Text and Graphics of PAGEOPH paper

Abstract

Official Russian sources in 1996 and 1997 have stated that 340 underground nuclear tests (UNTs) were conducted during 1961–1989 at the Semipalatinsk Test Site (STS) in Eastern Kazakhstan. Only 271 of these nuclear tests appear to have been described with well-determined origin time, coordinates and magnitudes in the openly available technical literature. Thus, good open documentation has been lacking for 69 UNTs at STS.

The main goal of our study was to provide detections, estimates of origin time and location, and magnitudes, for as many of these previously undocumented events as possible. We used data from temporary and permanent seismographic stations in the former USSR at distances from 500 km to about 1500 km from STS. As a result, we have been able to assign magnitude for eight previously located UNTs whose magnitude was not previously known. For 31 UNTs, we have estimated origin time and assigned magnitude – and for 19 of these 31 have obtained locations based on seismic signals. Of the remaining 30 poorly documented UNTs, 15 had announced yields that were less than one ton, and 13 occurred simultaneously with another test which was detected. There are only two UNTs, for which the announced yield exceeds one ton and we have been unable to find seismic signals.

Most of newly detected and located events were sub-kiloton. Their magnitudes range from 2.7 up to 5.1 (a multi-kiloton event on 1965 Feb 4 that was often obscured at teleseismic stations by signals from an earthquake swarm in the Aleutians).

For 17 small UNTs at STS, we compare the locations (with their uncertainties) that we had earlier determined in 1994 from analysis of regional seismic waves, with ground truth information obtained in 1998. The average error of the seismically-determined locations is only about 5 km. The ground truth location is almost always within the predicted small uncertainty of the seismically-determined location.

Seismically-determined yield estimates are in good agreement with the announced total annual yield of nuclear tests, for each year from 1964 to 1989 at Semipalatinsk.

We also report the origin time, location, and seismic magnitude of 29 chemical explosions and a few earthquakes on or near STS during the years 1961 – 1989.

Our new documentation of STS explosions is important for evaluating the detection, location, and identification capabilities of teleseismic and regional arrays and stations; and how these capabilities have changed with time.

Introduction

It has been reported in recent official Russian publications (Mikhailov et al., 1996; USSR Nuclear Tests, 1997) that a total of 340 underground nuclear tests (UNTs) were carried out on the Semipalatinsk Test Site (STS) from 1961 to 1989. Only 279 of them had been included in previously published lists of Soviet underground nuclear explosions that included purportedly accurate origin times and locations (specifically, the lists contained in Bocharov et al., 1989, Ringdal et al., 1992; and Lilwall and Farthing, 1990). For eight of these 279 explosions, the magnitudes have not been available. So accurate epicenter parameters of 61 UNEs and the magnitudes of 69 UNEs appear not to have been given previously for this test site (STS).

The main goal of this paper is to estimate the origin time and location, and to assign the magnitude, for as many of the 69 hitherto undocumented UNTs at STS as possible. Our analysis is based principally upon seismic observations using regional stations located in Kazakhstan and elsewhere in Central Asia and southwest Siberia. We also evaluate the accuracy of locations for small UNTs at STS, as determined from regional seismic signals.

Besides underground nuclear explosions, our paper includes information about chemical explosions and earthquakes which have been detected on and near STS. Their parameters also were obtained from data of regional stations, and in some cases teleseismically. It is of interest, that some of these earthquakes and chemical explosions were included in some lists of Soviet underground nuclear explosions published in the West in the mid 1980s before Russian announcements about Soviet nuclear explosions were made beginning in 1992.

It is important to develop thorough documentation of all nuclear explosions, and especially for small explosions, as an aid in evaluating the detection and identification capability of monitoring stations. Of course, explosion monitoring in the present and the future will typically be done using stations that differ from those we have used to document small explosions at STS. Nevertheless our database of small explosions (chemical and nuclear), and nearby earthquakes, can provide guidance in estimating the capability of current networks, which can be expected to be better than the capability that was available for much of the period of active nuclear testing.

We shall use the distinction between a nuclear test and a nuclear explosion that was adopted in the revised protocol of 1990 for the Threshold Test Ban Treaty. Thus, a single underground nuclear test (UNT) can consist of a number of different underground nuclear explosions (UNEs) provided these are carried out within a time interval not exceeding 0.1 s and within an area delineated by a circle whose diameter is less than 2 km. Explosions with a time interval longer than 0.1 s, or a distance greater than 2 km, are counted as separate tests. We note that this distinction between UNTs and UNEs has been followed in official Russian documentation of the Soviet test program at STS, but with one exception, namely the nuclear test which was conducted at Degelen in a tunnel on January 30, 1974. (It consisted of three separate explosions. Western lists typically have given two tests separated by 4.4 seconds, but the official Russian publication states, in translation, that "This test can be classified as one test with several subexplosions, however the time difference between subexplosions was more than 0.1 sec" and we have followed the Russian listing in counting this as one UNT.)

In the following sections, first we summarize the information from Russia officially available on STS UNTs. Second we describe the seismic data and methods of analysis that we have used in the study of small events. Third we describe our regional seismic detections of small events on and near STS, and their locations and magnitudes. Fourth we discuss the accuracy of their seismically-determined locations by making comparisons with ground truth information given by Leith (1998). Fifth we describe the agreement between seismically-estimated yields, and announced yield information; and we comment on detection thresholds for a small number of different networks.

1. Summary of available official information about UNTs from STS

1.1. General information about all UNTs from STS

The boundaries of the Semipalatinsk Test Site (STS) were defined and communicated to the US by the USSR in 1990 upon entry into force of the Threshold Test Ban Treaty. The information the USSR gave the US (now filed in the library of the US Department of State) consists of a set of 152 marker locations (latitude, longitude) on the perimeter of this test site, together with a map showing the boundary line from one marker to the next. (Go to detailed description of test site boundaries.) As indicated in Figure 1, STS is about 100 km from east to west, and 150 km from north to south.

