende’) and green actinolite and minor chlorite, with trace amounts of titanite, epidote and clinozoisite. White mica, if present, is very rare. Amphiboles replace and pseudomorph pyroxene but also occur as discrete and distinctive fine-grained mosaics intergrown with chlorite and enclosing very minor titanite, possible zircon and feldspar (perhaps albite). Texturally there is a strong suggestion of more than one generation of amphibole, with actinolite being the later.

Unzoned polysynthetically twinned plagioclase is variably altered from largely unaltered to highly altered, hence much relict feldspar is present. The main alteration is to fine-grained actinolite crystals, with this alteration initiated along cleavage and twin planes. Very minor amounts of chlorite, epidote with high interference colours and probable clinozoisite with low interference colours accompany actinolite. Plagioclase altering solely to fine-grained clinozoisite is rare; fine-grained white mica could not be positively identified.

Pyroxene displaying both high and low interference colours is the only primary mafic mineral; no olivine or pseudomorphs after olivine are present. Relict pyroxene is enclosed within lower-relief amphibole; both are in optical continuity. Much pyroxene is altered to colourless to grey-brown-green amphibole with a good cleavage and with very fine-grained titante lying along that cleavage; this amphibole is enclosed within green actinolite rims. Amphibole fringes about pyroxene are absent but total replacement is common. Other pyroxenes are altered to green actinolite along cleavage and fracture planes or to mosaics of fine-grained stubby actinolite. Locally pyroxene with low interference colours is altered to amphibole with unusual yellow interference colours.

Although a positive identification of the opaques is not possible in a normal thin section, their habit (equant and skeletal) and texture (abundant opaque ilmenite laths within a less opaque different phase; titanite replacing magnetite) strongly suggest the presence of altered titanomagnetite. Lobate opaques that would suggest ilmenite are absent. Opaques are replaced/pseudomorphed by actinolite and an opaques–actinolite association is widespread.

Pleochroic colourless to green actinolite (brown-grey-green amphibole is absent from this association) forms mosaics comprising euhedral to subhedral blocky crystals. Although some replace/pseudomorph pyroxene and opaques, most infill spaces between plagioclase laths; all mosaics are fine-grained but they vary in grain size between aggregates. Some just contain actinolite, others actinolite in minor chlorite, and a few are chlorite-rich with actinolite laths, often radiating, growing into the chlorite. Trace amounts of titanite, a high-relief accessory mineral within its pleochroic halo (zircon perhaps) and small twinned feldspar with fluid inclusions (possibly albite) are present but rare.

Colourless to very pale green pleochroic chlorite with blue and very rarely brown interference colours is the second most abundant secondary mineral after amphibole in amount, but is uncommon. It forms thin cross-cutting veinlets or occurs as a very minor secondary mineral in plagioclase. Most chlorite occurs as short stubby crystals surrounding stubby actinolite or as the main phase enclosing actinolite laths and enclosing minor titanite and possible albite. Some chlorite laths are intergrown with a phyllosilicate with high interference colours; although superficially it looks like muscovite, it may be limonite-stained chlorite. Biotite and positively identified muscovite were not recognised.

Trace amounts of titanite may replace primary iron–titanium oxides or lie along cleavage planes in pseudomorphing amphibole. Small euhedral rhombic titanite is present within stubby actinolite–chlorite segregations.

Very minor amounts of high-relief epidote form thin laths along twin planes in plagioclase; other high-relief epidote-group minerals with blue interference colours are visually identified as clinozoisite.

Trace amounts of sulphide, now altered to limonite, are associated with actinolite mosaics.

Possible zircon has a pleochroic halo when enclosed in actinolite.

This new petrographical description continues to allow the possibility that the erratic is Cornubian in origin, namely a microgabbro that has suffered contact metamorphism from the underlying granites. It cannot be matched to any of the main IPG Group Cornubian axe groups (Groups I–IV), but nor can many Cornubian axe-heads.

Stonehenge Access All Areas - all the way to 11- the final episode (for now)

Stonehenge 3D Model

Stonehenge 3D Model — deceptively simple

It looks like a toy. Grey boxes, a blueish ring, pink Altar, two bright silver station markers. Orbit with a finger. Tap for coordinates. Switch the sun and moon between today and c. 2500 BC.

