Chapter 2: Developing a Monitoring Strategy

Challenging Past Assumptions

Confidence Levels

Monitoring thresholds are based conventionally on 90% probability of detection at four or more stations. From a deterrence perspective it is overly cautious to assume that a country would secretly test if it had an 89% chance of being caught. More likely, a potential proliferator would want near certainty that they would not be caught.17 Monitoring threshold should be presented as a plot of threshold versus confidence level, so when the question is asked "How low can we go?", it can be answered in the context of "How sure do you want to be?" As we shall see in chapter four, the threshold for monitoring differs significantly depending on the level of confidence assumed to be required.

Yield Estimation

When we evaluate the importance of monitoring levels, we assume that a country would be able to test weapons with yields up to that threshold. For example, the United States and the former Soviet Union tested right up to the 150 kiloton limit of the 1974 Threshold Test Ban Treaty (TTBT). The military significance of testing slightly above the threshold limit therefore had to be weighed in the evaluation of the treaty. Efforts to improve monitoring were focussed on reducing the uncertainty with which we were able to estimate the yield of an explosion from the size of the seismic signal it produced. In monitoring a CTBT in the context of proliferation, however, an emerging nuclear state will not be able to test "right up to" the monitoring threshold - not only because of the high probability of being caught, but also because such a state probably would not have high confidence in predicting the exact yield of their newly developed nuclear device. In fact, yield determination might be the basic motivation for testing.

If a proliferator planned to test at two kilotons for example, they would have to assume that the yield might be as high as four kilotons or as low as zero.18 The depth of burial, containment structures, size of decoupling cavity, etc. would have to be set for the maximum credible yield that could be produced by the weapon, rather than the design yield or most likely yield. The maximum estimate, in turn, would have to be below the monitoring threshold at the lowest confidence level. As a result, if the proliferator thought they could hide a one kiloton explosion, they would probably be restricted to testing at a level that they thought would produce a yield approximately half that size. The exception would be the situation of a country that had managed to obtain nuclear weapons from another state, and was testing a sophisticated device of known yield to establish that they could detonate the device.

Decoupling

In the "1970's & 1980's 'How-Low-Can-We-Go' Algorithm ", it is the need to demonstrate a capability for detecting decoupled explosions that most controls the low-magnitude threshold for a monitoring network. Decoupling, however, requires a very specific geology and engineering capability.19 For example, to fully decouple a 5 kt explosion in salt, a spherical cavity with a radius of at least 43 meters would be required (see figure below). We assume that a nation conducting a test will readily undertake decoupling, and that they will have confidence that their attempts to reduce the seismic signal through decoupling would be successful. While this may be a legitimate assumption for established nuclear powers such as the U.S. and Russia, it is a much less credible assumption for a first-time tester.

Minimum Cavity Size Required to Decouple a 5 kt Nuclear Explosion
[graphic of Statue of Liberty in cavity]
Source: Office of Technology Assessment, 1988

While at one time it was estimated theoretically that decoupling might reduce the seismic signal by as much as 300 or even 1000 times, subsequent experiments and calculations have shown these earlier predictions to be too optimistic. Extrapolations from U.S. experiments indicate that at low seismic frequencies, a fully decoupled explosion may have a signal 70 times smaller than that of a fully coupled explosion. At high frequencies, the decoupling factor is probably reduced to somewhere between ten and seven, once again indicating the importance of including broadband seismic stations and high frequency sensors within a global monitoring system. Even these extrapolations (all of which have been made from explosions of a fraction of a kiloton) may still overstate decoupling potential. For example, data has recently become available from a ten kiloton partially decoupled explosion carried out by the Soviet Union in a salt dome in Western Kazakhstan. Data from this explosion indicate that the seismic signal amplitudes were reduced only by about a factor of ten relative to that which would have been expected from a tamped explosion of similar yield, suggesting that the decoupling factor drops off rapidly if the explosion is too large for full decoupling within the cavity.20 In contrast, computer simulations of explosions in spherical and ellipsoidal cavities indicate that the decoupling factor may decrease slowly as the yield increases beyond that associated with full decoupling.21 In any case, there are large uncertainties in predicting the size of seismic signals produced from decoupled explosions larger than a fraction of a kiloton.

