下载掌阅APP，畅读海量书库
立即打开
Monitoring Underground Explosions
D. Davies
Editor’s Note
During the Cold War, the testing of nuclear weapons proceeded in parallel with international discussions about whether they might be banned. One of the obstacles to a ban was the problem of verifying a nation’s adherence to it. The Soviet Union was particularly reluctant to agree to external inspections of its weapons facilities, so that verification would need to rely on a capacity for detecting tests from afar. Underground tests, still being conducted in the early 1970s, create seismic waves, but to use these for verification they would need to be distinguished from earthquakes. Here David Davies of the Massachusetts Institute of Technology reviews progress towards a sufficiently discriminating seismology, concluding that techniques were becoming adequate but still not infallible. 中文
Seismological means for detecting and identifying underground nuclear explosions have improved steadily during the past ten years. A technique exists for separating explosions from earthquakes. The problem now is lowering the threshold and understanding the occasional problematic event. 中文
AN underground nuclear explosion converts about 1 percent of its energy into seismic waves and these carry information about the time, size, and location of the event. They may also indicate that the event is indeed an explosion and not an earthquake. This information is somewhat more difficult to extract, and research in many countries for the past few years has been intensively directed towards discrimination between natural and artificial events. There is no other known way of identifying underground explosions that can be used on a world-wide basis. In this article, I shall describe the progress that has been made recently in this science, which has such obvious implications for arms control. The conclusions that I reach should not, however, be construed as a measure of the prospect of a total test ban. Many ingredients go into the making of a treaty, and ability to police it is only one of them. Seismological capability is thus not sufficient, but it may be necessary, at least down to some level of explosive yield. What would constitute an adequate level is a matter of some discussion at present. The science reported here is the result of the research of many seismologists. In order to reduce the number of references to a manageable level, however, I have cited only a few large compilations of results in which the genealogy of the various ideas and instruments may be found. 中文
The reason for testing nuclear weapons underground is not, of course a matter of public discussion, but a paper by Neild and Ruina 1 probably provides a reasonably complete list of purposes. Between 1968 and 1971 there was an average of about twenty-five underground tests a year announced by the United States compared with about ten presumed tests a year for the Soviet Union, one test per year for China and about five a year for France—all but one of the Chinese tests and all of the French tests being atmospheric. Britain has not announced a test for seven years. 中文
When the banning of nuclear tests first became an international issue in 1958, a conference on the technical problems in Geneva clearly indicated that the science of seismology would need substantial advances to reach a stage at which instrumental observations could indicate unambiguously that an explosion with a yield of a few kilotons (kton) had been detonated. Indeed the rather meagre data available during that conference and the subsequent negotiations (there had at that time only been two or three quite small underground tests) suggested that a first objective could reasonably be an international network of about 170 stations which could detect seismic signals down to a certain threshold but not necessarily identify their source. It was expected that most seismic events would clearly indicate their earthquake nature by their location and depth, their radiation pattern and the shape of their signal, but there would be a residue which would need further investigation. In the early days of international negotiations, inspections of the sites of a fraction of these suspicious events were discussed in detail—the number and nature of such inspections being particularly contentious topics. With time, the position of the Soviet Union on the inspection issue hardened to the assertion that inspections were unnecessary, and that purely national means of policing would be satisfactory. Thus the “Geneva network” was never built. 中文
The two issues which were seen as central in 1958 are central today. Background seismic noise (from wind, traffic, ocean waves, and so on) placed a limitation on the detection of events; and a certain number of earthquakes did not immediately reveal themselves as such. The advances in seismology have been substantial since 1958, but increasing the signal-to-noise ratio (s.n.r.) and finding improved methods of separating earthquakes and explosions are still the principal fields of research. The rather narrow frequency bands involved in seismology have controlled techniques rigorously, and because the work is concentrated in situations of low s.n.r., the problem has been that of finding the frequency at which the s.n.r. is highest, and aiming improvement at that frequency, rather than trying to encompass the whole spectrum of the seismic signal. This “mission-oriented” approach to seismology, when combined with the intellectually stimulating problems that have arisen on the way, has exerted a vital role in the development of seismology since 1958. 中文
