Silencing the Bomb: One Scientist’s Quest to Halt Nuclear Testing

NEW METHODS TO IDENTIFY UNDERGROUND TESTS: 1963–1973

Although large seismic events could be identified in 1963, few methods were available then to distinguish, or discriminate, the signals of small underground nuclear explosions from those of the many small earthquakes that occur worldwide every day. More specifically, this discrimination was needed for countries of special interest to the United States at the time—the Soviet Union and China.

In the 1950s, the U.S. government quickly recognized seismology’s potential for detecting and identifying underground nuclear tests. A panel of technical experts headed by physicist Lloyd Berkner recommended in 1959 that the United States should greatly expand funding of seismology to increase fundamental understanding and to develop better instrumentation.

Subsequent funding for the underground explosion part of the program came from what was called the Vela Uniform program. Run by the Advanced Research Project Agency (ARPA) of the Defense Department, it transformed seismology almost instantaneously from a sleepy, poorly supported scientific backwater to a field flooded with new funds, instruments, professionals, students, and excitement. Seismology and its instruments for recording earthquakes became the main technology for detecting, locating, and identifying underground nuclear tests. As a result of this investment, seismologists today can identify much smaller seismic events than they could in 1963.

DETERMINATION OF DEPTHS OF EARTHQUAKES AND EXPLOSIONS

No nuclear explosions have been set off deeper than 3 miles (5 km), and nearly all have been detonated within the upper 2 miles (3 km) of the Earth’s crust. Therefore, the reliable determination that a seismic event is considerably deeper means that it almost certainly cannot be a nuclear explosion. Holes drilled deeper than 6 miles (10 km) in hard rock are difficult to use for clandestine testing. An experimental deep Soviet scientific test hole on their Kola Peninsula deformed quickly in response to large stress differences.

Good determinations of depths greater than about 30 miles (50 km) readily identified events beneath the Kuril Islands (figure 6.1) and Kamchatka, the most active areas of the USSR, as earthquakes. Similar determinations also identified frequent deep events beneath the Hindu Kush and Pamir mountains of Central Asia as earthquakes.

FIGURE 6.1

Locations of earthquakes projected onto a cross section of the Earth extending from the Pacific Ocean at the right, across the Kuril subduction zone, to the eastern Sea of Okhotsk at the left.

Source: Sykes, Evernden, and Cifuentes, 1983.

Many earthquakes with depths shallower than 30 miles (50 km) occur well offshore of the Kamchatka Peninsula and the Kuril Islands of Russia (figure 6.1). They can be distinguished as earthquakes by their very small hydroacoustic (sound) waves in the Pacific Ocean. Explosions within the water column generate much larger acoustic waves that propagate readily to large distances in the oceans. Hence, earthquakes offshore of, and most earthquakes beneath, Kamchatka and the Kuril Islands have been identifiable as such for many decades.

Removing these and deeper shocks from consideration in 1963 still left a number of small shallow earthquakes in continental areas as difficult to identify (figure 6.2). With the breakup of the Soviet Union, however, many of the regions where these shallow events occur are now located in the independent countries of Armenia, Azerbaijan, Kazakhstan, Kirghizstan, Tajikistan, Turkmenistan, and Uzbekistan. The Russian Republic is less likely to test in those areas because secrecy would now be more difficult to maintain. In addition, much of the Russian Republic itself consists of old geological terrains, which contain few earthquakes.

FIGURE 6.2

Earthquakes in the former Soviet Union and surrounding areas from 1971 to 1986.

Source: U.S. Geological Survey; Office of Technology Assessment, 1988.

FIRST MOTIONS OF SEISMIC P WAVES

The Conference of Experts meeting in Geneva in July and August 1958 thought that the first motions of seismic P waves could be used to identify earthquakes by observing clear downward, or dilatational, first motions of P waves. Nuclear explosions do not generate dilatational first motions, only compressions. It soon became clear in late 1958, however, that the first motions of short-period P waves were not reliable for events comparable in size to the U.S. underground explosions Rainier and Blanca.