The 340 UNTs at STS listed by Mikhailov et al. (1996) and USSR Nuclear Tests (1997) were each associated with one of three sub-areas of the test site. Thus, 209 UNTs were in the Degelen sub-area, 105 at Balapan (sometimes referred to as Shagan), and 26 at Murzhik (sometimes referred to as Konystan). These explosions covered a wide range in yield, from less than 1 ton up to to 165 kilotons (kt). Among 96 UNTs with magnitude mb mb less than 5.0, 84 were at Degelen, only 7 at Balapan, and 5 at Murzhik, so about 88% of the smaller yield events were in tunnels at Degelen.

The origin time and coordinates (latitude, longitude and depth) of STS UNTs have so far been announced by Soviet/Russian sources only for a group of 96 events that were conducted during a period from October 1961 to December 1972 (Bocharov et al., 1989; see also Vergino, 1989). Among these 96 UNTs, 6 were small and were not mentioned by Lilwall and Farthing (1990), or by Ringdal et al. (1992). With the official Russian announcements of 1996 – 1997 it became clear that 20 small magnitude tests at STS during this 1961 – 1972 time period had not been included in the list of Bocharov et al. (1989).

In 1992, the Russian Federation declassified information about the dates on which Soviet UNTs had occurred, the number of tests and number of nuclear explosions carried out within one nuclear test, the yield range, the sub-area, and the purpose of these UNEs. However, the origin time, coordinates, and yield of most Soviet UNEs are still unavailable; and their seismic magnitudes, as determined from Soviet seismographic networks have not been announced, either by the network operated by scientists, known as ESSN, or by the military network, known as SSK.

A preliminary list of Soviet nuclear explosions was published by Gorin et al. (1994) in a scientific journal. Two official documents, Mikhailov et al. (1996) and USSR Nuclear Tests (1997), contain information about the date, name of the shaft or tunnel, purpose and yield range and identification numbers (from #1 to #715) for all 715 UNTs carried out by the Soviets, including for 340 UNTs conducted at the STS. In these latter two publications, yields were given for nuclear explosions conducted off the two weapons test sites at Novaya Zemlya and Semipalatinsk. Yields were also given for 27 UNTs at STS. For each of the remaining 313 STS UNTs, one of three yield ranges is given for the test: either, less than 1 ton (of TNT equivalent); or, from 1 ton to 20 kt; or, from 20 kt to 150 kt. Yields at STS are announced as greater than 150 kt, for only two tests: November 2, 1972 , with yield 165 kt and mb 6.16; and July 23, 1973, with yield range 150 – 1,500 kt and mb 6.17. These magnitudes were assigned by Ringdal et al. (1992) using the procedure of Lilwall et al. (1988) developed by the British Atomic Weapons Establishment, and we refer later to such magnitudes as mb(AWE). Both these tests took place prior to the date (March 31, 1976) given in the Threshold Test Ban Treaty, negotiated in 1974, for imposition of a 150 kt threshold.

Apart from the explosion locations given by Bocharov et al. (1989), Soviet and Russian publications have not listed UNT coordinates. But within the framework of Kazakhstan – US cooperation, coordinates of tunnel portals at Degelen have become available (Leith, 1998). Separately, the coordinates of Balapan shafts have been obtained through fieldwork conducted by the National Nuclear Centre of the Republic of Kazakhstan, and these locations are also now available (NNCRK, 1999). We use ground truth information in Tables below whenever these locations are available for specific explosions. When ground truth is absent (for example for chemical explosions) we give coordinates determined by seismological methods – which are shown to be quite accurate in a later section of this paper.

1.2. Small UNTs at STS previously undocumented by Western seismologists in the open literature

In this section we identify 61 UNTs at STS, out of the 340 now officially announced, for which accurate location information has not been given in openly available publications so far as we are aware; and we identify 69 for which accurate magnitude information has not been given. We assign each of these 69 UNTs to a category that indicates why their documentation has been poor (for example, low yield, or occurrence at the same time as another UNT). The following sections then report our own efforts to acquire and generate additional information, including locations and magnitudes, for as many of these 69 UNTs as possible.

Thus, the International Seismological Centre (ISC) has reported the seismically-determined location and magnitudes of 271 UNTs at STS. Many researchers have carried out additional levels of analysis based upon ISC data for subsets of the STS events listed by the ISC. One of the largest such efforts, by the British Atomic Weapons Establishment (AWE), has applied the Joint Epicenter Determination method described by Douglas (1967) to ISC data, using several UNTs at STS as master events for which ground truth information was given by Bocharov et al. (1989) (and in English translation by Vergino, 1989). The AWE location estimates are given by Lilwall and Farthing (1990). AWE has also obtained maximum likelihood mb's for 239 STS UNTs and has made them widely available on an informal basis. AWE mb's for 100 UNTs and one chemical explosion in the Balapan sub-area were published by Ringdal et al. (1992). An additional 8 UNTs, not mentioned by Lilwall and Farthing (1990) or Ringdal et al. (1992), are given with locations but not magnitudes by Bocharov et al. (1989) and Vergino (1989). On the basis of comparisons with SPOT locations (e.g., Thurber et al., 1993) and recently available ground truth information, we believe the AWE locations for 271 UNTs at STS, based on re-analysis of ISC data, are accurate to within a few km. In Figure 1, we show a map of the Semipalatinsk Test Site boundaries (as reported by the Soviet Union at the time of TTBT entry-into-force in 1990), together with the locations of 279 UNTs with coordinates given by the publications cited in this paragraph.

We can give three reasons why specific UNTs at STS were not included in the ISC and subsequent AWE listings. First, some UNTs have now been announced as having had yield less than 1 ton; such tests would generally be too small for either regional or teleseismic detection. Second, some UNTs were carried out at essentially the same time as another UNT and only one test was reported. Third, some UNTs have now been announced as having had yield greater than 1 ton, but they may still have been too weak for teleseismic detection with high confidence, given the networks in operation at the time. For these events, we can inquire as to the possibility of regional detection as discussed in the following sections. [Note that the papers by Lilwall and Farthing (1990), and Ringdal et al. (1992) were not intended as reports on teleseismic detection capability, and they characterized UNTs only for which the teleseismic data was of high quality. We comment below on papers by Sykes and Ruggi (1986, 1989) and Ringdal (1990) which reported the occurrence of several UNTs at STS additional to those listed by the ISC. Such detections were useful and important but they were not associated with accurate location estimates so we continue to include them in this paper with what we call "previously undocumented" events.]