It was clear from very early in the Geneva discussions (which lasted until 1962) that whatever international arrangement might finally develop, national interests required an improved seismic capability. The United States, Britain and Sweden, in particular, started to modernize their seismological capability for the explicit purpose of test-ban monitoring. Other countries, notably Canada and Japan, were also improving their seismic capability. The US programme “Vela Uniform” has been easily the largest, and so far more than $150 million has been spent on research and development. What has been bought 中文
In the first few years basic research was heavily supported and a network of 100 seismic stations, the World-Wide Standard Seismographic Network (WWSSN), was established—no stations were, however, sited on the territory of the Soviet Union or her allies. The network provided a huge data base, uniformly recorded, of short period (around 1 s) and long period (around 20 s) seismic signals. 中文
Certain “realities” soon emerged from this work. The decreasing prospects of an international network with access to the Soviet Union led to the dominance of teleseismic studies; seismic body waves (P and S waves) are well registered up to a few degrees from the source; indifferently and variably from there to about 25°; and then clearly and predictably out to 100°. Distances are measured in central angle degrees. Some examples of P waves are given in Fig. 1. This last, teleseismic, zone covers half the world and would be crucial if access were not permitted to particular countries. The zone of observation having been restricted, immediate limits can be placed on the detectability of explosions. A 5 kton explosion in hard rock, such as granite or salt, produces in the teleseismic zone a P wave of amplitude around 5 nm at a period of 1 s. Background noise at 1 s is about a third of this figure for the best observatories. Thus in the absence of techniques for noise rejection, a network would usually detect a 5 kton explosion in hard rock (an event cannot be said to have been detected until recorded with a signal-to-noise ratio of at least 1.5 at four widely spaced stations). The capability declines drastically below this yield. Further, explosions in softer rocks, notably tuff, a volcanic ash common in Nevada, produce a smaller signal by a factor of at least two than comparable explosions in hard rocks. Explosions in dry alluvium have even smaller seismic effects—1 kton in hard rock and 10 kton in alluvium are roughly equivalent. There is, however, a limit to the thickness of dry alluvium to be found anywhere on Earth, and as an explosion has to be adequately buried for safety’s sake, it is unlikely that explosions of more than about 20 kton can be fired and contained in known deposits of alluvium. 中文
Fig.1. Short period teleseismic observations of P waves from events located in a, Eastern Kazakh, USSR; b, Alaska; c, Turkey and d, South Sinkiang Province, China. The top and bottom events are presumed underground explosions. All events have body wave magnitudes in the range 5.0 to 5.7. Note the great difficulty there would be in identifying the direction of first motion for these events.
At the same time that the WWSSN was being equipped with conventional instruments and photographic recording, several experiments with arrays were being conducted. The aim was to suppress noise, presumed incoherent, by the addition of many channels. Another advantage accrued in the recording of data on analogue and later digital tape. Although practically all decisions in seismology are taken on the basis of observation by eye, direct visual recording suffers obvious disadvantages of dynamic range. Further, there is a strong noise peak at a period of about 6 s arising from microseisms generated in the oceans and the rejection of these by WWSSN instruments, peaked at 1 s and 20 s, is usually not entirely satisfactory. As a result, film records, in which the magnification is controlled by the ease of viewing, are often dominated by 6 s microseisms and are thus not running at the best magnification for 1 or 20 s signals. The data processing possibilities of tape recording remove this obstacle. 中文
The first arrays built in the United States had an aperture of 4 km. They were in many ways prototypes of possible arrays that the Geneva meetings had envisaged for an international network. Up to sixteen seismometers were spread out over this aperture, and the signals from all seismometers were added without phasing. Seismic signals from teleseismic distances arrive at steep angles—the signal sweeps across the ground with a horizontal phase velocity of at least 10 km s
–1
, and up to 24 km s
–1
and at typical sites this corresponds to angles of incidence of less than 20°. Thus straight addition does not seriously degrade the 1 s component of a signal seen in the small aperture. It was found with these small arrays that the noise was coherent in this frequency band for distances typically greater than the 0.5 km spacing of instruments, so the gain in s.n.r. was somewhere between one (if the noise were totally coherent and travelling in a near vertical direction) and
(if the noise were totally incoherent). Later arrays have used a seismometer spacing of around 2 km, at which distance local noise is substantially incoherent.