With the stations of the World-Wide Standardized Seismograph Network (WWSSN) in place after 1963, clear first motions for earthquakes larger than magnitude m_b 5.5 often could be identified on long-period (low-frequency) records. First motions are either up or down on vertical instruments from earthquakes. Because first motions were not viable in 1963 to reliably identify events of m_b 4.75 and smaller, other identification methods clearly were needed for those events.

SEISMIC ARRAYS

Seismic arrays are groups of instruments typically spaced a few to about 6 miles (10 km) apart that are connected to one common recorder. Since 1963, additional seismic arrays have been installed around the world to detect and identify short-period seismic P waves. The signals from the group of instruments in an array can be processed to enhance a specific seismic signal and to suppress earth noise. This allows seismologists to identify more small events and to determine the direction of the source relative to the array.

Prior to the U.S. Longshot nuclear explosion in the Aleutians in 1965, the British group working on nuclear verification rushed to install new seismic arrays at Yellowknife, Canada, and in central Australia and India. These were in addition to the existing UK array in Scotland. Since 1996, when the CTBT was signed, several countries have installed many new arrays. The seismic wave called pP (see figure 3.1), which is reflected from the Earth’s surface near the hypocenter and arrives after the P wave, can now be identified more readily using array processing and used to determine depths for many small to moderate shallow earthquakes.

DETECTION OF SEISMIC SURFACE WAVES FROM EXPLOSIONS

In 1958 Jack Oliver, my PhD advisor at Lamont, observed long-period seismic waves at Palisades, New York, generated by the 22-kiloton Blanca test. The long-period waves he observed were surface waves, not P or S waves. Over the next twenty years, his observation led to much study at Lamont and elsewhere on the use of surface waves to distinguish the signals of underground explosions from those of earthquakes.

When I arrived at Lamont in 1960, funding for the identification of underground nuclear explosions under the Vela program of the Defense Department had just started. I shared an office in Lamont Hall, originally the home of the Lamont family, with three scientists, including Paul Pomeroy, a graduate student who was several years ahead of me. For his PhD thesis, Pomeroy examined long-period recordings from nuclear explosions, mainly those in the atmosphere, as recorded at Palisades and at the long-period seismograph stations deployed by Lamont during the International Geophysical Year (IGY) in 1957 and 1958. In the early 1960s, Lamont and Cal Tech competed to discover methods for distinguishing the seismic signals of underground nuclear explosions from those of earthquakes.

USE OF SURFACE AND P WAVES FOR EVENT IDENTIFICATION

In 1963 Pomeroy and Lamont graduate student Robert Liebermann, now a retired professor at Stony Brook University, discovered a method called Ms-m_b that was to become one of the most reliable techniques to discriminate the signals of earthquakes from those of underground nuclear explosions. Magnitude Ms is a measure of long-period surface waves with periods of about twenty seconds. The period of a wave is the time it takes for one cycle of motion to occur—the time from one upward motion to the next—in this case, twenty seconds. Surface waves travel around the Earth’s circumference, not through its very deep interior (figure 3.1). Magnitude m_b is determined from measurements of short-period P waves, which arrive before the surface waves, taking a shorter time to travel to a station through the deep interior of the Earth. The difference between Ms and m_b is then determined and used to identify an event as either an earthquake or an underground explosion.

This method takes advantage of the differing nature of the two types of seismic sources. Underground nuclear explosions instantaneously crush a relatively small volume of surrounding rock, which absorbs the force of the blast and generates seismic waves that propagate outward in all directions. In contrast, earthquakes are caused by slip (displacement) along a fault. Rupture in an earthquake takes place more slowly than in a nuclear explosion. Hence, earthquakes generate seismic waves for longer periods of time. They also rupture larger areas and radiate seismic waves whose amplitudes vary in azimuth (direction) around the source of an earthquake. This effect is much like that of a radio antenna that beams signals preferentially in certain directions. For earthquakes and underground explosions of the same m_b, earthquakes generated much larger surface waves. These measurements are shown in figure 6.3. British, Canadian. and other U.S. scientists went on to do a great deal of work using this technique.