Using information from the official Russian publications, we can tentatively see for each of these three reasons how many UNTs were not included in lists of events accurately located by seismic methods:

1.2.1. Weak UNTs with yield Y announced as less than 1 ton

This category consists of the 15 UNTs listed in Table 1. They would not be detected by standard instruments at distances more than 100 – 150 km. One of these small UNTs, with yield Y less than 1 ton, was carried out at Balapan (1973 Sep 20); the other 14 were carried out in the Degelen sub area.

1.2.2. Pairs of UNTs exploded simultaneously

This category is concerned with pairs of tests carried out within a short time interval, or even simultaneously, but with a spatial and/or temporal interval that requires them to be listed as different tests. In order to discuss specific UNTs with the same date, we use the number for each test that appears on official Russian lists.

Thus, we note that in the official lists there are 19 pairs exploded on the same day. Two of them were on the same day but are known to be separated by a long time interval: # 414 and # 415 (December 16, 1974, Degelen) with more then three hours time interval; and # 440 and # 441 (April 21, 1976, at Degelen and Balapan and hence with a significant spatial separation) with a four minutes interval.

For the remaining 17 pairs of UNTs, only for four pairs were both tests detected and reported in the standard western publications. These tests were conducted on Dec 10, 1972 (# 376 and # 377, time interval 10 sec), on Oct 29, 1977 (# 473 and # 474, time interval 4.9 sec), on Aug 29, 1978 (# 493 and # 494, time interval 8.8 sec) and Nov 29, 1978 (# 506 and # 507, time interval 4.8 sec). In all four cases, the two tests were carried out in different subareas (Balapan or Degelen) at separations of 50 km and more, which significantly facilitated their identification as test pairs.

The last 13 pairs of UNTs are shown in Table 2. Among them are 12 pairs for which just one UNT (for each pair) was reported by ISC; and one pair (Dec 5, 1980, # 561 and # 562) for which neither test was reported. In the last column of Table 2 are the numbers of the 14 UNTs unreported by the ISC. Only for one pair of UNTs (1980 Dec 5) were both tests unreported by the ISC, so one event from this pair potentially can be detected. The remaining 13 UNTs could not be detected by standard methods.

1.2.3. Small UNTs with yield more than 1 ton, not reported by the ISC

These events, listed in Table 3, are most interesting for us because potentially they can be detected at regional distances. They are the main object of our investigation.

We note that a total of 61 previously undocumented UNTs are given in Tables 1, 2, and 3, namely, 15 UNTs with yield less than 1 ton; 13 UNTs that occurred at essentially the same time as another UNT, including one of the pair # 561 and # 562; and 33 UNTs which potentially can be detected at regional distances, including the other of the pair # 561 and # 562. In section 3 below, we report our locations and magnitudes, based on regional detections, for most of these 33 UNTs.

2. Regional data and methods of analysis for small-magnitude seismic events from STS

Our work on this subject was carried out in two stages. The first, in late 1993, resulted in a technical report (Khalturin et al., 1994), produced prior to the publication of the first Russian preliminary list of Soviet UNTs (Gorin et al., 1994). Information about the dates on which UNTs occurred, and ground truth locations, were not then available for us. We used regional data, and tried to detect and locate all seismic events at STS which could be UNTs, chemical explosions or earthquakes. The second stage was carried out in 1997 – 1999, in light of official information on UNT dates and acquisition of ground truth locations. Naturally, the magnitude threshold of detected signals in the first stage was significantly higher then in the second stage.

2.1. System of observations. Stations and instrumentation

Our results are based mainly on seismic data acquired by the Complex Seismological Expedition (CSE) of the Institute of the Physics of the Earth, Russian Academy of Sciences. Also we used bulletins of other regional stations of Central Asia including the Altai region. In total we used the records or bulletins of more than 50 seismographic stations (station information). Most of them were operated on a temporary basis, and typically we used data from about 3 to 6 stations for each event to estimate its location. Most useful for detecting and locating small magnitude UNTs, were seismograms of narrow-band short period instruments installed in several stations by CSE in North Kazakhstan at distances of 500 km to 1200 km from STS. Long-term CSE observations in this region show that high-frequency regional phases propagate very efficiently. Also very helpful for detection and discrimination were records of "multichannel frequency selected stations", known from their Russian acronym as "ChISS", installed at the base stations Zerenda, Talgar, Novosibirsk (all in the distance range 740 – 780 km from STS) and Garm (1350 km from STS). These stations included the ChISS set of 8 or 12 channels from 0.5 to 45 hz or from 0.025 to 45 hz. Base stations also operated instruments as characterized in Table 4 and Figure 2.

Two different types of high-gain short-period instruments – RVZT and CSE (see Table 4 and Figure 2) – were developed for detection of low-magnitude UNEs. Several places were found (the best ones were on the Kokchetav massif in Northern Kazakhstan) where magnification could be set as high as one million, with noise amplitudes only about 1 mm on the paper record, even though the attenuation of regional waves was very low. Figure 3 gives an example of such a high gain record, for an event not previously documented by standard western publications. It is the UNT of January 29, 1971. This signal was recorded by a CSE instrument located near the site of the present-day Zerenda broadband station (ZRN) on the Kokchetav massif, at a distance of 720 km with S/N ratio about 150 – 200.