中文
The next development was the medium aperture phased array implemented by the United Kingdom Atomic Energy Authority
2
. Up to twenty seismometers were laid out in two perpendicular lines. Arrays of this type are in operation in Canada, Brazil, Scotland, India and Australia. The aperture was large enough to require phasing of the signal after recording to steer the array in the appropriate direction, but small enough to ensure that the signal was still coherent. Signal-to-noise improvements approaching
(about 4.5) are obtainable.
中文
The first large aperture seismic array (LASA in Montana) was completed by the United States in 1965, and a second array (NORSAR in Norway) has been in operation about two years. LASA has an aperture of 200 km and 350 short period seismometers; NORSAR is somewhat smaller. LASA is steered (by a computer) to be face-on to incoming seismic signals and the angle of approach is sufficiently well determined to locate the source with an accuracy of one or two hundred km. The real time digital operations of LASA and NORSAR each occupy entirely two medium sized computers in hunting for P waves. The gain for large arrays does not meet the
expectation, however, for over an aperture of 200 km crustal geological conditions can vary widely and this leads to a degrading of the coherence of the signal. An improvement in s.n.r. of a factor of ten is possible for LASA, however, and the array is capable of detecting and locating explosions in the teleseismic zone with yields down to about 2 kton in hard rock. Several thousand earthquakes occur each year with P waves of comparable or greater amplitude, and these form the population against which explosions have to be discriminated.
中文
It became increasingly apparent in the mid 1960s that the registration of body waves was not sufficient to solve the discrimination problem; a broader spectrum of seismic data was necessary. The answer lay in the dominant signal on the long period seismic traces, the Rayleigh wave. This is a dispersed wave which has travelled over the Earth’s surface and is well excited at 20 s periods—there are short period Rayleigh waves and long period P waves, but they are unimportant for discrimination. As for short period recording, noise is the limiting factor. The 6 s oceanic microseismic background, longer period fluctuations arising from atmospheric effects and the interference of Rayleigh wave trains from other earthquakes all contribute to the background. In the absence of an interfering event, most seismic stations have a background noise level at 20 s period of 50 to 100 nm; there is a great variability in noise level both from place to place and in time. For the reasons already given, tape recording is most desirable. 中文
Rayleigh waves spread cylindrically, but also undergo other attenuation in the ground. The signal decays as distance (Δ) to the power 1.6. This rapid decay makes it imperative that the seismic station be sited as near to the event as possible. Within a political requirement of non-intrusive monitoring we have to consider a value of Δ of 20° as the minimum at which stations will be available. At 20°, the Rayleigh waves from a 10 kton explosion with 20 s period are somewhat less than 100 nm in amplitude, so individual stations at this distance have a relatively small probability of detecting surface waves. As with short period observations, arrays prove a valuable means of increasing the s.n.r. Instrument spacings have to be greater (typically 20 km) for noise to be incoherent. Small tripartite arrays are either under construction or recently completed in Sweden, Canada and India. LASA, NORSAR and ALPA (an exclusively long period array in Alaska) all have at least fifteen long period seismometers within them. Results from LASA indicate an improvement in s.n.r. of a factor of four—close to
. In addition, knowledge of the nature of the dispersed waveform allows matched filter techniques to be applied, and these contribute up to another factor of two to the signal-to-noise ratio. Thus a twenty element long period array can yield an order of magnitude improvement in the s.n.r. against incoherent noise. In addition an array has a certain amount of directional discrimination (Rayleigh waves have a wavelength of 80 km at 20 s period) which makes it possible to study an event in the presence of an interfering signal, provided that the two are coming from directions differing by more than 30° and that the interfering signal is no more than a factor of ten larger than the signal being studied.
