Jeffrey LewisGeolocating the Kalahari Test Site

I made an appearance at the end of Glenn Kessler’s fact check on Mark Kirk’s bizarre claim that Nelson Mandela abandoned South Africa’s nuclear weapons program — something we’ve been scratching our heads over for a while.  While I am officially against handing out Pinocchios, maybe this is the kick in the pants Kirk needs to lose the lame slide.

The South African case is really interesting — not just for the precedent, but also for the role of satellite imagery.

The discovery of South Africa’s Kalahari nuclear test site is one of my favorite case studies. The story is that Cosmos 922, a Soviet photo-reconnaissance satellite, photographed the test site on 3-4 July 1977.  The Soviets didn’t like what they saw, then took a second look with Cosmos 932 and concluded South Africa was preparing a nuclear weapons test.
That’s one version. Dieter Gerhardt, a South African military officer later arrested for spying for the Soviets, told Ronen Bergman that he was the source of the intelligence.

Whatever put Moscow on to Pretoria’s tail, the Soviet Embassy delivered, on August 6, 1977 , a letter from Soviet leader Leonid Brezhnev indicating that South Africa was preparing for a nuclear test, something that would “have the most serious and far-reaching aftermaths for international peace and security.”  Carter wrote back, asking the Soviets for the geographic coordinates.

The United States looked at the site, concluded it was a nuclear test site, and confronted the South Africans with the coordinates and other details.  South Africa’s bomb program was blown.  The scrutiny didn’t stop the program, but the events of 1977 are a good illustration of detection, pressure and so on.  And, in principle, the events of 1977 are now replicable using open source tools.  That’s a big reason that I have always wanted to geolocate the site myself.

I finally got around to it, only to discover that David Albright Paul Brannan,  Zachary Laporte, Katherine Tajer, and Christina Walrond had already done it in 2011.  As it turns out, though, we  used completely different methods.  I used a pair of declassified US documents; Albright et al had used information from the IAEA. We got the same answer, which is nice.  I also learned a few things, some of which may make for an interesting blog post.  You tell me.


Declassified Documents

Thanks to mandatory declassification –and the excellent work of the National Security Archive’s Bill Burr and Jeffrey Richelson — there are a bunch of declassified documents relating to Kalahari Test Site.  I used two in particular to geolocate it.

* Lawrence Livermore National Laboratory, Special Projects Division, South Africa: Motivations and Capabilities for Nuclear Proliferation, September, 1977.

* Lawrence Livermore National Laboratory, Special Projects Division, Geology and History of the Kalahari Drill Site, Proliferation Group Quarterly Report, January – March 1978  (June 1978)

The first document was prepared in rush following the Soviet allegations about the site.  It suffers for the haste in which it was prepared. The second report, which makes use of open source information ,is more limited in scope, but very interesting.  It also includes some satellite images.  These images, or at least their reproduction, is terrible (see the top of the post).  But it does contain some very important details spelled out in the text: the name and relatively location of two “nearby wells.” One is “Niete Min, 7.3 km N10°W from the drill site” and another is “Hop Hop, 13.3 km N25°W from the drill site.”

Both locations appear in the database of NGIA place names, which makes for short work.  The location, distance and heading of both sites agree reasonably well the sites identified by Albright et al as Shaft 2, which was the object of interest in the 1978 report.  Shaft 2 is located in the vicinity of  27°45’32″S, 21°27’43″E

The 1977 report describes other aspects of the site,  which allows us to piece together the whole place:

  • A so-called “head frame” over what was Shaft 1. (27°45’56″S,  21°28’6″E)  This is approximately 1 km from the drilling underway at what would become Shaft 2. The “head frame” was later replaced in the 1980s by a building referred to as a “shade” pictured right. The “shade” is now long gone.
  • A concrete pad that was considered as a possible “tower site”  for an atmospheric test. This is located 2.8 km SSE of the probable head-frame, as described in the 1977 document. ( 27°47’18″S,  21°28’52″E)
  • The support camp ( 27°50’25″S,  21°35’52″E) and the airfield (27°50’5″S 21°37’50″E). The description of the support area (72 m high antenna and about 720 m2 of housing) and the length of the airfield (1600 m) both match, of course.

