U.S. National Capabilities in Regions of Interest
Seismic Monitoring Improvements of the Past Decade
Since 1960, a comprehensive program to improve the monitoring of underground testing has been conducted by the United States and has involved the national laboratories, federal agencies, independent contractors, and the academic research community. The effort has produced major advances in all areas—the sensitivity of seismological instrumentation, the methodologies for analyzing the data collected from seismometers, and the geographic distribution of seismological data collections systems.
FIGURE D-1: Improvement in seismic monitoring over the last 20 years. Threshold values indicate statistically significant confidence of detection. Note that yield and seismic magnitude scales are logarithmic; each unit of improvement is a factor of ten. Seismic sensitivity to nuclear explosions has improved significantly due to increased deployment of seismometers and improved data analysis. For locations of interest, this allows regional monitoring at distances less than about 1,600 km (1,000 miles), which has a much lower threshold (~ 20 tons or 0.020 kilotons explosive yield) than does global monitoring (~ 200 tons or 0.20 kilotons) recorded at distances typically greater than 2,000 miles. The mention of a “local threshold” lower right in
figure, refers to the future possibility of bilateral monitoring agreements, distinctly separate from the CTBT. The magnitude estimates listed on this figure are from the IDC and other public sources. SOURCE: Seismology Subcommittee
As depicted in Figure D-1 (and Figure 2-7), over the past decade the capabilities of the U.S. and other countries to detect small tests (and, hence, to detect attempts at evasive testing) have improved greatly due to the implementation of regional monitoring methods, increases in data availability and quality and the application of new discriminants. Fundamental improvements in our ability to monitor underground nuclear testing have occurred in the following three areas:
1. Use of regional seismic waves
The routine use of “regional” waves—seismic waves that travel at higher frequencies within the earth’s crust and that travel distances up to 1,600 km (1,000 miles) have fundamentally changed the strategies used for nuclear monitoring. Regional seismic waves travel through the earth’s crust and uppermost mantle (the top few tens of kilometers of the earth’s interior) at high frequencies. Particularly important is the availability today of data from seismic stations throughout the Middle East, North Africa, Russia, Kazakhstan, Mongolia, China and South Korea. Access to these regions permits countries of special concern to the United States to be monitored at much closer distances and thus down to events of much smaller size. While research into regional seismic methods was underway when the CTBT text was finalized more than a decade ago, it was during the past decade that many of these new research products began to be implemented into the routine operational systems that continue today.
2. Increased data coverage, quality, and availability
Over the last decade, the amount of seismological data that is available to detect nuclear weapons tests (including evasively-conducted tests) has increased approximately 10 fold. Improvements in global digital communication networks have increased the capacity to transmit these large amounts of data from around the world in real or near-real time by approximately 100-fold. Computer power, data storage and retrieval increased approximately 10-fold. Furthermore, much of this newly available data is coming from nations and areas that were previously inaccessible to the United States.
3. Improvements in our ability to distinguish the seismic signals of a small nuclear explosion from those of naturally occurring earthquakes
The use of digital data now provides for ground motion to be recorded continuously with high sampling rates over broad frequency ranges. This has led to new methodologies that use the characteristics of the ground motion that are recorded at various frequencies to distinguish small explosions from naturally occurring earthquakes. Access to high-frequency seismic data has allowed for applications of new discriminants that have improved by about a factor of 10 or more our capability to distinguish explosions from naturally occurring earthquakes. The number of problem events that occur (fewer than about one per year) is now small enough that on-site inspections are feasible under the verification provisions of the treaty.
Monitoring the Russian Test Site at Novaya Zemlya
In ratifying the Threshold Test Ban Treaty with the United States in 1990, the Soviet Union declared two official sites for testing nuclear weapons—eastern Kazakhstan (Semipalatinsk) and Novaya Zemlya. From 1949 to 1989, at which time the Semipalatinsk test site was closed (Adushkin and Leith, 2001), eastern Kazakhstan was the primary location for Soviet nuclear tests. Kazakhstan became an independent country in 1991. The largest Soviet underground nuclear tests, however, were conducted at two sites on remote Novaya Zemlya because they would have produced unacceptable damage in mainland Asia or Europe. The Krasino site on southern Novaya Zemlya has not been used for testing since 1975.
Russia continued testing at its arctic test site near Matochkin Strait on Novaya Zemlya until 1990 (Khalturin et al., 2005). The logistics of work at that site are difficult because its latitude, 73o North, is farther north than the northernmost part of Alaska; 24 hour days of darkness and winter-like cold come early in the Fall. A glacier covers the northern part of the north island of Novaya Zemlya. The containment record of underground nuclear explosions in the hard rock at Novaya Zemlya is poor, with many shots leaking gaseous fission products and particulates (Adushkin and Leith, 2001). Today, much more sensitive measurements of xenon and other noble gases make containment an even more difficult challenge for a potential evader.
A number of public claims have been made about possible Russian testing at Novaya Zemlya since Russia signed the CTBT in September 1996. Seismic monitoring capabilities for the Russian test site on Novaya Zemlya place it among the best-monitored test sites in the world. Novaya Zemlya and its adjacent seas have very low levels of earthquake activity and chemical explosions (Figure D-2). Since 1995 a few small earthquakes on Novaya Zemlya and in the nearby parts of Kara and Barents seas have been claimed in media reports as being problem events (i.e., hard to verify as either explosions or earthquakes. The Seismology Subcommittee examined those claims and concluded that all of those events were small earthquakes (see Sykes, 1997 and the discussion following Figure D-4).