中文
Recent developments in instrumentation have been successfully aimed at reducing the non-seismic noise in instruments by more effectively sealing and isolating them, and at broadening their response to cover a wide spectrum. There is also promise in the siting of instruments down boreholes where long period noise is frequently substantially lower. 中文
Both P waves and Rayleigh waves are used to define a magnitude for seismic events. The body wave magnitude ( m b ) on the Richter scale is a measure of the excitation of P waves. The amplitude of the P wave signal on the record is converted to ground displacement ( A ) at a dominant period T , and then used in a formula
where B (Δ) is a term counteracting the variation of the signal with distance from the source. A ground displacement of 10 nm at 1 s period at a distance of 60° from a source indicates an event of magnitude 4.7, for example. 中文
The surface wave magnitude is obtained by measuring on the Rayleigh wave the ground displacement A at a period T (usually near 20 s) and using a formula
An event with an M s of 4.0 will have a ground displacement at 30° of 300 nm at 20 s period. 中文
For earthquakes, there is an empirical and very scattered relationship between M s and m b , the best fit to which is
M s =1.59 m b –3.9
I shall discuss this relation further later. 中文
It is possible to relate both m b and M s to the yield for explosions. Two kton in hard rock produces an event with an m b of 4.0, 20 kton an event with an m b of 5.0. For higher yields the equivalence of a factor of 10 in yield to a change of one unit in magnitude breaks down for several reasons. I have already noted that softer rocks produce smaller seismic signals; an m b of 4.0 is equivalent to 2 kton in granite, 3 or 4 kton in tuff and 20 kton in alluvium. 中文
The surface wave magnitude M s can equally well be related to the yield, and is a more useful measure, for it seems to obey a relation (for hard rock) of the form
M s =1.3 log 10 Y + 1.5
over a range from at least 1 kton to 1 mton. 中文
In the early development of the subject, the concentration on the detection of events by means of P waves led to a hope that the study of P waves alone might show differences between explosions and earthquakes. It was soon found that such differences did exist, but not to the extent that all earthquakes could be separated from all explosions; it was only possible to quote probabilities. Diagnostic aids, as they have come to be called, are the location and depth of the event, the direction of first motion of the seismic trace and the complexity of the signal. Clearly if an event can be unequivocally identified as having occurred deeper than a few km it is not an explosion—likewise if it occurs under deep water. The quality of depth determination declines as the depth decreases and it is frequently impossible to assign depths to events shallower than 30 km. The first motion criterion uses the difference in radiation pattern between a totally compressive explosion and an earthquake for which the radiation is successively compressive and rarefactional in adjacent quadrants. There are numerous problems—the uneven distribution of stations around the globe and the high signal-to-noise ratio required to make polarity measurements both render the technique of little use for lower yields. “Complexity”, a measure of the duration of the P wave signal, is a criterion which showed promise; signals that last more than a few seconds are likely to have come from earthquakes. Unfortunately there are many earthquakes which last for less than a second and these must be placed in the category of “suspicious events”. 中文
More recent work on the spectra of P waves has shown that even explosions of very small magnitude are significantly richer in high frequencies (2 to 3 Hz) than are earthquakes. This accords with generally accepted ideas of the nature of earthquake and explosion sources. Regional variations present problems, however; it seems at present that the attenuation (and hence the frequency transmission characteristics) of the Upper Mantle varies sufficiently from place to place that genuine spectral differences between earthquakes and explosions are overprinted with the frequency absorption of the Upper Mantle in such a way that P waves from explosions can occasionally look like those from an earthquake. 中文
The broadening of the spectrum that the joint study of Rayleigh and P waves brought with it provided the one indubitable “breakthrough” that the subject has seen. It had been realized for several years that explosions did not generate anything like the amount of Rayleigh waves that did earthquakes of comparable body wave magnitude. The WWSSN allowed these observations to be put on a more quantitative and global basis. It was noted that if the surface wave magnitude M s for each event was plotted against its body wave magnitude m b , the natural and man-made events clustered round two well separated lines. 中文
A typical example is shown in Fig.2. Clearly the earthquake population has a wider scatter than the explosion population, and this is unsurprising considering the diversity of earthquake types. Vagaries of propagation in the real Earth also contribute to the scatter. Nevertheless one point can be asserted—that the diagram has greatly increased our confidence in identifying events above a certain size. Of course the purist may reasonably object that a diagram such as Fig. 2 cannot offer certainty as it is only a modest sized population; nevertheless the diagram has been widely accepted. 中文
Fig. 2. An Ms : mb diagram for Eurasia based on WWSSN recordings by Marshall and Basham. ○, Earthquakes; ●, presumed explosions.