The nomenclature in the reports can be a little confusing at times, but after a while it all makes sense. There are some errors in the 1977 report.  But fortunately, we have a post-apartheid site visit from the IAEA to clear things up.



Albright et al, on the other hand, had a map drawn by the IAEA, along with a video released by the South Africans. And before you think that made it easy on them, the map is terrible.  There is no compass to warn you that North is inexplicably off to one side, the placement of the roads is impressionistic at best and the author seems to have transposed the locations of the first and second shaft. Other than that, it is a model of verisimilitude!  One benefit of my having found the site through a different method is that it supports Albright et al‘s various suspicions about inaccuracies in the map.

The video wasn’t much help either.  Lots of blurry shots of flat, featureless terrain. Yay. There is still some value to the ground truth images, but it is pretty limited.

What the video is good for, though, is dealing with a special problem.  Shaft 2 apparently never received a head frame, shade or other identifying structure — possibly because of the post-1977 pressure.  The South Africans just buried it.  That meant, when it came time to disable it, the IAEA and South Africans had to find it again.  Albright et al note that there is an area with ground scarring that seems like the probable location of Shaft 2.  The video is useful in confirming that, yep, there aren’t any obvious signatures that can allow for better geolocation.  The spot Albright et al picked is also consistent with the location described in declassified documents, when there were drill rigs, equipment and other structures on site.

We also, thanks to the IAEA, have a location for the second shaft relative to the first — “780 meters from the first one and 80 meters from the road.”  That is also consistent with the earlier US claim the two sites were about 1 km apart.  Without more identifying information our location for Shaft 2 is an approximate one, but it is accurate even if it is approximate.



The fun of doing an exercise like this is that you learn a lot about methods, as well the program itself.

The 1977 US intelligence estimate, prepared in the rush following the Soviet revelation, made a pretty embarrassing mistake.  The author (or authors) got the depth of the test shafts wrong — estimating the depth of the first shaft at 150 m based on the size of the head frame and the depth of the second shaft at a mere 77.5 meters based on the number of casing segments visible at the site.  This led to speculation that any South African nuclear test of respectable yield would crater due the shallow depth of the shafts.

This is one thing I’ve learned.  If you have a conclusion based on a single signature — like the number of casings — and that conclusion further requires that you assume a foreign nuclear program is staffed by utter morons, you may have misidentified the pertinent moron.

The 1977 estimate was totally wrong.  The actual depths of the shafts were 385 and 216 meters. The South Africans were not morons.  The IAEA actually measured them on site.

And lest you think I am beating up on the intelligence community, the author of the 1978 report did much better — and did it with open source data about the local geology!  The author(s) of the 1977 report missed something very important — the site was sand only down to about 75 meters, after which it was granite. Someone looked it up! (See: PJ Smit, “The Karoo system in the Kalahari of the northern Cape Province” Ann. Geol. Surv. S. Afr, 1972.  Only $33!)

In other words, the casings were just for the first 75 m or so of sand.  After that, the shaft kept going, drilled down into the granite.  The 1978 report looked at the number of drilling days and got a very wide estimate (190-345 m) of shaft depth.  Then, the 1978 report adds the  site had ordered six cutters for drilling granite — and returned one unused in December 1977. Each drill set can drill 22.9-30.5 meters. 75 + (5 x 30.5) = maximum shaft depth.

The author estimated the depth of the second shaft at between 169-227 m.  The measured depth was 216 m. That’s pretty damned good work, if you ask me.

There is a second mystery that I think we can clear up.  What is this thing?

