Three Cavities Within Six Kilometres. The Gypsum Dissolved And The Roof Held.
Two processing houses, working blind of each other, reprocessed four 1974 gas-industry seismic lines across Archuleta Mesa. Both image three voids within six kilometres of the east face. The largest is 240 metres across and 70 high.
Background written by the news staff. Contains no argument or recommendation.
DULCE, N.M. — The largest of the three is 240 metres across and about 70 metres high, and the rock over it is a sandstone bed 76 metres thick that will stand across 260 metres without support.
There is no shortage of ways to argue about Dulce. Almost all of them end in the same place: not in what anyone saw, but in whether the ground could hold a room that size at that depth at all.
For fifty years the answer has been given as a matter of common sense rather than measurement. This newspaper decided to measure it.
What we bought
In 1974 a gas contractor shot four vibroseis lines across the mesa and the benches east of it, looking for the Dakota. The tapes have sat in a commercial data library ever since. We bought all four for $18,600, with the observers' logs.
The 1974 crew was hunting gas at 2,000 metres. Everything shallower than about 1,200 was noise to them, and the processing of the day muted it out. That mute is the reason nobody has looked. The shallow section was never lost. It was discarded.
Two houses, neither knowing
We stripped every header off the tapes, rebased the survey into a coordinate frame that puts the mesa in the wrong hemisphere, and sent identical copies to Kestrel Datum in Calgary and Ordinal Geoscience in Houston. Each was briefed to image 0–1,200 metres and nothing else. Neither was told the other existed.
Both came back with three anomalies inside six kilometres of the mesa's east face.
On two of them the houses agree on position to within 40 metres. On the third — cavity 2, the smallest — they are 115 metres apart, and we print both picks rather than splitting the difference.
The mechanism is boring, which is the point
Under the mesa, between about 690 and 900 metres below the rim, sits a Pennsylvanian evaporite sequence: gypsum and anhydrite, laid down in a drying sea. Gypsum dissolves in circulating groundwater roughly a hundred thousand times faster than limestone does.
What it needs is a route for the water. The Dulce dike swarm provides one — a set of intrusions that fractured this ground thoroughly on their way up, and left a plumbing system behind them.
Above the evaporite, the Dakota Sandstone. Competent, thick, well jointed, and the reason the roof does not come down when the salt goes.
That is interstratal karst. It is on the second-year syllabus. It has made cavities of this size under Sorbas, under Ripon, and under a good deal of quiet farmland in Kansas.
The reading
Three geologists were given both stacks with no site name, no story, and no knowledge of each other. Dr. Cressida Ubeku-Feld wrote "classic interstratal karst, and a good one" on her first pass, before she was told what mesa she was looking at.
All three picked the same roof. All three put the base of the largest cavity between 750 and 766 metres below the rim.
In 1978 a stock well on the east bench lost its entire mud column at 754 metres and was abandoned. The driller's log is in the state engineer's files and has been since March 1979.
What the rock says, and where it stops
A room of the size described in PX-1930 is not exotic here. It is what this stratigraphy does on its own, at that depth, on the ordinary assumptions, with the ordinary numbers.
That is the whole finding, and it is worth being exact about its shape. The rock permits the chamber. The rock does not put anyone in it.
Professor Rustad-Nnoli's reply below takes our headline apart for exactly this, and she is right that a cavity is what water leaves and a chamber is what somebody makes. We have left both words standing where they were, with her objection under them, because quietly swapping one out would be the cheaper thing to do.
The next measurement is not a seismic line. It is one 200-millimetre hole and a camera on a cable, and we have been refused the ground for it twice.
The six o'clock edition
Three stories a morning, in your inbox before the coffee is made.
Sources & Method
We wanted rock, not testimony. We bought four 1974 gas-exploration lines out of a commercial library, stripped every header, rebased the geometry into a false coordinate frame and sent identical copies to two reprocessing houses who did not know the other existed. Both were briefed to image 0–1,200 metres and nothing else. The two stacks then went to three geologists who were told nothing about Dulce, about this newspaper's interest, or about each other.
Who we spoke to
- Kestrel Datum Ltd., Calgary, Seismic reprocessing house; 31 staff. Contracted by this newspaper to reprocess four field tapes for shallow-to-intermediate structure, with all headers stripped and the survey rebased into a false coordinate frame February–May 2026 Was not told where the lines were shot and did not ask until the deliverable was signed off.
- Ordinal Geoscience Inc., Houston, Second reprocessing house, working blind of the first. Given identical tapes and logs, the same false frame, the same one-line brief, and no knowledge that a second contract existed February–May 2026 Puts cavity 2 some 115 m north-east of where Kestrel puts it. Both deliverables are published with this story.
- Dr. Cressida Ubeku-Feld, Structural geologist, Bochum; works on evaporite dissolution. One of three readers given both stacks with no site name, no story and no idea what the other two had been sent June 2026 Wrote 'classic interstratal karst, and a good one' on her first pass, before she knew what mesa it was.
- Prof. Isolde Rustad-Nnoli, Professor of rock mechanics, Trondheim. Commissioned to run the stand-up span independently, then sent the full draft; replied in writing after sixteen days June–August 2026 Supplies the 260-metre figure and then objects, at length, to what we do with it. Printed entire under Right of Reply.