FIGURE 6.3

Seismograms of long-period waves from an earthquake in the Arctic near the USSR (top) and a Soviet underground nuclear explosion at Novaya Zemlya (bottom). Each was recorded at Eilat, Israel, at about the same distance from the two sources. The short-period P waves (not shown here) were of nearly the same magnitude, m_b, yet the surface waves are much larger for the earthquake.

Source: Sykes, Evernden, and Cifuentes, 1983.

In the Ms-m_b diagram in figure 6.4, when Ms is plotted on the vertical axis and m_b on the horizontal, a very good separation is obtained for earthquakes and underground explosions. One exception was a small earthquake in the southwest Pacific where the network used to measure those two magnitudes was weak. The network was much more sensitive for measuring seismic events in the northern hemisphere. For a given yield, m_b readings are larger for earthquakes and explosions in regions of old strong crust and uppermost mantle of the Earth than they are for younger regions like Nevada. The values of m_b were corrected for this effect in figure 6.4.

FIGURE 6.4

Magnitude Ms measured from long-period surface waves and magnitude m_b determined from short-period P waves. Note the clear separation of explosion and earthquake populations, with the exception of one earthquake in the southwest Pacific.

Source: Sykes, Evernden, and Cifuentes, 1983.

Unfortunately, the development and validation of the Ms-m_b technique were a few years too late to be considered during the test ban negotiations in 1963. At the time, the U.S. government concluded that effective identification of underground explosions was too difficult. In addition, the Soviet Union was not willing to accept U.S. and UK requests for on-site inspections and monitoring stations on its territory. Consequently, underground tests were excluded from the Limited Test Ban Treaty (LTBT) between the United States, the USSR, and the UK, which outlawed explosions in the atmosphere, in space, and underwater, but not underground. Unfortunately, the Defense Department initially classified the Lamont work like that shown in figure 6.4 even though it provided a good international tool for event identification vital to a test ban.

ADVENT OF PLATE TECTONICS

Plate tectonic theory did not exist in 1963, but by 1969 it provided a new foundation for understanding why most earthquakes, volcanoes, and young mountain belts occur where they do. At the same time, it helped us to understand and predict the type (styles) of earthquake mechanisms for regions of the USSR and China, allowing for better discrimination between earthquakes and nuclear blasts. About half of my second book is devoted to my involvement with the plate tectonics revolution.

Plate tectonics also furnished an understanding of why seismic P waves propagate efficiently in the old rocks beneath the Soviet, Chinese, and Indian test sites but not beneath the U.S. test site in Nevada and the French site in southern Algeria. Until as late as 1988, the United States government failed to appreciate or to acknowledge this difference and accused the Soviet Union of cheating on the Threshold Test Ban Treaty.

During these years, many people in the U.S. departments of Defense and Energy either ignored or resisted seismological evidence that took these differences in the structure of the Earth into account. They also insisted that seismic surface waves should not be used for yield estimation or calibration. Debates over Soviet yields and verifying a Comprehensive Test Ban Treaty continued for many years. Throughout this time, I, along with other colleagues, fought hard to give voice to the notion that the United States was overestimating the yields of Soviet nuclear explosions by large amounts. I testified about this before Congress several times between 1972 and 1986.

DEVELOPMENT OF NEW SEISMOGRAPHS FOR NUCLEAR VERIFICATION

In the late 1960s, Paul Pomeroy, George Hade, and others at Lamont developed seismographs that could detect even longer-period (lower-frequency) seismic surface waves than those with periods of about twenty seconds that were then being used in the standard Ms-m_b method for distinguishing underground explosions from earthquakes. These new instruments, called High-Gain, Long-Period seismographs (HGLPs), magnified signals in the Earth more than one hundred thousand times for periods between twenty and seventy seconds—about one hundred times greater than the long-period instruments of the World-Wide Standardized Seismograph Network (WWSSN), which were installed in 1962 and 1963.