Releases From U.S. Underground Tests22

	Major pre-1971 releases:
	Platte, 1962..............................1,900,000 Curies (Ci)
	Eel, 1962.................................1,900,000 
	Des Moines, 1962.........................11,000,000
	Baneberry,1970............................6,700,000
	26 others from 1958-1970..................3,800,000

	Containment Failures 1971-1988:
	Camphor, 1971...................................360 Ci
	Diagonal Line, 1971...........................6,800
	Riola, 1980.....................................690
	Agrini..........................................690

Containment

In addition to requiring that the explosion not produce a discernible seismic signal or other evidence of a test, the decoupling scenario is further complicated by the requirement that the preparations for the test, its detonation, and any after-effects also be kept secret. All radioactive materials produced by the explosion must be contained underground (i.e. the genie must stay in the bottle). One might assume that small, deeply-buried explosions are easily contained. Based on both U.S. and Soviet experience, this is not necessarily true. The table above is adapted from the Congressional Office of Technology Assessment (OTA) report "The Containment of Underground Nuclear Explosions".23 It shows all accidental releases of radioactive material from U.S. tests, excluding late-time seeps or deliberate tunnel purgings and operational releases.24

All four of the 1971-1988 containment failures (Camphor, Diagonal Line, Riola, and Agrini) had yields of "less than 20 kilotons". Of the pre-1971 releases, Platte was 1.8 kilotons; Eel was "less than 20 kilotons"; Des Moines was "less than 20 kilotons"; and Baneberry was 10 kilotons.25 In the case of the United States, essentially all accidental releases of radioactive material have been from tests smaller than 20 kilotons.26 Similar experiences seem to be true for the Soviet Union's testing program, which suffered more frequent releases.27

The United States and the Soviet Union developed their containment technology largely through trial and error. Most of what is known to cause problems for containment - carbonate materials, water, faults, scarps, clays, coaxial cable, etc. - was learned through repeated experience at well-studied test sites. In addition, features that are thought to assist containment - porosity of the rock to contain noncondensible gases (CO2, H2), chimney collapse to reduce steam pressure within the cavity created by the explosion, rebound of the surrounding rock to close off fractures - might be absent from the evasion scenario of a decoupled, low-yield explosion, conducted in hard rock under the guise of a mining operation. Late-time seeps may also occur days or weeks after a test as the noncondensable gases diffuse up through the overlying rock and are drawn to the surface by decreases in atmospheric pressure (called "atmospheric pumping"). In particular, fission byproducts include noble gases such as krypton and xenon that can filter through the soil column and reach the atmosphere.

Test site workers flee the area after the unexpected venting of a 1 kiloton nuclear test.
Photogrpah courtesy of J.E. Carothers, Lawrence Livermore Natioanl Laboratory

A one kiloton explosion will produce 41 billion curies one minute after detonation, and the amount will decrease to ten million curies in 12 hours. An accidental venting might release one to ten percent of the total radiation generated by the explosion, i.e. 100,000 to 1,000,000 Curies in the case of a one kiloton explosion. In the photograph on the previous page, test site workers are fleeing the area after the unexpected release of radioactive material from the Des Moines test on June 13, 1962. This one-kiloton test was one of the first carried out in a new tunnel at a depth of about 200 meters. Approximately 11,000,000 Curies were released unexpectedly into the atmosphere.

Collapse Crater

When gas pressure in the cavity declines to the point where it is no longer able to support the overlying rock, the cavity may collapse. The collapse occurs as overlying rock breaks into rubble and fills the cavity void. As the process continues, the void region moves upward as the rubble falls downward. The chimneying continues until either: a) the void volume within the chimney completely fills with loose rubble, or b) the chimney reaches a level where the shape of the void region and the strength of the rock can support the overburden material, or c) the chimney reaches the surface. If the chimney reaches the surface, the ground sinks, forming a saucer-like subsidence crater. Cavity collapse and chimney formation typically occur within a few hours of detonation but sometimes take days or months. Cavity collapse is useful for containment because the cool rock falling into the cavity condenses the steam. In general, surface cratering can be avoided by deep burial depth, although the unexpected formation of a collapse crater is not out of the question and would have to be considered by a proliferator when testing in certain geological formations.

Nuclear Testing and Nonproliferation

Return to:Table of Contents
Return to: Chapter II: Rethinking the Problem
Continue to:Chapter II: Overall Assesment for Clandestine Testing