A few people have noticed that it looks like the “shade” over Shaft 1, but doesn’t quite match.   I haven’t been able to track down the provenance of the image either, although Carey Sublette may know. (Update | See below.) Notice it has only two doors, not three.  And it is rectangular, not square.  It can’t be the shade the over Shaft 1. My initial I think it was a “shade” located on the concrete pad at  27°47’17″S,  21°28’52″E, about 2 km south of Shaft 1. There is an outline on that pad that matches the shape of the building.  When oriented properly, the roads and vegetation also match. I would say with moderate confidence that, like the other shade, it was probably built in the 1980s, then torn down and replaced with the structure now visible.

I am mildly shocked that the South Africans didn’t do a better job of camouflaging the site.  The author of the 1977 report was shocked, too. “One of the unusual aspects of the Kalahari site is that no attempt has been made to prevent the site’s discovery by airborne or satellite reconnaissance.”  It is one thing to read it; another thing to see it.  Some of the signatures visible in 1977 are no longer apparent, but others are.  Why did they do that?

The author of the 1977 report speculates that the South Africans did not care about international scrutiny, but their subsequent actions suggest otherwise.  The South Africans seemed surprised they got caught.  And, in the 1980s, the constructed the “shade” over Shaft 1 to hide future test preparations from prying spy satellites or red Commodore’s with cameras.

It is an enigma.  The South Africans apparently didn’t even bother to close the airspace over the site.  In October 1977, the US military attache cabled back that the airspace was open over the site, which tended to suggest it was not for nuclear testing.  The US intelligence community apparently took full advantage of this lapse.  In April 1979, the South Africans discovered that the US had outfitted an embassy aircraft with a secret camera to photograph sensitive sites while it ferried US diplomats around the country.  They expelled three U.S. military attaches from the country.  One U.S. official laughed off the expulsion as Pretoria “catching a couple of attaches doing what they are paid to do.”  (See: “Spy in the Sky Flap,” Newsweek, April 23, 1979.)  But the South Africans weren’t smiling.

That’s a useful lesson.  In theory, countries can do all sorts of things to defeat analysts, whether of the open source sort or the more traditional kind.  But sometimes they don’t.  It might be too expensive or cumbersome.  Or maybe they just didn’t think about it.  Or maybe they did and decided that the obfuscation might be detected more easily than the wrongdoing.  Whatever the reason, the 1977 detection of the South African nuclear test site in the Kalahari is a really interesting case.

Update | On the provenance of the black-and-white image of a shade at the Kalahari Test Site.  Carey Sublette email to say that the image was probably from Burrows and Windrem, Critical Mass.  Sure enough, the image is there and credited to The Johannesburg Star. I have a hunch that the photograph was taken during a visit to the site, which appeared as Mark Stansfield, “Terrain of Destruction,” The Sunday Star, March 28, 1993, page 27.  The full text is available via FBIS, but no pictures sadly.  Life sucked before the internet!  Based on the story told by the reporter, it would seem that the site personnel took the reporter to the wrong building.


  1. ChrisM (History)

    “although Carey Sublette may know.”

    If Carey Sublette doesn’t know it, it isn’t worth knowing.

  2. Derick S (History)

    A fascinating story.Two minor points: In the third paragraph, this sentence seems to be incomplete:
    “Whatever put Moscow on to Pretoria’s tail, the Soviet Embassy delivered, on August 6, 1977 , a letter from Soviet premier Leonid Brezhnev indicating that South Africa was preparing for a nuclear test, something that would be .”
    Also, Brezhnev never held the position of Soviet premier. In August 1977 he serving as both the party general secretary and as president of the USSR.

    • Jeffrey (History)

      Yeah, I never finished that sentence and the left the placeholder title in. Sigh. Fixing it now.

  3. Janet M. Simons (History)

    “Whatever put Moscow on to Pretoria’s tail, the Soviet Embassy delivered, on August 6, 1977 , a letter from Soviet premier Leonid Brezhnev indicating that South Africa was preparing for a nuclear test, something that would be .”

    Seems that there may be a word missing at the end of this sentence. Bad? Unacceptable? Catastrophic? Inquiring minds want to know …