Under the Minecraft-box look is a finished layout of the stones — each one placed and sized so lintels sit on uprights — on LiDAR-derived ground that rolls out to the horizon when you turn Terrain on. Tap or hover for OSGB / SU / WGS84 coordinates (same language as the stonehenge-plan page, without draping a plan drawing under the model). Sky controls aren’t just decoration: sunrise and sunset by day of year, moonrise and moonset on a most-northerly ↔ most-southerly dial, with an Epoch switch so the azimuths use the Earth’s obliquity for Modern or for around 2500 BC (flat-horizon geometry, stated plainly).

Phone-friendly controls that tuck away. Labels you can turn on. Coords you can pin. No app store, no login — just open the link.

I’m publishing it as a public baseline others can fork and improve (TimDaw37/stonehenge-block-3d). The hard part was getting the stones to sit true on real ground and the tools to stay quiet. The easy part is what you get to do with it.

I am seriously impressed by this model, there are layers of features built in to an accuracy that no other model has. Have a play.

Brimham Rocks and the Devil's Arrows

European Anthropic Megalithic Transport Interactive Gazetteer

 

European anthropic megalithic transport is a new interactive gazetteer of every case in Neolithic and Chalcolithic Europe where a published petrographic or quarry match demonstrates genuine human transport of architectural stone — as distinct from a stone that arrived by ice, or one simply raised on the rock it was quarried from.

What it shows

Each entry is a matched pair: a monument and an identified source, joined by an arrow whose weight scales with the published distance. The map filters by region and by evidence quality (high vs medium confidence in the underlying source-matching), and carries a schematic overlay of ice-sheet margins — a reminder that in previously glaciated regions, a "close" erratic and a genuinely local outcrop are not the same thing.

A distance histogram sits alongside the map: a large peak under a few kilometres, and a thin tail stretching out to the exceptional cases — Carn Goedog and Craig Rhos-y-felin to Stonehenge, the Orcadian Basin candidate source for the Altar Stone, Matarrubilla's palaeo-estuary haul at Valencina. Tapping a bar isolates that distance band on the map.

Stonehenge's four hauls can be switched off entirely with a "Not Stonehenge" toggle, so the rest of the European record — Brittany's orthogneiss, the Guadalquivir sites, Newgrange's quartz and greywacke, the Devil's Arrows' Millstone Grit — can be read without one monument dominating the picture.

Sourcing

Every line traces to a cited quarry-provenance study — petrographic, geochemical, or direct archaeological identification of an extraction site — not to tradition or inference from stone type alone. The underlying data is downloadable as CSV, and the full gazetteer lists sources and references for each entry. Corrections and additions, particularly for regions still thin on the map, are welcome via the GitHub repo.

Braemore, not Sarclet: at least they're updating

The Telegraph today has Stonehenge's Altar Stone "formed in Braemore, Caithness," with Dr Remy Veness saying it "seems likely it came from within a 50km [30-mile] radius of Braemore," and the body copy declaring that "the sandstone around Braemore is the closest match."

That is a striking upgrade from the peer-reviewed paper Veness co-authored earlier this year.

What the paper said

Clarke, Veness et al. (2026), From Highlands to Henge (J. Quaternary Science):

"the sample most similar to the Altar Stone zircon population is the ORS at Sarclet (p-value = 0.96)"

"The Altar Stone detrital zircon cargo is also indistinguishable from ORS at Braemore, Kirtomy and Portskerra"

"(i) Sarclet & Braemore, combined and henceforth named as Caithness"

"both approaches identify Sarclet as the closest match"

Conclusion: Caithness or Inverness–Black Isle — a region, not a village.

Sarclet was the headliner. Braemore was the supporting cast, unquantified beyond p > 0.05.

What this blog said (with numbers)

5 July 2026 — Independent audit of that Sarclet claim (sarsen.org):

"Independent replication confirms the broader claim (four of five localities are statistically indistinguishable) but does not reproduce Sarclet as uniquely strongest. Braemore matches comparably or better."