Earthquakes and explosions generate other seismic phases—shear or S waves and horizontally polarized surface waves called Love waves. In addition P waves have a long period component. All of these offer possibilities for discrimination, particularly because explosions in theory generate few S and Love waves. For large explosions and earthquakes, use of these waves works well, but it is found in general that the largest wave amplitude on the long period record is that of the Rayleigh wave. 中文
Fig. 2 is taken from a recent study 3 using WWSSN instruments only. Each point represents one event in Eurasia and it is necessary to have P wave detections at at least four widely separated stations before an event is listed. This places a lower limit on the body-wave magnitude for which events are reported. There are several reasons why this threshold should be rather blurred, but as a gross simplification one can say that the global network of seismometers (excluding arrays) will detect by P waves at four or more stations almost all events with an m b of 5.0 and greater and almost no events with m b less than 4.0. 中文
Each point also represents a determination of surface wave magnitude, and it is customary to require at least four stations to have detected the surface wave before a value is accepted (this is a very crude way of allowing for the radiation pattern of earthquakes). An event with an M s of 4.0 is certain to be detected by its Rayleigh waves, but an event with an M s of 3.0 is unlikely to be detected. This is partially a statement about noise, but the locations of the receivers are also important as they need to be as close to the event as possible. 中文
The thresholds of detection are such that an earthquake is likely to be seen both by its P and Rayleigh waves or by neither. This means that there are few earthquakes which are assigned an m b but cannot be placed on the diagram because M s is not measurable. On the other hand, there may be explosions which have, for instance, an m b of 4.5 and (by extrapolation) an M s of 2.5. If it is required that an event may be identified only if both parameters can be determined one can thus talk about a threshold for discrimination as the Rayleigh wave detection threshold. This threshold is statistical in nature, but may be defined in some such way as “If an event detected by P waves has an M s of 3.2 or greater there is a 90 percent chance that its surface waves will be detected at four or more stations and thus that it can be categorized as an explosion or earthquake”. This somewhat portly statement is necessary because events are routinely detected by body waves but the crucial detection for discrimination purposes is of surface waves. An alternative version is “Explosions in hard rock of 20 kton or more can be identified as such; so can practically all earthquakes detected by the WWSSN”. 中文
It is obvious from Fig.2 that any attempt to lower the yield threshold should reasonably start from an attempt to improve the detection of the surface waves, for it is demonstrable that there are events in Russia with an m b of less than 5.0 which have no surface waves visible on present instruments but which may be presumed to be explosions. Obviously as a programme for improvement of surface wave capability developed it would be highly desirable to boost up body wave detection in order that almost all events with m b greater than 4.0 be reported. Such a strategy is discussed later. 中文
Fig. 2 clearly indicates a difference in excitation between 20 s Rayleigh waves and 1 s P waves for the two types of event. A natural question is how this difference arises, and the question is not purely academic, because in attempting to lower the threshold (that is to bring in points with lower M s and m b ) it is valuable to know exactly when the two populations merge. In Fig. 2 there is a clear separation, but least squares straight lines may not be the best representations of the relationships at lower magnitudes. Theoretical models for earthquakes and explosions should be extendable to any magnitude and thus should indicate whether capital investment to improve detection capability would be justified. 中文
Unfortunately knowledge of the nature of both types of event is still shadowy. An earthquake is known to be a propagating rupture on a fault plane that leads to a finite offset across the fault; and an explosion as seen by seismometers is equivalent to the sudden application of an elastic stress on a notional surface of radius less than 1 km. It is also known in part how P waves and Rayleigh waves are generated from both types of source. This turns out not to be sufficient to answer the question of why the discriminant works. The question can be subdivided by recognizing that the discriminant may depend on the different partitioning of energy at source into body and surface waves; or differences in spectra at source that are reflected in differences in excitation at 1 and 20 s periods regardless of wave type; or some combination of these two. 中文