ChISS instruments were installed in Talgar and Garm in September 1961, prior to the first UNT conducted at STS. Later, ChISS instruments were installed by the Complex Seismological Expedition (CSE) in Zerenda and Novosibirsk. They operated until the end of 1990. Figure 4 gives a ChISS record of the first Soviet underground nuclear explosion, at STS on October 11, 1961. These historic data are of remarkably high quality. ChISS records such as these are very effective even today for quantifying frequency-dependent features of regional seismic wave propagation. After 1969, ChISS instruments with ink-pen recording were installed by CSE for military seismic surveys in several places including Semipalatinsk, Makanchi (East Kazakhstan), Mongolia, Mayli-Say (East Fergana), and Malin (Northern Ukraine). The main goal was detection and prompt discrimination of the signals from underground and atmospheric nuclear explosions from non-Soviet test sites (Chinese nuclear testing in the atmosphere continued up to 1980). Discrimination was based on measurement of P-wave amplitudes on high frequency channels, and S, Lg, and Rg on low frequency channels. Figure 5 shows regional phases on the ChISS record at Talgar of a Degelen UNT, mb = 5.08. On the different narrowband channels the calibration signal can be clearly seen, with frequency changing slowly from 40 – 45 hz down to 0.3 – 0.5 hz and thus appearing on different channels at different times.

2.2. Regional phases observed from UNEs at STS

The following regional phases are observed from the Semipalatinsk underground nuclear tests at distances up to 1200 km:

•Pg (6.15 km/s), Sg (3.55 km/s) and Rg (2.6 – 2.8 km/s) observed up to 230 – 250 km;

•Pn, appearing beyond 230 – 250 km as the first arrival with velocity 8.1 km/s. Its velocity stays constant up to distances of 800 – 900 km and then starts to increase slightly. Following Pn out to 800 – 900 km, Pg is observed with velocity 6.2 km/s.

•Sn is also observed (4.7 km/s) beyond about 240 km, but it is weak and can be clearly detected only on 60-70% of the records.

•A very intensive Lg group is observed beyond about 240 km with an impulsive onset. Usually the amplitude of Lg is 4 to 8 times bigger than amplitudes of Pn or Pg. Beyond 400 – 500 km the Lg wave train consists of two groups, denoted as Lg₁ and Lg₂, with velocities 3.55 and 3.40 km/s.

•An intensive Rg wave train is observed for UNEs with magnitudes greater than 5 on the long-period records. The train is short and consists of 1.5 to 2 cycles without clear dispersion. The velocity of the apparent first arrival is 3.0 km/s for periods 7 – 10 s.

Figure 5, showing five ChISS channels recorded at distance 730 km for a Degelen UNT of magnitude 5.08, indicates that the phases Pg, Pn (marked as P1), Sn, Lg1 and Lg2 can be clearly seen, often with impulsive arrivals.

Khalturin et al. (1994) describe how these regional travel time – distance relations were used to estimate the location and location accuracy of small seismic events at STS.

2.3. Magnitude determination

Two methods were used to assign magnitudes (mb) from regional data. The first method was based on the K scale (energy class), which is still used in the former Soviet Union by analysts at all local networks, for characterizing the size of seismic events using data acquired at distances up to 3000 km. The energy class K is calculated from the sum of max P amplitude, and max S or Lg amplitude, on records of the short-period instrument SKM. For UNEs and chemical explosions at STS, Khalturin et al. (1998) report the following relationship between K values, and mb values as given by the British Atomic Weapons Establishment and by NORSAR:

mb(K) = 0.46 K – 0.64.

We used this relationship in the present study, to assign mb values when the K value was available.

The second method to assign mb was based on calibration of a measurement of the maximum amplitude of Lg waves. This scale was typically applied to the narrow-band records. The resulting magnitude is denoted as mb(Lg).

3. Detection of small events from STS from regional recordings in Kazakhstan and Central Asia

3.1. Monitoring of seismic signals from STS: detection and identification

The seismographic network operated by CSE was used to acquire observations in the Kazakhstan region throughout the period of UNT activity at STS – from 1961 to 1989. During the long-term monitoring effort, besides the well-known intermediate and large magnitude UNTs from STS, several tens of small magnitude events were detected that were not mentioned by Lilwall and Farthing (1990), or Ringdal et al. (1992). These events can be UNTs, or they could be chemical explosions used for military experiments and for construction. Few of these signals can be from earthquakes, which are very rare in the Semipalatinsk region since it is located on the far western flank of the Altai seismic zone.

Our first stage of study (Khalturin et al., 1994) examined data for 57 of these events that were on or near STS; estimated their coordinates, origin time, and magnitude; and made a preliminary identification as to the nature of each event (nuclear or chemical explosion, or earthquake). We now know that these 57 events consisted of 19 UNTs, 27 chemical explosions, 8 small magnitude UNTs known from Bocharov et al. (1989), and three earthquakes. Our first stage identified all of the UNTs and earthquakes correctly, but wrongly listed two of the chemical explosions as UNTs, and two other chemical explosions as "either UNE or chemical explosion".

Our second stage of study, in this paper, done following the release of UNT date information, has examined data for 71 events on or near STS, and has resulted in estimates of the origin time and magnitude of an additional 12 small UNTs which were missed in the first stage. So, from the 33 previously undocumented UNTs of Table 3, our first stage of study uncovered 19 UNTs and the present paper documents another 12. The present paper also includes two more chemical explosions than our first study: that of September 15, 1984 (previously listed by some papers as a UNT), and the chemical explosion carried out in 1987 as part of a Joint US-Soviet Experiment of the USSR Academy of Sciences and the US Natural Resources Defense Council (see Given et al., 1990).

3.1.1. Earthquakes near the Semipalatinsk Test Site

Among the 71 regionally-recorded events were three earthquakes, given in Table 5, which occurred near the border of STS or in the surrounding area. This Table also lists three more recent earthquakes in the region.

In Table 5, event #1 was described by Khalturin et al. (1994). Event #2 was widely detected, and it is still sometimes listed as a UNT though the seismological basis for identifying it as an earthquake was given several years ago (Pooley et al., 1983). Event #3 was reported by the ISC (47.9N 83.5E, mb 4.5) and listed by Sykes and Ruggi (1989) as a Soviet UNE at (50N, 79E) with mb 3.6. Events ##4 – 6 occurred near STS after the period of our study. Event #5 was reported in the US Geological Survey's Preliminary Determination of Epicenters. Events #4 and 6 have been reported by the Altay-Sayan Seismological Expedition, based in Novosibirsk.