- Four 1974 field tapes and the observers' logs, Lines 74-A, 74-C, 74-D and 74-F, San Juan Basin. Bought from a commercial data library for $18,600, with a licence to publish the reprocessed sections but not the raw shot gathers January 2026
Documents
- PX-1930 — The Bennewitz material — 340 pages given to a private citizen by persons unknown, 1979–82 contested
What we could not confirm
- Whether anything is inside them. Seismic images the contrast between rock and not-rock; it cannot distinguish a flooded void from an air-filled one, let alone an occupied one. The next door is a borehole camera, and it needs one 200-millimetre hole. The mineral-rights holder on the east bench has refused us access twice, in writing, without giving a reason.
- Whether the three connect. Our line spacing is 800 metres. A drift 40 metres wide running between cavity 2 and cavity 3 would sit entirely inside our blind ground and neither house would see it. We are not saying there is no connection. We are saying we bought the wrong survey to answer that.
- How many there are. The 1974 programme does not cross the mesa's west face — the permit was refused then, and our own application to shoot four new lines there was refused in April, by the same office, in eleven words. Three is a floor. It is not a total.
How Others Covered This
The same events, as reported elsewhere on the same day. We list what each outlet had that we did not, as well as what we had that they did not — including where we come off worse. Why we print this.
-
The Meridian TelegraphThe Cavern Under Archuleta Is Real, And It Is The Size Of A Cathedral
Took our largest cavity and furnished it — floors, lifts, a rail head — none of which appears in either deliverable.
Had that we did not
The 240 metres, correctly, and the fact that two houses worked blind.
Left out
That a seismic line images an impedance contrast, not a room, and that the top 300 metres of the mesa are muted out of both stacks.
-
ApertureThe Mesa That Wants To Be Believed
Read the whole exercise as an essay about wanting, with the correspondent as its subject.
Had that we did not
The clearest lay account of what a vibroseis sweep actually is that we have read anywhere, ours included.
Left out
The arithmetic. The stand-up span, which is the entire load-bearing number in this story, appears nowhere in four thousand words.
-
They Buriedthis newspaperThree Cavities Within Six Kilometres. The Gypsum Dissolved And The Roof Held.
Bought four old gas lines, paid two houses to reprocess them blind, and had three geologists read the results cold.
Had that we did not
Both deliverables entire, the disagreement on cavity 2's position, and the two depth bands where we can see nothing at all.
Left out
Our headline says the roof held. It has held for eleven thousand years without anyone putting a load on it, and 'held' is doing work our own rock-mechanics source expressly declines to do. Her objection runs at the foot of this page. It belongs in the first line. — V. Ashcombe-Doyle, standards editor
Right of Reply
They Buried contacted Prof. Isolde Rustad-Nnoli, rock mechanics, Trondheim on 24 July 2026, in writing, with the full draft, both deliverables and her own span calculation attached. Replied 9 August 2026. Printed in full and unedited, including the parts about us.
I gave you a number and you have built a sentence on it that the number cannot hold.
First, what I will sign for. The voids are there. I read those stacks before I knew whose mesa it was, and I would have said the same thing to a gas company that had drilled a dry hole: that is interstratal dissolution in an evaporite sequence under a competent clastic roof, it is the least surprising thing in the section, and anyone who tells you a 240-metre span at 700 metres is geologically impossible has confused impossible with unfamiliar. The Sorbas basin has bigger. So does Ripon. So, quietly and expensively, does the ground under a good deal of Kansas.
Now the objection. The 260 metres I calculated is a stand-up span: the width at which a beam of that lithology, that thickness and that joint spacing will remain self-supporting under gravity for a geological interval, given no seismic loading, no excavation, no drawdown of the water table and no idiot with a drill. It is a statement about a beam. You have turned it into a statement about a room.
A cavity is not an excavation. Nothing in my arithmetic tells you the void formed with that roof intact, and the collapse breccia banked up the walls in both stacks tells me plainly that at some point it did not. What I can say is that the roof is intact now and was not lifted or bolted by anybody. That is a smaller claim than 'the roof held' and it is the only one I will make.
On your word 'chamber'. I notice you use 'cavity' in the diagram, where I checked it, and 'chamber' in the headline, where I did not. Those are not synonyms and you know they are not. A cavity is what water leaves behind. A chamber is what somebody makes. Every reader who takes the second word from your front page and the first word from your figure will come away believing this newspaper established something it did not.
And on the drill. You lean on a stock well that lost its mud column at 754 metres in 1978 as though it were a probe. It is a driller's log. It says a man lost circulation and gave up, and there are nine ways to lose circulation in karst, eight of which are boring. It happens to sit within twelve metres of a depth two independent contractors picked from a different physical measurement, and I accept that this is a real and quite striking coincidence. It is still a coincidence until someone puts a camera down a hole.
Get the hole. Everything else is conversation.
Published unedited under our right-of-reply guarantee.
How was this story?
We publish the result, whatever it is. Reader verdicts appear on the front page and in our newsroom metrics.
Readers' Letters 0
Printed at once under the name you give and read by the desk afterwards; anything unfit is removed, with a note saying so, and nothing else is ever deleted — only corrected. Letters that changed something in the story carry a mark saying so, and there are 5 of those across the archive.