The United States and Norway have worked together for more than 40 years in deploying seismic arrays for monitoring seismic activity on and near Novaya Zemlya and in developing better seismic methods for detection and identification. Data are available in real time from array stations in northern and southern Norway, Spitsbergen (a Norwegian island), and Finland (Figure D-3) and arrays in other parts of Europe. With short time delays, data are available from other stations in Scandinavia and Finland, including the IMS auxiliary station Hagfors in Sweden. The detection capability for the four arrays in Figure D-3, when their continuous data streams are used for analysis, is better than that of the IMS because the station SPITS is only an auxiliary IMS array. Thus, even though it is usually the most sensitive station for signals from Novaya Zemlya, it can only be used by the IMS for signals from events already detected by the other arrays. SPITS is particularly useful for locating Novaya Zemlya events because its azimuthal coverage for that region is so different from that of the other arrays.
FIGURE D-2: Location of seismic events and stations in the vicinity of the Russian Test Site at Novaya Zemlya. Overlapping red stars show the locations of past nuclear tests since 1977, including the last test on October 24, 1990. Seismic events from 1999 to 2009 with magnitude greater than 2.0 from the Norwegian organization NORSAR reviewed regional bulletin are shown as blue circles.2 IMS primary stations (triangles), auxiliary station (squares) and some of the other publicly available stations (pentagons) are also shown. Some of the NORSAR events in Scandinavia (including Finland) and mainland Russia are associated with mine blasts. Overlapping orange diamonds locate seismic signals associated with the submarine Kursk disaster in 2000. Note that the area around the test site has a low level of some natural earthquake activity, as indicated by the blue circles. SOURCE: William Walter, Seismology Subcommittee member
FIGURE D-3: Map of Novaya Zemlya and locations of four seismic arrays in Norway, Finland and Spitsbergen. SOURCE: Kværna et al., 2002
Seismologists at the Norwegian organization NORSAR have developed a program for monitoring the magnitude thresholds of specific sites on a continuous basis. It is particularly good for source areas of low earthquake activity as shown for Novaya Zemlya (Figure D-4) for a 24-hour period on February 9, 1998 (Kværna et al., 2002). Noise levels and spikes of energy received from earthquakes during that day (in the azimuthal beam focused on the test site at Novaya Zemlya) are indicated for each of the arrays in the bottom four panels and in the top panel for the network that consists of those four arrays. This is a “smart” network that for every possible event time at Novaya Zemlya favors the array (s) that have the least background energy at the corresponding (known) arrival delay for the explosion (P) waves. Magnitude thresholds at 90 percent confidence for detection on the vertical axes of each panel vary as a function of time and consist of background noise levels with superposed spikes of energy from earthquakes that are larger than background for brief time intervals. On the date in question the background noise is highest at NORES (approximately magnitude 2.9), and lowest at SPITS (magnitude 2.2). For the whole smart network (top panel), the background noise level for possible events at Novaya Zemlya is reduced to about magnitude 1.8. Because P waves from seismic events that are not at Novaya Zemlya arrive at different times, they are suppressed in the network trace. Hence, the number of spikes and their magnitudes are reduced for the network.
The largest excursions above background noise are associated with a distant earthquake (labeled 1) in the Sea of Okhotsk. On the network panel, that excursion reaches magnitude 3.1 but only for about 10 minutes. Hence, the threshold capability of the network as a detector of a realistic nuclear test on that day is about magnitude 1.8. Kværna et al. (2002) examined data for the last two months of 1997 and found a threshold capability most of the time at magnitude 2.0. For those two months and for all of 1998 they found some time periods in which those thresholds are as high as magnitude 2.5. The detection and identification thresholds for monitoring Novaya Zemlya are therefore estimated by the Seismology
Subcommittee to be in the range of magnitudes 2 to 2.5. Corresponding yields are discussed below.
FIGURE D-4: Example of site-specific threshold (“smart network”) monitoring for seismic events from Novaya Zemlya for 24 hours on February 9, 1998. SOURCE: Kværna et al., 2002
For well-coupled nuclear explosions at the main Russian test site at Novaya Zemlya, magnitude = 4.3 + B log Y (where Y is yield in kilotons)3. B is about 0.75 for explosions larger than 1 kt and results from larger explosions being detonated at increasingly greater depth to ensure containment. For very small nuclear explosions, the Seismology Subcommittee assumes that explosions would be at least as deep as previous 1 kt tests and hence take B = 1.0. It should be realized that at very small magnitudes, the calculation of yield is more uncertain because few calibration data exist for yields below 1 kt. Also, at low yields it can become practical to influence the environment of the nuclear device in ways that somewhat change its coupling to seismic energy, which can introduce additional uncertainty in yield estimates.
Hence, for magnitudes 2.0 and 2.5 one obtains Y = 5 to 15 tons (0.005 to 0.015 kt) if an underground explosion is fully coupled. For a decoupling factor (DF) of 40, Y = 200 to 630 tons. Assuming a potential evader is more conservative and uses a DF = 20 for hard rock in his assessment of yield associated with the detection limits at Novaya Zemlya, Y = 100 to 320 tons.
3 This magnitude-yield relation uses that of Murphy (1996) for hard rock at eastern Kazakhstan minu s 0.15.
These yields for either fully coupled or fully decoupled nuclear tests are a fraction of a kiloton. See Appendix E for a discussion of decoupling factors in hard rock versus salt.