3.1.2. Chemical explosions at the Semipalatinsk test Site

The remaining 60 regionally-recorded events have all the characteristics of explosions. Comparing their origin times with the list of UNT dates contained in Russian official publications issued in 1996 – 1997 we now conclude that 29 were chemical explosions at STS, and 31 were indeed UNTs at STS. Parameters of the 29 chemical explosions are given in Table 6. More than half of these chemical explosions were assumed in some publications during the late 1980s to be UNTs from STS. It is of course appropriate to list all candidate UNEs in projects that set out to evaluate monitoring capability.

In Table 6, origin time, coordinates and K values were obtained from CSE observations; the magnitude value m(NOR) refers to magnitudes obtained from F. Ringdal [personal communication, 1994]; magnitudes in parentheses ( ) were assigned by Hagfors (HFS) and are known typically to be significantly higher than m(NOR) values; and mb(K) is the body wave magnitude calculated from energy class K using the relationship mb(K) = 0.46 K – 0.64.

Notes on Table 6:

A – Fully contained explosion with a yield of 600 tons of TNT, carried out in a Degelen Mountain tunnel prior to the start of nuclear testing (Sultanov et al. 1995). Goals included calibration of a seismic network and estimation of the expected seismic signal strength at different distances.

B – Large experimental chemical explosion on the surface with a yield of 5000 tons.

C – This chemical explosion has been wrongly listed as an underground nuclear test, in some cases with mb = 5.04; for example see Lilwall and Farthing (1990), and Ringdal, Marshall, and Alewine (1992). For this chemical explosion, mb(ISC) = 4.7, mb(HFS) = 5.2.

D – These two experimental chemical explosions, each of 500 tons of TNT, were conducted by the Institute of Dynamics of the Geospheres (IDG) at the same place on the surface of Degelen Mountain near the mouth of tunnel #160 (49.7841°N, 77.96722°E). See Adushkin et al. (1997).

E – Chemical explosion (yield 20 tons, depth 25 m) carried out in the Degelen sub-area during the Joint US-Soviet Experiment of the USSR Academy of Sciences and the US Natural Resources Defense Council (see Given et al., 1990).

F – Hansen et al. (1990) assumed this chemical explosion was an underground nuclear test in their study of the stability of RMS Lg.

SR – Sykes and Ruggi (1986, 1989) list all these 15 chemical explosions as UNTs from STS with the following mb values:

Some if not all of these magnitudes are from the Hagfors Observatory (HFS). On average HFS magnitudes are larger than NORSAR magnitudes by about 0.6 — 0.8 magnitude units.

A very weak regional signal originating on October 20, 1989 at 13:22:45 from about (50°N, 78°E) may be a chemical explosion or a collapse of the cavity from a previous UNE.

More recently than the period of our study, chemical explosions ranging in size from a few kilograms up to 100 tons have been carried out in Kazakhstan in a cooperative program between the National Nuclear Centre of the Republic of Kazakhstan and the US Defense Threat Reduction Agency. Times and locations of such explosions larger than one ton are listed in Table 7. Both calibration explosions listed with Y = 100 tons were carried out at Degelen in the tunnels. Their mb values (3.7 - 3.8) correspond to the upper limit of mb(Y) relationship described in Khalturin at al. (1998). All other explosions were made in Balapan in shafts.

3.2. Detection and analysis of small UNTs from STS

Here we give the origin times, locations and magnitudes of small UNTs whose detection and or location was not previously well-documented by western seismologists in the open literature. We also give magnitude values based on regional detections, for 8 small events whose hypocenter coordinates have been available from Bocharov et al. (1989).

3.2.1. Main parameters of previously undocumented small UNTs

For the 33 small UNTs listed in Table 3, we have found regional seismic detections at temporary and permanent stations of CSE for all but two events. Parameters for the 33 events are listed in Table 8. Origin times were estimated for all 31 of them. For the 19 largest of these small events, we obtained location estimates based on seismic signals, and K values and hence mb(K). For the 12 smallest events we give estimated mb(Lg) values – which range from 2.2 to 3.7.

During the course of our work, ground truth information on locations became available from Leith (1998) and NNCRK (1999) for 30 of the UNTs, and in section 4 we discuss the accuracy of our seismically-estimated locations. Thus, origin time, K and mb(Lg) of all UNTs, and coordinates of the 1976 Aug 4 UNE, are our estimates based on regional observations of CSE. Coordinates of all other UNTs are ground truth values.

One relatively large UNT, with mb(K) 5.1 (February 4, 1965), was not reported by standard western publications as it was obscured teleseismically by a swarm of Aleutian earthquakes. We can be sure this was a coincidence rather than an effort to obscure the event, because the origin time (06:00:00) was typical for UNTs of the mid-1960s. But even if we exclude this large event, the mb value (calculated from K) for missed events ranges up to 4.55, and during 1964 – 1989, about 10 Soviet UNTs at STS, with magnitude 4.0 or more, had teleseismic signals that were too weak or too noisy to lead to publication of good location estimates. Some of these events were detected teleseismically at particular arrays (Ringdal, 1990).

Notes on Table 8:

mb(K) – calculation of mb from K using the relationship: mb(K) = 0.46 K — 0.64.

m(NOR) – from F. Ringdal (pers. comm., 1994), based on teleseismic signals at NORSAR.

A – This event was obscured by many Aleutian earthquakes, up to mb 6.4 on that day.

B – The yield of this explosion has been announced as 0.23 kt (USSR Nuclear Tests, 1997)

C – This event was mentioned by Ringdal (1990).

SR – These three events are listed by Sykes and Ruggi (1986, 1989) with the following coordinates and magnitudes mb:

1. 55.8N and 75.1E; mb = 4.6.

2. 49.9N and 77.7E; mb = 4.1.

3. 50.0N and 78.0E; mb = 4.0.