I do not believe that we yet have a simple answer to this issue. This reflects the inability as yet to describe satisfactorily the processes at work at the focus of an earthquake. I think it is reasonable to say that it is not a depth dependence nor a size difference that controls discriminative capability. Beyond this it is difficult to go. 中文
A more direct and at present more successful approach to the problem of the smaller magnitude events which are as yet beyond the global network’s accessibility is to do pilot work in a region where there is the opportunity to get close in to events. Accordingly much research has been concentrated on observations of explosions at the Nevada test site and earthquakes in the western United States. Results have been reported by Evernden 4 and, even using standard definitions of M s and m b , separation is good down to the lowest yields observed. When careful allowance is made for regional propagation effects separation can probably be improved. 中文
These results show that, within the context of relatively unrestricted access to the area, explosions and earthquakes in the western United States can be discriminated to substantially lower levels than the global threshold of 20 kton in hard rock. There seems to be an order of magnitude of improvement possible without encountering insuperable problems of overlap. It is of the utmost importance that this be understood in context. Monitoring of the Soviet Union, for instance, would involve forgoing the nearby stations and attempting to make up for the loss by high grade instrumentation, arrays and digital processing. Further, there is no guarantee that earthquakes and seismic wave propagation in the Soviet Union are comparable to those in the western United States (although there are grounds for optimism). Lowering the threshold of the monitoring network could, however, reasonably be expected to yield a lower discrimination threshold, and a large investment in such an improvement would be justifiable on scientific grounds. 中文
Several instrumental developments and deployments are under way at the moment. A small number of high quality, long period seismometers are now set up (in Spain, Israel, Thailand, Australia, Alaska and New Jersey). Others, suitable for deep boreholes where noise may be substantially reduced, are in production. Two large new long period arrays, in Alaska and Norway, and three three-element long period arrays in Canada, Sweden, and India, are now operating. The hope is that within two years there will be sufficient data from these instruments to assess a new level of discrimination. It seems reasonable to expect at least a diminution of a factor of two, to 10 kton in hard rock, on the basis of these investments. This is equivalent to saying that in monitoring the Soviet Union almost all events with an M s of 2.8 or greater will show detectable surface waves at four or more widely spaced stations. As a goal this is quite modest; the two large arrays plus the new stations in Thailand and Israel alone should accomplish this. 中文
A considerably lower threshold seems, however, to be attainable. Extrapolations from the western United States suggest that it is not at all unreasonable to expect discrimination to be feasible at an M s of 2.0 (2 to 3 kton in hard rock). How would this be achieved? If a monitoring network is allowed the luxury of expanding, the country being monitored must be allowed the luxury of moving its test-site to the most remote spot possible. It would be possible for the Soviet Union to test at a distance of 30° from the nearest seismometers in a monitoring network, and, at this distance, surface waves from an explosion with an M s of 2.0 are less than 10 nm peak to peak. There are few stations with a noise level as little as a factor of ten higher than this; and a factor of ten might be the upper bound on what the largest possible array might contribute to the improvement of signal-to-noise ratios. 中文
At this stage, economics may control the threshold more than seismology. For instance, Professor J. N. Brune of the University of California, San Diego, recently estimated that a thorough programme to upgrade seismic capabilities to such a high level would cost at least $200 million. 中文
How might a major programme reasonably develop? The present operation of WWSSN is decidedly non-optimal, as the network was not designed for, nor is it dedicated to, discrimination. Resiting of some seismometers, particularly removal from centres of population, would yield reduced noise levels. Better protection from environmental changes, particularly pressure and temperature fluctuation, could not but be beneficial. The aim should be to record ground noise only, and it is worth a substantial amount of time and money to ensure that this is all that is coming out of the instruments. Placing some instruments down boreholes could certainly reduce noise further. Digital recording, in parallel with photographic recording, is highly desirable, preferably with prefiltering to eliminate 6 s microseisms. 中文