The HGLP instruments took advantage of the fact that natural earth noise drops off significantly for periods longer than twenty seconds and reaches a stable minimum at a period of thirty to forty seconds. Reduced earth noise at those longer periods enabled surface waves from even smaller earthquakes to be observed and used to distinguish signals from underground explosions from those of earthquakes.

Installing those seismometers in thick hemispherical steel tanks and deep underground in mines reduced earth noise, especially on horizontal-component HGLP instruments. One of the most important findings from the first HGLP station, in a deep mine at Ogdensburg, New Jersey, was that the discrimination of earthquakes and underground explosions was enhanced using forty-second surface waves compared to using the standard twenty-second waves. The period of a surface wave is the time it takes for one cycle of wave motion to occur—say, up to down and back up again on vertical records. It is not the length of time a surface wave lasts on a seismogram, which is many periods.

Recording quiet horizontal motions allowed another type of surface wave, called Love waves, to be detected. Love waves, like light, have horizontal vibrations perpendicular to the direction of wave propagation. Named for applied mathematician A. E. H. Love, they could be employed when Rayleigh waves, the more common surface waves used to measure Ms, were not recorded well. Rayleigh waves can be unusually small for certain types of focal mechanisms of earthquakes that occur at depths of about 25 miles (40 km). The use of Love waves also provided an excellent separation between earthquakes and explosions on an Ms-m_b plot for periods of twenty to forty seconds.

The results from Ogdensburg in distinguishing the signals of explosions from those of earthquakes were spectacular enough that in 1970 the U.S. Defense Department funded the deployment of additional HGLP stations around the world set up by Lamont, the University of Michigan, and the Albuquerque office of the U.S. Geological Survey. The data from these stations were recorded digitally, a first for university seismologists. Many Lamont graduate students and staff worked on the installation of seven additional HGLP stations around the world and on the analysis of the data. Peter Ward, a postdoctoral student in seismology, headed the installation and operation of the Lamont HGLP stations. I was the principal investigator and was involved in analyzing the data to test their value for event discrimination.

The seven HGLP stations operated by Lamont showed similar very positive results for seismic discrimination to those at Ogdensburg. A minimum of earth noise was observed from all of these stations, which, along with the use of digital filtering, improved the detection levels for surface waves.

The HGLP stations operated in the 1970s at a time when the United States and the Soviet Union detonated a number of very large underground nuclear explosions, which the bilateral Threshold Test Ban Treaty forbade after March 31, 1976. (Today, modern digital recording stations around the world record seismic waves of similar long-period as well as shorter-period waves.)

John Savino, Peter Ward, and I of Lamont presented papers on the first results of the Ogdensburg HGLP experiment at the ARPA Conference on Nuclear Test Detection at Woods Hole, Massachusetts, in July 1970. Jack Evernden, then at the Advanced Research Project Agency (ARPA) of the U.S. Defense Department, organized the conference. In his written summary of the meeting, Evernden indicated that the CTBT detection and identification problem for underground explosions and earthquakes was solved. His summary and other papers presented at the conference were openly distributed in 1971. Evenden, however, failed to get his written summary cleared by Steven Lukasik, the director of ARPA. Lukasik exclaimed that the policy position in Evernden’s summary was not approved by the Defense Department and demanded the return of all copies. Senator Clifford Case of New Jersey, a proponent of the CTBT, asked to see a copy of the entire document. When he found shreds of paper where the summary was torn out from the rest of the papers, he raised hell in the Senate.

JACK EVERNDEN AND CARL ROMNEY

This is a good time to discuss the major efforts Evernden made on behalf of the CTBT, as well as the role of Romney. I worked with Evernden on a number of occasions and published several papers with him on test ban verification. He was born in 1922 and received his PhD in geophysics from Berkeley in 1951. Along with physics professor John Reynolds, Evernden developed instrumentation for measuring and dating volcanic rocks younger than several million years. To date the time those rocks were formed (solidified), they made precise measurements of one isotope of the element potassium and of its radioactive decay product, an isotope of the element argon.