3.2.2. Magnitude estimation of small known UNTs

Among the analysed signals were 8 small UNEs known from Bocharov et al. (1989) but listed there without magnitudes. Four of these events had been reported using teleseismic signals by Sykes and Ruggi (1986, 1989), who also listed a magnitude for three of the events.

For these 8 events the energy class K is known from regional records at several stations, allowing us to give values of mb(K) which we list in Table 9. The other mb values are from Sykes ad Ruggi (1989).

Notes on Table 9:

A - the first underground nuclear explosion conducted by the USSR was not included in many western lists of USSR explosions prior to publication of Bocharov et al. (1989), but it was reported (without coordinates and magnitude) by Bolt (1976), and by Sykes and Ruggi (1986, 1989).

B - detection and approximate location reported by Sykes and Ruggi (1986, 1989):

1 - no magnitude estimation;

2 - wrong time (03:05:00);

3 - coordinates with 150 km error.

4 - coordinates with 200 km error.

C - These UNTs were not published in western lists of USSR explosions prior to publication of Bocharov et al. (1989).

4. Comparison of ground truth and seismologically-determined locations of small magnitude UNTs from STS

Our earlier study (Khalturin et al. 1994) determined coordinates of 18 small-magnitude UNTs (one additional UNT was detected only by one station) on the basis of arrival times of regional waves, and estimated the location uncertainty, which was typically an area of the order of 100 km2. One of those UNTs was at Murzhik for which we do not yet know the ground truth coordinates. Thus for 17 UNTs (magnitudes 3.8-4.6), which occurred at Degelen, we can now compare the seismically-located epicenters with ground truth recently obtained for that sub-area by Leith (1998).

For these small UNTs, regional signals were acquired at CSE stations located at distances in the range 500–1400 km from STS. We mostly used data from bulletins but did read waveforms ourselves in some cases. Thus for these 17 UNTs we had 20 records and 49 station bulletin data from stations to the south; and 37 station bulletin data from stations to the east or west. So on average for the location of one event we had about one record and about three pieces of data from station bulletins located to the south of STS, and about 2 data from stations located to the east or west. On each record, 2–3 regional phases were measured (typically Pn, Sn, Lg). To obtain a preliminary estimate of location and origin time, we usually used (if they were available) three values of time intervals such as t(Lg) – t(Sn); t(Lg) – (Pn) and t(Sn) – t(Pn) from each station record or bulletin. Having estimated the origin time (t₀) in this way, the next step for location was to use time intervals such as t(Pn) – t₀; t(Sn) – t₀ and t(Lg) – t₀.

For event location we used travel times of regional phases as given by Nersesov and Rautian (1964), based on a Pamirs-Baikal profile, slightly adapted by Khalturin for Northeast Kazakhstan. Our locations, and the comparison with ground truth information, are given in Table 10. On average, the seismically-determined location error was only about 5 km. The ground truth location was found to lie within the interval specified by Khalturin et al. (1994) as the location uncertainty in almost all cases, and only marginally outside that interval in the few cases where it was outside. The average of absolute errors for all 17 UNTs is only 3.2 km in latitude, and 4.4 km in longitude. The average of signed errors is only 0.53 km in latitude and 0.45 km in longitude (i.e., real epicenters systematically lie 0.53 km south and 0.45 km west of our estimated locations). Since the average length of the seismic paths was 750 km, the systematic error is remarkable small – about 0.07%, corresponding to an error in velocity of about 0.005 km/s.

We have thus been able to demonstrate the utility of regional seismic waves for purposes of accurate estimation of UNT locations, even when only a few records are available per event. The location uncertainty is so small in our case, because of the availability of good information, appropriate to the region, on travel times. The strongest constraint typically came from values of the time interval between Lg and Pn.

5. Comment on magnitude distribution and yields

Now that we have obtained a fairly complete set of magnitudes for the nuclear tests at Semipalatinsk, it is of interest to see how they are distributed, and how well the announced information on yield is in accord with yield estimates based on seismic magnitude.

5.1. Comparison of seismically determined yields, and announced information on annual total yield

USSR Nuclear Tests (1997) gave the total yield each year at each test site, and so we can compare these announced totals at STS for the years from 1964 to 1989 with the total obtained by summing the seismically determined yields. (The announced total yield at STS for the years 1961 and 1962 included several atmospheric tests; and in 1963 there were no tests. After 1963 all STS nuclear tests were underground.) Using the relation mb = 4.45 + 0.75 log Y advocated by Murphy (1990) and Ringdal et al. (1992), we have estimated the yield of all the STS tests for which we have a magnitude. Figure 6 compares the annual total of yields determined seismically with the officially announced yield information. The yield of tests for which we do not have a magnitude is insignificant in comparison with the well-documented tests. Note that the vertical axis for this histogram is linear in yield, rather than in logarithmic units. The agreement between seismically determined yield totals, and announced total is remarkable. The differences are somewhat greater than 10% in the earlier years, but are less than 10% for most of the last ten years of testing.

5.2. Detection threshold and magnitude distribution

The small underground nuclear tests at STS provide some practical experience with certain aspects of open monitoring, albeit for years in the past, especially in the early years of underground nuclear testing, when capabilities were not as good as they are today. Thus, we can comment on three different levels of preliminary information about detected events:

(a) monitoring in the teleseismic zone (distances greater than 3,000 km) without any information, other than the known position of the test site;

(b) the same as case (a) but with stations located in the regional interval of epicentral distances (700 – 3,000 km); and

(c) the same as case (b) but when the dates of UNTs (and preferably even narrower time intervals) are known.

As our example of case (a) we can take the 271 UNTs at STS reported as seismic events by the ISC during 1961 – 1989 (note, not all of these events were reported at the time as nuclear in origin). However, this is monitoring of events large enough to enable good estimates of location using seismological methods. As noted above, in discussing the magnitude of these ISC events we prefer to use mb(AWE) values rather than mb(ISC) since the latter are known to be routinely too large, particularly for events with mb less than about 5.0 (when the number of stations reporting magnitude is significantly lower, so that the ISC mb is biassed by reliance on such sensitive stations – see Ringdal, 1976). Case (b) corresponds to the first stage of our analysis of regional observations (Khalturin et al., 1994), using stations located in the 700 – 3000 km distance range, but with unknown origin times. Case (c) corresponds to the analysis presented in this paper, when dates of UNTs at STS were known.