It should be realized that the above are 90 percent confidence limits for detection. From the viewpoint of a potential evader not wanting to be detected, a lower detection confidence level likely would be appropriate. For instance, a 10-percent probability of detection could result in the above yields for fully coupled and decoupled cases being reduced by a factor of about 2.5. In that case, for a risk-averse evader determined to conduct nuclear explosion a maximum yield for fully decoupled tests using a DF of 20-40 would be between 0.04 and 0.42 kilotons.
Because salt is not present in any appreciable thicknesses at Novaya Zemlya, decoupled tests would have to be conducted in hard rock. Somewhat smaller threshold yields than those above are obtained for the old Krasino test site in southern Novaya Zemlya where magnitude = 4.45 + B log Y, a relationship that also is appropriate for past tests at eastern Kazakhstan (Murphy, 1996), at Lop Nor, China, and in northern India.
Several small earthquakes have been detected, located, and identified in the vicinity of the Novaya Zemlya test site during the last 25 years. The 90 percent confidence limit for location of one of these, a magnitude 2.7 earthquake in 1992, included the test site. This event, events in the Kara Sea in 1986 and 1997, and two on the north island of Novaya Zemlya in 1995 and 1996 were claimed in the media to be either Russian nuclear tests or possible tests. Subsequent special studies showed that each of these was an earthquake (see Sykes, 1997).
The Kara Sea seismic event (magnitude about 3.5) of August 16, 1997, received the greatest amount of attention in the U.S. media. It was described soon after as having explosive characteristics (Gertz, 1997). The results of initial efforts to evaluate the location of the event were poor, but analysis of data from stations in mainland Norway, Finland and Spitsbergen quickly showed that it occurred in the Kara Sea, well to the southeast of the test site. Further work later confirmed that the event was a small earthquake.
The Seismology Subcommittee drew three lessons from the handling of the 1997 event 1) use all available data for accurate location estimates and event characterization 2) avoid the use of just a narrow range of azimuths such as those to southern Norway and Finland, and 3) provide a mechanism for new information to be updated to policy makers as it becomes available for occasional “problem” events of this type. For example, scientists at the UK Atomic Weapons Establishment who work on nuclear verification regularly publish papers about once a year in leading scientific journals about “problem” seismic events such as those in the Kara Sea in 1986 and 1997. Their thorough analyses have convinced nearly all seismologists that those events were earthquakes. In general, the occasional problem events that have arisen have—through the research they have motivated—resulted in significant improvements in operational monitoring.
Seismic waves from two explosions of local seismic magnitude 1.5 and 3.5 that led to the sinking of the Russian Kursk submarine in the Barents Sea on August 12, 2000 (Figure D-2) are described by Ringdal et al. (2000). The larger explosion was well recorded at stations in northern Europe and as far away as Kazakhstan. Subsequent depth charges set off by the Russian navy were recorded by the seismic array ARCES in northern Norway—another indication of how well the Barents Sea and Novaya Zemlya can be monitored.
Explosions in and around Novaya Zemlya can be identified by the usual seismic techniques, including high-frequency P-wave-to-S-wave amplitude ratios (e.g., Richards and Kim, 1997; Hartse, 1998). An example of typical differences is shown in Figure D-5 where the high frequency seismograms of the nuclear test have larger P-waves and smaller S-waves than do nearby earthquakes. By correcting for distance effects, we can compare P/S ratios for many events around Novaya Zemlya, as shown in Figure D-5. Note the 1997 Kara Sea event separates cleanly from the nuclear tests, as do more recent earthquakes in 2007 and 2009. The magnitude 2.8 (NORSAR ML) June 26, 2007 earthquake was located by NORSAR to within 50 km (31 mi) of prior nuclear tests. By combining these single-station results with similar
measurements at other stations and with other discriminant measures, a very high degree of confidence in identifying small explosions in this region has been achieved. An earthquake of magnitude 4.5 (according to NORSAR) occurred close to the northern main island of Novaya Zemlya on October 11, 2010, as this report was being finalized. It was detected by many IMS stations and will be useful in calibrating teleseismic and regional travel times for the vicinity of the Russian test site.
FIGURE D-5: Example showing how ratios of high frequency P-wave to S-wave amplitudes discriminate Novaya Zemlya nuclear tests from earthquakes. The left-hand side compares 6-8 Hz seismograms of the 1997 Kara Sea earthquake (blue) with the last nuclear test in 1990 (red). The right-hand side shows MDAC distance-corrected P/S values for five nuclear tests (red stars), Kursk related explosions in the water (orange diamonds), earthquakes around Novaya Zemlya (light blue circles) and other earthquakes around the Barents Sea (dark blue circles). The 1997 Kara Sea and more recent earthquakes in 2007 and 2009 are called out. SOURCE: William Walter, Seismology Subcommittee member
In summary, the use of high-frequency seismic waves has come of age in the last 10 years as a valuable discriminant for earthquakes and underground explosions. The Russian test site at Novaya Zemlya is one of the world’s best-monitored places of high concern to the United States. At 90 percent confidence, seismic events can be detected there down to magnitudes in the range of 2.0 to 2.5 corresponding to fully coupled explosions of about 5 to 15 tons (0.005 to 0.015 kilotons). As salt is not present on Novaya Zemlya in any appreciable thicknesses, any decoupled testing would have to be done in hard rock for which maximum decoupling factors of 20 to 40 are most appropriate. For decoupled explosions, detection at high confidence therefore corresponds to yields of about 100 to 600 tons. From the viewpoint of a potential evader not wanting to be detected, maximum decoupled yields of about 40 to 250 tons (0.04 to 0.25 kilotons) may be more appropriate. The low level of earthquake activity on and near Novaya Zemlya also makes monitoring the Russian test site easier.
Monitoring the Chinese Test Site at Lop Nor
China conducted all of its nuclear tests near Lop Nor in the sparsely populated northwestern part of the country. China stopped testing in 1996 just prior to signing the CTBT.
As a consequence, newer stations of the International Monitoring System (IMS) and other modern stations recorded more Chinese explosions in the 1990s than they did for other countries that stopped testing earlier. The Lop Nor area is seismically active when compared with the former Soviet Union’s test site in eastern Kazakhstan (KTS) or the Russian test site on Novaya Zemlya, with several earthquakes of magnitude 4 and greater having occurred near the testing area over the past few decades. In this section we describe several techniques for distinguishing (discriminating) seismic waves of nuclear explosions from those of earthquakes since they are well illustrated for this test site.
Lop Nor is well monitored by arrays and seismic stations in Kazakhstan, Kyrgyzstan, Mongolia, Russia, Afghanistan, Pakistan, and Thailand. Eleven modern digital seismic stations in China are operated in conjunction with the U. S. Geological Survey. Most of these stations send data with a short time delay (within 30 minutes) to the IRIS data center in the United States and are openly available. The Chinese station Urumqi (assigned the station code WMQ by seismologists) is about 250 km (156 mi) from Lop Nor, closer than any other open station. Waveforms from WMQ are archived at IRIS on about a one-month delay. WMQ recorded a number of nuclear tests from eastern Kazakhstan before the Soviet Union halted testing there in 1989. IRIS archives also include a WMQ recording of one Lop Nor test, in 1988. Many other seismic stations at greater distances also recorded past Chinese nuclear explosions. Lop Nor is located near the southeastern side of the Tien Shan, a region of moderate earthquake activity and contemporary horizontal compressive stress in the earth’s crust, which proves to be of significance in the discrimination of earthquakes from nuclear explosions in that region.
Sykes and Nettles (2009) found that more than half of the total numbers of earthquakes in the Reviewed Event Bulletin (REB) of the IMS that occurred within 100 km (62 mi) of six test sites from 2000 through 2008 occurred near Lop Nor. All seismic events near Lop Nor down to magnitude 3.4, the smallest event reported in the REB, were identified as earthquakes (Figure D-6), based on the ratio of P to Lg waves at frequencies from 4 to 16 Hz (Kim et al., 2009).
This high-frequency discriminant, which utilizes data from seismic stations around Lop Nor at regional distances, is one of the most important recent advances in nuclear verification. Hartse et al. (1997) published a regional discrimination study using earthquakes from northwest China and nuclear explosions from Lop Nor and KTS, demonstrating the effectiveness of the P/S ratio for frequencies at 3 Hz and greater. A method for improving regional discrimination performance was first applied to the Hartse el al. (1997) data set by Taylor and Hartse (1998). This magnitude and distance amplitude correction procedure (MDAC) was further refined by Taylor et al. (2002) and Walter and Taylor (2001) and is now routinely used for regional discrimination processing. Figure D-7 shows discrimination plots with MDAC corrections applied to earthquakes in the Lop Nor area and four nuclear explosions recorded at station MAKZ (780 km, or 488 mi) northwest of Lop Nor in eastern Kazakhstan.
Determining accurate depths is important in identifying many seismic events as earthquakes. This can be done using the time delay between the first-arriving P waves and slightly later waves that reflect off the surface of the earth, which are commonly called “depth phases.” Seismic arrays provide a powerful tool for identifying depth phases, especially for events smaller than about magnitude 4.5.
FIGURE D-6: Measurements of amplitude ratio P to Lg waves as a function of frequency for seismic events within 100 km (62 mi) of Lop Nor test site. Known nuclear explosions (triangles, with full amplitude range marked in pink) consistently have higher values than most earthquakes (circles, with full range in yellow). Figure D-6 includes 27 earthquakes of magnitude greater than 3.4, and 6 underground nuclear explosions. SOURCE: Kim et al., 2009 (reproduced with permission of the American Geophysical Union)
FIGURE D-7: High frequency Pg/Lg and Pn/Lg discrimination plots after application of the MDAC procedure. Stars are Lop Nor nuclear explosions from the 1990s recorded at station MAKZ in Kazakhstan. Green circles are earthquakes from within 100 km (62 miles) of the test site, and yellow circles are earthquakes from other regions of northwestern China. The “error rate” is the least-squares probability of misclassification with a discriminant line midway between the two populations. SOURCE: Modified from Hartse, 2000
Centroid Moment Tensor (CMT) solutions, which utilize the entire long-period seismogram, enable an analyst to determine if an event is an earthquake, explosion, or mine collapse and to estimate its depth. During the last 10 years, CMT solutions for the entire world have been extended downward from about magnitude 5.6 to 5.0 and are available over the
Internet within a few hours. This technique has been utilized for seismic events as low as magnitudes 3.3 for the Nevada Test Site and as small as 4.0 for those near Lop Nor (Sykes and Nettles, 2009).
FIGURE D-8: Seismic events identified as earthquakes within 100 km (62 miles) of the Lop Nor test site in northwestern China as reported in Reviewed Event Bulletins (REB) of the CTBTO from 2000 through 2008. Red circles denote locations of selected large earthquakes from 1987 through 1999 from International Seismological Centre (ISC). Sites of tunnels and shafts used for nuclear tests from Waldhauser et al. (2004). Four “beach balls” in color indicate focal mechanisms (CMTs). Black numerals beside earthquakes denote their depths in km. SOURCE: Adapted from Sykes and Nettles, 2009
Figure D-8 shows seismic events located near the Lop Nor test site in the Reviewed Event Bulletin of the CTBTO IDC in Vienna. The four CMT solutions, shown as colored “beach balls,” indicate that those events were earthquakes. Each of the four is characterized by a large component of horizontal compression, as is found farther northwest along the Tien Shan in China and central Asia. In contrast, no seismic activity was located in the triangular southwestern one-third of Figure D-8 within the old and rigid crust of the Tarim basin.
All of the well-determined depths of earthquakes (numerals in Figure D-8) are between 5 and 35-60 km (between 3 and 22-38 mi), much deeper than past underground nuclear explosions at either Lop Nor or other test sites. Hence, at Lop Nor, the determination of depth is an important seismic discriminant. Most earthquakes within 100 km (62 mi) of the Nevada Test Site are shallower. They are identified better with the CMT methodology than with depth phases.
Figure D-8 also shows the locations of past underground nuclear explosions at Lop Nor. Their locations are very well known and are concentrated in two small areas.
Sykes and Nettles (2009) examined additional seismic discriminants for events near Lop Nor. Clear downward first motion of P waves at the Urumqi station for 7 of 20 events indicates that those 7, in fact, were earthquakes. Earthquakes generate downward motion at some takeoff angles from their sources and upward first motion from others. (The colored and white parts
of the four “beach balls” in Figure D-8 denote upward and downward first motions.) In contrast, an explosion produces only upward (compressional) first motion at all stations (a “beach ball” that would be one solid color). First motion was proposed in the late 1950s as an important discriminant between earthquakes and explosions but was found to be not very useful when data were available only at great (teleseismic) distances from an event. For example, distant stations typically record only upward first motions from earthquakes near Lop Nor; hence, those signals alone do not classify a seismic event near Lop Nor as either an earthquake or an explosion. The availability of data from closer regional stations, however, permits different takeoff angles to be sampled. The types of “beach balls” in Figure D-8, which result from strong horizontal compression in the earth’s crust, indicate that downward P wave motion should be observed at some regional stations, which it is. Clear downward first motion of several P waves is another instance in which data from regional stations in Asia are invaluable for identifying seismic events as earthquakes, not explosions.
A seismic event near Lop Nor of magnitude 4.3 on March 13, 2003, was recognized as difficult to identify using teleseismic methods. Special studies of this “problem event” indicate that it was, in fact, an earthquake. Five distinct techniques (Selby et al., 2005; Sykes and Nettles, 2009) confirm this (1) its high-frequency P-to-Lg ratio (Figure D-7), (2) downward first motion of the Pg wave at Urumqi, (3) depth phases, (4) its CMT mechanism, and (5) some determinations of Ms-mb. Because earthquakes will occur in the future, there will be an ongoing need to make special studies of the occasional “problem event” like this and to update policy makers as more definitive results become available.
Detection Capability and Yield Estimation
A number of studies have examined the detection and location of small events near Lop Nor. Hartse et al. (1998) examined earthquakes in western China as small as magnitude 2.5. Hartse found that the catalogs of the Chinese State Seismological Bureau from 1973 to 1989 were complete to near magnitude 2.5. Other studies using the Makanchi station and its nearby array in eastern Kazakhstan, the closest seismic monitoring of Lop Nor outside of China, had a detection capability down to magnitude 2.8. Each of those studies is more than 10 years old, and the capabilities are likely better today. The Kazakhstan National Data Centre has a web-published bulletin that reports small events within a few hundred kilometers of Lop Nor.
An appropriate formula for explosions at Lop Nor that are smaller than one kiloton is magnitude = 4.45 + log (Yield in kt). That formula is appropriate to hard rock and good transmission of P waves to distant stations and applies also to southern Novaya Zemlya, the Indian test site, and Azgir, Kazakhstan. It was derived originally for eastern Kazakhstan (see Ringdal et al., 1992; Murphy, 1996). Magnitude 2.8 translates into a yield of 0.02 kilotons (20 tons) fully coupled in hard rock. For decoupling factors of 20 and 40, the corresponding yields are 0.4 and 0.9 kt, respectively.
Seismic monitoring of the Chinese test site at Lop Nor during the past decade has improved greatly with the availability of data from stations and seismic arrays in many surrounding and nearby countries. This has permitted smaller events to be located and high-frequency techniques to be used for identification of an event as being either an earthquake or an underground explosion. Since nuclear-explosion testing stopped in 1996, many seismic events near Lop Nor can be identified as earthquakes from a variety of other discrimination techniques including depth, first-motion of P waves, regional CMT focal mechanism, and the difference in amplitudes of long-period surface waves and short-period P-waves (Ms-mb).
Unclassified information indicates that events of magnitude 2.8 and larger can be detected. This translates into a yield of about 20 tons fully coupled in hard rock. For decoupling factors of 20 and 40, the yield is about 0.4 and 0.9 kt.
Monitoring North Korea
North Korea has not signed the CTBT. North Korea originally signed the Nuclear Non-Proliferation Treaty (NPT) and ratified it in 1985. However following a number of indications and accusations that North Korea was not abiding by the NPT, North Korea withdrew from the NPT in 2003. In 2005, North Korea declared that it possessed nuclear weapons.
On October 9, 2006, North Korea announced that it had conducted a nuclear test. Seismic signals from the event registered magnitude 4.1 (IDC-REB), and radionuclides from the test were detected. On October 16, 2006, the U.S. National Director of Intelligence released a statement: “Analysis of air samples collected on October 11, 2006 detected radioactive debris which confirms that North Korea conducted an underground nuclear explosion in the vicinity of P’unggye on October 9, 2006. The explosion yield was less than a kiloton” (DNI news release, 2006).
On May 25, 2009, North Korea declared it had conducted a second nuclear test. Seismic signals registered a magnitude 4.5 (IDC-REB), but radionuclides were not detected. On June 15th, the U.S. Director of National Intelligence released a statement that “North Korea probably conducted an underground nuclear test…(t)he explosion yield was approximately a few kilotons” (DNI news release, 2009).
The number of natural earthquakes per year in North Korea is lower than in some of the surrounding regions of China, Japan, and South Korea, as shown in Figure D-9. On average, North Korea has a natural earthquake over magnitude 4 only once every few years, so the nuclear tests stood out as unusual, just because of their location and size. There are several IMS primary and auxiliary seismic stations within 1,200 km (750 mi) of North Korea. Within the same distance, additional seismic stations, which are also shown in Figure D-9, are publicly available and were used by seismic researchers to analyze the tests in the days immediately afterward. Because of their concerns about earthquake hazards, China, Japan, and South Korea each maintain additional dense networks of seismic stations not shown in Figure D-9. Although these stations are not all publicly available, researchers in these countries have used them in analysis of the North Korean tests (e.g., Hong et al., 2008).
Earthquakes in this region are routinely reported in local and global catalogs at levels of magnitude 3 and below. There are also mine blasts and industrial chemical explosions in the region, but these are usually smaller than magnitude 3.5. Mine blasts and other non-earthquake seismic events are mostly excluded from the seismic catalogs by design, as the principal use of the catalogs is for earthquake hazard.
FIGURE D-9: Map showing the location of seismic events (blue dots) and stations in the vicinity of North Korea. The seismic locations of the 2006 and 2009 North Korean declared nuclear tests are shown by overlapping red stars. Other crustal (depth < 35 km, or 22 mi) seismic events from 1999 to 2009 with magnitude > 3 from several catalogs (REB, NEIC, ISC and the Korean Institute of Geosciences and Mineral Resources [KIGAM]) are shown as blue circles. IMS primary stations (triangles), auxiliary stations (squares) and some of the other publicly available stations (pentagons) are shown in white. SOURCE: William Walter, Seismology Subcommittee member
A number of researchers noted that 2009 nuclear test seismic waveforms look very similar to seismic waveforms from the 2006 event if that test is scaled up by a factor of around 4-6 (e.g., CTBTO Press Centre, 2009; Ford et al., 2009). This similarity between waveforms has been used by a number of different research groups to get very accurate relative locations between the events, and the best indications are that the two tests were located within about two-and-a-half kilometers of one another (e.g., Wen and Long, 2010). The absolute location of the two tests is less precisely known, but because the 2009 test was recorded by a large number of stations the IMS estimate of location error is relatively small: the 90 percent confidence region is roughly circular and about 10 km (6.2 mi) in diameter (CTBTO Press Centre, 2009).
The seismic signals from the two tests clearly indicated they were explosions and not earthquakes. Primarily this was shown by the regional high frequency P/S ratios from the 2006 test (e.g., Kim and Richards, 2007; Richards and Kim, 2007; Walter et al., 2007) and by moment tensor analysis (e.g., Ford et al., 2009). The 2006 nuclear explosion was an excellent real-world test of empirical seismic methods for a sub-kiloton explosion in a new region, and the
discrimination methods worked very well as shown in Figure D-10. Because the 2009 test had such similar seismic signals to the 2006 test, the discrimination results for the two tests are very similar. One important difference is that the 2009 test generated a very small infrasound signal (e.g., Che et al., 2009). The small size of the infrasound signal compared with the seismic signal can be used to determine that this event was not a surface explosion, such as those that might occur in an open pit mine, although this event was much larger than the usual mine-related blasts.
FIGURE D-10: Example showing how ratios of P-wave to S-wave amplitudes discriminate the 2006 and 2009 nuclear tests in North Korea (red seismogram represents data from the 2006 test, and stars on the right indicate 2006 and 2009 test data) from earthquakes in the region (blue seismograms and symbols). At distances of a few hundred kilometers the Earth separates P-waves into two groups, a mantle path (Pn) and crustal path (Pg). S-waves are similarly separated into Sn and Lg. As expected, the 2006 explosion shows stronger P-waves and weaker S-waves than do nearby earthquakes. When we measure these P/S amplitudes at high frequencies (e.g., 6-8 Hz here) and correct for path effects, we get the plots shown on the right (stations TJN and MDJ averaged), showing that the explosions stand out from the earthquakes. Seismologists can statistically combine such measures to achieve excellent explosion identification capability down to very low magnitude in this region. SOURCE: Adapted from Walter et al., 2007
An unclassified discussion of the various techniques to detect, locate, and identify nuclear tests is given in Chapter 2. Interestingly, although there is separation between the North Korean tests and nearby earthquakes when using the older Ms-mb earthquake-explosion discrimination method (e.g., Bonner et al., 2008; Bowers and Selby, 2009), the separation is not as good as previous tests. This may be related to depth of burial or other emplacement conditions, which have not been released by the North Koreans. Because of this lack of information about the emplacement conditions, uncertainty also remains concerning the yields of the two tests, although (as is apparent from the DNI statements) they can be estimated approximately as less than a kiloton for the 2006 test and a few kilotons for the 2009 test. The relative location of the two tests, and the ratio of their yields, can be determined more accurately than their absolute locations and absolute yields (e.g., Kim et al., 2009).
A topic of much discussion for the 2006 and 2009 North Korean tests is why 2006 had detectable radionuclides and 2009 did not. Containment of radionuclides following a test is complex, but it is thought that it may be harder for a smaller test than for a larger one, because for the larger test it may be more straightforward to establish a “stress containment cage” that prevents the formation of cracks through which radionuclides escape to the atmosphere. An underground nuclear explosion creates a cavity that shrinks slightly from its maximum size. In this shrinking process, compressive stresses are generated that act to close down cracks—by which (if they stayed open) radionuclides would have opportunities to escape to the atmosphere. The shrinking back from the maximum helps to provide containment (OTA Report, 1989, pp. 34-35, 48).
One result the two North Korean tests made very clear is that if multiple nuclear tests occur in the same region, then relative methods of detection, location, discrimination, and yield estimation can be brought to bear on the verification problem. These relative techniques often are more precise than absolute methods, and they provide additional confidence that the verification task is made easier in a region with multiple tests.
Monitoring the Middle East, Iran and South Asia
Several countries in the Middle East and North Africa have or are considering nuclear reactors for research and/or nuclear energy generation. A subset of these countries is listed in the CTBT Annex 2; as such, they must ratify the CTBT before it can enter into force. Information on some of these countries is given in the Table D-1.
There are no declared possessors of nuclear weapons in the Middle East region. However, there are ongoing concerns over activities in Iran and Syria, and Israel maintains a policy of nuclear ambiguity. Given the expanding commercial nuclear energy facilities in the region and longstanding non-proliferation concerns, the Middle East is a major area of concern for CTBT monitoring.
TABLE D-1: Partial list of Commercial Nuclear Reactor Plans for the Middle East and North Africa.
|Country||Nuclear Reactor Plans||CTBT Annex 2 States?|
|Algeria||Plans for first commercial reactor around 2020 (two research reactors, 1989 and 1993)||Yes|
|Egypt||Announced plan in 2007 to build several reactors (two research reactors, 1958 and 1998)||Yes|
|Israel||No definite commercial plans (two research reactors, 1960 and 1962)||Yes|
|Iran||First commercial plant fueled in 2010 (several research reactors since 1967)||Yes|
|Jordan||Plans to build a nuclear power plant by 2017||No|
|Kuwait||Considering developing nuclear power||No|
|Libya||In talks with Russia to build a plant (research reactor)||No|
|Qatar||Aqreement with France for cooperation||No|
|Saudi Arabia||In talks with France to develop nuclear power||No|
|Syria||Expressed interest in 2007 (small research reactor)||No|
|Turkey||Discussions of seeking bids for four new plants (three research reactors)||Yes|
|UAE||South Korea consortium to build 4 plants 2017-2020||No|
*SOURCE: Adapted from Fineren et al., 2009; NRC, 2009; Nuclear Threat Initiative Research Library, 2010; U.S. Department of State, 1997; and World Nuclear Association, 2009.
There have been no nuclear tests in the broad Middle East region running from Egypt and Turkey in the West to Iran in the East. However, Pakistan and India, just to the east of this region, in Southwest Asia, both tested nuclear devices in 1998 (see Figure D-11). The number of natural earthquakes in this region is quite high, as shown in Figure D-11. Compared with the North Korean region shown in Figure D-9, the broad Middle East and Southwest Asia region shown below has about twenty times as many earthquakes per year. Whereas North Korea has a magnitude 4 earthquake only every few years, Iran has a magnitude 4 earthquake on average once a week. In practice, the earthquakes occur more in concentrated main-shock and aftershock sequences than constantly over time, but the overall burden to identify these seismic events as earthquakes, and not of concern under the CTBT, is significant in this region. On the other hand, the large number of natural events provides a considerable amount of data for use in calibrating this region.
FIGURE D-11: Map showing the location of seismic events and stations in the broad Middle East and Southwest Asia region. The seismic locations of the 1998 Indian and Pakistan declared nuclear tests are shown by red stars. Other crustal (depth < 35 km or 22 mi) seismic events from 1999-2009 with magnitude > 3.5 from several catalogs (REB, NEIC, ISC, KOERI, IRSC) are shown as blue circles. IMS primary stations (triangles), auxiliary stations (squares), and some of the other publicly available stations (pentagons) are shown in white. SOURCE: William Walter, Seismology Subcommittee member
The large number of earthquakes reflects the active tectonics of this region. The Indian, Arabian, and African tectonic plates are all being driven northward into Eurasia, pushing up mountains and causing large earthquakes. The significant earthquake hazard in this region has led many countries to install large numbers (hundreds) of seismic stations to monitor and study the situation. Turkey, Israel, Saudi Arabia, Azerbaijan, and Iran, among others, maintain dense seismic networks and although only some of these data are available via publically accessible websites, they are not restricted and can often be acquired by scientists who know how to request them. There are a number of existing and planned IMS stations in the region shown in Figure D-11. Several, however, have not yet been built (e.g., Egypt, Saudi Arabia), and ones in Iran that have been built and certified by the IMS are not sending data. Iran’s stated position is that it will not send data until entry into force (EIF) of the CTBT. Pakistan and India have not signed the CTBT and are not actively participating in the work of the CTBTO at this time.
The active tectonics and alterations of the earth’s crust and upper mantle complicate the propagation of seismic waves in this region. The geological activity is associated with the rapid attenuation of seismic signals for some paths, raising detection thresholds. The structural complications also degrade the location accuracy of seismic events and explosion identification unless they are accounted for. For these reasons, the Middle East has been the focus of much seismic research and development effort aimed at using the large number of earthquakes to calibrate the region.
Seismic location can be improved by taking advantage of past earthquakes that have very good location accuracy based on a local network, which may have been temporarily
deployed after a large event. Seismologists have defined criteria for how well these reference events (also called “ground truth” events) are located (e.g., Bondar et al., 2004). When good ground truth reference events are available, statistical techniques such as “kriging” can be used to locate nearby events very accurately (e.g., Schultz et al., 1998). For example, reference events have been used to develop source-specific station corrections (SSSCs) that account for the variation in Earth structure and improve location accuracy (e.g., Richards et al., 2003).
To cover broad areas where such ground truth reference events might not be available, tomographic methods are used. As in medical tomography where x-rays are used to image inside the human body, seismic waves are used to image inside the earth. As seismic waves travel through complex geology their speed and amplitude are altered. By examining many paths crossing through a region one can quantify these variations and correct for their effects on observed signals, for purposes of using these waveforms to obtain improved information on the location and size of seismic sources.
A very large tomographic study of Eurasia using about 600,000 regional P-wave (“Pn” phase) paths to improve event locations has been completed by the national laboratories and USNDC staff (Myers et al., 2010). Based on tests using well-located reference events and regional Pn arrivals at four stations, this model improves the median location error from about 32 km (20 mi) using the best 1-D Earth model (ak135) to about 15 km (9 mi). Four stations are close to the minimum number needed for location and so this represents a small event near the monitoring limits. With additional station detections the location error can be reduced further. Through this model can be applied anywhere in Eurasia, it is particularly useful in a geologically complex region such as the Middle East and in places where more accurate SSSCs are not available.
Identification of explosions in the Middle East and Southwest Asia can be accomplished by the usual methods of Ms-mb, depth, moment tensor analysis, and regional high frequency P/S ratios. For example in Figure D-12, the May 11, 1998, Indian nuclear test (shown in red) is easily distinguished by its large P/S amplitude ratio compared to a nearby earthquake (shown in green) at station NIL (an IRIS GSN publicly available station), located very close to the planned primary IMS seismic station in Pakistan.
FIGURE D-12: The 1998 Indian nuclear test seismogram (red) compared with two earthquakes at station NIL in Pakistan. The nearby earthquake (green) shows the characteristic small P-waves at the beginning and large S-waves in the middle of the seismogram. These P/S differences are used to discriminate explosions from earthquakes. An earthquake in Kyrgyzstan to the north is filtered by the Earth as it travels to the station so that it ends up looking very similar to the explosion. By accounting for these path effects using amplitude tomography as shown in the upper right in terms of Qs (Pasyanos et al., 2009), these path effects can be removed. In the bottom plots, P/Lg ratios (Lg is a type of S-wave) for many earthquakes all over the region are compared and after correcting for the path effects, the explosion (red star) stands out and is discriminated from all the earthquakes (blue circles), including both the nearby Indian earthquake (green circle) and more distant Kyrgyz earthquake (light blue circle). SOURCE: Adapted from Pasyanos and Walter, 2009
However, there may not always be a good reference earthquake with which to compare a new unknown event, and the complex Earth structure in some regions can make an earthquake seismogram look like an explosion. For example in Figure D-12 the Earth has filtered the seismogram of an Kyrgyz earthquake, shown in blue, at NIL in such a way as to resemble the 1998 explosion seismogram. By taking advantage of the many previous earthquakes recorded at the many tens of available seismic stations deployed for earthquake hazard purposes, we can map out these path effects and correct for them using tomographic methods. The result, as shown in Figure D-12, is the ability to quickly identify explosions using their individual regional seismic signals in a large and tectonically complex area, without the necessity of having a good nearby reference event in each case.
In addition to tomography, the amplitudes of nearby events can be used via methods such as kriging to predict the amplitude variations due to propagation, and such corrections have been used to improve P/S discriminations (see, e.g., Bottone et al., 2002). In practice, kriged corrections from past events can be combined with tomographic corrections to get even better corrections (see, e.g., Pasyanos, 2000). Such path corrections are important when identifying explosions and discriminating them from earthquakes in a complex region like the Middle East and southwest Asia.
Another way to take advantage of the large number of seismic events in the Middle East is to use waveform correlation. Tests in California and China have found that a very large number of the earthquakes correlate with each other (e.g. Schaff, 2009). In this methodology, the waveforms of all events are compared and clusters are created of very similar waveforms. It can be shown that seismic waveforms that match each other with a high degree of correlation must be located close to each other and have similar sources (see, e.g., Schaff, 2008). So once a new event is correlated with a well-characterized cluster, the event can be detected, located and identified through the correlation process. These correlation techniques are also particularly useful for detecting and identifying mine blasts (see, e.g., Harris, 1991).
This page intentionally left blank.