Locality	Replication p	Clarke et al. (2026) reported
Sarclet	0.865	p = 0.96 (strongest claimed)
Braemore	0.889	p > 0.05 (not quantified)

"Braemore matches as well as or better than Sarclet across discordance thresholds and resampling scenarios."

3 September 2026 — fragment-by-fragment check (sarsen.org):

"Two of the three [fragments] are a closer match to Braemore than to Sarclet — the same ranking the July audit found on the pool."

"The group-level result is a Caithness neighbourhood (Sarclet, Braemore, Kirtomy, Portskerra), not a single harbour."

Also as Daw (2026), Screening the Orcadian Basin (10.13140/RG.2.2.10365.12008).

What the Telegraph said today

Standfirst: formed in Braemore, Caithness.

Veness: "within a 50km … radius of Braemore."

Body: "The sandstone around Braemore is the closest match for the Altar Stone."

Sarclet does not get a look-in. The paper's own closest-match locality has been quietly demoted in favour of a site the paper treated as one of several.

Meanwhile: Doggerland still fails Ockham

The provenance tweak is the interesting bit of today's coverage. The transport story wrapped around it is not.

The Telegraph retells the glacial-assist narrative: ice picks the stone up in Caithness, dumps it on Doggerland, Mesolithic people then somehow salvage and move it before the North Sea takes the place, and Neolithic people later finish the job to Salisbury Plain. That chain fails Ockham's razor in several large ways — and the paper itself already admits most of them.

The ice does not want to go there. Modelled flow from Caithness is predominantly north-east. A south-east stream to Dogger Bank needs a narrow, special-case pathway (often with an extra nudge into the Moray Firth first). You are not riding the main current; you are picking the one corridor that keeps the glacial story alive.

The clock does not work. Dogger Bank was inundated by post-glacial sea-level rise well before the Altar Stone is likely to have reached Stonehenge. So the model needs not only ice delivery, but a rescue from rising water, an intermediate storage site on dry ground, and a second anthropogenic haul centuries later. Each step is an unforced complication.

You still need people for hundreds of kilometres. Even under the assist model, glacial transport alone "cannot account for the final emplacement on Salisbury Plain." The hard problem — moving a six-tonne shaped block a very long way — remains. The glacier only relocates where the human journey is supposed to start, and adds a drowned landscape and a time gap in between.

The simpler hypothesis is still on the table. Neolithic people moved Scottish stone (and cattle, and other heavy things) by sea and river. One intentional long journey from a Caithness neighbourhood beats: rare ice vector + inundation race + Mesolithic salvage + later Neolithic re-transport. Ockham prefers the journey you already have evidence people could organise.

The paper's abstract is more careful than the press: glacial transport may have been an intermediate stage; substantial anthropogenic transport would still have been necessary. Today's coverage sells the ice. That is a choice.

Who is selling the ice

It may not be an accident of journalism. Dr Remy Veness is co-lead author, a glaciologist (Sheffield Hallam), and the modeller behind the ice-flow trajectories. He is the natural press voice for a glacial story. He is a less natural sole arbiter of which Caithness outcrop the zircons prefer — that is a detrital-geochronology argument, already published with Sarclet as closest match, now orally revised toward Braemore.

Having a glaciologist as the public face of the paper puts a predictable slant on which half of the work gets the standfirst. Provenance (zircon neighbourhood in Caithness) is the result that actually tightened. Glacial assist to Doggerland is the speculative half — interesting to model, weak under Ockham, and still incomplete without people. When the same voice updates the source pin and re-promotes the ice narrative, readers should separate the two. One is a ranking of open grain lists. The other is a preferred transport romance.

Braemore will not be the final word

One more thing, from this side of the desk. Even if today’s press line is catching up with the July zircon ranking, Braemore will not be the final word, in my view. The open audit never made it a quarry: it only showed Braemore matching better than Sarclet inside a neighbourhood. Later screening — thermal maturity against the Altar Stone’s expandable clays, facies, and a multi-proxy desk filter — moved the priority off that Lower ORS pin. By mid-July the same exercise had narrowed to a single field-accessible target; I published that conclusion and withheld the place-name pending a final check (Algorithm Result, 14 Jul 2026). So when Veness now sells “within 50 km of Braemore,” he is updating an intermediate ranking, not the end of the search. Braemore is a useful waypoint. It is not, in my opinion, where this settles.