At present the WWSSN produces data, but no one is required to analyse them. A small group of seismic analysts, having available both the superior data and also information from the large arrays and other seismic stations that cared to contribute, would greatly increase our number of body-wave detections of events and would develop knowledge about which stations were most useful both in body and surface wave observations. This knowledge would help decide what the next step should be. 中文
It is likely that certain stations would be worth even further improvement. Spotting these sites involves more than just measuring the noise. Seismology is not a wholly predictable subject and, in some locations, signals are unusually high for little known reasons. This may work in favour of sites which would be rejected on the grounds of noise level alone. The improvement might take the form of development of arrays centred on good stations. The establishment of new stations in favourable locations would also make sense. 中文
Beyond this stage it is not easy to speculate. Once one has started reinstrumenting the globe there is no limit to the ingenuity that could be employed. Political and economic realities, however, restrain the monitoring network from extending too far. The development sketched out might be accomplished within five years. 中文
The relative simplicity of Fig. 2 may give the impression that discrimination can be made into a list of instructions for seismic record readers. In most cases this is true, but there are nagging problems—some natural and some man-made. 中文
In considering a population of events in a certain region it turns out that some events cannot be categorized on the basis of an M s : m b discrimination because the surface waves from them are masked by surface waves from other unrelated earthquakes. These cannot be discriminated by the M s : m b technique unless a way can be found to separate the two events. An array is highly desirable for this, and even then there are prospects of success only under the circumstances discussed earlier. If the masking event is very large (with an m b of 7.5 or greater) the global network is blotted out to such an extent that detection of even P waves from practically any event for an hour or more afterwards may be impossible because of the reverberations within the Earth. This opens up the possibility for evasion by firing explosion immediately after a large earthquake. It is difficult at present to see any seismic means of improving the prospect of detecting such evasions. 中文
Another natural hazard is that of explosion-like earthquakes. If the time interval for which Fig. 2 had been compiled were expanded, earthquakes of quite high magnitude (an m b of 5.0 for instance) but which produce barely discernible surface waves would eventually be found. These earthquakes, which come from very limited regions on the Earth’s surface, would have to be classed as explosions according to the M s : m b criterion. It is believed that they can be understood in terms of earthquakes in which the stress drop is abnormally high. Whether with improved instrumentation any earthquake-like features of these events will reveal themselves remains to be seen. It is not yet known whether, with improved body wave detection capability and hence larger populations of smaller events, more of these earthquakes would be encountered. 中文
The chief man-made problem (apart from deliberate masking by earthquakes) is that of decoupling. The radiated seismic signal from an explosion fired in a sufficiently large hole can be at least a factor of 100 smaller than in the fully tamped case. Thus a 50 kton explosion could be moved below the bounds of detectability if fired in a deep hole of radius 80 m, and partial decoupling would occur if the radius were less than this. Although a removal enterprise of this magnitude might well excite attention if started after a treaty had been signed, there may at present be cavities in salt mines which could be used for partial decoupling. There seems no seismic way of detecting decoupled events, and entrants into a treaty would have to weigh the risks accordingly. 中文
Although there are prospects of reducing the yield threshold to a figure of a few kton in hard rock, it must be borne in mind that a violator is unlikely to fire his explosions in granite. As a threshold of 2 kton in hard rock would be very difficult to achieve, this might be turned into a statement that all tests under 20 kton fired in alluvium would not be discriminable. This depressing situation is fortunately unlikely to be valid. The thickness of alluvium deposits limits the depth of the shot, but, on the other hand, the shot must be buried deep enough not to form a subsidence crater. These craters have formed more often than not after a test in alluvium, and there would be a clear risk of visual detection. Thus a violator would have to judge the risks of alluvium firing and might well conclude that a yield much smaller than 20 kton was all that could safely be tested. 中文
