Mapping a Geochemical Fingerprint: Using Open BGS Data to Screen for the Stonehenge Altar Stone's Source

A note on how this post was made: the code, the geospatial analysis, the clustering, and every map below were written and run by Claude, Anthropic's AI model, working directly with me in conversation. I gave the direction and asked the questions; Claude did the data engineering, the coordinate maths, and the geological cross-checking. I'm saying that plainly up front rather than writing "I built a map" and quietly letting an AI's work pass as mine.

The Altar Stone — the six-tonne sandstone megalith lying at the heart of Stonehenge — was shown in 2024 to come not from Wales, as a century of scholarship assumed, but from the Orcadian Basin of northeast Scotland. That single result reopened an old question in a new place: where, within a basin that stretches from the Moray Firth to Shetland, did this stone actually come from?

The published case against Orkney itself rests on careful, multi-technique laboratory work — portable XRF, automated SEM-EDS mineralogy, clay XRD, Raman spectroscopy — applied to a small number of hand samples. What it can't easily do is tell you, at a glance, how that local result sits against the regional geochemical backdrop. That's a job for a map, and it's the job I set Claude to do.

Round one: barium

The British Geological Survey's Geochemical Baseline Survey of the Environment (G-BASE) publishes free, OGL-licensed national 500 m kriged grids for 18 elements, derived from roughly 110,000 stream-sediment samples across Great Britain. I handed Claude the barium grid as a zipped ESRI ASCII raster (Ba_grid.zip).

Claude parsed the six-line header, loaded both the concentration and uncertainty grids into NumPy, then used pyproj to convert every site I cared about — Yesnaby, the Ring of Brodgar, Westray, Caithness, Shetland — from latitude/longitude into the grid's native OSGB36 British National Grid coordinates, so it could index straight into the array:

col = (easting  - XLLCORNER) / CELLSIZE
row = NROWS - 1 - (northing - YLLCORNER) / CELLSIZE

Rather than just colouring the map by concentration, Claude built it around the Altar Stone's own published numbers: 105 of 106 in-situ pXRF analyses of the stone exceed 1025 ppm barium, with a mean above 2750 ppm. Every grid cell clearing the first threshold was rendered in flat green; cells clearing the mean, in magenta — so a real match jumps out as a solid colour rather than a slightly warmer shade.

Counted up, the regional contrast was stark: Orkney returned a 0.03% hit rate against the Altar Stone's barium floor (one cell, at Yesnaby), against 10.8% for the mainland Caithness/Moray/Black Isle area and 2.8% for Shetland — roughly a 350x difference between Orkney and the mainland basin.

Claude was equally clear about the catch: the one Orkney hit, at Yesnaby, is explained in the literature as vein-infill baryte tied to base-metal mineralisation, not the diagenetic cement that's actually in the Altar Stone. Stream sediment records whatever eroded out of an entire catchment — it can't distinguish the two. Claude also flagged that the smallest islands (Westray, Eday, the Hoy Sandstone outcrop) returned no data at all at their exact coordinates, so their apparently unremarkable background readings are weak evidence, not an exclusion.

Round two: adding rubidium

I asked Claude what other published characteristic of the Altar Stone could be screened the same way. Its answer: rubidium, because it's the other half of a discriminant Bevins and colleagues actually plot. In their Anglo-Welsh Basin comparison, the high-barium look-alike samples that failed to match the Altar Stone all had roughly 3x higher Rb; the one sample that did track the Altar Stone on barium also matched it on Rb. Ba-high paired with Rb-low is a much closer approximation of the real test than barium alone.

Rb is in the same free G-BASE grid list, so I downloaded it and handed it over. Claude loaded it, confirmed the grid geometry matched the barium file cell-for-cell, and was upfront about a limitation before building anything: none of the published papers give an absolute Rb concentration for the Altar Stone itself, only the relative "3x higher" comparison. So instead of an invented absolute number, Claude used the basin's own lowest quartile — 76 ppm — as a data-driven stand-in for "low Rb," and said so rather than presenting it as a published figure.

Claude then ran the obvious sanity check before trusting the new filter: did Yesnaby and the other known false positive — a barium anomaly near Tongue in North Sutherland, sitting on basement schist rather than any sandstone — actually carry high Rb, the way the Anglo-Welsh look-alikes did? They didn't. Both came back at 74–77 ppm, comfortably inside the "low" band. Claude's conclusion was direct: vein baryte is close to pure barium sulphate with effectively no rubidium-bearing mineral attached, so it dilutes Rb locally regardless of whether the baryte sits in a vein or a sedimentary cement. Rubidium doesn't separate the two. That's a real negative result, and Claude reported it as one rather than quietly dropping the rubidium idea.

What the combined Ba+Rb map mostly turned up instead was a large cluster in interior Sutherland — over 190 km² across more than a dozen patches — sitting well outside any mapped Old Red Sandstone, in Moine basement country. Claude's read: almost certainly an unrelated base-metal/baryte vein province, or possibly a peat artefact (blanket bog is known to retain barium while leaching rubidium). It's the single biggest feature on the map and the least relevant one — which Claude called out as the real lesson of the exercise: a two-element open-data filter can manufacture an impressive-looking anomaly just as easily as a one-element one.

Just the spots

To cut through that, Claude clustered every patch that passed both the barium and rubidium thresholds and checked each cluster's coordinates against the known Old Red Sandstone footprint.

Clusters	Area	On mapped ORS?
14 of 19	>190 km² total	No — Sutherland interior, basement/granite country. Discard.
1	23.75 km² (mean Ba 1303, mean Rb 69)	Yes — Melby/Walls, West Shetland
4 small clusters	~7.75 km² total (mean Ba 1046–1231, mean Rb 73–75)	Yes — just southeast of Inverness, around Nairn/Culloden (not previously flagged in this search)

One genuinely notable negative: Caithness and Orkney produced zero composite hits. Both have patches of elevated barium on their own, but none of those patches also fall into the lowest rubidium quartile. Claude was careful to phrase this as inconclusive rather than disqualifying — it doesn't rule Caithness out, it just means this particular two-element test doesn't add a tick mark there the way it does for Shetland. The cleaner published discriminant, strontium, isn't in the free national grid, so that test is still out of reach without licensed point data.

What two rounds of this actually add up to

• Shetland's Melby/Walls area has now passed two independent screens rather than one, which is a real if modest strengthening.
• The Inverness/Nairn corridor is a new lead this exercise surfaced that hadn't come up in the published literature search.
• Caithness is unresolved by rubidium, not contradicted by it.
• The most visually dramatic result of the entire two-element analysis — the Sutherland cluster — is the one Claude flagged as almost certainly irrelevant. That's worth sitting with: the biggest-looking signal in open geochemical data is not reliably the most meaningful one.

None of this replaces the actual rock. It's a free, fast, repeatable way to rank where someone should point a portable XRF analyser next — and, just as usefully, where not to bother. Claude did the engineering and the geological sense-checking behind every number in this post; I did the asking.

---

All data used is BGS G-BASE barium- and rubidium-in-stream-sediment grids, © BGS/NERC, used under the Open Government Licence. Analysis, code, and maps were produced by Claude (Anthropic). This work is independent and not affiliated with or endorsed by BGS or any of the cited research groups.