Links

• Telegraph, 12 Sep 2026
• Clarke et al. (2026) JQS: https://doi.org/10.1002/jqs.70080
• Independent Audit, 5 Jul 2026
• One Sample at a Time, 3 Sep 2026
• Daw (2026) RG: https://doi.org/10.13140/RG.2.2.10365.12008

Britain, 5000–2000 BC: no ice, no catastrophe, a familiar climate

An interactive, fully referenced viewer of published palaeoclimate proxies for Britain across the Stonehenge-building millennia is now online: 

Climate of Britain, 5000–2000 BC.

It draws on twelve independent proxy records: peat water tables, cave speleothems, lake chironomids, ice cores, ocean sediment, a pollen-based reconstruction and tree rings, each carrying its own citation, archive link and stated uncertainty. Three findings fall out of it directly.

Stonehenge - 2013 - Tim Daw

The Avon would not reliably have frozen thick enough to drag stones over

One proposed route for moving sarsens to Stonehenge has them dragged across the River Avon on a surface of thick winter ice. The pollen-based reconstruction gives a direct check on this: mean January temperature for Wessex across all four millennial slices in the window comes out at 2.7–4.2 °C — above freezing, and close to the modern Boscombe Down January average of 4.0–4.6 °C. A monthly mean sitting a few degrees above 0 °C does not rule out individual cold spells or short-lived ice, but it argues against the sustained, reliable deep freeze that dragging multi-tonne stones across a river would need. On this proxy evidence, an ice-covered Avon load-bearing enough for sarsen transport looks like the exception rather than something that could be counted on.

Glacial transport of the stones themselves is a separate and much earlier question, belonging to the last glaciation many thousands of years before this window opens; it isn't addressed by this dataset one way or the other.

No coherent climate catastrophe, and no population-ending event

The most-cited candidate for a climate-driven population collapse in this period is the "4.2 ka event," dated to around 2200 BC. Roland et al. (2014) tested this directly against the British and Irish peat record and found no regionally coherent, prolonged shift to wetter or colder conditions at that date — the four records that do show any change disagree with each other in timing, duration and structure. The one genuine regional exception is a centennial cold interval in eastern England dated to roughly 4320–4210 cal BP, identified from lake isotopes at Diss Mere and independently from tree-ring isotopes in the Fenland — real, but local to eastern England, about a century long, and not an island-wide catastrophe. Nothing in the compiled proxies supports a climate event severe or widespread enough to end a population.

The climate was not very different from today's

The pollen-based reconstruction (Mauri et al. 2015) gives four millennial temperature and rainfall estimates for the Wessex grid cell containing Stonehenge, referenced against the Boscombe Down climate normals — the nearest long Met Office record. The central estimates:

Slice	Annual mean °C	July °C	January °C	Annual rainfall mm
5050 BC	9.6	17.4	2.7	~730
4050 BC	10.1	16.7	4.2	~815
3050 BC	9.6	16.7	3.8	~870
2050 BC	9.6	16.5	4.0	~840
Boscombe Down, 1971–2000	9.85	16.75	4.0	736
Boscombe Down, 1991–2020	10.42	17.1	4.6	783

Every one of the four central annual-mean estimates sits within half a degree of the modern 1971–2000 average, and all are a little cooler than the current (1991–2020) average — Wessex has warmed slightly since, rather than the Neolithic and Bronze Age being markedly colder than today. Including the full published uncertainty (±1σ, propagated across the window), the plausible annual-mean band is roughly 8.6–11.1 °C: comfortably straddling both modern baselines. Winters may have run a little cooler in the earliest slice (5050 BC: ~2.7 °C January vs 4.0–4.6 °C today), but by the Beaker period the estimate is close to indistinguishable from now. Rainfall is centrally similar to modestly wetter than today, with an uncertainty band wide enough to include "about the same." None of this supports either an unusually harsh or an unusually benign climate: mid-Holocene Wessex was, within the resolution these proxies allow, a maritime temperate climate much like the one standing at Stonehenge today.

Full series, citations, archive links and uncertainty notes for every proxy are in the interactive viewer.

The Devil Is In the Detail

A new paper sources Yorkshire's Devil's Arrows to Brimham Rocks, 18 km west. The zircon evidence for that is sound, but the stratigraphy behind it needs a correction, and the case against glacial transport is stronger than the paper itself makes it.

---

Clarke, Leary and Kirkland have a new paper in Proc. R. Soc. A (20260504). Peel samples were taken from the three standing stones at Boroughbridge, and roughly a kilogram of rock each from Brimham Rocks and Plumpton Rocks. Detrital zircon U–Pb ages from the Arrows match Brimham (Kolmogorov–Smirnov p = 0.77) and do not match Plumpton (p = 6.7 × 10^–5). Historic England's long-standing attribution to Plumpton is the wrong outcrop.

That result stands. The stratigraphic framing around it does not. The paper describes Brimham and Plumpton as "both Upper Plumpton Grit outcrops... the same Namurian stratigraphic interval." The British Geological Survey's own lexicon says otherwise, and checking it against the 1:50k mapping changes what the zircon match is actually evidence for.

What they actually dated

The Devil's Arrows stand on Triassic Sherwood Sandstone and are not made of it — the paper establishes this cleanly, on both lithology and zircon spectrum, and it isn't in dispute. What they are made of is Kinderscoutian Millstone Grit, deposited around 318 Ma by the Pennine river system. The paper states that its two comparison outcrops, Brimham Rocks and Plumpton Rocks, "are both Upper Plumpton Grit outcrops... the same Namurian stratigraphic interval." Checked against the field literature for each site individually, that statement is wrong, and it's wrong for reasons independent of each other — the two outcrops aren't misidentified by the same error, they're simply two different units.

Brimham. BGS's own excursion guide to the site, describing the rock as sampled in the field rather than as read off a map, states that "the spectacular natural cliffs and tors of Brimham Rocks are carved out of a plateau of Lower Brimham Grit (mid-Namurian)." Soltan and Mountney's 2016 study of the outcrop's channel architecture — a paper Clarke, Leary and Kirkland themselves cite, for basin-scale heterogeneity — captions their own photograph of the site the same way: "Lower Brimham Grit succession exposed at Brimham Rocks." The BGS lexicon records Lower Brimham Grit under the alternative name Lower Plompton Grit (LPG), part of the Hebden Formation, sitting above the Ure Shell Bed and below the Eccup measures. Three independent descriptions, converging on the same unit: Brimham Rocks is the Lower leaf.

Plumpton. The BGS memoir for the Harrogate sheet (62) puts its cover photograph exactly where the dispute sits: Lover's Leap, Plumpton Rocks, captioned as "naturally sculptured crags of Upper Plompton Grit at its type locality." Plumpton Rocks isn't merely similar to the Upper leaf — it's the section BGS uses to define it. Cooper's Knaresborough Gorge excursion notes, and the Denys Smith memorial trip report covering the same ground, both describe the Lower Plompton Grit as exposed separately, a few kilometres away at Knaresborough Gorge, beneath the Upper leaf in the local succession, not at Plumpton Rocks itself.

So the two outcrops the paper compares are the lower and upper leaves of the Plompton Grit, divided by the Eccup Marine Band — not, as stated, the same stratigraphic interval. This isn't a map-reading disagreement over which of two adjacent colours a dot falls in; it's the paper's named comparison localities being described, independently and repeatedly, as different units in the literature it already draws on for other purposes. It also isn't an incidental distinction. The Hebden Formation's lower part is "a turbiditic facies... with laterally impersistent and locally thick, massive, coarse to very coarse-grained sandstones" — channelised, not a blanket sand — while the upper part is "sheet-like, laterally persistent." Individual channel bodies of that kind can carry genuinely distinct detrital zircon signatures, which is the mechanism the paper needs for Brimham and Plumpton to differ in the first place. Correcting the stratigraphy doesn't weaken the zircon result — Arrows vs Brimham, KS p = 0.77; Arrows vs Plumpton, p = 6.7 × 10^–5 — it explains it.

Plumpton Rocks — the outcrop the paper samples and compares against — is consistently described in BGS excursion literature (Cooper and Burgess's 1993 Harrogate memoir, and the published Knaresborough Gorge excursion guides) as the Upper Plompton Grit. So the two sampled outcrops are not "the same Namurian interval": they are the lower and upper leaves of the Plompton Grit, separated by the Eccup Marine Band. The paper's own Kolmogorov–Smirnov numbers already say this — Arrows vs Brimham not significantly different, Arrows vs Plumpton clearly different — the caption just doesn't correctly name why.

The place is Plumpton Rocks, with a U, used throughout the paper and by Historic England. The adjacent park and houses are Plompton as is the geological unit Plompton Grit, with an O, Both the Lower and Upper leaves carry it: LPG and UPG. Reading "Plumpton Rocks" against "Plompton Grit" without noticing the two are spelled differently makes it easy to miss that the lexicon is also carrying two separate codes under that name, not one.

It also isn't an incidental distinction. The BGS lexicon's lithological description of the Hebden Formation splits it into a lower part — "a turbiditic facies... with laterally impersistent and locally thick, massive, coarse to very coarse-grained sandstones" — and an upper part of "sheet-like, laterally persistent" sandstone. The Lower Brimham/Plompton Grit is channelised, not a blanket sand. Individual channel bodies of that kind can carry genuinely distinct detrital zircon signatures, which is the mechanism the paper needs for Brimham and Plumpton to differ in the first place. Correcting the stratigraphy doesn't weaken the zircon result; it explains it.

Click to enlarge

 BGS Geology 50k bedrock + superficial, © UKRI 2026. Frame BNG 415–445 E, 448–500 N km.

The map

The 50k sheet covering Brimham, Boroughbridge and Plumpton (dots are GetFeatureInfo point samples, not filled polygons) shows the Lower Brimham/Plompton Grit at outcrop in one cluster: the Brimham Moor tor field and the ground immediately west of it. The same leaf reappears elsewhere on the sheet, but under cover — beneath Devensian till north of Hack Fall, and beneath Vale of York Formation roughly 12 km west of the Arrows — rather than exposed. The Upper leaf is a separate storey above it.

A closer inset around Plumpton shows both leaves printed in the same yellow on the standard 50k colour scheme, distinguished only by lexicon code: Lover's Leap itself sits on Upper Plompton Grit (UPG), with a strip of Lower Plompton Grit (LPG) — the same unit as Brimham and the Arrows — running immediately to its west, separated by thin intervening measures. That strip was not sampled.

6 × 6 km, BNG 432–438 E, 450.5–456.5 N. Same WMS yellow, two lexicon codes.

What the air photos add

Brimham Moor is an open plateau of exposed stacks — individual joint blocks, already the size of pillars, visible from the path and from the air (Geograph squares SE2064 and SE2065). Plumpton, by contrast, is a lake set in woodland with a single crag along one shore — Lion's Den, Lover's Leap, Needle's Eye — roughly 10–12 m above the water, part natural and part quarried, matching Historic England's listing description. The fields west of the lake, where the 50k map places the Lower Plompton Grit at outcrop, are smooth pasture with no visible scatter of grit blocks.

Place	What the air photo shows	Leaf	Plausible as a stone source
Brimham Moor	Open plateau of stacks	Lower (LOBM/LPG)	Yes — blocks already detached along natural joints
Plumpton lake crag	One wooded cliff, 10–12 m, part quarried	Upper (UPG)	Wrong leaf on the zircon evidence; already excluded
Fields west of the lake	Smooth farmland	Lower (LPG), per the map	Not visible as exposed rock from the air

Brimham Rocks, "Idol Rock" — Penny Mayes, Geograph 1184, CC-BY-SA 2.0.

Against glacial transport

Late Devensian ice in the Vale of York moved NNW to SSE, down-valley — the wrong direction to have carried anything off the Pennine edge onto the Boroughbridge ridge, and no reconstruction maps a westerly conveyor off Brimham Moor at all. The excavated stone sockets — 1.5 m deep, flat-bottomed, and of matching dimensions at all three stones — already show the row was set by hand, not left in place by melting ice. Two further lines point the same way: the Arrows carry deep, weathering-fluted grooves and an elongate, naturally joint-bounded shape matching Brimham's own fracture pattern (fig. 6B), not the rounding expected of a boulder that has travelled inside moving ice; and all three stones share a single, statistically indistinguishable zircon population (table 1), closer to what one quarried outcrop produces than to the mixed assemblage typical of glacial till.

Once human erection is established — and the sockets establish it regardless of source — the question of transport is no longer whether people moved the stones, only how far. A group that can lever a pillar off a joint face and walk it into a socket a hundred metres away has already demonstrated the method; eighteen kilometres from Brimham is a longer version of the same task, not a different one. Glaciation, by contrast, still needs an ice path the reconstructions don't map, and still leaves the fluting, the jointing and the single-source signature unexplained. What the ice-flow reconstructions cannot do, and the paper's own discussion concedes, is resolve or exclude a specific 18 km path between Brimham and Boroughbridge at the resolution the abstract implies — but by that point glaciation has already lost the argument on capability, and an unresolved ice model doesn't reopen it.

What still needs checking

Three things follow directly from the mapping and are not addressed in the paper:

1. The Lower Plompton Grit strip immediately west of Lover's Leap — the same unit as the Arrows on the 50k map, closer to Boroughbridge than Brimham, and neither dated nor visibly exposed as a tor field from the air.
2. Whatever Carboniferous sandstone actually lies at or near surface roughly 12 km west of the monument, under the Vale of York Formation.
3. A second storey at Brimham itself — Soltan and Mountney (2016) already log two distinct channel systems within the tor field, which the single 1 kg sample cannot distinguish between.

Until those are checked, Brimham is the best-supported source among the outcrops that have actually been dated, not a uniquely identified quarry. The zircon evidence is good enough to rule out Plumpton and rule out the local Sherwood Sandstone; it is not yet enough, on its own, to rule out the other exposures of the same facies belt that lie closer to the monument and were never sampled.

Addendum: what the supplementary data show

The zircon U–Pb data released with Clarke, Leary & Kirkland (2026) as Dataset 1 (rspa.2026.0504) allow the provenance statistics to be rebuilt from the raw analyses. The Brimham attribution survives the exercise. Several aspects of how it was reached do not.

1. The grain counts are as published

The monument dataset comprises 67 dated zircons from Stone 1, 64 from Stone 2 and 105 from Stone 3, totalling 236. Brimham Rocks yields 75 concordant analyses and Plumpton Rocks 76. The subsets of 29 and 46 Brimham grains plotted in Figure 4 are display panels (<510 Ma and >510 Ma). They are not the samples tested. All KS and Kuiper comparisons in Table 2 use the full datasets. The Table 1 pairwise results between the stones reproduce exactly (Stone 1 vs Stone 2, KS p = 0.138; Stone 2 vs Stone 3, p = 0.834).

2. The Brimham statistic does not reproduce

Table 2 reports D = 0.079, p = 0.77 for the Arrows against Brimham. Recomputed from Dataset 1, D ranges from 0.10 to 0.12 and p from 0.39 to 0.56, depending on which age (preferred, concordia or conventional) is used. The published pair is also internally inconsistent: for n = 236 and 75, D = 0.079 corresponds to p ≈ 0.87, not 0.77. None of this alters the inference, because no difference is detectable at any of these values. The Plumpton comparison reproduces closely (D = 0.29, p ≈ 10⁻⁴ to 10⁻⁵).

3. Monument and outcrop data were filtered differently

The paper refers readers to Dataset 1 for its filtering criteria. The Methods sheet of that file is empty. The data themselves show that the outcrop analyses were screened at ≤10% discordance: exactly 75 Brimham and 76 Plumpton grains pass that cut. The monument analyses were not screened the same way. Thirty-six accepted monument grains exceed 10% discordance, and 38 rejected grains fall below it.

Four accepted analyses, all from Session 2, are extremely discordant, and their preferred ages match none of the ages calculated in the same row:

• Stone 1, spot 17: 174% discordant, preferred age 2868 Ma
• Stone 3, spot 15: 2860% discordant, preferred age 2723 Ma
• Stone 3, spot 49: 83.5% discordant, preferred age 448 Ma
• Stone 3, spot 124: 116% discordant, preferred age 423 Ma

Applying the same ≤10% filter to every sample leaves 232 monument grains and changes little: Brimham p = 0.26, Plumpton p = 2 × 10⁻⁵. The conclusion is robust. The data handling is not, and it warrants a correction to the supplementary file.

4. Stone 1 is not quite like its neighbours

Sample	n	380–520 Ma	900–1300 Ma	1300–1900 Ma	>2500 Ma
Stone 1	67	40%	3%	25%	27%
Stone 2	64	38%	20%	23%	17%
Stone 3	105	36%	15%	27%	16%
Brimham	75	39%	12%	31%	16%
Plumpton	76	63%	8%	9%	17%

Stone 1 carries more Archaean zircon and almost no Grenvillian grains. With 64–67 grains per stone, the KS test cannot separate it from Stones 2 and 3 (p = 0.14–0.26). Tested individually, each stone is indistinguishable from Brimham (p = 0.55, 0.43 and 0.66) and distinct from Plumpton (p < 0.001). The pattern fits the stones coming from different beds or channel bodies within a single grit unit, which is the third of the open questions above. It does not require separate sources.

5. The Brimham sample has no locality

The reference sample is described only as "approximately 1 kg" collected from the outcrop. Neither the paper nor the supplement gives a grid reference, bed or stratigraphic height. Supplementary Figure S2 does not show the sampled rock. It shows the scheduled cup-and-ring panel west of Brimham (SM 1014974), which is cited as evidence of prehistoric activity nearby. Given the channel-scale heterogeneity the authors themselves invoke, one undocumented sample from a tor field cannot fix the source to that tor field.

6. Binchester Crags was never tested

The paper states that the least-dissimilar published dataset, with an MDS uncertainty ellipse overlapping the Arrows, is the Langsettian Lower Coal Measures sandstone at Binchester Crags, about 70 km to the north. Table 2 gives no KS or Kuiper result for it. The paper reads the match as a shared Pennine Basin signature, but a northern source is the direction that the cited Vale of York ice-flow reconstructions favour. The glacial alternative is dismissed for Brimham and never evaluated for Binchester. The coarse, pebbly Namurian lithology of the Arrows argues against a Coal Measures source, but that argument has to be made, and the paper does not make it.

Summary

Brimham remains the best-supported of the dated outcrops, and Plumpton is firmly excluded. The supplementary data add three qualifications: an inconsistently applied concordance filter, a published Brimham statistic that cannot be reproduced, and heterogeneity between the stones. The data also underline the two gaps already noted: the reference sample has no location, and the nearest literature match was not tested.

References

1. Clarke, A.J., Leary, J. & Kirkland, C.L. 2026. Deliberate prehistoric sourcing of the Devil's Arrows, Britain's tallest stone row. Proc. R. Soc. A 482, 20260504. doi.org/10.1098/rspa.2026.0504
2. BGS Lexicon of Named Rock Units: Hebden Formation (HEBD); Millstone Grit Group (MG).
3. BGS Lexicon: Lower Brimham Grit (LOBM) / Lower Plompton Grit (LPG).
4. BGS Earthwise. Carboniferous rocks of upper Nidderdale — an excursion, Locality 1, Brimham [SE 212 670].
5. Cooper, A.H. 2008. Permian and Carboniferous, Knaresborough — excursion notes. Denys Smith Memorial Trip, BGS Open Report OR/08/044.
6. Soltan, R. & Mountney, N.P. 2016. Interpreting complex fluvial channel and barform architecture: Carboniferous Central Pennine Province, northern England. Sedimentology 63, 207–252.
7. Cooper, A.H. & Burgess, I.C. 1993. Geology of the country around Harrogate. Memoir, sheet 62.
8. Wilson, A.A. & Thompson, A.T. 1965. The Carboniferous succession in the Kirkby Malzeard area. Proc. Yorkshire Geol. Soc. 35, 203–227.
9. Bateman, M.D. et al. 2015. Last glacial dynamics of the Vale of York and North Sea lobes.
10. Clark, C.D. et al. 2022. The BRITICE-CHRONO reconstruction. Boreas.
11. Historic England NHLE 1014705, Devil's Arrows.
12. Historic England NHLE 1000535, Plumpton Rocks.

Contains British Geological Survey materials © UKRI 2026. Geograph photographs CC-BY-SA as credited on source pages.