It is reasonable to expect that the magnitude threshold for event reporting will decrease from case (a) to case (b) to case (c), and this expectation turns out to be correct. As a simple definition of the magnitude threshold, we chose the mb value at which 50% of the signals are unreported. For determination of magnitude threshold, we worked with magnitudes for three sets of UNTs:

(1) mb(AWE) for all UNTs from STS reported by the ISC;

(2) our determination of magnitudes of 27 small UNTs which were detected in the first stage of study, when days and times of explosions were not known;

(3) magnitudes of the 12 smallest UNEs which were detected later, when the date of each explosion was known. (In practice, it also proved helpful in searching for these data to know the times of day that were commonly used for UNTs at STS.)

For each of the three sets of magnitudes, we made a histogram giving the number of events in cells of width 0.2 magnitude units. We found that in the magnitude interval 4.35 – 4.54, AWE assigned magnitudes for 8 events and 2 were missed by ISC; and in the magnitude interval 4.15 – 4.34, AWE assigned magnitudes for 4 events and 8 were missed by ISC. On this basis, the magnitude at which there was a 50% chance of an STS event being reported by the ISC corresponded to about mb(AWE) = 4.28. For CSE regional observations, case (b), inspection of the intervals 3.15 – 3.34 and 2.95 – 3.14 indicates the magnitude corresponding to 50% chance of detection was about mb 3.25.

The magnitude threshold for NORSAR in case (c), when the origin time is known, is very low for teleseismic signals from STS, mb 3.3 ± 0.1. Among the 29 chemical explosions recorded regionally by CSE, NORSAR reported about 23 events using teleseismic data. NORSAR also reported all six UNTs recorded by CSE but not reported by ISC since 1973 (when NORSAR started to operate). (The NORSAR Semi-annual report for April – September 1984 indicated a detection threshold for STS in the range mb 2.5 to 3.0.)

The study of Sykes and Ruggi (1989), published prior to the list of Bocharov et al. (1989), used numerous sources of information on teleseismic detections, including LASA and Hagfors detections of events that potentially could have been STS UNTs. An earlier and longer version of this paper is Sykes and Ruggi (1986), which gave references for the locations that were used, and stated that identification and size determination are much more questionable for events of equivalent yield less than 5 kt. Sykes and Ruggi's success was impressive in that they found six small magnitude UNEs which had not previously been reported in the open literature. But the price for these six detections was large, in that these authors also reported 39 events as UNEs which turned out to be false alarms, namely, 17 chemical explosions at STS, together with 15 chemical explosions in the Azgir region, 6 chemical explosions in different regions of the USSR, and one earthquake. The most important point to draw from this history is that detection capability has been very good for recent decades, since all but two UNTs at STS with yields announced as greater than 1 ton have now been associated with detections. Detections have also been very good using only teleseismic data. But for accurate location and confident identification, additional data is often needed. In practice, such additional data can often be provided by regional stations.

We conclude this section with Figure 7 showing the magnitude distribution for the Balapan region, and for the Degelen and Murzhik regions. Again, it is clear that Balapan was the preferred location for the largest tests, and Degelen for the smallest. Out of the total of 340 UNTs, more than 40 UNTs had mb < 4.4 and thus were probably sub-kiloton.

6. Conclusions

We have found and analysed regional seismic data for underground nuclear tests (UNTs) at the Semipalatinsk Test Site (STS) that enable us here (Table 8) to report the origin time and magnitude for 31 nuclear tests at this test site, that had not previously been documented in the open literature. Seismic detections for three of these UNTs were previously reported by Sykes and Ruggi (1986) and one by Ringdal (1990), but detections for the remaining 27 were not reported in the open literature before our study.

For 19 of these events we also obtained seismically-determined estimates of location, and location uncertainty. By comparison with ground truth location information that became available after our seismic determinations of location, we found that our location estimates were accurate to within a few km and our uncertainty estimates included the ground truth location in almost all cases. We conclude that regional waves can be used to provide accurate locations even when few stations are available, provided regional travel times are well calibrated.

There are only two UNTs at STS, announced as having yields greater than on ton, for which we have been unable to find detections. Both occurred in the 1960s.

Yield estimates based upon seismic magnitudes give values for the estimated total annual yield at that differ by less than 10% from officially announced values of these yield totals, for most of the last ten years of testing at the Semipalatinsk Test Site.

The information we have been able to report here, on 67 of the 69 UNTs that were previously not well documented, can be used to assess monitoring capability for this major nuclear test site and how that capability has improved with time. The information can also be used to identify small UNTs, and chemical explosions and earthquakes near STS, suitable for evaluation of methods of discriminating between small seismic events. We recommend that efforts be supported, to build up the database of regional waveforms for small seismic events on and near STS, since these are the types of waveform that monitoring programs must be designed to detect and identify.

Acknowledgements

We thank Dr. Frode Ringdal of NORSAR and Dr. Yuri Kopnichev of CSE, who made considerable efforts to find detections and assign magnitudes for particular events reported in this paper. We also thank Dr. Bill Leith of the US Geological Survey, and Dr. Natalya Mikhailova of the National Nuclear Centre of the Republic of Kazakhstan, for assistance in obtaining ground truth information. Our work was sponsored by contract DSWA01-97-C-0156 with the Defense Threat Reduction Agency. This is publication number YYYY of the Lamont-Doherty Earth Observatory.

References

Adushkin, V., Pernik, L., and Spivak, A. (1997), Degelen Mountain chemical explosion cratering experiment, Institute for Dynamics of the Geospheres, Technical Report No 97-3011 to DSWA, 110 p.

Bocharov, V.S., Zelentsov, S.A., and Mikhailov, V.I. (1989), Characteristics of 96 underground nuclear explosions at the Semipalatinsk test site, Atomnaya Energiya, 67, 210 – 214.

Bolt, B.A. (1976), Nuclear explosions and earthquakes: the parted veil, W.H. Freeman and Co., San Francisco.

Douglas, A. (1967), Joint epicentre determination, Nature, 215, 47 – 48.

Given, H.K., Tarasov, N.T., Zhuravlev, V.V., Vernon, F.L., Berger, J., and Nersesov, I.L., (1990), High-frequency seismic observation in Eastern Kazakhstan, USSR, with emphasis on chemical explosion experiments, J. Geophys. Res., 95, 295 – 307.

Gorin, V.V, Krasilov, G.A., Kurochkin, A.I., Kurkin, A.I., Maltsev, A.L., Matushenko, A.M., Orlov, S.L., Pichugin, A.V., Smagulov, S.G., Strukov, V.G., Fillipovsky, V.I., Kharitonov, K.V., and Chernishev, A.K., (1994), Semipalatinsk Test Site: Chronology of underground nuclear explosions and their primary radiation (1961-1989). Bulletin of the Center of Public Information, #4, pp 21 – 32, Moscow, Central AtomInform Institute.

Hansen, R.A., Ringdal, F., and Richards, P.G. (1990), The stability of RMS Lg measurements and their potential for accurate estimation of the yields of Soviet underground nuclear explosions, Bull. Seism. Soc. Am., 80, 2106 – 2126.

Khalturin, V.I., Rautian, T.G., and Richards, P.G. (1994), A study of small explosions and earthquakes during 1961-1989 near the Semipalatinsk Test Site, Kazakhstan, technical report for subcontract B239589 with Regents of Univ. of Calif., 64 p.

Khalturin, V.I., Rautian, T.G., and Richards, P.G. (1998), The seismic signal strength of chemical explosions, Bull. Seism. Soc. Am., 88, 1511 – 1524.

Leith, W. (1998), Degelen Nuclear Test and Tunnel Data, Technical Memorandum dated June 26, 1998 to the Defense Threat Reduction Agency, US Geological Survey; and Technical Reports to the Defense Threat Reduction Agency (formerly Defense Special Weapons Agency; formerly Defense Nuclear Agency), under Contract DNA001-95-C-1079.

Lilwall, R.C., Marshall, P.D. and Rivers, D.W. (1988), Body wave magnitudes of some underground nuclear explosions at the Nevada (USA) and Shagan River (USSR) test sites, AWE Report No. O 15/88, H.M. Stationery Office, London.

Lilwall, R.C., and Farthing, J. (1990), Joint epicentre determination of Soviet underground nuclear explosions 1973-1989 in Eastern Kazakhstan, AWE Report No. O 12/90, H.M. Stationery Office, London.

Marshall, P.D., Bache, T.C., and Lilwall, R.C. (1985), Body wave magnitudes and locations of Soviet underground nuclear explosions at Semipalatinsk Test Site, AWRE Report No. O 16/84, H.M. Stationery Office, London.

Murphy, J.R. (1990), A new system for seismic yield estimation of underground nuclear explosions, in Proceedings of the 12th annual DARPA/GL Seismic Research Symposium, Key West, Florida.

Mikhailov, V.N. (editor) and 14 co-authors, (1996), USSR Nuclear Weapons Tests and Peaceful Nuclear explosions, 1949 through 1990, RFNC-VNIIEF, Sarov, 96 p.

Nersesov, I.L., and Rautian, T.G. (1964), Kinematics and dynamics of seismic waves at epicentral distances less than 3500 km, Proceedings (Trudy) of the Institute of Physics of the Solid Earth, Academy of Sciences of the USSR, #32.

NNCRK (1999), Technical documentation, contributed by the National Nuclear Centre of the Republic of Kazakhstan, Oslo Workshop on IMS Location Calibration, Jan 1999.

Pooley, C.I., Douglas, A., and Pearce, R.G. (1983), The seismic disturbance of 1976 March 20, East Kazakhstan: earthquake or explosion? Geophys. J.R. Astr. Soc., 74 621 – 631.

Ringdal, F. (1976), Maximum-likelihood estimation of seismic magnitude, Bull. Seism. Soc. Amer., 66, 789 – 802.

Ringdal, F. (1990), Teleseismic event detection using the NORESS array, with special reference to low-yield Semipalatinsk explosions, Bull. Seism. Soc. Am., 80, 2127 – 2142.

Ringdal, F., Marshall, P.D., and Alewine, R.W. (1992), Seismic yield determination of Soviet underground nuclear explosions at the Shagan River test site, Geophys. J. Int., 109, 65 – 77.

Sultanov, J.J., Adushkin, V.V., Danilova, T.V., Kaazik, P.B., Kuznetsov, O.P., Nedoshivin, N.I., and Rubinshtein, N.D. (1995), Completion of data base of PNE and large-scale chemical blasts conducted inside the territory of FSU and analysis of characteristics of seismic waves in order to improve the methods of seismic event identification. Institute for Dynamics of Geospheres Report to IRIS, subcontract 201, 144 p.

Sykes, L. and Ruggi, S. (1986), Soviet Nuclear Testing. Natural Resources Defense Council, Working Paper NWD 86-4, November 1986.

Sykes, L. and Ruggi, S. (1989), Soviet Nuclear Testing. Chapter 10, pp. 332 – 375 in Nuclear Weapons Databook, Vol.IV, edited by Cochran, T.B., Arkin, W.M., Norris, R.S., and Sandes, J.I.

Thurber, C.H., Quin, H.R., and Richards, P.G. (1993), Accurate locations of nuclear explosions in Balapan, Kazakhstan, 1987 to 1989, Geophysical Research Letters, 20, 399–402.

USSR Nuclear Tests (1997), Mikhailov, V.N. (editor) and 29 co-authors, UZDAT Publishing House, Moscow, 248 p.

Vergino, Eileen S. (1989), Soviet Test Yields, EOS, 70, 1511 – 1513 (with corrections and additions, EOS, 70, 1569).