One may justifiably feel vaguely uneasy about the M s : m b discrimination technique. It tends to enshrine a lot of science and technology in one diagram and it may be used misleadingly. For instance although the population of earthquakes in Fig. 2 is fairly representative, there can only be a limited number of explosions; most of these come from one test site and are thus not truly independent samples. Further, test sites could be moved around at will, and so the discriminant can only represent the past, not the future. These qualms can only be assuaged if the science of discrimination is seen to be well founded. 中文
On the other hand, there may also be some uneasiness that the M s : m b discriminant is not optimistic enough, and this is worth examining. When an event has both detectable P waves and detectable Rayleigh waves it can be placed on an M s : m b diagram. If, however, P waves are detected but no Rayleigh waves are seen, it is unacceptable. For instance, an explosion in the Soviet Union with an m b of 4.5 will not generate detectable Rayleigh waves. Could it be called an explosion on the basis of its lack of Rayleigh waves? Obviously this question depends on whether all earthquakes with an m b of 4.5 have detectable Rayleigh waves (we eliminate the masked event problem). The answer is that on present experience they do, with the exception of a limited number of events mentioned earlier. The spread of M s values for earthquakes with m b >4.5 is substantial, but it does not extend down to 3.2. 中文
Up to the present so called “negative evidence” has always been looked on with some suspicion. The reason for this is probably that, although to a scientist the absence of a signal can be as important as its presence, to the politician with a different mode of thinking “negative evidence” sounds as unconvincing as would absence of fingerprints to establish not being present at the scene of a crime. It may well be that the M s : m b diagram, for all its neatness, has placed an undue emphasis on the necessity for determining both M s and m b . 中文
The following would be a “negative evidence” type of discrimination statement. “When an event has been detected by P waves and assigned a body wave magnitude of 4.5 or greater it can be categorized as an earthquake if M s ≥ ( m b –1.3) and an explosion if M s <( m b –1.3) or if M s cannot be determined, but on the basis of noise observations must be less than m b －1.3”. It would switch the threshold to a body wave magnitude of 4.5 or about 5 kton in hard rock, a figure that with the M s : m b discriminant alone would take several years and much expenditure to achieve. Although this statement would let through as explosions those unusual earthquakes described previously, this would equally apply to the M s : m b technique. For these events a pragmatic approach is necessary. A very detailed survey with a large population is necessary to establish the general validity of negative evidence. Such work is at present being pursued. 中文
I shall conclude with a description of recent activity beyond the laboratory. In 1968 the Stockholm International Peace Research Institute (SIPRI) convened a conference of seismologists from ten countries, east and west, at which technical progress since 1958 was reviewed. Subsequent SIPRI Progress Reports, the latest dated September 1971 (ref. 5), have kept the technical content up to date. 中文
In the United States, numerous technical meetings have been held. Papers presented at the Woods Hole meeting in 1970 (ref. 4) have been widely used in this discussion. A recent meeting in Cambridge, Massachusetts, has allowed some of the projections to be strengthened and a growing appreciation of the problem events. 中文
The Conference of the Committee on Disarmament of the United Nations (which has lived under several names) meets regularly at Geneva, and the issue of seismology and a test ban is frequently aired there. In 1970, at Canadian instigation, a register of seismic stations with guaranteed available data was compiled by the United Nations and subsequently a Canadian survey 6 examined the theoretical potential of this network. In the same year Britain submitted a proposal 7 to Geneva on a new network which would improve discriminative capability. In 1971 a United Nations technical meeting was held in Geneva against a background of increased international interest in a test ban. 中文
It is clear that this subject will continue to remain alive before a more general public for the foreseeable future. During this time seismological advances are unlikely to be spectacular, but a continuing investment in seismology will undoubtedly produce a steady reduction of the threshold. Where this threshold ultimately comes to rest and whether a treaty will be signed will be determined by a multiplicity of factors, only one of which is seismology. 中文
This work was sponsored by the Advanced Research Projects Agency of the Department of Defense. 中文
Fig. 2 was kindly supplied by P. W. Basham and P. D. Marshall, and represents work done by them 3 at the Department of Energy, Mines and Resources, Ottawa, Canada. 中文
( 241 , 19-24; 1973)
David Davies
Lincoln Laboratory, Massachusetts Institute of Technology, Lexington, Massachusetts 02173
References: