Forensic Explosion Seismology Technologies and Applications Kim

Forensic Explosion Seismology
@seismicisolation
@seismicisolation
@seismicisolation
@seismicisolation
Forensic Explosion Seismology:
Technologies and Applications
Edited by
So Gu Kim and Yefim Gitterman
@seismicisolation
@seismicisolation
Forensic Explosion Seismology: Technologies and Applications
Edited by So Gu Kim and Yefim Gitterman
This book first published 2020
Cambridge Scholars Publishing
Lady Stephenson Library, Newcastle upon Tyne, NE6 2PA, UK
British Library Cataloguing in Publication Data

A catalogue record for this book is available from the British Library
Copyright © 2020 by So Gu Kim, Yefim Gitterman and contributors
All rights for this book reserved. No part of this book may be reproduced, stored in a retrieval system, or transmitted, in any
form or by any means, electronic, mechanical, photocopying, recording or otherwise, without the prior permission of the
copyright owner.
ISBN (10): 1-5275-4474-5

ISBN (13): 978-1-5275-4474-1
@seismicisolation
@seismicisolation
CONTENTS
Prologue .................................................................................................................................. viii
Chapter One ................................................................................................................................1

Surface-wave Magnitudes of Eurasian Nuclear Underground Explosions
Otto W. Nuttli, So Gu Kim and Hui-Yuin Wen
Chapter Two ..............................................................................................................................24

Discrimination of Anomalous Earthquakes from Underground Nuclear Explosions
Otto W. Nuttli, So Gu Kim, Hui-Yuin Wen and John A. Wagner
Chapter Three ............................................................................................................................56

Spectral Classification Methods in Monitoring of Small Local Events by Israel
Seismic Network
Yefim Gitterman, Vladimir Pinsky and Avi Shapira
Chapter Four .............................................................................................................................78

Difference between Micro-earthquakes and Artificial Explosions on the Basis
of Frequency Contents
So Gu Kim and Yong-cheol Park
Chapter Five ............................................................................................................................108

Uncertainties of Seismic Source Determination Using a 3-Component Single Station
So Gu Kim and Zhongliang Wu
Chapter Six ..............................................................................................................................120

Study on Some Characteristics of Earthquakes and Explosions Using the Polarization Method
So Gu Kim and Fuchun Gao
Chapter Seven .........................................................................................................................135

Spectral Discrimination Analysis of Eurasian Nuclear Tests and Earthquakes Recorded
by the Israel Seismic Network and the NORESS Array
Yefim Gitterman, Vladimir Pinsky and Avi Shapira
Chapter Eight ..........................................................................................................................155

Signal Processing for Indian and Pakistan Nuclear Tests Recorded at IMS Stations Located
in Israel
Yefim Gitterman, Vladimir Pinsky and Rami Hofstetter
@seismicisolation
@seismicisolation
vi Contents
Chapter Nine ...........................................................................................................................174

Overview and Review of Moment Tensor Inversion for North Korean Underground
Nuclear Explosions
So Gu Kim
Chapter Ten .............................................................................................................................207

GT0 Explosion Sources for IMS Infrasound Calibration: Charge Design and Yield Estimation
from Near-source Observations
Y. Gitterman and R. Hofstetter
Chapter Eleven ........................................................................................................................230

Secondary Shock Features for Large Surface Explosions: Results from the Sayarim Military
Range, Israel and Other Experiments
Yefim Gitterman
Chapter Twelve .......................................................................................................................252

Spectral Modulation Effect in Teleseismic P-waves from North Korean Nuclear Tests
Recorded in Broad Azimuthal Range and Possible Source Depth Estimation
Yefim Gitterman, So Gu Kim and Rami Hofstetter
Chapter Thirteen .....................................................................................................................270

The Ratio between Corner Frequencies of Source Spectra of P- and S-waves—
A New Discriminant between Earthquakes and Quarry Blasts
Galina Ataeva, Yefim Gitterman and Avi Shapira
Chapter Fourteen .....................................................................................................................285

Overview and Review on Analysis for Hydroacoustic Studies in Underwater Explosions
So Gu Kim
Chapter Fifteen ........................................................................................................................326

Near-Source Audiovisual, Hydroacoustic, and Seismic Observations of Dead Sea Underwater
Explosions
Yefim Gitterman
Chapter Sixteen .......................................................................................................................342

Underwater Explosion (UWE) Analysis of the ROKS Cheonan Incident
Chapter Seventeen...................................................................................................................358
Estimating Depth and Explosive Charge Weight for an Extremely Shallow Underwater
Explosion of the ROKS Cheonan Sinking in the Yellow Sea
So Gu Kim, Yefim Gitterman and Orlando Camargo Rodriguez
Chapter Eighteen .....................................................................................................................377

Forensic Seismology and Boundary Element Method Application vis-à-vis ROKS Cheonan
Underwater Explosion
So Gu Kim
@seismicisolation
@seismicisolation
Forensic Explosion Seismology: Technologies and Applications vii
Chapter Nineteen.....................................................................................................................399
Reply to Comment on “Underwater Explosion (UWE) Analysis of the ROKS Cheonan
Incident” by K. S. Kim
Chapter Twenty .......................................................................................................................406

Depth Computation and Source Characteristics DPRK’s 2016-01-05, 2016-09-09
and 2017-09-03 Nuclear Tests Using Body and Rayleigh Waves
So Gu Kim , Yefim Gitterman, Seoung-Kyu Lee and Sang-Mo Koh
Chapter Twenty-One ...............................................................................................................434

Source Depth and Characteristics for the 6 January 2016, the 9 September 2016 and the 3
September 2017 North Korean Nuclear Tests Using Body and Rayleigh Wave Spectra
and Polarization of Surface Wave
S. G. Kim, Y. Gitterman, S. Lee and H. Bae
Chapter Twenty-Two ..............................................................................................................461

Depth Estimate of the DPRK’s 2006-10-09, 2009-05-25 and 2013-02-12 Nuclear Tests
Using Spectral Nulls of Body Waves and Fundamental-mode Rayleigh Wave Amplitude
Spectra
S. G. Kim, Y. Gitterman, S-K. Lee and V. Vavryčuk
Chapter Twenty-Three ............................................................................................................499

On Computation of Depth and Source Characteristics of the DPRK Nuclear Tests
(2006, 2009, 2013, 2016J, 2016S and 2017) Using Regional and Teleseismic Arrays
So Gu Kim, Yefim Gitterman, Seoung-Kyu Lee, Hyung-Sub Bae and Gill Jae Lee
@seismicisolation
@seismicisolation
PROLOGUE
This book contains three major physical phenomena for active source seismology: underwater
explosions, underground nuclear explosions and large-scale on-surface chemical explosions.
In particular we attempted to demonstrate how to use the technologies and applications in active
source seismology and seismo-acoustics rather than the theoretical approach for the resolution as
forensic explosion seismology in the light of an application for Defense Sciences. We also presented
general seismological studies on discrimination between earthquakes and man-made explosions
including the past global underground nuclear explosions.
Since October 9, 2006, North Korea conducted six underground nuclear tests including one
thermonuclear test (mb=6.3) report on September 3, 2017. There was also a problematized issue to
identify the underwater explosion for the ROKS Cheonan Sinking on March 26, 2010 in the Korean
Peninsula. In this book those two artificial explosions are intensively analyzed.
Our special interests are to find and verify causes of the underwater explosions such as the ROKS
Cheonan Sinking and Kursk Submarine disaster as well as the North Korean underground nuclear
explosions and disastrous ammunitions store explosion accidents and terroristic attacks, taking into
account some scientific results in estimating source parameters of underwater explosions, on-surface
detonations and underground nuclear explosions. The main goal of this study is to answer this
curiosity for a professional audience as well as lay readers in forensic explosion seismology. This
book is based on published papers in the international journals as well as the original articles related
to observations, basic theories and conclusions for Forensic Explosion Seismology.
We would like to express our appreciation to our international colleagues and collaborators: Dr.
Zhongliang Wu (Director of Institute of Geophysics, China Earthquake Administration, China),
Prof. Dr. Orlando C. Rodriguez (University of Algarve, Portugal), Prof. Dr. Robert B. Herrmann
(Saint Louis University, USA), Prof. Dr. Paul Richards (Columbia University, USA), Dr. Václav
Vavryčuk (Institute of Geophysics, Czech Republic), Prof. Dr. Anne Trehu (Oregon State
University, USA), Prof. Dr. Khoo Boo Cheong (National University of Singapore, Singapore), Dr.
Zhang Aman (Harbin Engineering University, China) and Dr. Olivier Hyvernaud (Laboratoire de
Géophysique,Tahiti, French Polynesia). One of authors (SGK) also appreciates CTBTO’s funding
to participate in SnT2015, SnT2017 and SnT2019, CTBTO, Vienna Austria as well as collaboration
in providing IDC data during this work. We acknowledge the copyright permissions for some articles
from some publishers such as Springer-Nature, Elsevier, Geophysical Journal International, Journal
of Seismology, AGU, Journal of Physics of the Earth, etc.
So Gu Kim
Yefim Gitterman
November 2019
@seismicisolation
@seismicisolation
CHAPTER ONE
SURFACE-WAVE MAGNITUDES OF EURASIAN

NUCLEAR UNDERGROUND EXPLOSIONS 1
OTTO W. NUTTLI, SO GU KIM AND HUI-YUIN WEN
Global Nuclear Explosions since 1945 from Bill Rankin (2007) using Johnston’s Archive of
Nuclear Weapons which were modified including six North Korean nuclear tests.
ABSTRACT
Body-wave magnitudes, mb, and surface-wave magnitudes, MS, were determined for approximately
100 Eurasian events which occurred during the interval August through December 1971. Body-
wave magnitudes were determined from 1-see P waves recorded by WWSSN short-period, vertical
component seismographs at epicentral distances greater than 25°. Surface-wave magnitudes were
determined from 20-see Rayleigh waves recorded by long-period, vertical - component WWSSN
and VLPE seismographs. The earthquakes had mb values ranging from 3.6 to 5.7.
1
Some parts presented in Bull. Seism. Soc. Am., (1975). 65 (3), 693-709. Surface-wave magnitudes of Eurasian
earthquakes and explosions by Otto W. Nuttli and @seismicisolation
So Gu Kim.
@seismicisolation
2 Chapter One
Of 96 presumed earthquakes studied, six lie in or near the explosion portion of an mb:MS plot. The
explosion mb:MS curve was obtained from seven Eurasian events which had mb values ranging
from 5.0 to 6.2 and MS values from 3.2 to 5.1. All six anomalous earthquakes were located in the
interior of Asia, in Tibet, and in Szechwan and Sinkiang provinces of China. In general, ocean-
margin earthquakes were found to have more earthquake-like mb:MS values than those occurring in
the continental interior. Neither focal depth nor focal mechanism can explain the anomalous events.
INTRODUCTION
The research described in this report is concerned with the mb:MS discriminant between
earthquakes and underground explosions, principally as applied to small-magnitude events
occurring in Eurasia. Several related problems are investigated, namely: 1) The development of an
MS formula which makes use of the amplitude of 20-sec period surface waves at short distances.
This formula, when used in conjunction with data from the high-gain, long-period seismograph
net-work, provides a low threshold for reliable MS determination. 2) A determination of mb:MS
values for Soviet explosions of intermediate to high yield. The resultant mb:MS curve is found to
be identical to that obtained for Nevada Test Site events. 3) A search for shallow-depth Eurasian
earthquakes which have explosion-like mb:MS values.
BODY-WAVE MGNITUDE (Mb)
The mb values are determined by use of the Gutenberg-Richter (1956) equation
mb = log (A/T) + Q (h, ¨) (1)
where (A/T) is taken as one-half the maximum peak-to-peak motion (amplitude) (in
microns/seconds) in the first three cycles of the vertical component of the P-wave ground motion
in the period range of 0.7 to 1.3 seconds. As a further restriction we determine mb values only
from stations at teleseismic distances, i. e. greater than or equal to 25°, because at lesser distances
the lateral variations in the upper mantle structure make the function Q dependent on geographical
region. Q (h, ¨) indicates the geometrical spreading and absorption (Richter, 1956).
SURFACE-WAVE MAGNITUDE (Ms) FORMULA FOR DISTANCES LESS THAN 20°
The standard MS formula
MS = 3.30 + 1.66 log ǻ (°) + log A/T (microns/sec), (2)
where T is approximately 20 sec, is not used for ǻ < 20° because: 1) from theoretical
considerations it is not applicable at ǻ < 20° (Nuttli, 1973) and 2) the amplitudes of 20 sec period
surface waves are small on conventional seismograms at these distances. The first limitation can
be overcome by developing a formula appropriate to the distance range which applies for 20-sec
period waves, and the second by the use of data from the high-gain, 1ong-period seismo-graph
stations. Previously, attempts to determine MS for small events from the data of near stations used
the amplitudes of shorter period surface waves (e.g. Evernden, et al, 1971; Basham, 1971;
Marshall and Basham, 1972; Karnik et al, 1962; Nuttli, 1973). All of these formulas suffer to the
extent that they are dependent on the upper-crustal geology, which varies regionally. Therefore the
mb:MS relations for earthquakes and explosions derived from them differ with geographic region.
However, we shall show that if mb is determined from the amplitudes of 1-sec period P waves at
distances greater than 20°, and if MS is determined exclusively from the amplitudes of 20-sec
period surface waves, then the mb:MS curve established for Nevada Test Site events by Evernden
et al (1971) satisfies the mb:MS values for Soviet explosions in Novaya Zemlya, Kazakh and three
different sites in the Ural Mountains.
Deviation of the Formula. The empirical curve of Gutenberg and Richter (1936), which they used
to define the surface-wave magnitude, can be fitted by a theoretical attenuation curve for which the
@seismicisolation
@seismicisolation
Surface-wave Magnitudes of Eurasian Nuclear Underground Explosions 3
coefficient of absorption is 0.015 deg-1 (Nuttli, 1973). This curve is shown as the solid-line curve
in Figure 1. Over the distance range 25° d ' d 140° the theoretical curve can be closely
approximated by a straight line with slope -1.66 (see figure 1), which conforms with the standard
MS formula given previously as Eq. (1). A straight-line fit to the theoretical curve will have a
different slope at the shorter distances, which from Figure 1 is shown to be -1.07 for 10° d ' d 30°.
Therefore, an equation for MS using the amplitude of 20-sec period surface waves at the near
distances, 10° d ' d 30°, is the Nuttli-Kim formula (Nuttli and Kim, 1975):
MS = 4.16 + 1.07 log ' (°) + log A/T (microns/sec) (3)
where the constant 4.16 is determined by requiring that Eqs. (1) and (2) give identical MS values
for ' between 25° and 30°.
In practice, the amplitude of the largest wave motion in the period range 17 sec to 23 sec is used,
rather than the amplitude at exactly 20 sec.
Figure 2 is a reproduction of the vertical-component seismograms at EIL (Israel), QUE (Pakistan)
and KBL (Afghanistan) for an earthquake in the western Caucusus of MS= 4.2. The purpose of this
figure is to show that 20-sec period surface waves are well developed and measurable at distances
of about 10° to 20° for small to moderate size earthquakes.
Comparison with Other Formulas. Figures 3 to 7 compare MS values obtained using the newly-
derived Eq. (2) with those obtained using the Basham (1971), Marshall and Basham (1972) and
Karnik et al. (1962) formulas. The standard value of MS is taken to be that obtained using the
conventional formula (Eq. 1) for stations at distances of 30° and greater. Figures 3 to 7 show that
the Basham and Karnik et al. formulas generally overestimate MS, the former by 0.5 to 1.0 units
and the latter by about 0.2 to 0.7 units. Both the Marshall-Basham and the newly-derived Eq. (2)
give values much more consistent with those obtained from teleseismic distances; Figs. 3 and 6
show little preference between the two formulas, but Figs. 4, 5 and 7 suggest that Eq. (2) gives
better results. A further significant advantage of Eq. (2) is that it can be applied on a global or
world-wide basis, whereas for the Marshall-Basham formula one must have a set of tables of P
functions versus period, one table for each geographic region. This requires an accurate knowledge
of the crustal structure in all the regions.
MS Threshold Values as a Function of Epicentral Distance.
Under ideal conditions of low-noise background, a 20-sec period surface wave with a 1 mm zero-
to-peak seismogram amplitude can just barely be resolved. Accepting this as the minimum
amplitude signal that can be read to determine MS from the high-gain, long-period seismographs,
we calculated the smallest MS value that can be detected by these instruments, which are taken to
have a magnification of 20,000 at 20-sec period. Eq. (1) was used for ' t 30° and Eq. (2) for 30°
d ' d 10°. The results are:
(deg) Threshold Ms
10 2.6
15 2.8
20 2.9
25 3.1
30 3.2
40 3.4
50 3.5
75 3.8
100 4.0
130 4.2
@seismicisolation
@seismicisolation
4 Chapter One
The above values are the MS thresholds for a single seismograph (not an array) with a
magnification of 20,000 at 20-sec period, operating under quiet conditions so that a trace
amplitude of 1 mm can be detected. In the present study we have been able to determine MS values
as low as 2.9, under favorable conditions.
mb:MS VALUES FOR SOVIET EXPLOSIONS
We have determined mb:MS values for all the Soviet explosions known to us for the interval
August through December 1971. The dates, times and locations of the ten events, as taken from
the NEIC bulletins, are given in Table 1. Five of the ten underground explosions were at the
Eastern Kazakh site, but none of the remaining five were at the same location. Thus the data
represent six source regions, separated very widely geographically.
mb and MS values determined by us, as well as by the National Earthquake Information Center
(Boulder, Colorado), International Seismological Centre (Edinburgh, United Kingdom), Moscow
(U.S.S.R.) and Hagfors Observatory (Sweden) are given in Table 2. No MS value could be
determined by us for the 19 Sept 1971 event because the surface waves of
an earthquake interfered with those of the explosion. We could not identify any surface waves
from the explosions of 04 Oct 1971 and 22 Oct 1971, but from the background noise of the
seismograms we determined that the MS for each event could not have been any larger than 2.9,
and may have been smaller.
Figure 8 is a plot of our mb:MS values for the Eurasian explosions. On the same figure is plotted an
mb:MS curve for Nevada Test Site events, as given by Evernden et al (1971), where mb was
determined from teleseismic P waves and MS from 20-sec period surface waves. A remarkable
feature of Figure 8 is that the Soviet explosion data, from 5 widely scattered source locations, fit
remarkably well with the NTS data over the entire range of intermediate to high-yield events. This
strongly suggests that the mb:MS relation for underground explosions is the same for all regions of
the world, if mb is determined from 1-sec teleseismic P waves and MS from 20-sec surface waves.
The regional dependence of mb:MS, as found by other investigations, most likely results from their
use of an MS formula which had built into it a regional bias. In other words, there is a question of
the reliability of their MS values for small events.
SHALLOW-DEPTH EURASIAN EARTHQUAKES
We have selected for study a fairly large and representative set of Eurasian earthquakes of mb
between approximately 4 and 5 that occurred from August through December 1971. We
determined mb and MS for each of these earthquakes, with mb obtained exclusively from 1-sec
period teleseismic P-wave amplitudes and MS exclusively from 20-sec surface-wave amplitudes.
In addition to the data from the eight high-gain, long-period seismograph stations we used the
seismograms from 20 standard WWSSN stations.
Table 2 summarizes our mb:MS studies for the Eurasian earthquakes and explosions in the interval
August through December 1971. The results are also presented graphically in Figure 9.
In Figure 9 we can see that a curve (dashed line) displaced one-half mb unit to the left of the NTS
or Eurasian explosion curve will serve as an envelope to almost all the mb:MS values of the
Eurasian earthquakes. There are three, and possibly four, exceptions to this statement. These are
the presumed earthquakes of 16 August, 24 October, 4 December and 24 November 1971, which
are identified by date in Figure 9. Because no surface waves could be seen for the latter
earthquake, all we can say concerning its MS value is that it is no greater than 3.2. From pP and sP
phases we were able to establish that all four earthquakes had their focus in the crust, so that
greater-than-normal focal depth cannot be called upon to explain the relatively small MS values.
@seismicisolation
@seismicisolation
Although we have not made a detailed study of this point, it is our general impression that
earthquakes in the interior of Eurasia more often have mb:MS values close to that of the dashed
curve in Figure 9, whereas Pacific Coast earthquakes generally have a relatively large MS value,
which puts them to the left in Figure 9. We caution, however, that there likely will be exceptions
to this statement.
CONCLUSIONS
The newly-derived formula (Eq. 2) for MS, which is applicable on a world-wide basis and gives
magnitudes which agree with those obtained by the conventional formula, enables one to lower the
threshold of MS down to about 2.8 when data from the high-gain, long-period seismographs are
used. When the MS values of Eurasian explosions are determined by use of this new formula, the
mb:MS curve for intermediate to high-yield Eurasian explosions at five different sites is found to be
identical to that for Nevada Test Site explosions.
All of the Eurasian earthquakes studied, with the exception of three and possibly four, have an mb
value at least 0.5 units less than that of an underground explosion of the same MS value, over the
range of mb from 4.0 to 5.6. Two definitely anomalous earthquakes (24 Oct and 04 Dec) had
mb:MS values identical to those found for nuclear explosions. The earthquake of 16 Aug lies
between the explosion and limiting earthquake curves. No surface waves could be identified for
the third anomalous earthquake; the maximum possible value it could have, as determined from
the background noise level, places it on the fringe of the earthquake population. A smaller value
would place it in the explosion population. All of these anomalous earthquakes are shallow events.
REFERENCES
Basham, P. W. (1971). A new magnitude formula for short period continental Rayleigh waves,
Geophysical Journal of the Royal Astronomical Society, 23, 255-20.
Evernden, J. F., W. J. Best, P. W. Pomeroy, T. V. McEvilly, J. M. Savino and L. R. Sykes (1971).
Discrimination between small-magnitude earthquakes and explosions, Journal of Geophysical
Research, 76, 8042-8055.
Karnik, V., N. V. Kondorskaya, Y. V. Riznichenko, E. F. Savarensky, S. L. Soloviev, N. V.
Shebalin, J. Vanek and A. Zatopek (1962). Standardization of the earthquake magnitude scale,
Studia Geophysica et Geo-detica, 6, 41-47.
Marshall, P. D. and P. W. Basham (1972). Discrimination between earthquake and underground
explosions employing an improved MS scale, Geophysical Journal of the Royal Astronomical
Society, 28, 431-458.
Nuttli, O. W. (1973). Seismic wave attenuation and magnitude relations for eastern North America,
Journal of Geophysical Research, 78, 876-885.
@seismicisolation
@seismicisolation
6 Chapter One
Table 1. Dates, times and locations of soviet explosions studied

Date Origin Time Lat. (N) Long. (E) Location
19-09-71 11-00-06.8 57.8 41.1 Western Russia
27-09-71 05-59-55.2 73.4 55.1 Novaya Zemlya
04-10-71 10-00-02.0 61.6 47.1 Western Russia
09-10-71 06-02-57.1 50.0 77.7 Eastern Kazakh
21-10-71 06-02-57.3 50.0 77.6 Eastern Kazakh
22-10-71 05-00-00.4 51.6 54.5 Western Russia
29-11-71 06-02-57.1 49.8 78.1 Eastern Kazakh
15-12-71 07-52-58.6 50.0 77.9 Eastern Kazakh
22-12-71 06-59-56.3 47.9 48.2 Western Russia
30-12-71 06-20-57.7 49.7 78.1 Eastern Kazakh
@seismicisolation
@seismicisolation
Table 2. Magnitudes and locations of events studied

Datea Origin Timeb Lat. (°N)b Long. (°E)b Depth(km)c mbc M Sc
01-08-71 18-55-10.6 44.4 148.9 4.3 4.0
(33)-N (4.5)-N
(39)-I (4.5)-I
05-08-71 22-37-10.9 12.6 94.8 5.0 5.0
(31)-N (5.0)-N
(21)-I (4.9)-I
(110)-M (4.5)-M
07-08-71 15-21-52.5 36.1 77.7 4.5 4.2
(33)-N (4.8)-N
(10)-I (4.9)-I
(5.2)-M
08-08-71 02-42-20.5 8.2 58.4 4.6 4.4
(33)-N (5.1)-N
(179)-I (4.5)-I
(4.6)-M
09-08-71 01-03-16.6 42.1 83.4 4.3 3.4
(33)-N (4.2)-N
(33)-I
02-08-71 04-17-05.6 12.6 95.1 5.1 5.0
(40)-N (5.3)-N (5.1)-N
(20)-I (5.3)-I
(5.5)-M
12-08-71 20-57-57.7 49.7 156.0 4.6 4.2
(45)-N (4.5)-N
(42)-K (4.6)-I
(5.1)-M
15-08-71 12-14-29.3 21.9 121.8 5.2 4.7
(28)-N (4.8)-N
(19)-I (5.0)-I
(5.0)-M
16-08-71 04-58-00.3 28.9 103.7 5.7 4.8
(33)-N (5.5)-N
(32)-I (5.5)-I
(5.6)-M
16-08-71 13-29-24.8 28.8 103.7 4.8 4.0
(33)-N (4.8)-N
(61)-I (4.8)-I
(5.0)-M
@seismicisolation
@seismicisolation
8 Chapter One
Table 2 (Cont'd)
16-08-71 18-53-54.7 28.9 103.7 17 5.4 5.4
(33)-N (5.3)-N (5.6) -N
(64)-I (5.3)-I
(5.6)-M
16-08-71 22-37-33.6 28.8 103.6 13 5.3 4.7
(33)-N (5.4)-N
(25)-I (5.3)-I
(20)-M (5.1)-M
17-08-71 04-29-33.1 37.1 36.8 4.8 4.4
(33)-N (5.0)-N
(35)-I (5.0)-I
(4.6)-M
17-08-71 17-07-40.4 28.9 103.7 4.9 4.2
(33)-N (4.9)-N
(88)-I (4.7)-I
(4.8)-M
19-08-71 11-12-40.7 21.9 121.8 5.0 4.9
(15)-N (4.8)-N
(11)-I (4.9)-I
(5.3)-M
20-08-71 19-06-28.6 61.8 5.0 4.6 3.8
(35)-N (4.5)-N
(35)-I (4.2)-I
21-08-71 19-34-23.2 81.9 118.9 4.4 4.4
(33)-N (4.6)-N
(84)-I (4.5)-I
(4.7)-M
22-08-71 17-54-14.6 30.1 50.7 4.8 4.5
(33)-N (5.1)-N
(51)-I (4.9)-I
(20)-M (5.1)-M
24-08-71 09-52-52.0 45.3 151.3 4.8 4.0
(33)-N (4.7)-N
(40)-I (4.7)-I
(5.5)-M
24-08-71 16-33-22.7 52.2 91.4 4.6 5.0
(33)-N (5.2)-N
(12)-I (5.4)-I
(20)-M (5.7)-M (5.6)-M
(5.8)-H (5.5)-H
@seismicisolation
@seismicisolation
Table 2 (Cont'd)
25-08-71 00-30-44.5 28.2 52.3 -d 3.7
(33)-N (4.1)-N
(43)-I
25-08-71 18-08-44.6 28.4 128.1 4.5 4.1
(40)-N (4.4)-N
(21)-I (4.4)-I
(4.8)-M
26-08-71 06-55-08.7 30.0 50.7 4.3 4.4
(45)-N (4.8)-N
(42)-I (4.7)-I
(4.7)-M
26-08-71 11-55-51.0 44.5 9.6 e -e
(33)-N (4.0)-N
(33)-I
28-08-71 16-34-44.4 37.6 55.8 4.7 4.0
(33)-N (4.8)-N
(33)-I (4.8)-I
(4.7)-M (4.5)-M
(4.9)-H
29-08-71 15-16-56.9 36.5 78.5 3.8 3.0
(33)-N (5.0)-N
(101)-I
(4.4)-M
01-09-71 10-54-04.6 48.4 154.9 4.8 3.8
(50)-N (4.9)-N
(50)-I (4.9)-I
02-09-71 12-24-22. 8 30.5 50.3 4.8 4.1
(33)-N (4.8)-N
(61)-I (4.8)-I
(4.3)-M
03-09-71 18-42-16.0 28.9 103.7 15 4.7 3.3
(33)-N (4.8)-N
(110)-I (4.7)-I
03-09-71 21-33-08.5 48.2 6.5 4.3 <3.6f
(26)-N
(26)-I
04-09-71 01-10-33.0 29.0 103.7 4.9 3.8
(33)-N (5.0)-N
(97)-I (4.7)-I
(4.6)-M
@seismicisolation
@seismicisolation
10 Chapter One
Table 2 (Cont'd)
05-09-71 14-55-21.0 56.0 165.1 5.0 4.8
(33)-N (5.1)-N (4.9)-N
(26)-I (5.1)-I
(25)-M (5.2)-M
06-09-71 00-33-25.9 33.2 69.9 5.2 4.5
(37)-N (4.9)-N
(26)-I (5.0)-I
(5.2)-M
06-09-71 04-21-43.9 46.8 141.4 4.7 3.8
(26)-N (4.8)-N
(16)-I (4.7)-I
07-09-71 04-02-24.2 46.1 12.4 4.1 3.2
(25)-N (4.2)-N
(28)-I (4.1)-I
07-09-71 07-54-28.7 43.2 0.2 4.1 3.3f
(33)-N (4.0)-N
(33)-I
07-09-71 15-28-32.8 24.0 123.2 4.6 3.9
(33)-N (4.6)-N
(49)-I
08-09-71 22-35-15.8 41.1 43.8 4.8 4.2
(33)-N (4.8)-N
(33)-I (4.8)-I
(5.0)-M (4.6)-M
(5.3)-H (4.3)-H
09-09-71 06-51-08.8 38.2 20.1 4.3 3.0
(5)-N (4.3)-N
(3)-I (4.4)-I
09-09-71 21-06-20.1 46.6 140.9 4.5 4.4
(33)-N (4.5)-N
(10)-I (4.5)-I
11-09-71 02-03-09.9 38.9 22.1 4.5 3.8
(4)-N (4.5)-N
(5)-I (4.5)-I
11-09-71 06-28-11.1 15.1 96.3 5.0 4.4
(28)-N (5.3)-N
(30)-I (5.0)-I
(5.4)-M
@seismicisolation
@seismicisolation
Table 2 (Cont'd)
14-09-71 03-11-04.3 22.9 100.8 5.3 5.4
(33)-N (5.4)-N
(42)-I (5.3)-I
14-09-71 06-56-31.8 42.4 144.7 4.9 5.0
(39)-N (4.9)-N
(54)-I (4.9)-I
(5.1)-M
19-09-71 11-00-06.8 57.8 41.1 4.4 -g
E
(33)-N (4.5)-N
(33)-I (4.5)-I
(5.3)-H
21-09-71 09-13-51.5 32.4 91.8 4.8 4.7
(33)-N (5.0)-N
(34)-I (4.6)-I
25-09-71 10-34-04.3 44.2 8.6 4.4 3.5
(19)-N (4.1)-N
(3)-N (4.0)-I
27-09-71 05-59-55.2 73.4 55.1 6.2 5.1
E
(0)-N (6.4)-N (5.2)-N
(0)-I (6.5)-I
(!6.4)-H (5.6)-H
28-09-71 14-13-09.3 40.2 143.4 4.8 4.5
(45)-N (4.2)-N
(26)-I (4.6)-I
29-09-71 07-18-51.6 47.1 9.0 4.4 3.7
(24)-N (4.5)-N
(24)-I (4.4)-I
29-09-71 15-47-58.3 55.4 163.6 5.0 4.1
(33)-N (5.0)-N
(0)-I (5.2)-I
01-10-71 16-27-47.7 38.6 69.8 4.8 4.1
(36)-N (4.9)-N
(21)-I (4.9)-I
(10)-M (5.1)-M
(4.9)-H (4.4)-H
03-10-71 18-38-22.0 41. 5 142.4 4.1 4.4
(48)-N (4.1)-N
(45)-I (4.2)-I
04-10-71 10-00-02.0 61.6 47.1 4.5 2.8f
E
(13)-N (5.1)-N
(13)-I (4.6)-I
(4.9)-H
@seismicisolation
@seismicisolation
12 Chapter One
Table 2 (Cont'd)
14-10-71 16-43-31.8 42.9 13.1 4.6 4.1
(33)-N (5.0)-N
(33)-I (4.5)-I
(5.1)-M
04-10-71 22-21-57.2 43.1 12.9 4.4 3.2
(43)-N (4.4)-N
(43)-I
05-10-71 18-31-17.7 27.2 55.8 5.0 4.1
(39)-N (5.1)-N
(44)-I (5.1)-I
(4.9)-M
06-10-71 01-46-38.3 38.3 30.2 4.5 3.9
(19)-N (4.6)-N
(19)-I (4.4)-I
(4.1)-M
09-10-71 06-02-57.1 50.0 77.7 5.2 3.2
E
(0)-N (5.4)-N
(0)-I (5.3)-I
(5.8)-H
10-10-71 18-25-14.6 23.0 96.0 5.1 5.2
(33)-N (5.1)-N
(46)-I (4.9)-I
(5.3)-M
12-10-71 11-44-42.2 44.5 11.0 4.1 3.6
(25)-N (4.6)-N
(9)-I
12-10-71 16-45-35.0 52.6 174.2 4.7 3.4
(29)-N (4.4)-N
15-10-71 14-19-31.6 37.3 54.6 4.5 3.8
(39)-N (4.7)-N
(41)-I (4.6)-I
(25)-M (5.0)-M (4.0)-M
(5.3)-H (3.8)-H
15-10-71 17-08-06.3 41.4 48.6 4.9 3.7
(33)-N (4.9)-N
(54)-I (4.8)-I
(5.0)-M
(5.2)-H (3.5)-H
@seismicisolation
@seismicisolation
Table 2 (Cont'd)
20-10-71 08-40-19.0 21.9 121.4 5.6 5.3
(35)-N (5.5)-N (5.3)-N
(42)-I (5.6)-I
(35)-M (5.9)-M
21-10-71E 06-02-57.3 50.0 77.6 5.5 3.5
(0)-N (5.6)-N
(0)-I (5.5)-I
(5.7)-H (3.5)-H
22-10-71E 05-00-00.4 51.6 54.5 5.2 2 .8f
(6)-N (5.3)-N
(6)-I (5.2)-I
(5.8)-H (3.3)-H
24-10-71 08-59-04.6 28.2 87.2 20 4.8 2.9
(44)-N (5.1)-N
(57)-I (4.8)-I
(5.0)-M
28-10-71 13-30-57.1 41.9 72.4 5.3 4.9
(22)-N (5.5)-N
(15)-I (5.4)-I
(5.7)-M
(5.3)-H (5.2)-H
29-10-71 17-16-52.l 34.1 86.3 5.7 4.2
(33)-N (5.0)-N
(6)-I (4.9)-I
(4.7)-M
30-10-71 20-48-48.0 23.0 121.4 5.3 5.3
(35)-N (5.3)-N (5.5)-N
(47)-I (5.3)-I
(5.6)-M
31-10-71 15-54-47.9 26.2 90.7 4.1 3.2
(33)-N (4.6)-N
(33)-I (4.7)-I
01-11-7 05-29-57.2 44.0 85.0 4.9 4.7
(33)-N (5.0)-N
(30)-I (5.0)-I
(30)-M (5.2)-M
03-11-71 09-42-50.4 28.3 57.0 4.5 4.2
(33)-N (4.7)-N
(102)-I (4.7)-I
(4.9)-M
04-11-71 20-12-20.5 28.8 103.7 4.9 3.9
(34)-N (5.0)-N
(43)-I (4.9)-I
(4.7)-M
@seismicisolation
@seismicisolation
14 Chapter One
Table 2 (Cont'd)
05-11-71 14-55-48.8 24.7 63.3 4.7 4.1
(33)-N (5.1)-N
(50)-I (4.9)-I
(50)-M (5.0)-M
08-11-71 23-24-43.8 63.0 5.1 4.2 4.1
(33)-N (4.5)-N
(33)-I
11-11-71 04-40-56.7 21.4 93.9 4.9 4.0
(48)-N (4.8)-N
(55)-I (5.0)-I
(5.5)-M
13-11-71 15-47-41.5 11.0 39.7 5.2 4.9
(24)-N (5.3)-N
(39)-I (5.1)-I
(15)-M (5.6)-M
18-11-71 07-31-32.8 38.3 66.8 5.3 5.0
(30)-N (5.3)-N
(27)-I (5.2)-I
(40)-M (5.6)-M 5.4(M)
(5.5)-H 5.3(H)
19-11-71 01-00-01.0 41.9 72.4 4.9 4.5
(33)-N (4.9)-N
(39)-I (4.9)-I
(4.9)-M (4.9)-M
(5.0)-H (4.4)-H
23-11-71 17-40-05.8 28.8 103.7 4.8 3.4
(33)-N (4.8)-N
(33)-I (4.8)-I
24-11-71 08-23-24.6 38.7 73.3 40 4.6 3.2f
(33)-N (5.1)-N
(95)-I
(110)-M (4.8)-M
(5.1)-H
29-11-71E 06-02-57.1 49.8 78.1 5.4 3.8
(0)-N (5.5)-N
(0)-I (5.4)-I
(5.7)-H
04-12-71 08-38-00.7 27.9 87.9 5.9 3.2
(32)-N (5.0)-N
(29)-I (5.2)-I
(5.4)-M
@seismicisolation
@seismicisolation
Table 2 (Cont'd)
07-12-71 22-34-14.7 35.8 67.7 4.6 3.9
(43)-N (4.3)-N
(34)-I
(4.3)-M
12-12-71 13-41-39.8 41.4 79.2 4.7 3.5
(33)-N (4.7)-N
(21)-I (4.7)-I
(4.1)-M
(4.9)-H (3.7)-H
12-12-71 22-27-41.1 39.5 73.2 4.9 4.4
(33)-N (4.8)-N
(33)-I
(20)-M (4.5)-M (4.4)-M
(4.6)-H (4.6)-H
15-12-71E 07-52-58.6 50.0 77.9 5.0 3.3
(0)-N (4.9)-N
(0)-I (4.9)-I
(5.0)-H
16-12-71 00-03-06.3 55.7 164.0 4.9 4.8
(41)-N (4.8)-N
(28)-I (4.8)-I
(4.8)-M
16-12-71 18-35-45.5 77.9 17.8 5.0 4.8
(33)-N (5.0)-N (4.9)-N
(33)-I (4.9)-I
(5.1)-M
19-12-71 07-50-27.8 55.9 163.1 5.1 5.2
(33)-N (5.0)-N (5.3)-N
(7)-I (5.0)-I
(5.7)-M
20-12-71 01-29-18.5 41.2 48.3 5.3 5.2
(33)-N (5.0)-N
(2)-I (5.1)-I
(15)-M (5.5)-M (5.3)-M
(5.4)-H (5.0)-H
20-12-71 01-41-04.9 41.1 48.4 5.2 5.1
(33)-N (5.2)-N (5.2)-N
(28)-I (5.1)-I
(5.4)-M (5.2)-M
(5.8)-H (5.3)-H
@seismicisolation
@seismicisolation
16 Chapter One
Table 2 (Cont'd)
20-12-71 05-05-24.0 41.1 48.2 4.7 4.1
(33)-N (4.5)-N
(27)-I
(15)-M (4.9)-M (4.0)-M
(4.9)-H
20-12-71 07-53-11.4 41.2 48.3 5.0 4.1
(33)-N (4.8)-N
(4)-I (4.8)-I
(4.8)-M (4.7)-M
(4.8)-H (4.3)- H
20-12-71 16-23-25.9 55.8 163.6 4.9 4.7
(33)-N (4.5)-N (4.8)-N
(2)-I (4.5)-I
(35)-M (5.1)-M
20-12-71 21-42-00.4 55.5 162.9 4.6 4.1
(33)-N (4.4)-N
(85)-I (4.3)-I
(4.0)-M
22-12-71E 06-59-56.3 47.9 48.2 5.8 4.2
(0)-N (6.0)-N
(0)-I (6.0)-I
(6.7)-H (4.3)-H
23-12-71 01-11-43.7 56.0 163.7 5.1 4.1
(13)-N (4.8)-N
(24)-I (4.7)-I
(4.3)-M
27-12-71 00-18-34.2 46.5 142.2 4.9 4.0
(39)-N (4.5)-N
(3)-I (4.5)-I
(4.9)-M
27-12-71 20-59-34.1 35.1 73.l 4.8 4.3
(10)-N (5.4)-N
(55)-I (5.2)-I
(4.5)-M
28-12-71 19-35-55.5 55.8 163.9 5.0 4.1
(33)-N (5.0)-N
(27)-I (4.9)-I
(4.6)-M
29-12-71 09-41-51.7 55.2 164.5 5.0 3.9
(35)-N (5.0)-N
(13)-I (4.9)-I
@seismicisolation
@seismicisolation
Table 2 (Cont'd)
29-12-71 22-27-02.0 25.1 94.7 5.5 5.0
(33)-N (5.5)-N
(46)-I (5.6)-I
(60)-M (5.6)-M
30-12-71E 06-20-57.7 49.7 78.1 5.5 3.9
(0)-N (5.8)-N
(0)-I (5.7)-N
a An “E” after the date indicates a presumed underground explosion.

b The values given were determined by the National Earthquake Information Center (NEIC),
Boulder, colorado.
c The values not enclosed in parentheses were determined in the present study. Those in
parentheses and indicated N, I, M or H were determined by the NEIC, International
Seismological Center (ISC), Moscow, U.S.S.R. and Hagfors Observatory, Sweden,
respectively.
d The P-wave amplitudes were too small to allow us to determine mb.
e The P-wave and surface-wave amplitudes were very small, and possibly interfered with by
waves from another earthquake.
f The P waves and/or surface waves were too small to determine their amplitudes. The values
for mb and/or MS given are upper limits, determined from the background microseismic level
on the seismograms.
g The surface waves are interfered with by those of an earthquake, so that MS could not be
determined for this explosion.
@seismicisolation
@seismicisolation
18 Chapter One
Figure 1. Theoretical attenuation curve and magnitude formulas for 20-sec period surface waves.
Figure 2. Seismograms showing 20-sec period waves at small distances.

@seismicisolation
@seismicisolation
Figure 3. Surface wave magnitudes for earthquake of 27 September 1971. The X's are MS values obtained from the
conventional formula (eq. 1), the solid triangles from the formula presented in this report (eq. 2), the pluses from the
formula of Marshall and Basham (1972), the rectangles from the formula of Karnik et al. (1962), and the circles from
the formula of Basham (1971).
@seismicisolation
@seismicisolation
20 Chapter One
Figure 4. Surface wave magnitudes for earthquake of 21 September 1971. The symbols are the same as in Figure 3.
@seismicisolation
@seismicisolation
@seismicisolation
@seismicisolation
22 Chapter One
Figure 8. mb:MS values for Eurasian explosions. The solid-line curve is for Nevada Test Site events.
@seismicisolation
@seismicisolation
Figure 9. ms:Mb values for Eurasian earthquakes. The dashed-line curve, obtained by displacing the explosion curve
by 0.5 mb units, is an envelope for most of the earthquakes.
ACKNOWLEDGMENTS
The authors wish to acknowledge helpful discussions with F. E. Followill and B. J. Mitchell. They
also wish to thank P. W. Basham and an anonymous referee for constructive criticism. This
research was supported by the Defense Advance Research Projects Agency (Contract F 19628-73-
C-0269) under the technical cognizance of Air Force Cambridge Research Laboratories.
@seismicisolation
@seismicisolation
CHAPTER TWO
DISCRIMINATION OF ANOMALOUS EARTHQUAKES

FROM UNDERGROUND NUCLEAR EXPLOSIONS 1
OTTO W. NUTTLI, SO GU KIM, HUI-YUIN WEN AND JOHN A. WAGNER
2-lb TNT detonation at a depth of 3 m at the foot of mountain showing seismic and air waves
including sonic boom on DR-2000 recorder about 2 km away.
ABSTRACT
A number of main shock-aftershock sequences in the Eurasian interior contain some aftershocks
whose mb: MS values are close to those of underground explosions which are explosion-like
earthquakes aka anomalous earthquakes. Anomalous earthquakes are often found among the
aftershocks of mainshock-aftershock intraplate earthquake sequences. This paper is concerned
with a study of the amplitude spectra of the P waves and Rayleigh waves for earthquakes of those
main shock-aftershock sequences. It is found that for any given sequence studied, there is little if
any variation in focal depth or focal mechanism. This rules out variations in these quantities as
being the cause of anomalous mb:MS values for the anomalous earthquakes. A study of the P-wave
spectra establishes that one or both of the corner periods of anomalous earthquakes are smaller
than those of non-anomalous earthquakes of the same moment. Thus the cause of anomalous
mb:MS values of the earthquakes studied is a relative enrichment of the short-period portion of the
spectrum of the anomalous events, which cannot be attributed to focal depth or focal mechanism.
Our results suggest that it would be especially difficult to discriminate between a small-magnitude
aftershock of an intraplate earthquake and a small to intermediate yield explosion detonated within
a few hours or days of the mainshock.

1
Some parts presented in Bull. Seism. Soc. Am., (1977). 67 (2), 463-478. Spectral and magnitude characteristic
s of anomalous Eurasian earthquakes by So Gu @seismicisolation
Kim and Otto W. Nuttli.
@seismicisolation
Discrimination of Anomalous Earthquakes from Underground Nuclear Explosions 25
INTRODUCTION
For a long time it has been known that large underground nuclear explosions have small MS values
compared to earthquakes having the same mb values as those of the explosions. This conclusion
has been found also to be generally true for intermediate and smaller sized explosions. That is,
when a plot of the mb values as abscissae and MS values as ordinates is made, the earthquakes
separate from the explosions, with the latter lying below and to the right of the earthquake points.
However, 1t has been found that some earthquakes plot in or near the explosion population. These
are called anomalous earthquakes.
The primary purpose of this research is to search for shallow-depth Eurasian earthquakes that have
anomalous mb: MS values. Furthermore, because it is for intermediate to small magnitude events
that it is most difficu1t to discriminate between earthquakes and explosions, events were selected
which had mb values in the range of 4 to 5.5.
Other aims of the research are to identify geographical regions where anomalous earthquakes
occur, to determine the effect, it any, of focal depth and focal mechanism on observed mb:Ms
values, to investigate differences in the P-wave spectra of anomalous and non-anomalous
earthquakes, and to determine Love-wave MS (MS,L) values for anomalous earthquakes.
METHODOLOGY
Standard techniques were used for determining the mb and MS values of the earthquakes and
explosions studied. That is, for Mb evaluation the largest amplitude in the first 3 cycles of the P-
wave motion on the vertical-component, short-period seismogram was used, along with the
Gutenberg-Richter (Richter, 1958) calibrating function. Average mb values, along with their
standard deviations and the number of stations used, are reported, as well as average mb values and
the number of stations as given by the National Earthquake Information Center and the
International Seismological Centre.
To determine MS the largest amplitude of the vertical component Rayleigh-wave motion in the
period range of 17 to 23 seconds was used, along with the formulas
MS = 3.30 + 1.66 log ǻ + log A/T for 25° ǻ 140°
and
MS = 4.16 + 1.07 log ǻ + log A/T for 10° ǻ 25°
where ǻ is the epicentral distance in degrees, A is the maximum ground motion in microns, and T
is the period in seconds. We examined the use of MS formulas derived for waves of periods
substantially less than 20 seconds as recorded at small to regional distances, and found in general
that they were not as satisfactory as the second of the formulas given above (Nuttli and Kim,
1975). Mean MS values, along with their standard deviations and number of stations used, are
presented. In general the National Earthquake information Center does not determine MS values
for small to intermediate magnitude events, and the International Seismological Centre never
determines MS values.
The seismograph stations whose data were used to determine mb, and MS values were the World-
Wide Standard Seismograph Network (WWSSN) stations ALQ, AQU, BLA, BUL, CHG, COL,
JCT, DUG, GOL, KOD, KON, NIL, KBL, LPS, MAT, KBS, NDI, NHA, OGD, QUE, SEO, SHI
and SHL and the Very Long Period Experiment (VLPE) stations FBK, ALQ, TLO, CHG, CTA
and EIL.
@seismicisolation
@seismicisolation
26 Chapter Two
Focal depths were determined for a selected number of earthquakes, principally by means of the
depth phases pP and sS. When depths of earthquakes in an aftershock sequence were compared,
we also used the period of the minimum in the Rayleigh-wave spectrum. The latter method is
principally of value in establishing that two or more earthquakes of the same region have the same
depth, rather than in making an absolute determination of that depth.
DATA
Much of the basic mb:MS data have already been presented in Scientific Report No. 1 (Nuttli and
Kim, 1974) and Scientific Report No. 3 (Nuttli, Kim and Wen, 1975). In Report No. l data for 96
earthquakes and 10 underground explosions were given, and in Report No. 3 the mbMS values for
155 earthquakes and 4 explosions were presented. Inasmuch as these reports are readily available,
a relisting of these data will not be given here.
Table 1 presents the hypocentral coordinates of 37 earthquakes and 9 explosions whose mb:MS
values have not been included in any previous report. Table 2 gives the mb and MS values of the
events described in Table 1.
The 23 nuclear explosions took place at the testing sites in Novaya Zemlya, eastern Kazakhstan
and western Kazakhstan and at scattered sites in western Russia. Figure 1 is a plot of the mb:MS
values of these explosions. The dashed line in the figure is the curve obtained by Evernden et al
(1971) from Nevada Test Site data when MS was determined from 20-second period Rayleigh
waves. The two solid-line curves mark the envelopes or outer bounds of the Eurasian data. From
the upper of these two curves we can see that all the explosion data satisfy the relation
mb – MS 1.2 for mb 4.6
The single point mb = 4.4, MS = 3.3 of Figure l suggests that for mb < 4.6 the difference mb - MS
may become smaller than 1.2, with the difference. decreasing as mb decreases. However, much
more data in the small magnitude range than we have are required to justify such a conclusion.
Table 3 contains a list of 33 earthquakes, out of the 278 studied, which had mb minus MS values of
1.2 or greater, and which thus would overlap the explosion population on an mb:MS diagram. The
mb:MS values of these earthquakes are plotted in Figure 2, along with the explosion bounds as
determined from Figure l. From Figure 2 it can be seen that the anomalous earthquakes listed in
Table 3 have mb:MS values that spread over the range of explosion values.
Considering next the non-anomalous earthquakes, our data indicate that shallow oceanic margin
earthquakes (Aleutians, Kamchatka, Japan, Taiwan, Philippines, New Hebrides, New Britain) all
have mb:MS values to the left and above the upper-bound curve of the explosion population. Thus
they present no problem in being distinguished from explosions. Figure 3 is a plot of mb:MS values
of oceanic margin earthquakes.
All of the anomalous earthquakes that are listed in Table 3 and plotted in Figure 2 are intraplate
earthquakes. Not all intraplate earthquakes, however, are anomalous. Thus, for example, of the 128
intraplate earthquakes that we studied for the year 1972, only 23 had anomalous mb:MS values.
For the 1972 earthquakes, surface-wave magnitudes were calculated using Love-wave amplitudes
as well as the amplitudes of the vertical-component Rayleigh waves. Of the 122 earthquakes for
which Love-wave MS values could be determined, only 19 had |MS,R - MS,L| > 0.3. Thus, in most
cases, the MS,R and MS,L values were essentially the same. Of the 19 which showed a significant
difference, only 6 had MS,L > MS,R. Thus there is little evidence of small MS,R values compared to
MS,L values, as might be expected to occur if the focal depth were such that there would be a
minimum in the Rayleigh-wave spectrum at a period near 20 seconds. From this we conclude that
focal depth (unless it is very large, i.e. greater than the crustal thickness) cannot generally be called
@seismicisolation
@seismicisolation
upon to explain small MS values of earthquakes, and thus anomalous mb:MS values.
Evernden (1976) discussed the use of MS,L values to discriminate between earthquakes and
explosions. At places where the release of regional tectonic strain by the explosion is low, the
resulting MS,L values are small compared to the MS,R values. Of the 23 explosions considered in
this report, Love waves of 20-second period could be identified for only two. The first is the East
Kazakhstan explosion of 16 August 1972, for which a very small amplitude Love wave was
tentatively identified at a single VLPE station (KON). For this event MS,R is 3.7 and MS,L is 3.0.
The second event occurred in Novaya Zemlya on 28 August 1972. It was a large explosion, and
20-second period Love waves were identified at 18 stations. For it MS,R was 5.0 and MS,L was 4.6.
MAINSHOCK-AFTERSHOCK SEQUENCES
One of the most important findings of this study was that some aftershocks of a sequence can have
anomalous mb:MS values, whereas the mainshock and other aftershocks have non-anomalous
values. This finding is particularly important because it enables us to obtain a better understanding
of the cause of anomalous earthquakes, by eliminating as an explanation certain characteristics that
are often proposed as the causes of anomalous mb:MS values.
Among these are the source and receiver crust and the transmission path, which are identical for all
events of a sequence for a given seismograph station, and the focal depth and focal mechanism,
which are shown to be the same for selected events of selected sequences.
A more complete discussion of the data and conclusions is given in Appendix 1, which is a
manuscript of a paper submitted for publication. In this section of the report we merely present
some of the data and summarize the conclusions.
Mainshock-aftershock sequences in Szechwan province of China, Tibet, the Iraq-Iran border
region and the New Hebrides islands were investigated. Table 4 lists the hypocentral coordinates
and the mb and MS values of the earthquakes studied. For the events which were of sufficiently
large magnitude, focal mechanisms and P and Rayleigh-wave spectra were determined. The focal
mechanism solutions verified that the earthquakes of a given sequence had nearly identical
mechanisms, and the Rayleigh-wave spectra, along with the time differences pP-P and sS-S,
established that the events of a given sequence had nearly identical focal depths.
From the P-wave spectra it was found that for earthquakes of the same MS or seismic moment,
MO, the anomalous earthquakes had smaller spectral corner periods than the non-anomalous ones.
This indicated that there is a relative enrichment of the short-period part of the spectrum for the
anomalous earthquakes. Observations in the time domain showed that for earthquakes of the same
mb the duration of the short-period P-wave motion was smaller (more pulse-like) for anomalous
than for non-anomalous earthquakes, even though the maximum amplitude of the P-wave motion
was the same.
Although we have no conclusive explanation as to the cause of the anomalous aftershocks, the
observations in both the frequency and time domains suggest they are related to the time history of
the fault rupture and possibly to the stress drop.
RELEVANCE OF THE RESEARCH TO THE DISCRIMINATION OF UMDERGROUND
EXPLOSIONS
One point of immediate significance is the finding that anomalous earthquakes do not occur at
Plate margins, but rather only within the Eurasian plate. (We did not look at earthquakes in other
continental areas.) Because earthquakes occurring along plate margins constitute the great majority
of earthquakes, our finding suggests that for this large class of earthquakes the mb:MS criterion
will readily separate explosions from earthquakes.
@seismicisolation
@seismicisolation
28 Chapter Two
A second point of significance is that the anomalous earthquakes of Eurasia are not restricted to
any single or few geographic areas. Table 3 demonstrates that they occur anywhere from southern
Europe to eastern Asia.
Although from theory one might expect a minimum in the Rayleigh-wave spectrum at 20-second
period for a certain range of focal depths within the crust, and thus anomalous mb:MS values for
these earthquakes, our data suggest that for actual earthquakes recorded at a sufficient number of
stations this is not an important phenomenon. Nor does the character of the fault motion, whether
it be strike-slip or thrust faulting, lead to anomalous mb:MS values by itself if the Rayleigh waves
are recorded over a sufficient range of azimuth.
The mainshock-aftershock studies indicate that for some cases, at least, it is not focal depth, focal
mechanism, source crust, crust-mantle transmission path or receiver crust which is the cause of
anomalous mb:MS values. Whatever is the explanation, it causes the P-wave spectra to be enriched
at the short periods for the anomalous earthquakes by shifting the spectral corner period (or
periods) to smaller values. Our finding that anomalous mb:MS earthquakes fairly often occur as
aftershocks of intraplate earthquakes suggests that it might be a difficult task to distinguish
between a small magnitude explosion and an aftershock of an intraplate earthquake, if the
explosion were deliberately detonated after the mainshock.
REFERENCES
Evernden, J. F. (1976). Study of seismological evasion Part I. General discussion of various
evasion schemes, Bulletin of the Seismological Society of America, 66, 245-280.
Evernden, J. F., W. J. Best, P. W. Pomeroy, T. V. McEvilly, J. M. Savino, and L. R. Sykes (1971).
Discrimination between small-magnitude earthquakes and explosions, Journal of Geophysical
Research, 76, 8042-8055.
Nuttli, O. W. and S. G. Kim (1974). Research in Seismology: Earthquake Magnitudes, Scientific
Report No. 1, AFCRL-TR-74-0326, Air Force Cambridge Research Laboratories, Hanscom
AFB, Massachusetts.
Nuttli, O. W. and S. G. Kim (1975). Surface-wave magnitudes of Eurasian earthquakes and
explosions, Bulletin of the Seismological Society of America, 65, 693-709.
Nuttli, O. W., S. G. Kim, and H. Y. Wen (1975). Research in Seismology: Earthquake
Magnitudes, Scientific Report No. 3, AFCRL-TR-75-0433, Air Force Cambridge Research
Laboratories, Hanscom AFB, Massachusetts.
Richter, C. F. (1958). Elementary Seismology, W. H. Freeman and Company, San Francisco.
PUBLICATIONS RESULTING FROM RESEARCH PERFORMED UNDER CONTRACT

FI9628-73-C-0269
Nuttli, O. W. and S. G. Kim (1974), Research in Seismology: Earthquake Magnitudes, Scientific
Report No. 1, AFCRL-TR-74-0326, Air Force Cambridge Research Laboratories, Hanscom
AFB, Massachusetts.
Duda, S. J. and O. W. Nuttli (1974), Earthquake magnitude scales, Geophysical Surveys, l, 429-
458.
Nuttli, O. W., S. G. Kim and H. Y. Wen (1975). Research in Seismology: Earthquake magnitudes,
Scientific Report No. 3, AFCRL-TR-75-0433, Air Force Cambridge Research Laboratories,
Hanscom AFB, Massachusetts.
Nuttli, O. W. and S. G. Kim (1975). Surface-wave magnitudes of Eurasian earthquakes and
explosions, Bulletin of the Seismological Society of America, 65, 693-709.
Kim, S. G. (1976). Spectral Scaling of Earthquakes in Some Eurasian Aftershock Sequences,
Doctoral Dissertation, Saint Louis University, 203 pp.
Kim, S. G. and O. W. Nuttli (1976). Spectral and magnitude characteristics of anomalous Eurasian
earthquakes, submitted to the Bulletin of the Seismological Society of America.
@seismicisolation
@seismicisolation
Table 1. Hypocentral coordinates of events studied
Depth(km)
Event No.a Date Origin Timeb Lat. (°N)b Long. (°E)b Location
NEIC ISC
1 03-07-72 02-10-00.4 30.1 50.8 38 27 Iran
2 03-07-72 21-38-22.2 30.0 51.0 43 35 Iran
3 05-07-72 01-09-52.9 44.6 81.1 33 35 N. Sinkiang
4 05-07-72 04-09-49.0 43.6 87.9 33 33 N. Sinkiang
5E 06-07-72 01-02-57.7 49.7 78.0 0 0 E. Kazakhstan
6 07-07-72 12-04-11.6 20.5 98.1 27 37 Burma
7E 09-07-72 06-59-57.9 49.8 35.4 0 0 S. W. Russia
8 10-07-72 19-03-33.0 43.4 88.6 33 36 N. Sinkiang
9 16-07-72 02-20-23.6 32.5 95.9 33 19 Tibet
10 16-07-72 03-39-59.8 32.6 95.8 33 37 Chinghai
11 21-07-72 16-08-12.1 22.6 121.5 21 47 Taiwan
12 22-07-72 05-10-39.5 44.9 36.9 33 0 S. W. Russia
13 22-07-72 16-41-04.0 31.4 91.5 33 17 Tibet
14 22-07-72 21-00-08.6 31.4 91.4 33 17 Tibet
15 09-08-72 19-42-17.3 53.0 107.5 33 33 Lake Baikal
16 10-08-72 21-06-40.1 32.4 93.5 33 34 Tibet
17 11-08-72 02-22-14.2 44.7 102.0 33 26 Mongolia
18 12-08-72 23-47-57.3 41.1 22.7 12 12 Yugoslavia
19E 16-08-72 03-16-57.2 49.8 78.1 0 0 E. Kazakhstan
20E 20-08-72 02-59-57.9 49.5 48.2 0 0 W. Kazakhstan
21E 26-08-72 03-46-56.9 50.0 77.8 0 0 E. Kazakhstan
22E 28-08-72 05-59-56.5 73.3 55.1 0 0 Novaya Zemlya
23 30-08-72 15-14-09.9 36.7 96.5 33 17 Chinghai
24 31-08-72 14-03-16.3 52.3 95.4 33 21 central Russia
25E 02-09-72 08-56-57.6 50.0 77.7 0 0 E. Kazakhstan
26 02-09-72 10-37-39.4 39.9 53.7 33 50 Turkmenyia
27 03-09-72 17-46-17.2 36.0 73.3 33 62 N. W. Kashmir
28 03-09-72 20-32-18.4 35.9 73.5 33 49 N. W. Kashmir
29E 14-09-72 07-00-03.6 67.7 33.4 7 7 W. Russia
30 10-09-72 20-57-57.1 39.2 81.3 33 38 S. Sinkiang
31E 21-09-72 09-00-01.2 52.1 52.0 28 28 W. Russia
32 27-09-72 00-08-29.9 30.3 101.7 33 0 Szechwan
33 27-09-72 02-03-39.1 33.9 72.7 46 41 Pakistan
34 29-09-72 13-56-59.5 39.7 77.8 33 36 S. Sinkiang
35 29-09-72 16-21-38.0 30.4 101.5 33 3 Szechwan
36 29-09-72 20-24-42.0 30.3 101.7 33 33 Szechwan
a An “E” indicates a presumed underground nuclear explosion
b The values given were determined by the National Earthquake Information Center (NEIC), Boulder, Colorado
@seismicisolation
@seismicisolation
30 Chapter Two
Table 2. Body-wave (mb) and surface-wave (ms) magnitudes of events studied
Event No.a ms(NEIC)b ms(ISC)b mbb,c MS(NEIC)b MS,RZb,c MS,Lb,c

1 5.0(8) 4.9(20) 4.79±0.18(5) --- 4.06±0.41(18) 3.99±0.26(b)
2 5.1(7) 4.9(23) 4.85±0.19(5) --- 4.66±0.65(23) 4.60±0.40(4)
3 4.6(2) 4.2(4) 4.01±0.40(2) --- 4.08±0.35(23) 4.24±0.40(8)
4 4.3(3) 4.2(4) 4.02±0.17(2) --- 3.31±0.34(5) ---
5E 4.4(4) 4.4(6) --- --- 3.35±0.01(3) ---
6 5.0(7) 4.8(21) 5.17±0.39(9) 5.5(2) 5.04±0.30(25) 5.30±0.31(12)
7E --- 4.8(5) 4.33±0.47(6) --- 3.43±0.33(5) ---
8 4.7(2) 4.5(4) 3.50(1) --- 3.87±0.34(22) 3.65±0.24(11)
9 5.2(10) 5.1(30) 5.22±0.18(7) --- 4.38±0.53(22) 4.52±0.55(14)
10 4.7(4) 4.6(5) 4.37±0.06(2) --- 4.01±0.48(20) 4.17±0.40(17)
11 4.8(2) 4.5(3) 4.30±0.14(3) --- 3.97±0.50(14) 3.81±0.20(4)
12 4.6(10) 4.9(11) 4.38±0.26(3) --- 4.21±0.60(21) 3.44±0.08(2)
13 5.8(20) 5.4(36) 5.09±0.16(5) --- 5.29±0.26(30) 5.47±0.43(16)
14 4.7(4) 4.6(4) 5.06(1) --- 3.49±0.47(11) 3.92±0.23(6)
15 5.1(10) 5.0(21) 5.01±0.52(10) 4.7(2) 4.62±0.41(26) 4.44±0.42(11)
16 5.2(9) 5.0(23) 4.89±0.22(5) 4.8(1) 4.43±0.25(23) 4.34±0.30(18)
17 5.0(8) 5.0(15) 4.64±0.41(6) 5.1(1) 4.69±0.29(26) 4.42±0.36(9)
18 4.9(7) 4.6(6) 4.56±0.31(5) --- 4.07±0.39(20) 4.03±0.43(3)
19 E 5.2(18) 5.0(30) 4.89±0.29(11) --- 3.48±0.42(3) ---
20 E 5.7(29) 5.7(46) 5.73±0.29(11) --- 3.66±0.34(4) ---
21 E 5.5(14) 5.3(37) 5.09±0.29(8) --- 3.73±0.37(9) 3.02(1)
22 E 6.3(23) 6.3(71) 6.33±0.23(9) 4.7(1) 5.02±0.32(27) 4.64±0.24(18)
23 5.5(22) 5.5(39) 5.07±0.32(6) 5.3(3) 5.08±0.33(25) 4.90±0.28(16)
24 5.5(34) 5.5(51) 5.29±0.36(11) 4.9(1) 4.74±0.21(28) 4.66±0.40(13)
25 E 5.1(6) 4.9(10) 4.46±0.40(6) --- 2.76±0.57(3) ---
26 4.9(4) 4.8(6) 4.26±0.11(2) --- 3.90±0.52(7) 3.40(1)
27 5.1(2) 4.6(6) 4.53±0.32(4) --- 4.99±0.33(19) 5.13±0.19(8)
28 5.2(5) 5.2(5) 4.89(1) --- 3.77±0.18(2) ---
29 E 4.6(6) 4.6(6) 4.66±0.45(3) --- 3.32±0.21(3) ---
30 5.1(14) 5.0(25) 5.03±0.34(6) --- 3.96±0.27(8) ---
31 E 5.1(12) 5.0(19) 4.67±0.33(10) --- 3.65±0.32(3)
32 5.0(8) 5.0(14) 4.89±0.25(5) 5.5(2) 5.04±0.28(23) 5.07±0.25(11)
33 4.9(7) 5.1(17) 4.83±0.40(6) --- 4.55±0.40(20) 4.30±0.29(7)
34 5.1(3) 4.7(14) 4.36±0.26(3) ---- 3.38±0.52 (8) ---
35 5.1(9) 5.1(20) 5.00±0.33(5) 5.6(2) 5.06±0.28(23) 5.14±0.24(11)
36 5.1(7) 5.1(17) 4.67±0.20(8) 5.4(2) 5.00±0.28(24) 5.00±0.30(11)
a The hypocentral coordinates of each event are given in Table 1.
b The numbers in parentheses indicate the number of stations whose data were used.
c The number following the “±” sign is the standard deviation.
@seismicisolation
@seismicisolation
Table 3. Earthquakes for which mb (isc)-ms 1.2
Date Origin Time Lat. ( °N) Long. (°E) mb୉MS Location

03-09-71 18-42-16.0 28.9 103.7 1.4 Szechwan
09-09-71 06-51-08.8 38.2 20.1 1.4 Greece
04-10-71 22-21-57.2 43.1 12.9 1.2 Central Italy
24-10-71 08-59-04.6 28.2 87.2 1.9 Tibet
31-10-71 15-54-47.9 26.2 90.7 1.5 Eastern India
23-11-71 17-40-05.8 28.8 103.7 1.4 Szechwan
24-11-71 08-23-24.6 38.7 73.3 2.2 Tadzhikistan-Sinkiang
04-12-71 08-38-00.7 27.9 87.9 2.0 Nepal
20-12-71 05-05-24.0 41.1 48.2 1.4 E. Caucasus
27-12-71 20-59-34.1 35.1 73.1 1.6 Pakistan
06-01-72 09-41-33.2 30.3 50.5 1.4 Iran
25-01-72 20-24-38.9 43.8 13.4 1.3 Central Italy
25-01-72 23-22-17.1 43.8 13.4 1.6 Central Italy
04-02-72 14-08-21.7 30.3 84.6 1.2 Tibet
06-02-72 07-30-11.4 41.6 82.2 1.2 S. Sinkiang
20-02-72 03-02-14.0 34.6 80.3 1.3 Tibet
28-02-72 02-04-35.0 40.4 29.1 1.2 Turkey
03-03-72 21-26-51.3 44.7 18.4 1.5 Yugoslavia
04-03-72 08-22-16.6 42.1 83.3 1.2 S. Sinkiang
08-04-72 06-42-13.3 29.7 89.5 1.4 Tibet
09-04-72 10-43-56.3 42.0 84.6 1.5 S. Sinkiang
06-05-72 22-05-19.9 28.3 102.3 1.3 Szechwan
07-05-72 06-36-03.0 40.2 78.9 1.2 S. Sinkiang
08-05-72 09-20-54.5 41.6 23.5 1.3 Greece-Bulgaria
16-05-72 10-59-52.6 28.4 52.6 1.2 S. Iran
20-05-72 06-44-26.1 28.3 52.8 1.7 S. Iran
23-05-72 03-14-28.2 41.7 23.6 Greece-Bulgaria
23-05-72 18-17-14.1 38.5 70.2 1.2 Tadzhikistan
30-05-72 06-38-16.8 38.3 69.5 1.5 Tadzhikistan
14-06-72 00-49-54.4 40.1 51.9 1.3 Caspian Sea
17-06-72 09-02-47.5 48.3 14.5 1.4 Austria
03-09-72 20-32-18.4 35.9 73.5 1.4 N.W. Kashmir
29-09-72 13-56-59.5 39.7 77.8 1.3 S. Sinkiang
@seismicisolation
@seismicisolation
32 Chapter Two
Table 4. Hypocentral coordinates of foreshocks, mainshocks and aftershocks studied
Region Date Origin Time Lat.(°N) Long.(°E) Depth(km) mb MS

Szechwan 16-08-71 04-58-00.3 28.9 103.7 19 5.7 4.8
16-08-71 13-29-24.8 28.8 103.7 4.8 4.0
16-08-71 18-53-54.7 28.9 103.6 17 5.4 5.4
16-08-71 22-37-33.6 28.9 103.6 13 5.3 4.7
17-08-71 09-36-15.5 28.9 103.7 5.0 4.3
17-08-71 17-07-40.4 28.9 103.7 4.9 4.2
03-09-71 18-42-16.0 28.9 103.7 15 4.7 3.3
04-09-71 01-10-33.0 29.0 103.7 4.9 3.8
04-11-71 20-12-20.5 28.8 103.7 4.9 3.9
23-11-71 17-40-05.8 28.8 103.7 4.8 3.4
Iraq-Iran 14-01-72 22-10-03.7 32.8 46.9 5.0 4.1
10-06-72 19-31-41.8 32.9 46.3 4.7 3.6
12-06-72 13-34-00.7 33.1 46.3 16 5.3 4.9
12-06-72 13-39-58.8 33.1 46.2 5.0 4.5
13-06-72 00-55-37.3 33.1 46.3 12 5.1 4.5
14-06-72 04-34-28.l 33.0 46.1 15 5.3 4.1
23-06-72 08-39-35.8 32.9 46.2 4.5 3.8
Tibet 21-09-71 09-13-51.5 32.4 91.8 4.8 4.7
24-10-71 08-59-04.6 28.2 87.2 20 4.8 2.9
29-10-71 17-16-52.1 34.1 86.3 18 4.7 4.2
04-12-71 08-38-00.7 27.9 87.9 4.9 3.2
04-02-72 14-08-21.7 30.4 84.6 5.0 4.2
20-02-72 03-02-14.0 34.6 80.3 4.8 3.5
15-03-72 06-00-32.4 30.4 84.5 5.1 4.1
08-04-72 06-42-13.3 29.7 89.5 4.4 3.4
21-04-72 21-19-29.5 35.0 81.0 4.9 4.0
28-04-72 00-52-56.8 31.3 84.9 5.2 4.2
14-07-73 04-51-21.0 35.2 86.5 19 6.0 6.6
14-07-73 13-39-30.0 35.3 86.6 22 6.0 5.4
New 23-01-72 17-18-39.4 -13.1 166.3 25 5.3 5.4
Hebrides
23-01-72 18-04-00.2 -13.2 166.6 5.5 5.5
23-01-72 21-17-52.1 -13.2 166.4 28 5.9 6.8
24-01-72 00-18-33.5 -13.2 166.3 4.9 5.1
24-01-72 03-55-42.5 -13.0 166.4 26 5.9 5.9
24-01-72 09-27-07.3 -13.2 165.9 4.7 4.9
24-01-72 15-42-51.9 -13.2 166.5 4.5 4.4
26-01-72 01-44-45.5 -13.0 166.3 5.1 4.9
@seismicisolation
@seismicisolation
Figure l. mb:MS values of Eurasian underground explosions. The mb values are those of the International
Seismological Centre, and the MS values are those determined in the present study. The solid-line curves mark the
upper and lower bounds of the explosion data. The dashed-line curve is the NTS explosion curve, where MS is
determined from 20-second period surface waves, as given by Evernden et al (1971).
@seismicisolation
@seismicisolation
34 Chapter Two
Figure 2. mb:MS values of anomalous Eurasian earthquakes, for which mb-MS 1.2. The solid-line curves are the
upper and lower bounds of the explosion data, as given in Figure 1.
@seismicisolation
@seismicisolation
Figure 3. mb:MS values of Eurasian oceanic margin earthquakes of 1972. The solid-line curves are the upper and lower
bounds of the explosion data, as given in Figure 1.
@seismicisolation
@seismicisolation
36 Chapter Two
Appendix
SPECTRAL AND MAGNITUDE CHARACTERISTICS OF ANOMALOUS EURASIAN
EARTHQUAKES 2
By So Gu Kim* and Otto W. Nuttli
ABSTRACT
A number of main shock-aftershock sequences in the Eurasian interior contain some aftershocks
whose mb:MS values are close to those of underground explosions. This paper is concerned with a
study of the amplitude spectra of the P waves and Rayleigh waves for earthquakes of those
mainshock-aftershock sequences. It is found that for any given sequence studied there is little if
any variation in focal depth or focal mechanism. This rules out variations in these quantities as
being the cause of anomalous mb:MS values. A study of the P-wave spectra establishes that one or
both of the corner periods of anomalous earthquakes are smaller than those of non-anomalous
earthquakes of the same moment. Thus the cause of anomalous mb:MS values of the earthquakes
studied is a relative enrichment of the short-period portion of the spectrum of the anomalous
events, which cannot be attributed to focal depth or focal mechanism.
INTRODUCTION
A recent study of mb and MS values of approximately 300 intermediate-size Eurasian shallow-
focus earthquakes revealed that most anomalous earthquakes (events whose mb:MS values are
closer to those of explosions than to those of ordinary earthquakes) occurred in the continental
interior rather than at the oceanic margins (Nuttli and Kim, 1975). Additional study has established
that the majority of anomalous events occurred as aftershocks, or in some cases foreshocks, of
non-anomalous main-shocks. Not all foreshocks or aftershocks, however, were found to be
anomalous.
The validity of using mb:MS values in distinguishing between earthquakes and explosions has been
explained (Molnar et al., 1969; Lieberman and Pomeroy, 1970) in terms of a number of factors,
namely source-time functions, source size, properties of the travel path, source mechanism and
focal depth. Tsai and Aki (1971) argued against the first two factors as being the primary cause of
the difference in mb:MS values for earthquakes and explosions. Rather they maintained that the last
two are dominant after corrections due to differences in travel path are taken into account.
Evernden (1975) also concluded that anomalous events can be explained principally by differences
in source mechanism and focal depth.
On the other hand, Kanamori and Anderson (1975) demonstrated that inter-plate earthquakes
which occur along or parallel to major plate boundaries show low stress drop compared to intra-
plate earthquakes which occur within the plates. Forsyth (1975) concluded that anomalous mb:MS
values observed for a foreshock-mainshock-aftershock sequence in the Kirgiz-Sinkiang border
region resulted from the source finiteness or the rupture-time duration rather than from the focal
depth or the focal mechanism.
EARTHQUAKES STUDIED
In this paper we restrict our discussion to single mainshock-aftershock sequences in the Szechwan
province of China, in Tibet, in the Iraq-Iran border region and in the New Hebrides Islands. Table
1 gives the latitude, longitude, origin time and focal depth as determined by the National
Earthquake Information Service for the earthquakes whose focal mechanisms and P-wave and

2
Some parts presented in Bull. Seism. Soc. Am., (1977). 67 (3), 463-478. Spectral and Magnitude Characteristics of
Anomalous Eurasian Earthquakes by So Gu Kim and Otto W. Nuttli.
* Present address: Seismograph Service Corporation, Tulsa, Oklahoma USA.
@seismicisolation
@seismicisolation
Rayleigh-wave spectra were determined in this study. Also included are the mb and MS values as
determined by us.
An inspection of Table 1 will show that the hypocentral coordinates of the earthquakes of a given
sequence show little variation. The time differences, pP - P and sS - S, when they could be
observed, indicated a nearly constant focal depth for all the events of a sequence. Further evidence
for the same focal depth (about 16km) is given by the periods of the minimum in the Rayleigh-
wave spectra, which for a given station are almost constant at around 40 s for all the events of a
sequence. Sample Rayleigh-wave spectra for two stations for the Szechwan and New Hebrides
earthquakes are given in Figures 1 and 2, respectively.
For all four earthquake sequences the focal mechanisms of individual earthquakes of a given
sequence are remarkably similar. This is demonstrated in Figures 3a and 3b, in which solutions
based on the sign of the onset of the long-period P motion are presented.
By selecting for study events of mainshock-aftershock sequences which have the same hypocentral
coordinates, the same focal mechanism and the same transmission path, we eliminate many of the
quantities which have been proposed in explanation of anomalous mb:MS values. We can thus
restrict our attention to quantities such as source finiteness, rupture velocity and stress drop.
Although we shall not present a physical explanation as to why some earthquakes have anomalous
mb:MS values, we wish to point out certain ways in which the spectra of anomalous earthquakes
differ from those of non-anomalous ones. We shall leave to others the interpretation of these
spectral differences, and merely note that differences in the corner periods and slopes of body-
wave spectra have been interpreted in terms of source finiteness (Hanks and Wyss, 1972; Forsyth,
1975) and of fractional stress drop (Brune, 1970).
TIME-DOMAIN OBSERVATIONS
Direct observation of the seismograms led us originally to a study of mainshock-aftershock
sequences in a search for anomalous events. That is, by looking at the seismograms we could see
that events which produced nearly the same P-wave amplitudes at a station produced noticeably
different Rayleigh-wave amplitudes. Figure 4 gives some examples. on the left-hand side are the
seismograms for the Szechwan earthquakes. The upper left shows the P-wave motion at two
stations, BUL and SHK, for the three events. Maximum amplitudes in the first few cycles are
about the same (the onset of the P motion at SHK for event no.3 is obscured by the minute marks;
its peak and trough almost touch the trace above and below, respectively). The Rayleigh-wave
amplitudes for event no.3 as recorded at station SHI, are noticeably less than for events 1 and 2.
The P waves and Rayleigh waves for the two Tibetan events are given in the right-hand side of
Figure 4. The P-waves are shown for stations NUR, STU and KTG and the Rayleigh waves for
station KTG. Although the duration of the P waves of event no. 1 is greater than of no. 2, the
amplitudes of the first few cycles of P of no. 1 are almost the same as the amplitude of no.2, which
would result in nearly similar mb values for the two earth-quakes. The surface-wave amplitudes of
event no. 1, however, are much greater than of event no. 2, as can be seen from the lower right-
hand figure.
Figures 5 and 6 show as examples mb:MS plots for the Tibet and New Hebrides earthquakes,
respectively. The points labeled 1 and 2 in Figure 5 refer to the Tibetan earthquakes no. 1 and 2 in
Table 1. The other points refer to earthquakes for which mb:MS values could be determined but
which were too small for spectral analysis. Curve a in Figure 5 is a least-square linear fit to mb:MS
data from 64 non-anomalous Eurasian earthquake values, and curve c a similar fit for 11 Eurasian
underground explosion values. The data sets were taken from Nuttli and Kim (1975). Curve b is a
linear fit to the Tibetan data points, indicated by x's in the figure. Note that the mainshock (event
no. 1) as well as two of the other Tibetan events would be considered non-anomalous. Event no. 2,
@seismicisolation
@seismicisolation
38 Chapter Two
as well as 6 others, would be considered suspicious by the mb:MS criterion. The two remaining
events definitely fall in the explosion population.
Curves a and c of Figure 6 are the same as described for Figure 5. If event no. 2 is considered to be
the mainshock, its MS value is even larger than for the ordinary Eurasian earthquake of mb = 5.9.
The foreshock (no.1) and all the aftershocks also scatter around curve a. Curve b is the linear fit to
the eight data points of the figure. Thus none of the New Hebrides earthquakes would be
considered to have anomalous mb:MS values.
RAYLEIGH-WAVE SPECTRA
Rayleigh-wave ground-motion spectra were obtained from analysis of the long-period, vertical-
component seismograms. Stations were selected so that the path from epicenter to station was pure
continental (Figure 1) or pure oceanic (Figure 2). The observed spectra were adjusted for the
effects of geometric spreading and anelastic attenuation, to reduce them as if the seismograph
station were 1000 km from the epicenter.
Sample spectra are shown in Figures 1 and 2. Values below 10 sec are unreliable due to both small
instrument magnification and low signal amplitude. The spectra can be used for two purposes: 1)
to determine whether differences in log spectral amplitude at 20-sec period are equal to differences
in MS values, i.e. to test the equivalence of amplitude differences in the frequency domain with
amplitude differences in the time domain, 2) to determine the seismic moment and depth from
holes at around 40s estimating depth at about 21 km (16 km reported) using Rayleigh wave
amplitude spectra for the pure continental path of NDI and QUE by comparing spectral levels of
actual earthquakes with those calculated for a unit-moment, double-couple source at the
appropriate azimuth for a distance of 1000 km.
Table 2 presents the MS differences and the differences in the log spectral amplitude at 20-sec
period for the 11 events in the 4 geographic regions. The number n refers to the number of stations
whose spectral amplitudes were used to obtain an average log10 ᶭA. In general the differences
between ᶭMS and log10 ᶭA are small, of the order of 0.1 to 0.2 MS units. The exception is the set
of Szechwan earthquakes; either MS of event no. 1 is too small or its 20-sec period spectral
amplitude is too large. From studies of the P-wave spectra we believe that the MS value of this
event is about 0.3 units too large, which corresponds to one standard deviation.
Table 3 compares the seismic moment, as obtained from the Rayleigh-wave spectrum, with the
body-wave magnitude. From the table it can be seen that non-anomalous earthquakes of a given
body-wave magnitude have larger moments than anomalous earthquakes of the same magnitude,
sometimes by more than a power of ten.
P-WAVE SPECTRA
P-wave ground-motion spectra were determined from the LPZ and SPZ seismograms for stations
at teleseismic distances. As they were not normalized to any particular epicentral distance, in
comparing them one should compare the spectra of an individual station for an individual
mainshock-aftershock sequence. Figures 8, 9, 10 and 11 give sample spectra for a few of the
stations whose data were utilized. The small open circles refer to spectral values obtained from
analysis of the short-period seismograms, and the continuous curve to spectral values obtained
from the long-period seismograms. Because of low signal-to-noise ratios the spectral values
obtained from the long-period seismograms are not valid for periods less than 3 sec.
The spectra can be approximated by three straight-line segments, namely: a flat portion extending
from an infinite period to a corner period called T01, a sloping straight line extending from the
corner period T01 to a second corner period T12, and a third straight line, of slope greater than the
second, extending from corner period T12 to the short periods. The first and second of these
@seismicisolation
@seismicisolation
curves, intersecting at period T01 are superimposed on the spectra in Figures 7 through 10. From
the figures it can be noted that in general the non-anomalous earthquakes have larger value of T01
than the anomalous ones.
Figure 11 shows simplified P-wave spectra for each of the four earthquake sequences, as
determined from the seismograms of a single station for an individual sequence. In most cases the
part of the spectrum between the corner periods T01 and T12 has a slope of 1.3 to 2.0, and the part
between T12 and 1-sec period has a slope of 3 to 4. Differences in spectral level at 1-sec period are
in general in good agreement with mb, differences as obtained from time-domain measurements of
the seismograms. Differences in spectral level at the long periods agree in a qualitative, if not
always an exact quantitative, way with differences in seismic moment and MS.
The curves for NUR-Tibet of Figure 11 show that although the two earthquakes have the same mb,
(1-sec spectral amplitude), the non-anomalous earthquake (no. 1) has longer corner periods and a
larger seismic moment. The curves for TIK-Szechwan of Figure 11 also show approximately the
same spectral levels at 1 sec, but lesser corner periods and long-period amplitudes for anomalous
events no. 3 and l. Similarly, anomalous Iraq-Iran events no. 2 and 3 have about the same spectral
amplitudes at 1 sec, but lesser corner periods than the non-anomalous event no. 1. If we consider
the seismic moment, proportional to the long-period spectral amplitude, to be a more fundamental
property of earthquakes than the mb value, then we can conclude that for a given seismic moment
an anomalous earthquake has smaller corner periods and a larger 1-sec spectral amplitude than a
non-anomalous one.
DISCUSSION AND CONCLUSION
We have shown that earthquakes with anomalous mb:MS values occur in some of the foreshocks
and aftershocks of sequences in Szechwan Province, Tibet and Iraq-Iran earthquakes of the
Eurasian interior. The earthquakes in a given sequence, non-anomalous and anomalous alike, have
similar focal depth and focal mechanism. Thus neither differences in these focal parameters nor in
the transmission path from source to station can be called upon to explain the anomalous
earthquakes.
In all, we considered 29 earthquakes in the regions mentioned above plus 8 in the New Hebrides
region. Although there is some ambiguity as to what constitutes an anomalous mb:MS value, as
many as 24 of the 29 can be considered to be anomalous or near-anomalous events. All of the 8
New Hebrides events, on the other hand, are non-anomalous.
We found that a common feature of the anomalous events was that the corner periods of their P-
wave spectra were less than those of non-anomalous events which had the same mb value. We also
observed in the time domain, that for earthquakes of the same mb, the duration of the short-period
P-wave motion was less (more pulse-like) for the anomalous than for the non-anomalous events.
The findings of this study have some bearing on the problem of nuclear test detection, in addition
to the obvious one that some intra-plate foreshocks and aftershocks possess anomalous mb:MS
values. Because these anomalous values can be explained in terms of shorter corner periods and
thus an enrichment of the short-period portion of the P-wave spectrum, rather than in differences
in the long-period portion or the seismic moment, it follows that methods of discrimination that
compare the long-period portions of explosion and earthquake-generated waves, such as relative
excitation of Rayleigh and Love waves, will be unaffected by the results discussed here. Thus the
methods of discrimination which utilize exclusively the long-period portion of the surface and/or
body-wave spectra should be useful in discriminating between explosions and those relatively few
earthquakes which are classified as anomalous by the conventional mb:MS analysis .
@seismicisolation
@seismicisolation
40 Chapter Two
ACKNOWLEDGMENTS
We thank Dr. Robert B. Herrmann for his critical reading of the manuscript and comments. We
also wish to thank Dr. V. N. Vadkovsky, Chief of the Solid Earth Group, WDC-2, Moscow for
sending us copies of U.S.S.R. seismograms.
This research was supported by the Defense Advance Research Projects Agency (Contract P
19628-73-C-0269) under the technical cognizance of Air Force Cambridge Research Laboratories.
DEPARTMENT OF EARTH AND ATMOSPHERIC SCIENCES, SAINT LOUIS UNIVERSITY
Manuscript received
REFERENCES
Brune, J. N (1970), Tectonic stress and the spectra of seismic shear waves from earthquakes, J.
Geophys. Res. 75, 4999-5009.
Evernden, J. F. (1975), Further studies on seismic discrimination, Bull. Seism. Soc. Am. 65, 359-
391.
Forsyth, D. (1975), Causes of MS:mb, variation within a central-Asian earthquake sequence
(Abstract), Trans. Am. Geophys. Un. 56, 1024.
Hanks, T. C. and M. Wyss (1972), The use of body-wave spectrum in the determination of
seismic-source parameters, Bull. Seism. Soc. Am. 63, 561-589.
Kanamori, H. and D. L. Anderson (1975), Theoretical basis of some empirical relations in
seismology, Bull. Seism. Soc. Am. 65, 1073-1095.
Liebermann, R. C and P. W. Pomeroy (1970), Source dimensions of small earthquakes as
determined from the size of the aftershock zone. Bull. Seism. Soc. Am. 60, 879-890.
Molnar, P., J. Savino, L. R. Sykes, R C. Liebermann, G. Hade, and P. W. Pomeroy (1969), Small
earthquakes and explosions in Western North America recorded by high gain, long period
seismographs, Nature 224, 1268-1273.
Nuttli, O. W. and S. G. Kim (1975), Surface-wave magnitudes of Eurasian earthquakes and
explosions, Bull. Seism. Soc. Am. 65, 693-709.
Tsai, Y. B and K. Aki (1971). Amplitude spectra of surface waves from small earthquakes and
underground nuclear explosions, J. Geophys. Res. 76, 3940-3952.
@seismicisolation
@seismicisolation
Table 1 Hypocentral coordinates and magnitudes of earthquakes studied

Region No. Date Origin Time Lat.(N) Long.(E) Depth mb MS
(GMT) (km)
Szechwan 1 16 Aug 71 04-58-00.3 28.9 103.7 19 5.7 4.8
2 16 Aug 71 18-53-54.7 28.9 103.6 17 5.4 5.4
3 16 Aug 71 22-37-33.6 28.8 103.6 13 5.3 4.7
Iraq-Iran 1 12 Jun 72 13-34-00.7 33.1 46.3 16 5.3 4.9
2 13 Jun 72 00-55-37.3 33.1 46.3 12 5.1 4.5
3 14 Jun 72 04-34-28.1 33.0 46.1 15 5.3 4.1
Tibet 1 14 Jul 73 04-51-21.0 35.2 86.5 19 6.0 6.6
2 14 Jul 73 13-39-30.0 35.3 86.6 22 6.0 5.4
New
1 23 Jan 72 17-18-39.4 -13.1 166.3 25 5.3 5.4
Hebrides
2 23 Jan 72 21-17-52.1 -13.2 166.4 28 5.9 6.8
3 24 Jan 72 03-55-42.5 -13.0 166.4 26 5.9 5.9
@seismicisolation
@seismicisolation
42 Chapter Two
Table 2 Comparison of ೗ms and log ೗a·values
Region Earthquake Pair ᶭMS log ᶭA

Szechwan Province No. 2 - No. 1 0.6 0.21 (n = 6)
No. 1 - No. 3 0.1 0.49 (n = 6)
No. 2 - No. 3 0.7 0.70 (n = 6)
Iraq-Iran Border No. 1 - No. 2 0.5 0.51 (n = 4)
No. 1 - No. 3 0.8 0.80 (n = 4)
No. 2 - No. 3 0.4 0.29 (n = 4)
Tibet No. 1 - No. 2 1.20 1.45 (n = 2)
New Hebrides Islands No. 2 - No. 1 1.39 1.26 (n = 4)
No. 3 - No. 1 0.53 0.37 (n = 4)
No. 2 - No. 3 0.86 0.99 (n = 4)
@seismicisolation
@seismicisolation
Table 3 Comparison of mb and mo for anomalous and non anomalous earthquakes
Earthquake mb Mo (dyne-cm)
Anomalous Non-anomalous
Iraq-Iran No. 2 5.1 5.5 × 1023
Iraq-Iran No. 1 5.3 1.9 × 1024
Iraq-Iran No. 3 5.3 2.8 × 1023
Szechwan No. 3 5.3 3.2 × 1023
New Hebrides No. 1 5.3 3.6 × 1024
Szechwan No. 2 5.4 1.8 × 1024
Szechwan No. 1 5.7 1.1 × 1024
New Hebrides No. 3 5.9 8.5 × 1024
New Hebrides No. 2 5.9 9.0 × 1025
Tibet No. 1 6.0 3.3 × 1025
Tibet No. 2 6.0 1.2 × 1024
@seismicisolation
@seismicisolation
44 Chapter Two
Figure 1. Rayleigh-wave spectra of Szechwan earthquakes for pure continental-path at NDI and QUE and normalized
to a distance of 1000 km showing spectral holes at around 40 s corresponding to about 16-21 km.
@seismicisolation
@seismicisolation
Figure 2. Rayleigh-wave spectra of New Hebrides earthquakes for oceanic-path at JCT and OGD and normalized to a
distance of 1000 km showing unclear holes with sharply upward spectra at 20 s.
@seismicisolation
@seismicisolation
46 Chapter Two
Figure 3a. Focal mechanism solutions for Szechwan earthquakes (above) and Iraq-Iran earthquakes (below). Compres-
sions are indicated by octagons, dilatations by triangles and small amplitude arrivals by X's.
@seismicisolation
@seismicisolation
Figure 3b. Focal mechanism solutions for the Tibet earthquakes (above) and the New Hebrides earthquakes (below).
@seismicisolation
@seismicisolation
48 Chapter Two
Figure 4. (above) P-wave motion on SPZ records for the Szechwan earthquakes (left) and the Tibet earthquakes (right).
For the Szechwan earthquakes the station on the left is BUL, at an epicentral distance of 87.3° and an azimuth of 19.9°,
and the station on the right is SHK, at an epicentral distance of 25.9° and an azimuth of 265.2°. For the Tibet
earthquakes the station on the top is NUR (ᶭ= 40.4°, BAZ = 94.1°), in the middle is STU (ᶭ= 50.7°, BAZ = 73.9°)
and on the bottom is KTG (ᶭ= 63.1°, BAZ = 0.0°). (bottom) Rayleigh-wave motion at station SHI (ᶭ= 44.3°, BAZ
= 77.8°) for the Szechwan earthquake (left) and at station KTG(ᶭ = 63. 1°, BAZ = 60. 6°) for the Tibet earthquake
(right).
@seismicisolation
@seismicisolation
Figure 5. mb: MS diagram for the Tibetan earthquakes. Nos.1 and 2 refer to the earthquakes listed in Table 1. Curve a
is the linear least-square fit to 64 non-anomalous Eurasian earthquakes. Curve b is the least-square fit to the Tibetan
earthquake data, indicated by X's. Curve c is the least-square fit to 11 underground Eurasian explosion data.
Figure 6. mb: MS diagram for the New Hebrides earthquakes, indicated by X's. Nos.1, 2 and 3 refer to the earthquakes
listed in Table 1. Curve b is the linear least-square fit to the New Hebrides data. Curves a and c are as in Fig.5.
@seismicisolation
@seismicisolation
50 Chapter Two
Figure 7. P-wave amplitude spectra at stations TIK and COL for the Szechwan earthquakes. The continuous curve
represents the spectrum obtained from the long-period, vertical-component seismogram and the circles the spectral
values obtained from the short-period, vertical-component seismogram.
@seismicisolation
@seismicisolation
Figure 8. P-wave amplitude spectra at stations STU, NUR and KTG for the Tibetan earthquakes.
@seismicisolation
@seismicisolation
52 Chapter Two
Figure 9. P-wave amplitude spectra at stations AAE and SHL for the Iraq-Iran earthquakes.
@seismicisolation
@seismicisolation
Figure 10. -wave amplitude spectra at stations MAT and SHL for the New Hebrides earthquakes.
@seismicisolation
@seismicisolation
54 Chapter Two
Figure 11. Simplified P-wave spectra for the Szechwan earthquakes as recorded at station TIK (upper left), the Iraq-
Iran earthquakes as recorded at station SHL (upper right), the Tibet earthquakes as recorded at station NUR (lower left)
and the New Hebrides earthquakes as recorded at station SHL (lower right). The differences in amplitude at 1-sec
period scale as differences in mb, and the differences in amplitude at 20-sec period as differences in MS. The corner
period T12 for Szechwan earthquakes no. 1 and 3 is less than 1 sec.
@seismicisolation
@seismicisolation
Table 1 Hypocentral coordinates and magnitudes of earthquakes studied
Region No. Date Origin Time Lat. Long. Depth mb MS

(GMT) (°N) (°E) (km)
Szechwan 1 16 Aug 71 04-58-00.3 28.9 103.7 19 5.7 4.8
2 16 Aug 71 18-53-54.7 28.9 103.6 17 5.4 5.4
3 16 Aug 71 22-37-33.6 28.8 103.6 13 5.3 4.7
Iraq-Iran 1 12 Jun 72 13-34-00.7 33.1 46.3 16 5.3 4.9
2 13 Jun 72 00-55-37.3 33.1 46.3 12 5.1 4.5
3 14 Jun 72 04-34-28.1 33.0 46.1 15 5.3 4.1
Tibet 1 14 Jul 73 04-51-21.0 35.2 86.5 19 6.0 6.6
2 14 Jul 73 13-39-30.0 35.3 86.6 22 6.0 5.4
New 1 23 Jan 72 17-18-39.4 -13.1 166.3 25 5.3 5.4
Hebrides 2 23 Jan 72 21-17-52.1 -13.2 166.4 28 5.9 6.8
3 24 Jan 72 03-55-42.5 -13.0 166.4 26 5.9 5.9
Table 2 Comparison of ೗ms and log ೗a values
Region Earthquake Pair ᶭMS log ᶭA

Szechwan Province No. 2 - No. 1 0.6 0.21 (n = 6)
No. 1 - No. 3 0.1 0.49 (n = 6)
No. 2 - No. 3 0.7 0.70 (n = 6)
Iraq-Iran Border No. 1 - No. 2 0.5 0.51 (n = 4)
No. 1 - No. 3 0.8 0.80 (n = 4)
No. 2 - No. 3 0.4 0.29 (n = 4)
Tibet No. 1 - No. 2 1.20 1.45 (n = 2)
New Hebrides Islands No. 2 - No. 1 1.39 1.26 (n = 4)
No. 3 - No. 1 0.53 0.37 (n = 4)
No. 2 - No. 3 0.86 0.99 (n = 4)
Table 3 Comparison of mb and mo for anomalous and nonanomalous earthquakes
MO (dyne-cm)
Earthquake mb
Anomalous Non-anomalous
Iraq-Iran No. 2 5.1 5.5 u 1023
Iraq-Iran No. 1 5.3 1.9 u 1024
Iraq-Iran No. 3 5.3 2 .8 u 1023
Szechwan No. 3 5.3 3.2 u 1023
New Hebrides No. 1 5.3 3 .6 u 1024
Szechwan No. 2 5.4 1.8 u 1024
Szechwan No. 1 5.7 1.1 u 1024
New Hebrides No. 3 5.9 8.5 u 1024
New Hebrides No. 2 5.9 9.0 u 1025
Tibet No. 1 6.0 3.3 u 1025
Tibet No. 2 6.0 1.2 u 1024
@seismicisolation
@seismicisolation
CHAPTER THREE
SPECTRAL CLASSIFICATION METHODS IN MONITORING OF SMALL LOCAL

EVENTS BY ISRAEL SEISMIC NETWORK1
YEFIM GITTERMAN, VLADIMIR PINSKY AND AVI SHAPIRA
Controlled ripple-fired blast of 12 tons ANFO at the Kahal quarry, Israel, on August 2, 2001 (photo by Y. Gitterman)
ABSTRACT
We use the dense Israel Seismic Network (ISN) to discriminate between low magnitude
earthquakes and explosions in the Middle East region. This issue is important for CTBT
monitoring, especially when considering small nuclear tests which may be conducted under
evasive conditions. We explore the performance of efficient discriminants based on spectral
features of seismograms using waveforms of 50 earthquakes and 114 quarry and underwater blasts
with magnitudes 1.0–2.8, recorded by ISN short-period stations at distances up to 200 km.
The single-station spectral ratio of the low and high-frequency seismic energy shows an overlap
between explosions and earthquakes. After averaging over a subnet of stations, the resolving
power is enhanced and the two classes of events are separated. Different frequency bands were
tested; the (1–3 Hz)/(6–8 Hz) ratio provided the best discriminant performance. We also estimated
normalized r.m.s. spectral amplitudes in several sequential equal frequency windows within the 1–
12 Hz band and applied multiparametric automatic classification procedures (Linear
Discrimination Function and Artificial Neural Network) to the amplitudes averaged over a
subnetwork. A leave-one-out test showed a low rate of error for the multiparametric procedures.
An innovative multi-station discriminant is proposed, based on spectral modulation associated
1
Originally published in the Journal of Seismology 2: 237–256, 1998. © 1998 Kluwer Academic Publishers. Printed
in the Netherlands. @seismicisolation
@seismicisolation
Spectral Classification Methods in Monitoring of Small Local Events by Israel Seismic Network 57
with ripple-firing in quarry blasts and with the bubbling effect in underwater explosions. It utilizes
a distinct azimuth-invariant coherency of spectral shapes for different stations in the frequency
range (1–12 Hz). The coherency is measured by semblance statistics commonly used in seismic
prospecting for phase correlation in the time domain. After modification, the statistics applied to
the network spectra provided event separation.
A new feature of all the above mentioned procedures is that they are based on smoothed (0.5 Hz
window), instrument-corrected FFT spectra of the whole signal; they are robust to the accuracy of
onset time estimation and, thus well suited to automatic event identification.
Key words: regional seismic network, multiparametric event classification, spectral semblance, spectral ratios
INTRODUCTION
The monitoring of a Comprehensive Test Ban Treaty (CTBT) requires enhanced ability to
discriminate between small earthquakes and man-made explosions of different types. For example,
a kiloton clandestine nuclear test carried out in a large cavity may lead to a seismic decoupling
factor of 50 (Jarpe et al., 1996) decreasing magnitude to Mb 2.1, comparable with that of quarry
blasts. Regional dense seismic networks have a high potential for detecting and identifying small
(Mb<2.5) seismic events. Discriminants, based on spectral features of earthquakes and explosions
reflecting different source mechanisms and depths, appear to be efficient. Various types of spectral
ratios have been tested: ratios of wave phases (e.g. P/S); ratios of peak, r.m.s. or average amplitude
in low and high-frequency bands for a specific phase, mainly Pn and Lg (e.g. Taylor et al., 1988;
Bennett and Murphy, 1986; Walter et al., 1995).
Performance of the short-period discriminants, depending upon local crustal structure and
composition, varies from region to region. Most of the investigations mentioned above were
conducted for a single station or beamed array data. As shown by Kim et al. (1994), the
discrimination effect of their single-station discriminant (P/S spectral ratio) was enhanced when
averaged over the New York State network. For a specific dataset of earthquakes and underwater
explosions off the Levant coast, the spectral ratio RE of seismic energy of the whole signal in low-
frequency and high-frequency ranges, averaged over the Israel Seismic Network (ISN), provided
complete separation of the two populations (Gitterman and Shapira, 1993).
One of the main spectral features used in the identification of industrial explosions is spectral
modulation (SM) caused by ripple firing. The modulation is explained by constructive/destructive
interference of linearly superimposed waveforms from spaced charges fired at millisecond delays
(Gitterman, 1982; Stump and Reinke, 1988). Most of the studies of the ripple firing modulation are
based on time-independent patterns for a single station or array, presented in sonograms (e.g.
Hedlin et al., 1995; Kim et al., 1994), or in spectra of regional phases (e.g. Der and Baumgardt,
1995). These discriminants are concerned mainly with spectral maxima at high frequencies (more
than 10 Hz) corresponding to fired delay times. It has been argued that SM is not observable for
too short delays and the limited recording bandwidth (Bennett et al., 1989), and that it should be
extended to 80 Hz to resolve delays shorter than 25 msec (Baumgardt and Zigler, 1988).
A different multi-station Low Frequency Spectral Modulation (LFSM) approach based on the
azimuth-invariant SM caused by ripple firing considered jointly for spectra of the whole signal at
several network stations, is discussed by Gitterman and van Eck (1993). This approach was also
implemented for the identification of underwater explosions (UWE) in the Mediterranean,
demonstrating SM caused by the bubbling phenomenon (Gitterman and Shapira, 1993, 1994). The
method utilizes low-frequency (1–10 Hz) minima (or spectral nulls) and maxima and demonstrates
a distinct coherency of spectral shapes for broad ranges of azimuths and distances (10–300 km) in
the whole range of delay firing times (20–100 msec). Analysis of the whole signal window
provides accumulation of source spectral modulation features presenting in all wave phases. Using
@seismicisolation
@seismicisolation
58 Chapter Three
the low-frequency range for the analysis and application of SM for discrimination is explained not
only by the restricted recording bandwidth (1–12 Hz), but mainly by the fact that SM and spectra
coherency, most pronounced at low frequencies, vanish gradually at higher frequencies due to
random variance of blasting factors (in-particular, firing times) (Gitterman, 1982) and the Doppler
effect (Chapman et al., 1992). Variation of delays and delayed charges causes a reduction in
modulation depth (smoothing of sharp spectral troughs and peaks), increasing with frequency. The
same effect can be produced by differences in the propagation path of spaced charges (Stump and
Reinke, 1988).
An additional discrimination ability can be achieved by an integrative approach using jointly a
number of signal spectral features. Even a simple majority voting of several different physical
discriminants (Wüster, 1993) provides an improvement in classification results. Simultaneous
application of efficient, multi-dimensional procedures, such as linear or quadratic Fisher
discriminators (e.g. Tsvang et al., 1993) and Artificial Neural Network (e.g. Dowla, 1995; Pulli,
1995), to a suit of discriminants, enhances the resolving power as compared to an individual
discriminant or majority voting. The performance of the procedures was tested by the leave-one-out
procedure. In this study the advantage provided by ISN (dense and rather uniform station spacing) is
utilized for the new-developed discriminant based on coherency of spectral shapes for quarry blasts
at azimuthally distributed stations which is not observed in earthquake patterns. The coherency is
measured by ‘semblance’ and ‘cross-correlation’ statistics, used in seismic prospecting practice for
phase correlation in the time domain, and modified for spectral application.
INSTRUMENTATION AND DATA BASE

The ISN (Figure 1) consists of short period (1 Hz) seismometers (either Teledyne-Geotech S-13
and Mark Products L4C). The station records were transmitted via FM telemetry to the National
Seismology Center at the Geophysical Institute of Israel (GII, Holon), signals were bandpass
filtered (0.2 to 12.5 Hz), amplified and digitally recorded with a sampling rate of 50 samples per
second. All seismograms used in the study stem from vertical seismometers.
164 local seismic events recorded by the ISN were used to test different discrimination procedures.
The selected database covers a broad range of seismic events: ripple-fired quarry blasts,
underwater explosions (UWE) and local earthquakes occurring in Israel and adjacent areas like the
Galilee, off-coast Tyre, the Dead Sea basin, the Negev desert and the Jordan-Saudi Arabia border
(see Figure 1 and Table 1; detailed source information can be found in Gitterman et al., 1996).
Table 1. Description of various sources used in the study
# Region, dataset Number of Number of Comments

Earthquakes Explosions
1 Galilee, Northern Israel 30 39 Earthquake depth H = 0–21 km

Quarry blasts: W = 2000–14000 kg
2 Southern Dead Sea and 16 26 H = 0–24 km
Negev desert, Israel Quarry blasts: W = 1700–10500 kg
3 Tyre region, 4 (+4 from 16 H = 6–17 km
Mediterranean sea the Galilee) UWE – unauthorized fishing, no
Ground Truth Information UWE for
4 Dead Sea basin, seismic – 28 seismic prospecting:
refraction profile depth h = 70 m; W = 16–304 kg
5 Jordan/Saudi Arabian – 5 Jordanian quarry blasts:
border W = 15000–17000 kg
@seismicisolation
@seismicisolation
Figure 1. Epicenters of earthquakes ({), quarry blasts and underwater explosions (+), location of quarries (Ŷ) and ISN
stations (¨). The Dead Sea experimental UWE are located along a line.
@seismicisolation
@seismicisolation
60 Chapter Three
The analyzed data rely on ground-truth information which gives the total amount of explosives
and, occasionally firing delay times (usually 20–40 msec) are also provided. On seismograms for
two events on the Jordanian side of the Dead Sea, we observed clear sonic waves striking the
seismometers (see Figure 2), confirming their blasting character. Most of the man-made events
occurring off-coast Israel-Lebanon are unauthorized fishing explosions and some are naval tests,
so we were unable to obtain any additional information. Charge weights for UWE occurring in the
morning hours (sometimes at intervals of several minutes), possibly do not exceed a few dozen
kilograms of explosives. The water depth in the area of UWE varies from 50 to 300 m (Almagor
and Hall, 1984). We selected also underwater profiling explosions in the Dead Sea, executed for
studying the crustal structure in the Dead Sea transform. Ground-truth information here includes
accurate coordinates, charge weight, shot depth and local time of explosions. Five quarry blasts
conducted in southern Jordan near the border with Saudi Arabia were used in analysis.
The selected events were within the local magnitude range ML = 1.0-2.8 and distance range 10–
200 km. The magnitudes were determined from S-coda duration measurements (Shapira, 1988).
Over the years of ISN operation, quarry blasts were considered a nuisance and, as such, no
significant attempts were made either to locate them accurately or even determine their magnitude.
We relocated the blasts (estimated accuracy is 1–3 km), matched the computed locations with
reports from the quarries and re-evaluated their magnitudes. Owing to the relatively high density
of stations in the Galilee and Dead Sea/Negev region, we consider the earthquake hypocenters to
be accurately located. Nevertheless, we checked several earthquakes with zero depth estimates and
found better solutions for deeper (3–6 km) sources.
SEISMIC EVENT DISCRIMINANTS AND THEIR EXTRACTION FROM ISN

RECORDINGS
Procedures and parameters of preliminary spectral processing

We computed the Fourier spectra (FFT) of ground motions recorded by an ISN subnet of 4–10
stations with good quality seismograms. Computations are made for a time window of 20–40 sec.
The analyzed window for an event is chosen to provide the same frequency step Ï in FFT for all
stations and includes the whole signal. By using this window, we were able to avoid picking out
separate wave phases and accumulate information about source (and possibly propagation)
features being kept in all wave forms, thus enhancing the resolving power of the discriminants
considered. The spectra were instrument corrected and smoothed with a triangular operator
(Hanning window) in a fixed 0.5 Hz moving window to provide equivalent spectral resolution for
different stations. Removing instrument response from FFT causes a boundary effect of rising
spectral curves above the upper cut-off frequency 12.5 Hz; so the discrimination analysis in this
study was restricted by the frequency 12 Hz. The log-log plotting mode of a single event spectra
for a subnet of azimuthally distributed stations was used, facilitating clear identification of the
coherent LFSM extrema for quarry blasts in the higher part of the frequency range.
Energy spectral ratio
It is commonly observed in Israel that seismograms of explosions are richer in low-frequency
energy as pared to earthquakes (see Figures 3 and 4). It has been suggested that spall mechanisms
contribute to low-frequency seismic energy and the general tendency for quarry blasts to appear
deficient in high-frequency energy (Barker et al., 1993). We suspect that in our case the
phenomenon is caused mainly by the dominant surface waves (including fundamental Rayleigh
wave Rg, see Figure 3c), generated by open-pit quarry blasts (at zero depth) and associated with
the regional crustal structure of widespread unconsolidated subsurface sediments (e.g. Ginzburg
and Folkman, 1980), trapping of low-frequency energy and absorbing high-frequency radiation.
@seismicisolation
@seismicisolation
Figure 2. Clear sonic waves striking the seismometers are observed for an event on the Jordanian side of the Dead
Sea, confirming its blasting character.
(In most cases path length is more than 20 km, which corresponds to more than ten wave-lengths
of S-waves). This explanation is supported by similar observations for New England (Li et al.,
1996) and by the reversed situation for the Scandinavian shield: dominating high-frequency energy
in seismograms of quarry blasts (e.g. Baumgardt and Young, 1990).
We utilized this effect in the spectral ratio, RE, of seismic energy in the low-frequency range (f1, f2)
and the high-frequency range (f3, f4) (Gitterman and Shapira, 1993):
f2 f4
2 2 (1)
RE ³ S( f ) df ³ S( f ) df
f1 f3
where _S(f)_ is the smoothed spectrum of ground velocity for the whole seismogram. This energy
spectral ratio, implying evident physical meaning, is similar to the ratios of average spectral
amplitude (Bennett and Murphy, 1986) or power spectrum (Pulli, 1995).
Multidimensional automatic discrimination procedures
We tried a formal Integrative Approach, based on discrimination procedures, such as the Fisher
Linear Discrimination Function (LDF) (Tsvang et al., 1993) and Artificial Neural Network (ANN)
(Dowla, 1995).
The LDF, a statistical discrimination procedure commonly used in seismology, is based on the
assumption that the X vectors of observational discriminant parameters generated by two
stochastic (physical) mechanisms H1 and H2 (earthquakes and explosions) are described by a
Gaussian distribution with equal covariance matrices S and different means Mi. The optimal
decision rule, based on log-likelihood ratio function V(X), is as follows (Tsvang et al., 1993):
{assign X to H1, if V (X) >0 and assign
X to H2, if V (X) <0},
V (X)=F1 (X)íF2 (X) =
2(M2 í M1)TSí1XíM1T Sí1M1 + M2T Sí1M2. (2)
In practice, S and Mi are unknown and estimated from the training set during the learning stage.
@seismicisolation
@seismicisolation
62 Chapter Three
Figure 3a
@seismicisolation
@seismicisolation
Figure 3b.
@seismicisolation
@seismicisolation
64 Chapter Three
Figure 3. Examples of low SNR recordings at different ISN stations and appropriate spectra of quarry blasts EG33 (a)
and EG16 (b) in Galilee, and ES6 in Negev (c), showing low-frequency energy due to strong surface (in particular,
Rg) waves, and distinct azimuth-independent spectral modulation. The spectra, and spectral averages (shown by bold
lines) demonstrate clear spectral nulls. Any modulation and coherency of spectral shapes, or spectral nulls, are not
observed in pre-signal noise (a). Arrows on seismograms show analyzed time windows. FFT spectra are instrument
corrected and smoothed by a triangle operator in the 0.5 Hz window.
@seismicisolation
@seismicisolation
We used a multilayered (3 layers, including one hidden layer with two neurons) feed-forward
ANN architecture (Dowla, 1995). This network is a classic example of a supervised learning
network applying a back propagation learning algorithm to associate inputs with corresponding
outputs for all or most of the events in a training set. This algorithm uses a gradient descent
method to systematically modify the weights in the network so as to minimize the network output
error. The input feature vector was formed as follows: we selected smoothed spectra of event
records as described above, divided the whole frequency band (1–12 Hz) into several equal
intervals and computed the r.m.s. of spectral amplitude in each of them. The r.m.s. values were
normalized to the station spectral maximum to eliminate dependence on event magnitude and
distance and then averaged over the network stations, forming the vector X. The performance of
the procedures is characterized by error probability (ratio of misclassifications to the total number
of events), was estimated on the limited data set by the leave-one-out method.
New spectral coherency statistics
The efficiency of the LFSM approach in identifying explosions with a low signal-to-noise ratio
(SNR) was investigated. A number of quarry blasts recorded at remote stations with 615§ for P-
waves and 615§–2 for S and coda waves show clear spectral modulation patterns presented by
coherent deep minima (or spectral nulls) and maxima. The spectral averaging over selected
network stations emphasizes these features. Any modulation and coherency of spectral shapes, or
spectral nulls, are not observable in pre-signal noise spectra (Figure 3a).
Spectral scalloping depends on ripple-firing parameters. The frequency of the first null, dominant
in our frequency-limited recordings, is determined mainly by the total duration of firing delay
times (Gitterman and van Eck, 1993; Barker et al., 1993). These frequencies vary slightly for
different stations (Figure 3a and 3b), but leave SM and coherency of spectral shapes unaffected for
most selected quarry blasts. It has been hypothesized that the Doppler effect could be responsible
for this phenomenon (Blandford, 1995), while we consider these shifts to be caused mainly by
random slope of a station spectrum for a single shot (Gitterman, 1982).
Supposedly due to the directivity effect (Bakun et al., 1978), the interference pattern for an
earthquake is azimuth-dependent, i.e. maxima and minima in a spectrum are shifted significantly
in accordance with the direction to a station, therefore, for network with broad azimuthal coverage,
we observe an irregular character of spectral shapes and minima for different azimuths (Figure 4).
No clear regular modulation is present in these earthquake spectra, showing a quasi-flat shape
(Figure 4a).
The coherency of spectral shapes for different stations can be quantitatively assessed by
‘semblance’ and ‘cross-correlation’ statistics commonly used in seismic prospecting for phase
correlation of seismic traces in the time domain (e.g. Neidell and Taner, 1971). After some
modification the new discrimination statistics are written as
(3)
where
N 2
S1i ª ¦ ( S ki S k )º
«¬ k 1 »¼
N
2
S 2i ¦ ( S ki S k )
k 1 (4)
Ski = log10Sk(fi) í log spectral amplitude at station k; S k is the average spectral level; N is the
number of stations and [f1, f2] is the frequency range for calculation.
@seismicisolation
@seismicisolation
66 Chapter Three
Figure 4a.
@seismicisolation
@seismicisolation
Figure 4. Examples of earthquake recordings and spectra from Galilee, QG20 (a) and Dead Sea, QS3 (b),
demonstrating non-coherency of spectral shapes and high-frequency energy. Any modulation or spectral nulls, are not
observed in spectra, including the spectral average (marked by bold line), showing a quasi-flat shape (a).
@seismicisolation
@seismicisolation
68 Chapter Three
Figure 5. Energy spectral ratios for different Galilee events at individual stations of an ISN subnetwork. Diamonds
denote average values.
DISCRIMINANT PERFORMANCE RESULTS

Galilee region
This data set includes 39 quarry blasts and 30 earthquakes, including 8 earthquakes which
occurred in a small area of Lake Kinneret (Sea of Galilee; Figure 1). The energy ratio discriminant
was tested for different frequency bands. Initially we tried f1 = 1 Hz, f2 = 6 Hz, f3 = 11 Hz
(Gitterman et al., 1996) and obtained promising results. The ratios determined from a subnet of
ISN stations are presented in Figure 5. For some stations an overlap between quarry blasts and
earthquakes may be observed. If the ratios at different stations for a fixed event are averaged, the
resolving power is enhanced and the two populations of seismic events are separated. Several Lake
Kinneret earthquakes located exactly on the Dead Sea transform show anomalous high ratios. In
general, we would suggest that seismograms yielding an average ratio RE > 6.4 are associated with
quarry blasts.
Application of another (2–4 Hz)/(6–8 Hz) ratio, used by Li et al. (1996) for the Lg spectra
(definitely slightly different from the whole signal spectra we use in the analysis) to separate
small events in New England, showed an overlap (Figure 6a). However, after a shift in the low-
frequency range, the (1–3 Hz)/(6–8 Hz) ratio provided the best discriminant performance,
enhancing separation between quarry blasts and earthquakes (Figure 6b). It can be explained by
maximal energy for the local blasts at low frequencies (1–4 Hz), mainly due to surface waves, and
energy lack at high frequencies, beginning from 5–6 Hz (see Figure 3).
The leave-one-out cross-validation test was applied for the two multidimensional procedures. The
results (number of mistakes/mistake rate) are presented in Table 2. Both LDF and ANN give the
best performance when the frequency interval (1–9 Hz) is divided into four bands, although ANN
shows error-free performance in the four and five bands cases, while LDF has one mistake in the
best case. The results demonstrate that low-frequency bands (1–3 Hz and 3–5 Hz) do not provide
correct event classification and that higher frequencies (5–7 Hz and 7–9 Hz) are the most
informative.
The ‘semblance’ technique was applied to smoothed amplitude spectra of the subnetwork. We
tried two spectral intervals for the analysis: 1–7 Hz and 1–12 Hz; the results are presented in
@seismicisolation
@seismicisolation
Figure 6. The statistics obtained are much higher for quarry blasts (0.7–0.9) than for earthquakes
(0.1–0.4), and remain virtually unchanged when different windows of spectra smoothing are
applied (0.25 Hz and 1 Hz). The semblance statistics show a slightly higher resolving power than
cross-correlation; in general, better discrimination results are obtained when processing the whole
signal spectra in the frequency range 1–12 Hz. As expected, a strong correlation between
semblance and cross-correlation statistics is observed (see Figure 6d); therefore, in the following
we consider them as one discriminant. For the few earthquakes originating on the main fault at
Lake Kinneret, and showing increased energy ratio values (as mentioned above), we obtained
anomalous high semblance values of 0.4–0.65; the two sets of seismic events are, however,
completely separated. The same result is observed on the semblance versus energy ratio (Figure
6b). It should be noted that the spectral range 1–7 Hz showed ‘normal’ values of the statistics for
the earthquakes located in the lake (Figure 6c).
Dead Sea/Negev region
This data set includes 16 earthquakes and 26 quarry blasts from the southern Dead Sea basin and
Negev desert (see map on Figure 1). Spectra were calculated for about 10 selected ISN stations
shown on the map (distance range 10–150 km). Examples of typical blast and earthquake
recordings and appropriate spectra are presented on Figures 3c and 4b. For most quarry blasts a
clear spectral modulation is observed implying high regularity in blasting at Negev phosphate
quarries. A specific feature as compared to the Galilee events is presence of a distinct Rg wave
group on many blast seismograms (Figure 3c).
Table 2. Number of mistakes/mistake rate in the multidimensional automatic discrimination
test on the Galilee dataset
Number of Frequency intervals, Hz Procedure
bands LDF ANN
1 1–3 10/0.145 11/0.159

2 1–3, 3–5 8/0.116 10/0.145
3 1–3, 3–5, 5–7 3/0.043 1/0.014
4 1–3, 3–5, 5–7, 7–9 1/0.014 0/0
5 1–3, 3–5, 5–7, 7–9, 9–11 2/0.029 0/0
10 1–2, 2–3, 3–4, 4–5, 5–6, 6–7, 7–8, 8–9,
9–10, 10–11 2/0.029 1/0.014
Figure 6a. Figure 6b.
@seismicisolation
@seismicisolation
70 Chapter Three
Figure 6. Discrimination results for the Galilee data set showing semblance versus energy ratio with different ranges
for the ratio low-frequency energy: 2–4 Hz (a), 1–3 Hz (b), and a different range for semblance: 1–7 Hz (c).
Semblance versus cross-correlation is also shown (d) (<> - Lake earthquakes, ' – other earthquakes, z – quarry
blasts).
We included in the analysis five more Jordanian quarry blasts remote from the ISN, located near
the Jordan/Saudi Arabia border (see map in Figure 1). Seismograms and spectra of one of the
ripple-fired blasts are shown in Figure 7. Detailed information regarding patterns of blasting
practice is not available, but one may assume a high regularity of blast parameters along the bench
causing the deep coherent (for different stations) spectral minima (nulls).
The results of the joint application of two spectral discriminants (ratio and semblance) (Figure 8),
demonstrate almost full separation of the two classes of seismic events. There is only one evident
outlier, earthquake QS9 with an unusually high average ratio RE = 10.6, smb = 0.82 and cor =
0.78 and one marginal event QS15 (smb = 0.64 and cor = 0.60). The outlier, nocturnal event QS9,
is located in the southern Dead Sea basin near the shore (see Figure 1) and has a zero depth
estimate (error estimates are rather small: dH = 1.4 km; dX,dY < 1 km). An attempt to improve the
location did not change the hypocenter coordinates. The very shallow source of this earthquake led
to low-frequency content of seismic waves (nevertheless, Rg-excitation was not observed), and,
hence, caused anomaly values of the spectral discriminants.
It should be noted that prominent low-frequency Rg waves (included usually in FFT analysis
windows) at records of some Negev blasts provide a significant contribution in high energy
spectral ratios, and, consequently, in good discrimination results. However, these ratios have the
same range as for Galilee blasts (compare Figure 6b and Figure 8), where some surface waves are
present (see Figure 3a,b), but the classic Rg group (Figure 3c) hardly can be recognized.
The large values of the two discriminants for Jordanian ripple-fired blasts are close to those
estimated for the Galilee (see Figure 6) and Negev quarries.
Offshore events
The ratio and semblance spectral discriminants were applied to distinctive man-made marine
events: 16 UWE in the Mediterranean (illegal fishing and, therefore, no ground-truth information
is available) and a set of single UWEs in the Dead Sea detonated for seismic profiling (28 events)
(Figure 1). A subset of four stations was used to calculate the average ratio and semblance
statistics for Mediterranean UWE.
Detonations in deep water produce seismic waves bearing spectral features caused by a source
(bubbling phenomenon) and water path propagation effects (reverberations). Gitterman and
Shapira (1994) gave a detailed description of low-frequency azimuth-invariant spectral
modulation, observed from ISN recordings of UWE off the coast of the Levant, similar to ripple
firing. They also showed that this UWE source-effect is significant, but of a different physical
@seismicisolation
@seismicisolation
Figure 7. Seismograms (a) and spectra (b) of the ripple-fired blast EJ1 (W= 16500 kg) from a Jordanian quarry
(distance range 114–168 km, azimuth range 260–360º, time window for processing t~30 sec.).
nature, i.e. caused by the interference of gas bubble oscillations. This bubbling effect produces a
complete harmonic series of spectral maxima. The fundamental frequency f1b is dependent on the
detonation depth, d(m) and explosive yield, W (kg of TNT) (e.g. see Cole, 1948):
f1b = (d + 10)5/6/ (2.1W1/3
); fnb = nf1b (n = 1, 2, .. .).
@seismicisolation
(5)
@seismicisolation
72 Chapter Three
Figure 8. Discrimination results for the Dead Sea/Negev data set: semblance versus energy ratio (' – earthquakes;
z – quarry blasts; <> – Jordanian quarry blasts).
Remarkable examples of spectral maxima series up to the fifth order due to the bubbling
phenomenon were observed for UWEs in the Tyre region with high coherency of spectral shapes
at different ISN stations. These effects, not observed in the case of earthquakes (occurring off the
coast), can be utilized in the semblance discrimination procedure. Here we present only one
example of a UWE (see Figure 9), showing long trains of low-frequency surface waves owing to
bubbling and reverberations (and causing overestimation of local magnitudes).
Similar wave trains are observed on seismograms of Dead Sea explosions (e.g. EU1 on Figure 10).
The bubbling actually affects all types of recorded seismic waves throughout the whole
seismogram. However, it is most clearly manifested in the initial body-wave portion (P and P-
coda) of the seismogram which is virtually free from reverberations and water low-velocity
channeling effects: Figure 10b shows a complete harmonic series (3, 6, 9 and 12 Hz) due to the
bubbling effect. Spectra of the whole signal are similar to the P-portion spectra, but much more
complicated and mixed with reverberations maxima (Figure 10c), however, low-frequency SM
minima (4.5 and 7.5 Hz), corresponding to the bubbling interference, are expressed more
distinctly. (The decrease of seismic energy above 10 Hz is caused by the anti-alias 12.5 Hz filters
of the network channels). Using ground truth information (W = 304 kg, d = 70 m) and taking into
account the high density of salt water in the Dead Sea (ȡ |1.11 gr/cmí3) for a correction to greater
detonation depth, we estimated from Equation 5 the bubbling fundamental frequency for the
explosion EX1 at f1b | 2.9 Hz, fitting well to the observed value of f1b | 3.0 Hz (Figure 10b).
The discrimination results for offshore events are presented in Figure 11. For the sake of
comparison, we also included eight Mediterranean off-shore earthquakes located relatively close to
the UWEs (Figure 1). They show low ratio and semblance values, being well separating from
underwater explosions which have high semblance values. Low ratio values for Dead Sea UWEs,
as compared with off-shore Tyre UWE (supposed depth 10–30 m), can be referred to a higher
frequency content of radiated seismic waves due to deeper source (70 m), the higher density of
water and charges mostly 16 and 24 kg, which produce (according to Equation 5) the fundamental
frequency about 6.5 Hz, corresponding to observations.
DISCUSSION AND CONCLUSIONS

The high resolving power of the ‘semblance’ discriminant lies in the nature of the seismic sources
investigated. A ripple-fired blast may be considered a point source, therefore the interference
pattern is uniform in different directions, resulting in the azimuth-independent SM and high (0.6–
0.9) semblance values. The spectral discriminants show dependence on the source region.
@seismicisolation
@seismicisolation
Figure 9. Example of the Mediterranean UWE recordings EU7 from the Tyre region, showing the complete harmonic
series in the spectra of the whole signal (distance range 13–102 km, azimuth range 102–160º, time window t~30 sec.).
Semblance values as well as energy spectral ratios for most earthquakes from Lake Kinneret are
greater than for those from the adjacent Galilee region, but provide fair separation of earthquakes
from quarry blasts.
Our results provide a positive answer to questions such as whether short delay (20–30 msec)
ripple-fired explosions are capable of generating SM below 20 Hz and if this feature can be used
for discrimination of regional (0–400 km) events (Hedlin et al., 1995). The LFSM approach avoids
also the problem of disappearing high-frequency (more than 10 Hz) ripple-firing spectral
@seismicisolation
@seismicisolation
74 Chapter Three
modulation at distances greater than 100 km due to high crustal seismic attenuation (Beck and
Wallace, 1995).
The ISN regional seismograms and their corresponding spectra show two main features, caused by
source physics phenomena, which were utilized for the construction of discriminants:
1) As a rule, earthquakes are much richer in high-frequency (6–12 Hz) energy than explosions.
2) Smoothed spectra from explosions showed distinct minima and maxima, being coherent for
the ISN stations located at different azimuths and distances.
The first feature may be attributed to low-frequency S and surface waves generated by shallow
seismic events (mainly quarry blasts) and associated with the regional crustal structure of
widespread unconsolidated subsurface sediments as well as low-frequency reverberations
generated by underwater explosions in the sea. The second is related more to the ripple-firing
effect combined with symmetric radiation diagram characteristic of the explosion (actually, point)
source.
Based on these source characteristics, new discriminants having a large resolving power for Israel
and surrounding areas were constructed:
a) The seismic energy spectral ratio (RE) between the low-frequency (1–3 Hz) and high-
frequency (6–8 Hz) bands.
b) Spectral ‘semblance’ and ‘cross-correlation’ statistics measuring the coherency of smoothed
spectra at different ISN stations in the (1–12 Hz) band.
The discriminants were tested with a large amount of data, 50 earthquakes and 114 explosions, for
epicentral distances up to 200 km (Figure 1). Ground truth information were available for quarry
blasts and underwater explosions in different geological settings within the region. The statistics
take advantage of the ISN as a dense network. The RE parameter, estimated for individual ISN
stations, showed an overlap of the earthquake and explosion populations due to path and site
effects. Being averaged over a subnetwork of the ISN stations, its resolving power was enhanced
and the two populations were separated with a 98% success rate for 164 events. The spectral
semblance statistics, utilizing azimuthal invariance of explosion spectra, appeared to be even more
efficient with a 99% success rate.
Finally, our search for physics-based discriminants was complemented by testing the multivariate
formal procedures using the Integrative Approach. Two of them, LDF and ANN, were applied to
the averaged normalized r.m.s. of the Galilee events spectra in different frequency bands. The
optimal feature vector, based on the most informative spectral intervals, yields 96% (LDF) and
100% (ANN) success rates. From observations and testing of discriminants we may conclude that
much of the information related to difference of earthquakes and explosions in Israel is
concentrated in the seismogram spectra and may be effectively extracted using the discriminants
developed.
ACKNOWLEDGEMENTS
We are grateful to Dr F. Dowla (Lawrence Livermore National Laboratory, USA) who kindly
provided the program for Artificial Neural Network analysis. Our thanks are due to Prof. A.
Ginzburg (Tel-Aviv University) who kindly supplied us with ground truth information on the
underwater explosions in the Dead Sea and to Dr A-Q. Amrat (Jordan Seismological Observatory)
who provided data on quarry blasts conducted in southern Jordan. We thank the anonymous
referees for their valuable comments and suggestions. This study was supported by the U.S.
Department of Energy under Contract No. F19628-95-K-0006.
@seismicisolation
@seismicisolation
Figure 10. Example of the Dead Sea experimental underwater explosion EX1 with charge weight W= 304 kg: (a)
recordings (distance range 49–116 km, azimuth range 205–358º); (b) spectra of the first 8 sec of P and P-coda waves;
(c) spectra of the whole signal (t~60 sec.).
@seismicisolation
@seismicisolation
76 Chapter Three
Figure 11. Discrimination results (semblance versus energy ratio) for Mediterranean earthquakes (') and UWE (z)
and Dead Sea experimental explosions (<>).
REFERENCES
Almagor, G. and Hall, J. K., 1984, Bathymetric chart of the Mediterranean coast of Israel, Geol.
Surv. Bull., 77.
Bakun, W. H., Stewart, R. M. and Bufe, C. G., 1978, Directivity in the high-frequency radiation of
small earthquakes, Bull. Seism. Soc. Am., 68, 1253–1263.
Barker, T. G., McLaughlin, K. C. and Stevens, J. L., 1993, Numeral simulation of quarry blast
sources, SSS-TR-93-13859, S-Cubed, La Jolla, California.
Baumgardt, D. R. and Young, G. V., 1990, Regional seismic waveform discriminants and case-
based event identification using regional arrays, Bull. Seism. Soc. Am. 80, 1874–1892.
Baumgardt, D. R. and Zigler, K. A., 1988, Spectral evidence of source multiplicity in explosions:
application to regional discrimination of earthquakes and explosions, Bull. Seism. Soc. Am. 78,
1773–1795.
Bennett, T. J., Barker, B. W., McLaughlin, K. L. and Murphy, J. R., 1989, Regional discrimination
of quarry blasts, earthquakes and underground nuclear explosions, Final Report, GL-TR-
890114, S-Cubed, La Jolla, California.
Bennett, T. J. and Murphy, J. R., 1986, Analysis of seismic discrimination capabilities using
regional data from Western United States events, Bull. Seism. Soc. Am. 76, 1069–1086.
Blandford, R. R., 1995, Regional seismic event discrimination. In: Husebye, E. S. and Dainty, A.
M. (eds), Monitoring a Comprehensive Test Ban Treaty, NATO ASI Series, Series E: Applied
Sciences, Kluwer Academic Publishers, Vol. 303, pp. 689–719.
Beck, S. L. and Wallace, T. C., 1995, Broadband seismic recordings of mining explosions and
earthquakes in South America. In Proc. 17th Seismic Research Symposium on Monitoring a
CTBT, Phillips Laboratory, Hanscom Air Force Base, MA, 12–15 Sept., Scottsdale AZ, pp.
157–163.
Chapman, M. C., Bolinger, G. A. and Sibol, M. S., 1992, Modeling delay-fired explosion spectra
at regional distances, Bull. Seism. Soc. Am. 82, 2430–2447.
Cole, R. H., 1948, Underwater Explosions,@seismicisolation
Princeton University Press, 437 pp.
@seismicisolation
Dowla, F., 1995, Neural networks in seismic discrimination. In: Husebye E. S. and Dainty, A. M.
(eds), Monitoring a Comprehensive Test Ban Treaty, NATO ASI Series, Series E: Applied
Sciences, Kluwer Academic Publishers, Vol. 303, pp. 777–790.
Der, Z. A. and Baumgardt, D. R, 1995, Source finiteness, signal decorrelation, spectral scalloping
and identification of multiple delayed explosions, Proc. 17th Seismic Research Symposium on
Monitoring a CTBT, Phillips Laboratory, Hanscom Air Force Base, MA, 12–15 Sept.,
Scottsdale, AZ, pp. 723–732.
Ginzburg, A. and Folkman, Y., 1980, The crustal structure between the Dead Sea rift and the
Mediterranean Sea, Earth Planet. Sci. Lett. 41, 181–188.
Gitterman, Y., 1982, Assessment of dynamic parameters of seismic and air shock waves from
multiple shot systems, Ph.D. Thesis, Institute of Mining, Academy of Sciences of the USSR,
Novosibirsk.
Gitterman, Y. and van Eck, T., 1993, Spectra of quarry blasts and microearthquakes recorded at
local distances in Israel, Bull. Seis. Soc. Am. 83, 1799–1812.
Gitterman, Y. and Shapira, A., 1993, Spectral discrimination of underwater explosions, Isr. J.
Earth Sci. 42, 37–44.
Gitterman, Y. and Shapira, A., 1994, Spectral characteristics of seismic events off the coast of the
Levant, Geophys. J. Int. 116, 485–497.
Gitterman, Y., Pinsky, V. and Shapira, A., 1996, Discrimination of seismic sources using Israel
Seismic Network, PL-TR-96-2207, Phillips Laboratory, Hanscom Air Force Base, MA, 98 pp.
Hedlin, M., Vernon, F., Minster, J. G. and Orcutt, J. A., 1995, Regional small-event identification
using seismic networks and arrays, Proc. 17th Seismic Research Symposium on Monitoring a
CTBT, Phillips Laboratory, Hanscom Air Force Base, MA, 12–15 Sept., Scottsdale, AZ, pp.
875–884.
Jarpe, S. P., Moran, B., Goldstein, P. and Glenn, L. A., 1996, Implications of mining practices in
an open-pit gold mine for monitoring of a Comprehensive Test Ban Treaty, LLNL report
UCRL-ID-123017, 35 pp.
Kim, W. Y., Simpson, D. W. and Richards, P. G., 1994, High-frequency spectra of regional phases
from earthquakes and chemical explosions, Bull. Seis. Soc. Am. 84, 1365–1386.
Li, Y., Toksöz, M. N. and Rodi, W. L., 1996, Discrimination of small earthquakes and explosions.
In: Proceedings of the 18th Symposium on Monitoring a Comprehensive Test Ban Treaty,
Phillips Laboratory, Hanscom Air Force Base, MA, PL-TR-96-2153, 4–6 Sept., Annapolis, pp.
574–583.
Neidell, N. S. and Taner, M. T., 1971, Semblance and other coherency measures for multichannel
data, Geophysics 36, 482– 497.
Pulli, J. J., 1995, Extracting and processing signal parameters for regional seismic event
identification. In: Husebye, E. S. and Dainty, A. S. (eds), Monitoring a CTBT, NATO ASI
Series, Series E, Kluwer Academic Publishers, Vol. 303, pp. 741–754.
Shapira, A., 1988, Magnitude scales for regional earthquakes monitored in Israel, Isr. J. Earth Sci.
37, 17–22.
Stump, B. W. and Reinke, R. E., 1988, Experimental confirmation of superposition from small
explosions, Bull. Seism. Soc. Am. 78, 1059–1073.
Taylor, S. R., Sherman, N. W. and Denny, M. D., 1988, Spectral discrimination between NTS
explosions and Western United States earthquakes at regional distances, Bull. Seis. Soc. Am. 78,
1563–1579.
Tsvang, S. L., Pinsky, V. I. and Husebye, E. S., 1993, Enhanced seismic source discrimination
using NORESS recordings from Eurasian events, Geophys. J. Int. 112, 1–14.
Walter, W. R., Mayeda, K. M. and Patton, H. J., 1995, Phase and spectral ratio discrimination
between NTS earthquakes and Explosions. Part I: Empirical observations, Bull. Seis. Soc. Am.
85, 1050–1067.
Wüster, J., 1993, Discrimination of chemical explosions and earthquakes in central Europe – a
case study, Bull. Seism. Soc. Am. 83, 1184–1212.
@seismicisolation
@seismicisolation
CHAPTER FOUR
DIFFERENCE BETWEEN MICRO-EARTHQUAKES AND ARTIFICIAL

EXPLOSIONS ON THE BASIS OF FREQUENCY CONTENTS1
SO GU KIM AND YONG-CHEOL PARK*
Dynamite explosion for Petroleum Exploration in Wyoming, USA by courtesy of S. G. Kim

(SSC).
ABSTRACT
In this study, our purpose is to develop a technique to discriminate artificial explosions from local
microearthquakes on the basis of time-frequency domain. To obtain spectral features of artificial
explosions and microearthquakes, we used 3-d spectrograms (frequency, time and amplitude)
because this is a useful tool to study the frequency content of entire seismic waveforms observed
at local and regional distances (e.g., Kim et al., 1994). P and S waves from quarry blasts show that
frequency content of dominant amplitude appeared above 10 Hz, and Rg phases are observed at
near distance ranges. But P and S waves from microearthquakes have more broad frequency
content as well as below 10 Hz. And for discrimination, Pw'Lg spectral ratio is performed below
10 Hz. In order to select time windows, we computed group velocity using multiple filter analysis
(MFA) and removed free surface effects from all 3-components data for improving on data quality.
*Present Address, Korea Polar Research Institute, @seismicisolation

Incheon, Republic of Korea.
@seismicisolation
Difference between Micro-earthquakes and Artificial Explosions on the Basis of Frequency Contents 79
Using FFI', a log average spectral amplitude is calculated over seven frequency bands: 0.5 to 3, 2
to 4, 3 to 5, 4 to 6, 5 to 7, 6 to 8 and 8 to 10 Hz. The best separation is observed from 6 to 8 Hz.
INTRODUCTION
A growing interest in discriminating explosions from earthquakes at local and/or regional
distances is also associated with negotiations on a Comprehensive Test Ban Treaty (CTBT) as well
as identifying (looking for) local active faults. The CTBT will call for a monitoring seismic
network of international monitoring system (IMS) as well as local and/or regional monitoring to
detect man-made seismic activities, such as quarry and mine blasts that often consist of several
delayed blasts, sometimes referred to as "ripple firing" including clandestine underground
explosions. If two or more explosions are fired at nearly the same location but with small time
delays, the seismically recorded waveforms for each of the explosions should be very similar
although their amplitudes may vary if the sizes of delayed explosions are different. The amplitude
spectrum computed on a time window which incorporates all the delayed-explosion signals will
exhibit a characteristic modulation or scalloping pattem (Douglas et al., 1988). These rippled-fired
chemical explosions may produce a unique spectral signature, and recognition of these features
facilitates discrimination of chemical from nuclear explosions and microearthquakes.
This study presented the concept of spectral modulation and spectral features of explosions and
microearthquakes. We also tried to discriminate using Pg/Lg spectral amplitude ratio over seven
frequency bands.
THEORY OF SPECTRAL MODULATION IN RIPPLE-FIRED EXPLOSIONS
Douglas and Kathleen (l988) consider the theoretically expected spectra of seismic signals from
ripple-fired explosions, both single and multiple delays. They expressed the equation of multiple
signal y (t) with n delays, with nth delay being scaled by an (a real value) and delayed by Ĳn as:
They analyze the above equation in the cepstrum and this equation shows that the various
periodicities in the spectral modulations depend not only on the basic delay, Ĳn11, but also on the
interaction of delays, Ĳn - Ĳm . The required coarseness of the fragmentation governs the length of
the delays, with longer delays required for higher coarseness. For ground-vibration reduction, the
required delay interval is directly proportional to the period of the strongest vibrations (Chapman
et al.; Douglas and Kathleen, 1988).
@seismicisolation
@seismicisolation
80 Chapter Four
Smith (1989) also examined the effect of ripple firing on a sequence of shots at a single location.
He used u(t) as the seismic signal from a single shot and U(f) as its Fourier transform(or
spectrum), and assumed a constant time delay 8 1 between each shot or set of simultaneous
shots(e.g., a row of shots), then the signal s(t) from the ripple-fired blast is the sum of all N shots:
(4)
where r(m8 1) represents the number of charges at time m8 1 from the initiation of the blast. If
time delays between shots are not uniform, the above equation is represented as a continuous
integral:
࢚ࡺ
S(t)=‫ࢀ׬‬ୀ૙ ࢘(࣎)࢛(࢚ െ ࣎)ࢊ࣎ (5)
Where tN, is the total duration of the blast and r( r ) describes either the number of charges
exploding at a specific time r or the total size of the charges initiated at time r•
These equations represent the discrete convolution of two functions, the source function u (t) for a
single shot or row and the delay-time function r (t) for the shooting pattern:
s(t) = Ĳ(t) • u(t). (6)
The Fourier transforms is
S(f) = R(f) • U(f). (7)
If each of shots is equal size in similar conditions,
(8)
Using the above equations, the amplitude spectrum of U(f) becomes (Gitterman and Eck,, 1993):
(9)
Thus ripple-firing procedure can be assumed as a complex filter modulating the spectrum of single
shot, so to speak some frequencies are enhanced and others are suppressed. Figure 1 shows feature
of spectrum modulation that is proportional to the number of rows N. In this figure, we fmd
absolute maxima at frequencies fmax= k/¨W for k = 0, 1, 2, ···, minor maxima at frequencies fmin =
(2k + l)/(2N¨t) for k = 1, 2, ···, N-2, N+2, ··· and minima or zeros at frequencies fmin = k/ (N¨t)
for k = 1, 2, ···, N-1, N+l, ···. These principles allow us to predict the spectral modulations
introduced by ripple firing known blast patterns (Smith, 1989; Gitterman and Eck, 1993).
DATA ANALYSIS
The data used for this study were recorded by temporal SIHY seismic 81TllY, BHS, SMW,
KIGAM seismic array, IRIS (lnchon station) and KSRS seismic array. BHS and SMW stations
are operated by SIHY in the North of Seoul. BHS and temporal SIHY seismic array are equipped
with velocity seismometers which are Springnether 8-6000 and
@seismicisolation
@seismicisolation
Table 1. Station Locations
Station Name Latitude (°N) Longitude (°E) Place

BHS 37.65 126.96 Bukhan-san
SMW 37.75 1Z7.16 Soomok-won
BBK 35.58 129.44 Bangbang-kol
CHS 36.18 129.09
DKJ 35.94 129.11 Dukjung-ri
HAK 35.93 129.50 Hakge-ri
KJM 34.83 128.59
KMH 35.18 128.93
MAK 35.37 129.18 Makok-ri
MKL 35.72" 129.24 Myunke-ri
KSRS 37.48 1Z7.90 Wonju
IRIS 37.48 126.63 Inchon
04/25/96
SIHYl-1 37.58 128.28 Kudu-mi
SIHY2-3 37.49 128.00 Hoengsung-gun
SIHY3-1 37.58 128.54 Mapyung-ri
04/26/96
SIHYl-1 37.50 1Z7.63 Yongmun-san
SIHY2-3 37.50 128.07 Yongdun-ri
SIHY3-3 37.53 1Z7.83 Sungji-bong
Mark Product lAC-3D respectively. KIGAM seismic array is equipped with 3 component
seismometers either Markrand JV-100 or JV-200. JV-100 has 1 Hz natural frequency, and its
amplitude instrument response is flat to velocity from 1 to 12.5(50 SPS) or 25.0(100 SPS), and
JV-200 is a broad band seismometer. IRIS Inchon station is a broadband type of seismometers
(VBB, LP, VLP, ULP, VSP,...), but IRIS data used in this study are only two types: one is VBB
(very broad band seismometer) whose sampling rate is 20 SPS on 24-bit channels from STS-1
seismomters, the other is VSP (very short period seismometer) whose sampling rate is 80 SPS and
also 24-bit channels from 3 component STS-2 high gain seismometer. Table 1 shows
configurations of each seismic station.
Table 2 shows data of seismic events used in this study. Our data set of 28 small events, of which
10 are explosions (event nos. 1 to 10) and 16 are microearthquakes (event nos. 11 to 26). The
hypocentral parameters of all explosion data are immediately determined by previous studies (Kim
et al., 1995; Kim and Park, 1996). And the event of number 27 and 28 are unknown events
recorded at BHS on October 31, 1996, and December 25, 1996, respectively. Figure 2 shows
location of events and each seismic station. In this figure, open stars and dark stars represent
explosions and microearthquakes, respectively and closed circles represent unknown events.
Closed triangles indicate KIGAM seismic array, open triangles and squares indicate SIHY
temporal seismic stations and an open circle indicates IRIS (Inchon) station and KSRS Array
(Wonju). Figures 3 and 4 show vertical seismograms of explosions and microearthquakes,
respectively.
@seismicisolation
@seismicisolation
82 Chapter Four
Table 2. Explosions and Earthquakes Used in This Study
• no time correction
@seismicisolation
@seismicisolation
@seismicisolation
@seismicisolation
84 Chapter Four
In order to study the spectral characteristics of explosions and micro earthquakes, we computed
frequency-time displays of seismograms (spectrogram). This method is a useful tool to study the
frequency content of entire seismic waveforms observed at local and regional distances (e.g., Kim
et al., 1994). In this computation, spectral estimates were calculated for each time moving window
(usually use about 4 sec), with 75 per cent overlapping between each successive time window, by
applying the adaptive multi-taper spectral estimation method. Next, spectral estimates of all time
windows were presented in time-frequency space using a continuous curvature surface gridding
algorithm. All spectrograms shown in this study correspond to the signal velocity spectra.
QUARRY BLASTS (RIPPLE FIRED EXPLOSIONS) AND MICROEARTHQUAKES
We analysed three quarry blasts recorded on the SIHY temporal seismic stations from April 24 to
27, 1996. The first explosion was observed on April 25, 1996. The sticks of dynamite were loaded
into 80 drill holes with 7.5 cm diameter (vertical 57, horizontal 23; event number 8) in depth 25m
and total yield is 7,500 kg. On April 26, 1996, the second and third explosions were carried out in
the all horizontal holes (S.0-cm-diameter drill hole and depth Sm) and total yield is 2,790 kg and
2,629 kg, respectively (event number 9 and 10 in the table 2). Figure 5 shows spectrograms of
quarry blasts, and figures 6 and 7 show spectrograms of micro earthquakes. Example spectrograms
for the blast show that there are weak but clear spectrum modulation, and we can notice that body
waves form quarry blasts have high energy and narrow frequency content above 10 Hz due to
spectral modulation caused by ripple-firing of subshots independent of distance. The short-period
Rg waves are essentially confined to depths of about 1 to 5 km, and as such, Rg signals are
significant attenuation resulting from low Q materials in the shallow crust (Kim et aL,1996).
Hence, Rg phases are observed only at near distance ranges (Figure 5-a). S waves from
earthquakes are stronger than P waves below a low 10 Hz, and microearthquakes have lower
(below 10 Hz) dominant and more broad frequency contents (Figure 6 to 8). Figure 9 shows
summed amplitude spectra of P waves at three stations in the distance range 37 to 61 km(Fig 9-a)
and 85 to 124 km (Fig. 9-b and 9-c) from quarry blasts, and figure 10 is from earthquakes at three
stations (Fig. 10-a), four stations (Fig. 10-b) and five stations (Fig. 10-c). Dot lines indicate
spectrum of each station and thick lines indicate their mean spectra. We can notice a clear
difference in P-wave amplitudes and frequency contents. Above 10 Hz the figure 9 shows spectra
modulations over all quarry blasts, and P-wave spectra from earthquakes have large amplitude in
low frequency band (below 10 Hz) as well as high frequency band.
Figure 3. Seismograms of explosions used in this study.

@seismicisolation
@seismicisolation
Figure 4. Seismograms of microearthquakes used in this study.
THE REMOVAL OF FREE SURFACE INTERACTION & FROM THREE-COMPONENT

DATA
The key to the successful discrimination of various types of seismic sources is the clear
observation of the signals radiated from the seismic sources. one of the most important problems is
correction of the effects of the source to receiver paths on the observed regional signals (Kim and
Richards, 1996). Any theoretical description of the passage of seismic waves from source to
receiver has to include the effects of the structure near the receiver. The presence of the free
surface on which traction vanishes leads to reflection of upgoing waves back into the medium
when they may undergo further interactions before return to the surface. The first stage is to rotate
the components in the horizontal plane into a (Z, R, T) coordinate system radial and tangential to
the path from the source using the azimuth. In the rotated coordinate system, using the notation of
Aki and Richards (1980, chapter 5), the motion of the free surface itself is
[R(t), 0, -Z(t)] (10)

@seismicisolation
@seismicisolation
86 Chapter Four
where Į = surface P-wave velocity, ȕ = surface $-wave velocity, i = angle of incidence of P-wave,
j = angle of incidence of S-wave, p = slowness and sinLĮ = sinMȕ = p. Figure 11 shows a
theoretical sketch of ground motions (Kim and Richards, 1996) in a homogeneous half-space for a
P-wave incident on a free surface with unit amplitude and an angle of incidence i = 33.70° (=
0.1s/km), surface P and S velocities: Į = 5.5 km/s and ȕ =3.1 km/s, respectively. The total ground
motion of the free surface (shaded line) is shown above the surface level z = 0 and has amplitude
2.03 times the incident wave with an apparent angle of incidence i' = 36.12°. The incident P- and
SV-wave motions can then be recovered as a linear combination of the radial, R(t), and vertical,
Z(t), components available from three-component records. Thus the above equation can be used to
show that (Kim and Richards, 1996)
P(t) = cos2j/2cosi + Z(t) + ȕĮVLQM + R(t) (11)
Figure 5. Spectrograms from quarry blast on 4/25/96 recorded at SIHY temporal seismic array (Event No. 8 in table 2)
(left). Figure 6. Spectrograms from Munkyong earthquake on 08/14/96 recorded at KIGAM seismic array (Event
No.21 in table 2). (right)
@seismicisolation
@seismicisolation
OIIOIIM
(b)
Figure 7.. Spectrograms from samchok earthquakes on 08/03/96 recorded at KIGAM seismic array (Event No. 20 in
table 2). Samchok earthquake recorded at MAK and DKJ (b).
SV(t) = cos2j/2cosj * R(t) – sinj * Z(t) (12)

The estimate of the incident SH-wavefield is simply one half of the tangential component after
rotation and is thus quite sensitive to the azimuth employed (Kennett, 1991).
SH(t) = ½ T(t) (13)
Figures 12 and 13 show examples of free surface correction data. After correction of free surface
effect, P waves are mainly on the P-wavevector component, and Lg waves are dominantly on the
SV-wavevector component in case of explosions (see figure 13).
CALCULATION OF PG/LG SPECTRAL RATIOS
We calculated Pg/Lg ratios below 10 Hz in order to use data recorded by KSRS and IRIS broad-
band seismic stations with 20 SPS. To calculate Pg/Lg ratios, we removed free surface effects on
our data since this procedure can make direct comparison between the P-, SV- and SH-wave
amplitudes in a particular group velocity window and so get closer to the radiation characteristics
from the source, especially with regard to the ratio of SV to SH in the regional S-waves (Kennett,
1991). Figure 14 shows group velocity that is computed by the Multiple Filter Analysis (MFA).
The largest amplitudes of envelope for each frequency are plotted as squares, then circles,
triangles, and pluses for correspondingly smaller amplitudes. The seismogram on the right is the
waveform, while the inner trace represents the waveform scaled linearly versus apparent group
velocity. In this figure the breaks between different slowness bands are marked by horizontal bars,
and slowness (in !Vkm) used in each group velocity window is indicated. The Pwl,g ratios from
the free surface corrected P- and SV- and SH-seismograms of three-component regional records
are calculated from the average of the seven frequency bands: 0.5 to 3, 2 to 4, 3 to 5, 4 to 6, 5 to 7,
6 to 8 and 8 to 10 Hz and obtained for each station by defining
@seismicisolation
@seismicisolation
88 Chapter Four
(14)
where Pgp = spectral amplitude of P-wave on the P-seismogram, LgSV = spectral amplitude of the
Lg-wave on the SV-seismogram and LgSH = spectral amplitude of the Lg-wave on the SH-
seismogram. Figure 15 shows results of Log(Pg/Lg) ratios over seven frequency bands. In this
figure we can see the best separation is observed in 6 to 8 Hz. With this result we can presume that
the unknown events (event no. 23 and 24) are presumed earthquakes.
Figure 8. Spectrograms from earthquakes on 09/02/96 recorded at KIGAM seismic array (Event No. 21).
@seismicisolation
@seismicisolation

Through spectra analysis, it is found that the features of S waves from earthquakes are stronger
than those of P waves over a broad frequency band (up to 20 Hz). P and S waves from earthquakes
show that high spectral amplitudes appear in lower-frequency content (between 1 and 10 Hz; Figs
7 to 10), while P and S waves from quarry blasts have high amplitude in the part of high frequency
(above 10 Hz) because of spectral modulation caused by ripple firing subshots. It is shown in the
spectrograms and the summed spectra that the spectral modulation of ripple-fired blast is
independent of paths and station site responses (Figs. 5 and 9).
For the sake of discrimination, we tried to calculate Pg/l,g ratios after correction of free surface
effect. A very useful product of this procedure is that we can make direct comparison between the
P-, SV- and SH-wave amplitudes and so get closer to the radiation characteristics from source. On
the SH vector component, Lg waves have relatively small amplitude in case of explosions. This is
of potential significance for discriminating different types of sources. After correction of free
surface effect, we calculate Pg/Lg ratios over seven frequency bands. And we obtain the best
separation in 6 to 8 Hz band. The difference of the explosion Pg/Lg ratio with frequency is not
totally understood. But one effect that could account for this is understood. But one effect that
could account for this Rg to S (and Lg) scattering (Gupta et al., 1992; Steven, 1996) that would
boost the Lg amplitudes at low frequencies.
Figure 9. Summed spectra at three stations of quarry blasts. Dot lines indicate amplitude spectra of each station and
thick lines indicate summed spectra. (left). Figure 10. Summed spectra at four or five stations of earthquakes. Dot
lines indicate amplitude spectra of each station and thick lines indicate summed spectra (right).
@seismicisolation
@seismicisolation
90 Chapter Four
A key to the Pg/Lg spectral ratio is the correction of regional signals for their source to receiver
path effects, and the frequency contents of P and S waves depend on specific propagation paths
and local structure. So we have to carry out special study, for example the dispersive properties of
surface waves.
incident P
Figure 11. Theoretical sketch of ground motions in a homogeneous half-space
@seismicisolation
@seismicisolation
Figure 12. Example of free surface correction data. (Upper three traces) Row three-component records from Tongo-
san earthquake (event no. 18). (Lower three trace) composite incident wave vector component traces (P, SV and SH)
produced by applying free surface correction.
@seismicisolation
@seismicisolation
92 Chapter Four
Figure 12. continued
@seismicisolation
@seismicisolation
Figure 13. Example of free surface correction data. (Upper three traces) Row three-component records from quarry
blast (event no. 8). (Lower three traces) composite incident wave vector component traces (P, SV and SH) produced
by applying free surface correction.
@seismicisolation
@seismicisolation
94 Chapter Four
(c)
Figure 13. continued.
@seismicisolation
@seismicisolation
Figure 14. Group velocity is computed by the Multiple Filter Analysis (MFA). The largest amplitudes of envelope for
each frequency are plotted as squares, then circles, triangles, and pulses for correspondingly smaller amplitudes. And
the seismogram on the right is the waveform, while the inner trace represents the waveform scaled linearly versus
apparent group velocity
@seismicisolation
@seismicisolation
96 Chapter Four
Figure 15. Pg/Lg ratios for the seven frequency bands. Lines show average values for each event type.
ACKNOWLEDGMENT
This study was supported by the Korean Ministry of Education (BSRl-96-5420). This research
was also partly financed by the Korea Institute of Geology, Mining and Materials (KIGAM),
Daejeon, Republic of Korea.
REFERENCES
Aki, K. and P. G. Richards (1980). Quantitative Seismology, 2 vols, W. H. Freeman, San
Francisco.
Smith, A. T. (1989). High-frequency seismic observations and models of chemical explosions:
implications for the discrimination of ripple-fired mining blasts, Bull. Seism. Soc. Am., 79, 4,
1089-1110.
Chapman, M, C., G. A. Bollinger and M. S. Sibol (1992). Modeling delay-fired explosion spectra
at regional distances, Bull. Seism. Soc. Am., 82, 6, 2430-2447.
Bqumgardt, D. R. and K. A. Ziegler (1988). Spectral evidence for source multiplicity in explosions
: application to regional discrimination of earthquakes and explosions, Bull. Seism. Soc. Am.,
78, 5, 1773-1795. underground nuclear explosions at East Kazakh and Nevada Test Sites, BulL
Sei.sm. Soc. Am., 82, 352-382.
Kennett, B. L. N. (1991). The removal @seismicisolation
of free surface interactions from the three-component
@seismicisolation
seismogram&, Geophys. J. Int., 104, 153-163.

Kim, So Gu and Yong-cheol Park (1996). Discriminate between Small Explosions in Quarry Sites
and Microearthquakes using amplitude ratio, BulL Sei.sm. Assoc. of the Far East (SAFE), 2, 1,
47-63.
Kim, So Gu, Seoung-Kyu Lee, Sang-Yun Mah and Yong-cheol Park (1995). Detection and
analysis of the artificial underground explosions in N. Korea using KSRS data, J. of Eng.
Geo/,ogy, 5, 2, 181-192.
Kim, So Gu and Zhongliang Wu (1996). Seismological problems associated with decoupled
explosions: an overview, report of Science and Technology Policy Institute of Korea.
Kim, W. Y., D. W. Simpson and P. G. Richards (1994). High-frequency spectra of regional phases
from earthquakes and chemical explosions, Bull. Seism. Soc. Am., 84, 5, 1365- 1386.
Kim, W. Y. and Paul G. Richards (1996). Discrimination of regional earthquakes and explosions
using three-component high-frequency digital data, Modem Seismology- Proceedings of The
Korea-China International Joint Seminar and Seismological Workshop edited by S. G. Kim, 46-
61.
Taylor, S. R. (1996). Analysis of high-frequency Pwl-,g ratios from NTS explosions and Western
U.S. Earthquakes, BulL Seism. Soc. Am., 86, 4, 1042-1063.
Gitterman, Y. and T. V. Eck (1993). Spectra of Quarry Blasts and Microearthquakes Recorded at
Local Distances in Israel, Bull. Seism. Soc. Am., 83, 6, 1799-1812.
@seismicisolation
@seismicisolation
APPENDIX
DISCRIMINATION OF SMALL EARTHQUAKES AND ARTIFICIAL EXPLOSIONS IN
THE KOREAN PENINSULA USING Pg/Lg RATIOS
SO GU KIM, YONG-CHEOL PARK
AND WON-YOUNG KIM
SUMMARY
The purpose of this study is to develop a technique to discriminate artificial explosions from local
small earthquakes (Mч4.0) in the time–frequency domain. In order to obtain spectral features of
artificial explosions and earthquakes, 3-D spectrograms (frequency, time and amplitude) have been
used. They represent a useful tool for studying the frequency content of entire seismic waveforms
observed at local and regional distances (Kim, Simpson & Richards 1994). P and S(L g) waves
from quarry blasts show that the frequency content associated with the dominant amplitude
appears above 10 Hz and Rg phases are observed at close distances. P and S(L g) waves from the
Tongosan earthquake have strong amplitudes below 10 Hz. For the Munkyong earthquake,
however, a broader frequency content up to 20 Hz is found.
For discrimination between small earthquakes and explosions, Pg/Lg spectral ratios are used below
10 Hz, and through spectrogram analysis we can see different frequency contents of explosions
and earthquakes. Unfortunately, because explosion data recorded at KSRS array are digitized at 20
sps, we cannot avoid analysing below 10 Hz because of the Nyquist frequency. In order to select
time windows, the group velocity was computed using multiple-filter analysis (MFA), and free-
surface effects have been removed from all three-component data in order to improve data quality.
Using FFT, a log-average spectral amplitude is calculated over seven frequency bands: 0.5 to 3, 2
to 4, 3 to 5, 4 to 6, 5 to 7, 6 to 8 and 8 to 10 Hz. The best separation between explosions and
earthquakes is observed from 6 to 8 Hz. In this frequency band we can separate explosions with
log(Pg/L g) above о0.5, except EXP1 recorded at SIHY1-1, and earthquakes below о0.5, except
the Munkyong earthquake record at station KMH.
Keywords: artificial explosion, CTBT, free-surface effect, ripple firing, spectral ratio,
spectrogram.
INTRODUCTION
A growing interest in the discrimination between explosions and earthquakes at local and/or
regional distances is associated with the negotiation of a Comprehensive Test Ban Treaty (CTBT),
as well as with identifying and/or searching for local active faults. The CTBT will call for a
seismic network of international monitoring systems (IMS) as well as local and/or regional
monitoring to detect man-made seismic activities such as quarry and mine blasts (which often
consist of several delayed blasts, sometimes referred to as ‘ripple firing’), which may also include
clandestine underground explosions. If two or more explosions are fired at nearly the same
location but with small time delays, the seismically recorded waveforms for each of the explosions
should be very similar, although their amplitudes may vary if the sizes of the delayed
explosions are different. The amplitude spectrum computed for a time window which incorporates
all of the delayed explosion signals will exhibit a characteristic modulation or scalloping pattern
(Baumgardt & Ziegler 1988). Such ripple-fired chemical explosions may produce a unique spectral
signature, and recognition of these features facilitates the discrimination of chemical from nuclear
explosions and earthquakes.
________________
Geophys. J. Int. (1998). 134, 267–276
@seismicisolation
@seismicisolation
In the Korean Peninsula the first study of the discrimination between earthquakes and explosions
was undertaken by Kim (1994); he discriminated earthquakes from underground explosions (in
North Korea) using the characteristics of pure continental-path seismic waves and also showed that
seismic signals are more affected by source mechanism (double-couple earthquakes and single-
couple explosions) than path. Kim & Park (1996) tried to discriminate small earthquakes from an
amplitude–energy ratio; they obtained an index ratio (0.7) of summation of amplitude study. Our
data set contains 32 small events (Mч4.0), of for low-frequency (1~10 Hz) to high-frequency
(10~20 Hz) which 10 are explosions (event numbers 1 to 10) and 20 are energy. Measurements
above 0.7 are associated with ripple-small earthquakes (event numbers 11 to 30). The hypocentral
fired quarry explosions and below this value are associated parameters of all explosion data have
been determined in with small earthquakes. In this study we try to find spectral previous studies
(Kim et al. 1995; Kim & Park 1996). Events characteristics of small earthquakes and explosions
and dis-numbers 31 and 32 are unknown events recorded at BHS on criminate small earthquakes
from explosions using the Pg/L g October 31 1996 and December 25 1996, respectively. Fig. 1
ratio.
DATA
The data used for this study were recorded by the temporary SHY seismic arrays, BHS, SMW, the
KIGAM seismic arrays, IRIS (Incheon station) and KSRS seismic array. Stations BHS and SMW
are operated by the SIHY (Seismological Institute of Hanyang University) tos recorded on the
SIHY Institute of Hanyang University) to the north of Seoul. BHS and the temporary SIHY
seismic array are equipped with Springnether S-6000 and Mark Product L4C-3D velocity
seismometers, respectively. The KIGAM seismic array is equipped with 57, horizontal 23; event
number 8), a depth of 25 m and a three-component seismometers, either Markrand JV-100 or total
yield of 7500 kg. On April 26 1996, the second and third JV-200. JV-100 has a 1 Hz natural
frequency and its amplitude explosions were carried out in the horizontal holes (8.0 cm instrument
response is flat to velocity from 1 to 12.5 (50 SPS) diameter drill hole, 8 m depth), and the total
yields were 2790 or 25.5 Hz (100 SPS), and JV-200 is a broad-band seismometer. and 2629 kg,
respectively (event numbers 9 and 10 in Table 2). The IRIS Inchon station had broad-band types of
seismometers (VBB, LP, VLP, ULP, VSP channels, etc.) but IRIS data used in this study are of
only two types: VBE (very broad-All explosions have a 100 ms delay time between charges. band
seismometer), whose sampling rate is 20 SPS on 24-bit channels from STS-1 seismometers, and
VSP (very short-period seismometer), whose sampling rate is 80 SPS, also on 24-bit In order to
study the spectral characteristics of explosions channels from a three-component STS-2 high-gain
seismo- and earthquakes, we have computed frequency–time displays meter. The IRIS VBE
stations and KSRS seismic arrays have for the seismograms (spectrogram). This method is a useful
an 8 Hz cut-off frequency (high-cut filter). Thus we performed tool to study the frequency content
of entire seismic wave- instrument correction that divided amplitude spectra into forms observed at
local and regional distances (e.g. Kim et al. instrument response spectrum slopes in the frequency
domain 1994). In this computation, spectral estimates were calculated up to 10 Hz. Table 1 shows
the configuration for each seismic station. Table 2 shows the data for the seismic events used in
this study. Our data set contains 32 small events (M4.0), of which 10 are explosions (event
numbers 1 to 10) and 20 are small earthquakes (event numbers 11 to 30). The hypocentral
parameters of all explosions data have been determined in previous studies (Kim et al., 1995; Kim
& Park, 1996). Event numbers 31 and 32 are unknown events recorded at BHS on October 31,
1996 and December 25, 1996, respectively. Fig. 1 shows the locations of the events and each
seismic station. In this figure, open stars and solid stars represent explosions and earthquakes,
respectively., and solid circles represent unknown events. Solid triangles indicate the KIGAM
seismic array, open triangles and squares indicate the SIHY temporary seismic stations and open
circles indicate station IRIS, Incheon) and the KSRS array (Wonju). We have analysed three
quarry blasts recorded on the SIHY temporary seismic stations from April 24 to 27 1996. The first
explosion was observed on April 25 1996. Sticks of dynamite were loaded into 80 drill holes with
@seismicisolation
@seismicisolation
100 Chapter Four
a 7.5 cm diameter (vertical 57, horizontal 23; event number 8), a depth 25 m and a total yield 7500
kg. On April 26 1explosions have 1966, the second and third explosions were carried out in the
horizontal holes (8.0 cm diameter drill hole 8 m depth), and the total yields were 2790 and 2629
kg, respectively (event numbers 9 and 10 in Table 2). All explosions have a 100 ms delay time
between charges.
Table 1. Configuration of the seismic stations
Station Name Instrument Lat. (°N) Long. (°E) Place
BHS DR-2000/S6000 37.65 126.96 Bukhan-san
SMW 16ACT/J100 37.75 127.16 Soomok-won
BBK // 35.58 129.44 Bangbang-kol
CHS // 36.18 129.09
DKJ // 35.94 129.11 Dukjung-ri
HAK // 35.93 129.50 Hakge-ri
KJM // 34.83 128.59
KMH // 35.18 128.93
MAK // 35.37 129.18 Makok-ri
MKL // 35.72 129.24 Myunke-ri
KSRS // 37.48 127.90 Wonju
IRIS // 37.48 126.63 Inchon
04/25/96*
SIHY-1 PDAS 100/LAC 37.58 128.28 Kudu-mi
SIHY2-3 // 37.49 128.00 Hoengsung-gu
SIHY3-1 // 37.58 128.54 Mapyung-ri
04/26/96*
SIHY1-1 // 37.50 127.63 Yongmun-san
SIHY2-3 // 37.50 128.07 Yongdun-ri
SIHY3-3 // 37.53 127.83 Sungji-bong
* Date of seismic experiment for studying the velocity structure in the Central Korean Peninsula.
DATA ANALYSIS AND INTERPRETATION
In order to study the spectral characteristics of explosions and earthquakes, we have computed
frequency–time displays for the seismograms (spectrogram). This method is a useful tool to study
the frequency content of entire seismic waveforms observed at local and regional distances (e.g.
1994). In this computation, spectral estimates were for each moving time window (usually we use
75 per cent overlap between each successive by applying the adaptive multitaper spectral
estimation method. Spectral estimates of all time windows are presented in time-frequency space
using a continuous curvature algorithm. All spectrograms shown in this study correspond the
signal velocity spectra.
@seismicisolation
@seismicisolation
Table 2. Explosions and earthquakes used in this study.
* no time correction.
† UTC.
@seismicisolation
@seismicisolation
102 Chapter Four
Figure 1. Location of SIHY temporary seismic stations in the central Korean Peninsula (open squares: 04/25/96; open
triangles: 04/26/96), KIGAM seismic stations in the Kyongsang basin, KSRS at Wonju and station IRIS at Inchon
(open circles). Open stars are explosions, solid stars are microearthquakes and solid circles are unknown events.
Fig 2. Shows the spectrograms of the quarry blast (event 8); from microearthquakes by observing
the frequency contents there is weak but clear spectral modulation. Note also that the body waves
of the quarry blasts have high energy and a narrow frequency content above 10 Hz .The short-
period Rg are essentially confined to depths of about 1 to 5 km and waves such as Rg signals
suffer significant attenuation resulting from low-Q materials in the shallow crust (Kim & Richards,
+HQFH5JSKDVHVDUHREVHUYHGRQO\DWFORVHGLVWDQFHV )LJD¨ NP DQGZHFDQ
observe that the amplitudes of S (Lg) waves are very small compared to P-wave amplitudes (Fig.
2a). Figs 3 and 4 are seismograms of the Tongosan (event 19) and Munkyong earthquake (event
23), respectively. Note that the peak amplitudes od P and S (Lg) waves appear at about 3 to 5 Hz
and S (Lg) waves are stronger than the P waves. Small earthquakes have lower dominant
frequencies ( below 10 Hz) and broader frequency contents. By means of of spectrograms, one
can distinguish explosions from microearthquakes by observing the frequency contents of P and S
(Lg) waves..
We calculated Pg/L g ratios below 10 Hz in order to use narrow frequency content above 10 Hz in
order to use data recorded by the KSRS and IRIS broad-band seismic stations with 20 sps
@seismicisolation
@seismicisolation
sampling. To calculate Pg/Lg ratios, we removed free-surface effects in our data, since this
procedure can facilitate a direct comparison between the P- , SV- and SH-wave amplitudes in a
particular group-velocity window so become closer to the radiation characteristics of the source,
especially with regard to the ratio of SV to SH waves in the regional S wavefield (Kennett, 1991). .
The key to the successful discrimination of various types of seismic sources is the clear
observation of the signal radiated from the seismic source. One of the most important problems is
the correction of the effects of the source to receiver paths in the observed regional signals (Kim &
Richards, 1996). Any theoretical description of the passage of seismic waves from source to
receiver must include the effects of the structure near the receiver. The presence of the free
surface, on which traction vanishes, leads to the reflection of the upgoing waves back into the
medium, where they may undergo further interactions before returning to the surface.
Figure 5. (a) The group velocity of KSRS07 (event 2, recorded at KSRS) is computed by MFA. The largest amplitudes
of envelopes for each frequency are plotted as squares, then circles, triangles and pluses for correspondingly smaller
amplitudes. The seismogram on the right is the waveform, while the inner trace represents the waveform scaled
linearly versus apparent group velocity. (b) Group velocity of the Munkyong earthquake (event 22, recorded at MAK).
@seismicisolation
@seismicisolation
104 Chapter Four
Fig. 5 shows the group velocity computed by MFA. In order to determine the group velocity of Lg
waves, one can use MFA. The largest amplitudes of the envelope at each frequency are plotted as
squares, then circles, triangles and pluses for correspondingly smaller amplitudes (Herrmann,
1977). The seismogram on the right is the waveform, while the inner trace represents for the
waveform scaled linearly versus apparent group velocity.
Fig. 6 shows examples of free-surface corrections applied to data. After correction for the free-
surface effect, P waves are mainly on the P-wave-vector component, and Lg waves are dominantly
on the SV-wave-vector component in the case of explosions. Fig. 6 (b) in particular represents
seismograms of a shallow blast, indicating that there are some noisy signals, but the SH The
Pg/L g ratios from the free-sur component of earthquakes are stronger than those of explosions.
Figure 6. Example of free-surface correction data. The upper three traces are raw three-component records from
explosions. The lower three traces are composite incident-wave-vector component traces (P, SV and SH) produced by
applying the free-surface correction. The SH component is weaker than the SV component. (a) KSRS07 (event 2,
explosion); (b) EXP1 (event 8, explosion); (c) the Tongosan earthquake (event 19); (d) the Munkyong earthquake
(event 22).
@seismicisolation
@seismicisolation
Figure 7. Pg/L g ratios for the seven frequency bands. Dashed lines show average values of explosions, and dotted
lines show average values of microearthquakes. Stars and diamonds represent KSRS data and explosions, respectively,
and shaded circles represent microearthquakes. Thick crosses represent unknown events.
@seismicisolation
@seismicisolation
106 Chapter Four
Figure 8. Pg/L g ratios at 6 to 8 Hz with epicentral distances. The dashed line shows the discrimination value í
Stars and diamonds represent KSRS data and explosions, respectively, and circles represent microearthquakes. Thick
crosses represent unknown events.
The Pg,/Lg ratios from the free-surface-corrected P, SV and SH seismograms of three-component

regional record are calculated from the average of seven frequency bands: 0.5-3, 2 to 4, 3 to 5, 4 to
6, 5 to 7, 6 to 8 and 8 to 10 Hz and obtained for each station by defining
where PgP is spectral amplitude of the P wave P seismogram, LgSV is the spectral amplitude of
the Lg wave on the SV seismogram and LgSH is the spectral amplitude of the Lg wave on the SH
seismogram. Fig. 7 shows the log(Pg/L g ratios over seven frequency bands. In this figure, it
would appear that the best separation is observed in the 6 to 8 Hz range, although other frequency
bands are also valid for discrimination. We used many different kinds of digital records whose
sampling rates are 20 to 100 Hz in this study. Due to the limitation of the Nyquist frequency (10
Hz) for broad-band records, Fig. 7 shows the frequency band up to 10 Hz only. With this result we
can infer that the unknown events (events 31 and 32) are earthquakes. Fig. 8 represents Pg/L g
ratios at 6 to 8 Hz as a function of epicentral distance. In this figure we can separate explosions
from small earthquakes, except EXP1 recorded at SIHY1-1 and the Munkyong earthquake
recorded at KMH.
Through spectral analysis, it is found that the features of S waves from earthquakes are stronger
than those of P waves over a broad frequency band (up to 20 Hz). P and S waves from earthquakes
show high spectral amplitudes at lower frequencies (between 1 and 10 Hz Figs 2 to 4), while P and
S waves from quarry blasts have high amplitude at high frequencies (above 10 10 Hz) because of
the spectral modulation caused by ripple-fired subshots.
@seismicisolation
@seismicisolation
For the discrimination studies, we calculate Pg/Lg ratios after correction for the free-surface
effect. A very useful product of this procedure is that we can make direct comparisons between the
P-, SV - and SH-wave amplitudes and so get closer to the radiation characteristics of the source.
On the SH-vector component. Lg waves have relatively small amplitudes in the case of explosions;
this is of potential significance for discriminating different types of sources. After correction for
the effect of the free surface we calculate Pg/Lg ratios over seven frequency bands. The best
separation between earthquakes and explosions occurs in the 6 to 8 Hz band. We found that log
(Pg/L g)хо0.5 describes the explosion category in this study; this value also coincides with the
results of the study of Kim et al. (1997). The variation of the explosion Pg/Lg ratio with frequency
is not totally understood, but one effect that could account for this is Rg to S (and Lg) scattering
(Gupter, Chan & Wagner 1992; Taylor 1996), which would boost the Lg amplitudes at low
frequencies.
A key to the Pg/Lg spectral ratio is the correction of regional signals for their source to receiver
path effects since the frequency contents of P and S waves depend on specific propagation paths
and local structure. We thus need to carry out special studies, for example, using the dispersive
properties of surface waves.
ACKNOWLEDGMENTS
This work was supported by the Korean Ministry of Education (BSRI-96-5420) and the Korea
Research Foundation (1996). We express our thanks to C. Y. Wang, Director of the Geophysical
Institute, Central National University, Taiwan, who rented us the 13 sets of seismic instruments
(PDAS-100 and L-4C) including a Research Assistant graduate student for this study.
REFERENCES
Baumgardt, D.R. & Ziegler, K.A., 1988. Spectral evidence for source multiplicity in explosions:
application to regional discrimination of earthquakes and explosions, Bull. seism. Soc. Am., 78,
1773–1795.
Gupta, L.N., Chan, W.W. & Wagner, R.A., 1992. A comparison of regional phases from
underground nuclear explosions at East Kazakh and Nevada Test Sites, Bull. seism. Soc. Am.,
82, 352–382.
Herrmann, B., 1997. Computer Programs in Seismology, St. Louis University, St. Louis, MO,
P200.
Kennett, B. L. N. 1991. The removal of free surface interactions from three-component
seismograms, Geophys. J. Int., 104, 153-163
Kim, S.G., 1994. Discrimination of underground explosions from microearthquakes through the
pure-continental path, J. Eng. Geol., 4, 29–42.
Kim, S.G. & Park, Y.C., 1996. Discrimination between small explosions in quarry sites and
microearthquakes using amplitude ratio, Bull. seism. Assoc. Far East (SAFE), 2, 47–63.
Kim, S.G., Lee, S.K., Mah, S.Y. & Park, Y.C., 1995. Detection and component, Lg waves have
relatively small amplitudes in analysis of the artificial underground explosions in North
Korea using KSRS data, J. Eng. Geol., 5, 181–192.
Kim, W.Y., Simpson, D.W. & Richards, P.G., 1994. High-frequency spectra of regional phases
from earthquakes and chemical explosions, Bull. seism. Soc. Am., 84, 1365–1386.
Kim, W.Y., Aharonian, A.C., Lerner-Lam, A.L. & Richards, P.G.,1997. Discrimination of
earthquakes and explosions in Southern Russia using regional high-frequency three-component
data from IRIS/JSP Caucasus network, Bull. seism. Soc. Am., 87, 569–588.
Taylor, S. R., 1996. Analysis of high-frequency Pg/Lg ratios from NTS explosions and Western U.
S. Earthquakes, Bull. seism. Soc. Am., 86.
@seismicisolation
@seismicisolation
CHAPTER FIVE
UNCERTAINTIES OF SEISMIC SOURCE DETERMINATION
USING A 3-COMPONENT SINGLE STATION1
SO GU KIM AND ZHONGLIANG WU
Seismic exploration across the central part of the Korean Peninsula using 13 sets of PDAS-100
recorder and L-4C seismometer rent from National Central University, Geophysics Department,
Taiwan.
At present, most of the single-station location approaches are based on the estimation of azimuths
and epicenter distances. In this paper, the uncertainty of such kind of single-station location for
local and regional earthquakes is analyzed systematically. The uncertainty of the azimuth
estimation is proportional to the average noise level of the seismogram taken for analysis.
However, in the analysis, still a larger uncertainty comes from the fact that sometimes the selection
of waveforms is subjective, and the results may depend on that selection. To evaluate this kind of
uncertainty and possibly avoid the arbitrariness of the selection of waveforms, we proposed a
time-direction stacking approach. The uncertainty of the estimation of distances comes from the
travel time readings and the apparent velocity model. To reduce the uncertainties of distance
estimation, we proposed an apparent velocity model which depends on the range of distances. As
1
Originally published in the J. Phys. Earth (1997). 45, 1-11.
@seismicisolation
@seismicisolation
Uncertainties of Seismic Source Determination Using a 3-Component Single Station 109
an application, we present the results of expected uncertainty for the Korean Peninsula and its
vicinity.
INTRODUCTION
In recent years,single-station location has attracted renewed attention in seismology, mainly

because of the developments of digital seismological observation and automatic detection
techniques (e.g., Magotra et al., 1987; Jurkevics, 1988; Roberts et al., 1989; Roberts and
Christoffersson, 1990; Kim and Lee, 1995). In the Korean Peninsula, such a problem is of special
interest because the network coverage in this region is limited due to the geographical border, and
most of the earthquakes that have happened in its vicinity are outside the coverage of the
seismograph network, making the problem of earthquake location degenerate into a single-station
problem (see, e.g., Kim and Lee, 1995).
In the single-station location of earthquakes, estimation of the uncertainty is a fundamental

problem which needs a systematic analysis. Such a problem, while not being complicated in
mathematics, is essential for practical data processing. In this paper, taking the situation of the
Korean Peninsula as an example, we present a systematic analysis of this problem, in which new
approaches are also proposed.
In the analysis we take the station location as 36.30°N x 127.40°E, near the Daejeon station, as an
example. The local and regional earthquakes are defined as those earthquakes with epicenter
distances less than 30°, especially less than 5°, centered at the seismic station, including
earthquakes that have happened in the Korean Peninsula and its vicinity.
At local and regional distances, the location of earthquakes with a single three-component station
is often obtained by measuring the azimuth and the epicenter distance. Given the latitude and
longitude of the recording station (represented by and ), the latitude ( ) and longitude ( ) of
the earthquake can be obtained by
in which Az is the azimuth of the earthquake related to the recording station and is the
epicenter distance.
Figure 1 shows the effect of the uncertainty of the azimuth and epicenter distance on the accuracy
of the location result. From the figure, it may be seen that the uncertainty of the location has a
strong dependence on the uncertainty of the azimuth and epicenter distance. The further the
epicenter distance, the stronger the dependence will be. From this point of view, to estimate and
reduce the uncertainties of the azimuth and epicenter distance are the key elements in enhancing
the quality of single-station location.
AZIMUTH ESTIMATION AND ITS UNCERTAINTY
Usually the azimuth is estimated by a polarization analysis of seismograms (Jurkevics, 1988;

Kanasewich, 1990). For three-component seismograms; xj(i), where j= 1, 2, 3 represents the
@seismicisolation
@seismicisolation
110 Chapter Five
vertical, north-south, and east-west component, respectively, and i = 1, 2, …ヤm indicates the
sampling point within time window , where is the time step, the covariance matrix
may be used for the polarization analysis. The eigenvalue ( ) and eigenvector (uj) can be
calculated by solving the equation
which requires the determinant to be zero, leading to
The eigenvalues Ȝ1Ȝ2Ȝ3 represent the polarization properties of the samples under
consideration.
Consider the simple case of P-wave propagation, for which the eigenvalues are Ȝ1=Ȝ2, Ȝ2=Ȝ3=0
Therefore, for the ideal P-polarization,
As an approximation, assuming that the noise is “isotropic” (i.e. ࢾࣅ૛૚ + ࢾࣅ૛૛ + ࢾࣅ૛૜ §ࢾࣅ૛૚ )
one has
It may be assumed that the value ‫ܫߜڿ‬ଵ ‫ۀ‬/‫ܫ‬ଵ is of the same order as the average error of the
seismogram (į[[!). On the other hand,1 the azimuth of P-wave propagation can be estimated
from the horizontal orientation of the rectilinear motion given by the eigenvector corresponding to
so the uncertainty of the azimuth can be estimated by
implying that the uncertainty of the estimated azimuth is of the same order as the error level of the
seismogram taken for analysis.
@seismicisolation
@seismicisolation
Fig. 1. Effect of different factors on the uncertainty of single-station location, (a) Effect of the uncertainty of azimuth
( ) on the uncertainty of the estimated location of an earthquake. In the figure, the equivalent ellipse radius is
represented byඥߜ‫׎‬ா ߜߠா . From bottom to top in each figure, the lines correspond to = 1, 2, 3, 4, and 5°. The
azimuth ( ) is taken as 30 and is taken to be zero, (b) Effect of the uncertainty of epicenter distance ( ) on the
uncertainty of the estimated location of an earthquake. In the figure, the equivalent ellipse radius is represented by
ඥߜ‫׎‬ா ߜߠா . From bottom to top in each figure, the lines correspond to =0.5, 1.0, 1.5, and 2.0°. The azimuth (Az)
is taken as 30°and is taken to be zero.
UNCERTAINTIES FROM SUBJECTIVE CAUSES: SELECTION OF WAVEFORMS

AND THE TIME-DIRECTION STACKING APPROACH
From the above discussion, it may be seen that the uncertainty of the azimuth is proportional to the
average noise level of the seismogram taken for analysis. However, in the analysis, still a larger
uncertainty comes from the fact that sometimes the selection of waveform is subjective, and the
results may depend on that selection; naturally leading to the question whether it is possible to
estimate the uncertainties caused by the selection of waveforms, and furthermore, whether it is
possible to consider longer wa vet rains and more signals to make the selection more robust. For
this purpose, we propose a time-direction stacking technique.
The methodology is straightforward and similar to the philosophy of the stacking technique used
in other fields. It is assumed that the signals contained in local and regional seismograms mainly
consist of two types: the P-type motion parallel to the direction of the source-station vector and the
@seismicisolation
@seismicisolation
112 Chapter Five
S-type motion perpendicular to the source-station vector no matter what phases they belong to.
In this case, there is a unique P-direction and many S-directions. What this requires is to take the
average direction of motion for all samples (time stacking) and add the motion vectors in the
P-direction and all of the S-directions altogether (direction stacking).
For a time series of three-component ground motion (x1(i), x2(i), and x3(i), i=1, 2, • • •,
N), applying a sliding time window ( ), the average square roots over the sliding window can be
calculated as
k=1, 2, 3,
and one can calculate the azimuth (Az) and apparent incident angle ( ) at time by
The azimuth and apparent incidence angle are measured from the horizontal orientation of the
rectilinear motion and its vertical for the eigenvector ( ) corresponding to the largest eigenvalue.
The more accurately we estimate the azimuth and apparent angles, the more closely we approach a
single point on the plot for the azimuth vs. apparent incidence angle. Next, the number of samples
with certain Az and , represented by R(Az, ), is counted over the given time span ( ). Then,
direction stacking,
is undertaken. Seen in a 2-dimensional view, this algorithm is equal to stacking all eight directions
(P, S1, S2, a, b, c, d and e) together, as shown in Fig. 2. However, as directions
contain mainly apparent noise, stacking is only effective for directions P, S1, and S2. In practical
analysis, whether the final azimuth takes the value of Az or /2—Az
/ is determined by whether the
motions are mainly S-type or P-type, which may
be determined by the polarization analysis mentioned above. In the estimation of azimuth,
the value , is normalized by
@seismicisolation
@seismicisolation
Fig. 2. Direction stacking. See text for details.
and the centroid location of the distribution
is taken as the final result. Accordingly, the uncertainty of the estimation can be obtained by
@seismicisolation
@seismicisolation
114 Chapter Five
which give the upper limit of uncertainty caused by the selection of waveforms. The calculation
gives four possible solutions. The quadrant is determined by polarization analysis and/or P-wave
first motions.
As a test of the algorithm, Fig. 3(a) gives the location of a few earthquakes and/or explosions as
well as the recording stations. The events in the example distribute within four quadrants. The
parameters of the stations and events are listed in Tables 1 and 2. Figure 3(b) gives the results of
the time-direction stacking analysis.
A detailed example is shown in Fig. 4, which gives the 95-07-01 Kobe earthquake recorded at the
Daejeon station. In the figure, the first few seconds of P-wavetrains are shown. From the
seismograms, it may be seen that it is really hard to isolate a “seismic phase” without any
arbitrariness. Below the seismogram, the azimuth measured by polarization analysis is shown. It
may be seen that the azimuth is not stable. As a result, the selection of the “phases” for the analysis
is subjective to some extent, and the azimuth estimation depends on the selection of waveforms. In
fact, the problem that it is not easy to isolate a “seismic phase” for analysis is quite common in the
processing of local and regional earthquake data. What done by the time-direction stacking
approach is that an evaluation of the uncertainties caused by the selection of waveforms is given
together with an average result of different phases. The methodology, advantage and limit of the
time-direction stacking technique are similar to those of the stacking techniques applied in
exploration seismology. As a comparison to an ordinary result, in Fig. 4, the time-direction
stacking result is also shown. Above the seismogram, is the function within the
plane, in which the cross represents the centroid.
@seismicisolation
@seismicisolation
Fig. 3. Time-direction stacking results, (a) Stations and earthquakes in the example. Triangles denote stations and
points denote earthquakes. Earthquakes are numbered according to their azimuths. A line between station and
earthquake indicates that the seismogram of the earthquake recorded at the station is chosen for the analysis. The
parameters of the stations and earthquakes are given in Tables 1 and 2. The Kobe earthquake is out of the range of the
figure, yet its azimuth relative to the recording station is shown. (b) Azimuths obtained by stacking (vertical axis)
versus the “actual” azimuth calculated from network locations (horizontal axis). Circles give the uncertainty of the
stacking estimation.
Table 1. Parameters of stations in the example.
The stacking result gives that the azimuth is 99 ⦮ 6°, near the actual azimuth (103°) calculated
by network location. This implies that the arbitrariness of the selection of waveforms may cause
uncertainties up to 6°, being much larger than that estimated by the mean errors of the
seismograms taken for analysis. On the other hand, with the stacking process, the deviation of
measured azimuth caused by different phases may be smoothed out to some extent. As a result, the
stacking average is near the “actual” value within the uncertainty range.
@seismicisolation
@seismicisolation
116 Chapter Five
Table 2. Parameters of earthquakes and explosions in the example.
Fig. 4. The Kobe earthquake recorded at the Daejeon station, demonstrating the details of time-direction stacking. The
middle shows the P-wavetrains. From the seismograms, it may be seen that it is hard to isolate a “seismic phase”
without any arbitrariness. Below the seismogram, the azimuth measured by polarization analysis is shown. It may be
seen that the azimuth is not stable. Above the seismogram is the function within the plane, in which
the cross represents the centroid.
UNCERTAINTY OF THE EPICENTER DISTANCE
The epicenter distance ( ) is estimated from travel time readings. For the simplest case of S-P,
such estimation may be expressed by
@seismicisolation
@seismicisolation
where Va, being referred to as the apparent velocity, is a quantity with the dimension of velocity
and depends on the travel time difference for different distances.
The uncertainty of Va, and the uncertainty of the location caused by it, depends on the travel times.
In Fig. 5, based on the data from the Korean Meteorological Administration (KMA), an S-P vs.
epicenter distance diagram is shown. The result is mainly for crustal earthquakes with depths less
than 30 km. From the figure, it may be seen that, in the single-station location, the hypocenter
depth is hard to determine at first. As a result, the ignorance of the depth of the earthquake will
cause uncertainty in the distance estimation much larger than the uncertainty caused by the
uncertainties of the travel time readings. To minimize the uncertainties caused by the ignorance of
hypocenter depth, we proposed an apparent velocity model which depends on the range of
distances. According to the travel times around Korea, the distances are divided into four ranges:
0-5, 5-10, 10-20, and 20-30°. In the figure, for each distance range, three lines are used to
represent the model. The center line represents the function which is a straight line
passing the origin, and the other two lines give the upper and lower limits of the data. The
uncertainties are shown in the figure at the bottom. The figure gives =2° for =0-5°, = l°
for = 5-10°, =2° for = 10-20°, and = 3° for =20-30°. According to Fig. 1, this
implies that ඥߜ‫׎‬ா ߜߠா . ൎ 1.5° for =0-5°,ඥߜ‫׎‬ா ߜߠா ൎ 0.8° for =5-10°,and ඥߜ‫׎‬ா ߜߠா ൎ
1.5-2.0° for = 10-20°. For epicenter distances above 20°, 2.5-3.0°. On the other hand,
however, as long as the depth is known, for instance, for the case of explosions, the uncertainty of
can be considerably reduced.
Fig. 5. S-P travel time difference for the Korean Peninsula. The data comes from the Korean Meteorological
Administration (KMA). The distance has been divided into four ranges. See text for details.
@seismicisolation
@seismicisolation
118 Chapter Five
The uncertainty of the location caused by the uncertainty of travel time readings is easy to estimate
by Suppose and that the uncertainty of the travel time reading is
Such an accuracy is easy to reach using digital three-component seismic recordings. In this case,
we have as the order of 0.1°. The uncertainty caused by this is negligible compared to the
uncertainty caused by Va as discussed above.
Considering the effects of azimuth and epicenter distance, the results suggest that for the regions
of the Korean Peninsula and its vicinity, with the analyzing processes based on azimuth and
epicenter distance estimations, for explosions and earthquakes whose depths can be
obtained as a priori constraints can be located by a single station with an accuracy of
ඥߜ‫׎‬ா ߜߠா . ൎ 0.2-0.4°. For other earthquakes within this distance rangeඥߜ‫׎‬ா ߜߠா . ൎ 1.5°; for
=5-10°,earthquakes can be located by a single station with an accuracy of ඥߜ‫׎‬ா ߜߠா . ൎ 0.5-0.8°;
and for larger distances, we estimate ඥߜ‫׎‬ா ߜߠா . ൎ 1-2° for = 10-20° andඥߜ‫׎‬ா ߜߠா . ൎ 3.0-3.5°
for =20-30°.
CONCLUSIONS AND DISCUSSION
In this paper, we conducted a systematic analysis of the uncertainty of single-station location of

earthquakes at local and regional distances. We pointed out that the uncertainty of the azimuth
estimation is proportional to the average noise level of the seismograms taken for analysis.
However, in the analysis, still a larger uncertainty comes from the fact that sometimes the selection
of waveforms is subjective, and the results may depend on that selection. To estimate such kind of
uncertainty and to avoid the uncertainties caused by the arbitrariness of the selection of waveforms,
we proposed a stacking approach. On the other hand, the uncertainty of the estimation of epicenter
distances comes from the travel time readings and apparent velocity model. To reduce the
uncertainties of distance estimation, we proposed an apparent velocity model which depends on
the range of distances. Finally, for the Korean Peninsula and its vicinity, based on observational
data, we estimated the order of expected uncertainties of single-station location on the basis of
azimuth and epicenter distance estimations for different distance ranges: 1) for explosions
and earthquakes whose depths can be obtained as a priori constraints can be located by a single
station with an accuracy of 0.2-0.4°; 2) for =5-10o, earthquakes can be located by a
single station with an accuracy of 0.5-0.8o; 3) for = 10-20°, 1-2°; and 4)
for =20-30°. 3.0-3.5°. From the results, it seems that the time-direction stacking
approach and the distance-range-dependent apparent velocity model proposed in this paper might
be of help in evaluating and reducing the uncertainties in the single-station location of local and
regional earthquakes.
ACKNOWLEDGMENTS
This study was supported by the STEPI of Korea and the State Seismological Bureau of China.
The research was also financed in part by the Ministry of Education of Korea under contract
number BSRI-95-5420. One of the authors (Z.L.W.) thanks the Seismological Institute, Hanyang
University, Korea, for its support and help during his visit. Acknowledgments are also due to Mr.
S. K. Lee and Mr. F. C. Gao for helpful discussions.
@seismicisolation
@seismicisolation
REFERENCES
Jurkevics, A., Polarization analysis of three-component array data, Bull. Seismol. Soc. Am., 78,
1725-1743, 1988.
Kanasewich, E., Seismic Noise Attenuation, Pergamon Press, Toronto, pp. 161-192, 1990.
Kim, S. G. and S. K. Lee, Determination of the hypocentral parameters outside the seismic array
using a single station of three-component, J. Eng. GeoL, 5, 59-74, 1995.
Magotra, N., N. Ahmed, and E. Chael, Seismic event detection and source location using single
station (three-component) data, Bull Seismol. Soc. Am., 77, 958-971, 1987.
Roberts, R. G., A. Christoffersson, and F. Cassidy, Real-time event detection, phase identification
and source location estimation using single station three-component seismic data, Geophys. J.,
97, 471-480, 1989.
Roberts, R. G. and A. Christoffersson, Decomposition of complex single-station three-component
seismograms, Geophys: J. Int., 103, 55-74, 1990.
@seismicisolation
@seismicisolation
CHAPTER SIX
STUDY ON SOME CHARACTERISTICS OF EARTHQUAKES AND EXPLOSIONS
USING THE POLARIZATION METHOD 1
SO GU KIM AND FUCHUN GAO*
DPRK’s an artificial explosion on 01/29/1988 (mb=2.8, depth=0 km) recorded through the
pure-continental path on the 3-component of KSRS (Wonju) array, Republic of Korea.
ABSTRACT
In this paper, a case study has been made on the characteristics of earthquakes and explosions,
mainly through the polarization method, estimating the epicentral azimuths and incident angles in
the case of shallow- and deep-focus earthquakes as well as explosions based on three-component
signal-to-noise ratios. We have found that it is possible to determine the epicenter and incident
angle of an earthquake or an explosion by three-component single-station polarization analysis,
especially using body waves like Pn, and in some cases, stable and accurate results are obtained.
We made a theoretical discussion on the polarization properties of shear waves, and found that
multievents which take place occasionally within explosions can be discerned by azimuth and
motion pattern monitoring.
1
Originally published in the J. Phys. Earth (1997).@seismicisolation
45, 13-27.* Present Address: Rice University, Houston, USA
@seismicisolation
Study on Some Characteristics of Earthquakes and Explosions Using the Polarization Method 121
INTRODUCTION
Seismic waves are generated by sources that radiate energy with well defined polarization
characteristics, so polarization techniques which can be applied to three-component analysis of a
single sensor by solving the eigenproblem for a covariance matrix, originally proposed by Flinn
(1965), have been developed in the past.
Various attributes featuring particle motions are extracted from the motion ellipsoid. In this paper,
the polarization technique involves filtering the signals into a series of narrow frequency bands,
applying short sliding time windows and then computing the polarization ellipse from the
covariance matrix in each window of the band in which the highest three-component
signal-to-noise ratio (hereafter referred to as 3-C SNR) is gained. Azimuths, incident angles and
rectilinearities as a function of time are outputs (Kim and Lee, 1994).
The polarization method has been used in sediment structure survey (Kriiger,1994) and the on-line
analysis system (Kedrov and Ovtchinnikov, 1990) by being applied to many earthquake events,
mainly through P-wave polarization analyses. It is widely recognized that azimuths and incident
angles can be estimated reliably by P-wave polarization analyses. S and Lg waves are seldom used
for their unreasonable results if the same method is used. Usually, these shear waves are subject to
signal-generated noise, besides background noise. Signal-generated noise is the result of multiple
reflections and refractions of P and S body waves at crustal interfaces and in-homogeneities and
local conversion of body waves to surface waves. Therefore, they are not polarized as well as P
waves. We here only make a theoretical discussion on waves. Here, we determined the azimuths of
events using P waves in a traditional polarization method which has been widely applied
(Jurkevics, 1988; Roberts et al.,1989; Kanasewich, 1990; Roberts and Christoffersson, 1990).
The analyses are applied to several series of regional or local events (earthquakes and explosions)
recorded in the Korean Peninsula as well as its adjacent areas, with special attention paid to
different polarization properties between earthquakes and explosions. There are seismic zones of
Japanese Islands toward east Korea and the Tanlu seismic zone toward the west region.
Deep-focus events take place in regions of the East Sea (Japan Sea) occasionally due to subduction
of the Pacific plate into the Eurasian continent. Our objective here is to determine the azimuths and
incident angles of events, including earthquakes as well as explosions. Based on the motion
patterns, we found some different polarization characteristics between explosions and earthquakes.
To do this, we used programmes to process three-component seismograms interactively on a PC
computer. We attained azimuths and incident angles by applying the analyses to seismic phases in
earthquakes and explosions. Polarization properties n different cases are compared and discussed.
METHOD AND ANALYSIS
The polarization technique is a time domain algorithm. The polarization ellipse (or ellipsoid) is
computed with sliding time windows by solving the eigenproblem for the covariance matrix. If it
is assumed that there are N points (x 1i (i = 1, • • •, N)) taken from one of the three components NS
(north-south), EW (east-west), and UD (up-down) of a digitized seismic record with a sampling
rate of f, the sliding time window length is N- =N/f. The mean value over the whole time
window is
@seismicisolation
@seismicisolation
122 Chapter Six
(1)
The covariance between N observations of two components (X x and X2 ) is given by
(2)
where Av1 and Av2 are computed as in Eq. (1) and it is evident that
(3)
The quantity Cov(X1, X1) is defined as the autocovariance or the variance of X1. The matrix with
Cov(Xr, Xs ) in its r-th and s-th column (r, s = 1, 2, • • •, n) is the covariance matrix for the set of n
variables Xj, j =1, 2, • • •, n. If X is the vector of the variables and U the vector of the means for
each of these variables, the covariance matrix (V) is defined by a covariance matrix
V=E[(X-U) (X-U)t], (4)
where t denotes the transpose of the corresponding vector. For our case of three variables (UD, NS,
EW) over the time window , the covariance matrix is
(5)
By solving the eigenvalue problem of the matrix in Eq. (5), that is
(6)
we can obtain three eigenvalues ( ) and corresponding eigenvectors (E 1 E2 , E3 ).Thus,

three polarization principle axes (L 1, L2, L3),which have the same direction of E1 E2, and E3
accordingly,can be obtained (Jurkcvics, 1988).
If we assume L1 is the largest axis, L2 the next large axis,and L3 the smallest, the rectilinearity,
which indicates the quality of the data, is defined as
(7)
The P-wave azimuth is simply computed as the horizontal orientation of the eigenvector
corresponding to the largest eigenvalue of the covariance matrix, and incident angle (IC1) is the
angle between this eigenvector and vertical direction.
In the case of SV and Lg waves, where the movements are mainly in the vertical plane, we also
apply the above method to them to get three eigenvectors, and take the horizontal orientation of the
largest eigenvector as the azimuth. However, as we know, there is 180° ambiguity in the
orientation for planar waves, we use the first arrival as a reference to eliminate this ambiguity. The
@seismicisolation
@seismicisolation
incident angle is computed in the same way as the P waves.
As for SH waves, where movement is mainly in the horizontal plane and the motion pattern is
perpendicular to propagation, we calculate the incident azimuth as
Az SH = tan -1(EW/NS) + 90° . (8)
Again, there is 180° ambiguity of the azimuth and the reference is also used here. The incident
angle is computed as
(9)
where | UD | is the absolute value of the vertical component. It is understandable that IncSH is
larger than the P incident angle due to its smaller vertical component. The AzSH and IncSH are
averaged in a small time window, and RecSH is also from a polarization analysis.
Theoretically, both compressional (P type) and shear (SH or SV type) waves exhibit a high degree
of linear polarization. The particle motion of S waves is perpendicular to the direction of
propagation. But because of the effects of anelasticity and anisotropy, the particle motions of P and
S body waves are never exactly linear, and resemble highly eccentric elliptical shapes or even
ellipsoids.
If we consider an ellipsoid with three principle axes, L1 (largest), L2, and L3 (smallest) correspond
to the particle motion of one P wave. IC1, IC2, and IC3 are the incident angles corresponding to L1,
L2, and L3. The ellipse with axes L2 and L3 has a line of intersection with the horizontal plane. If
is the angle between the line of intersection and the axis L2 , we can estimate the error of
azimuth AZ as
(10)
where is the function of the spatial positions of L2 and L3; that is, = ( AZ2 , IC2, AZ3,
IC3). From Eq. (10), we can see that the smaller IC1 is, the larger AZ is for P waves, as will be
shown in later examples.
From Eq. (10), we can estimate the range of azimuth residuals ( AZ ) and find the relationship
between the rectilinearity and AZ . Because
(11)
and
(12)
@seismicisolation
@seismicisolation
124 Chapter Six
From we know
(13)
For an earthquake, generally IC1 scarcely has a very small value except for deep-focus earthquakes,
etc. Therefore, it is reasonable to estimate AZ just as
(14)
Here thresholds of rectilinearity should be applied to reduce the scatter in the azimuth estimates. If
we set the scatter of 1.0°, then rectilinearity should be larger than 0.98 which is rarely
reached practically. Later, we limit our polarization study mainly to rectilinearities larger than 0.80.
In the case of low rectilinearity, narrow band filters will be applied to enhance the 3-C SNR, and
then the rectilinearity. In fact in our practice, almost all data are prefiltered in a set of frequency
bands selected on the basis of their 3-C SNR.
In this study, first we filtered the data of three components in several narrow-frequency bands, and
the signal-to-noise ratio was computed in each band respectively. The resulting seismogram with
the highest 3-C SNR was chosen for polarization analysis. Second, we picked up a seismic phase,
such as Pn, etc., in one of the components and set the beginning of the time interval to be
processed. Then, the sliding time window length was defined, and lastly, we set the end of the time
interval. Generally, we set the sliding time window length a fraction of or similar to the phase
period for P-wave polarization. Results showed that the proper selection of time interval and
window length that include purely polarized motion of the phase are important factors affecting
the quality of results.
@seismicisolation
@seismicisolation
Table 1. Polarization results for earthquakes.
*1 NEIC (National Earthquake Information Center), USA, determined. *2 KMA (Korea Meteorology
Administration), Korea, determined. *3 Numbers of events correspond to those in Fig. 4.
Based on the method described above, we applied the polarization method to two deep-focus
earthquakes, the Vladivostok earthquake (7/22/1994) and East Sea earthquake (7/22/1994), and
four shallow ones, such as the Kobe earthquake (1/17/1995), Paekryong-do earthquake
(8/11/1995), etc. Their characteristics and polarization results are shown in Table 1.
All six earthquake records have high 3-C SNR (all 10.0) in their raw or filtered data. The Kobe
earthquake was recorded at Daejeon station (127.4°E, 36.3°N) with a sampling rate of 50 Hz. The
Paekryong-do earthquake, Puyo earthquake and Samchok earthquake were recorded at Inchon
station (126.6°E, 37.5°N) with sampling rates of 80 Hz for the former two and 20 Hz for the latter
one. The two deep-focus earthquakes were recorded at KSRS station (127.93°E, 37.48°N) by a
Teledyne ST-2 broadband digitized seismometer with both sampling rates of 20 Hz. The
instrumental response of the seismometer is flat in a broad band.
@seismicisolation
@seismicisolation
126 Chapter Six
Fig. 1. Polarization analysis using the Pn phase of (a) the Kobe earthquake (1/17/1995) and (b) the
Paekryong-do earthquake (8/11/1995). The three-component waveforms in (a) are butterworth filtered in a
band of 1-3 Hz at order 4. A small box indicates the time interval in which the polarization analysis, shown
below, is applied. Motion patterns are shown from up to down in NE, UN, and UE planes in three windows
on the right side. In the top two of the three windows listed on the left side, three values from up to down
correspond to three axes L1 (largest), L2, and L3 (smallest). The first value in the bottom window is
rectilinearity (see text for definition), the other two show the ratios of L2/L1 and L3/L1, respectively. There
are three lines on the left side. One of them intersects with the seismograms as well as the three windows,
and has 3 points of intersection with 3 curves in each window. The 3 points in each window correspond to
the three values listed in that window. All these values (9 in total) define polarization properties of seismic
waves in a small time window, which is defined by the other two lines (so called moving time window), (b)
is plotted in the same way as (a) except the seismogram is bandpass filtered in a band of 1-2.5 Hz at order
4.
In Table 1, the longitude and latitude of epicenters are determined by using the azimuth obtained
from polarization analyses of first arrival P phase and the S-P method, if possible, as compared
with epicenter reports from the NEIC or KMA. From Table 1, we can see that source azimuths can
be determined reliably using the first arrival P phases in the case of deep-focus earthquakes as well
as in the case of shallow-focus earthquakes whose Pn phase polarization results are plotted in Figs.
1 and 2.
@seismicisolation
@seismicisolation
Fig. 2. Polarization analysis using the Pn phase of (a) the Puyo earthquake (9/16/1995) and (b) the Samchok
earthquake (10/6/1995). The three-component waveforms in (a) are butterworth filtered in a band of 1-2.5
Hz at order 4 while those in (b) are in a band of 1-3 Hz at order 4. They are plotted in the same way as in
Fig. 1.
As for the two deep-focus earthquakes (see Fig. 3), we identified converted sP and pS phases right
after the first P arrival, which made the waveforms complicated and contaminated. Therefore, the
key point to estimating azimuth reliably is to select the proper waveforms. For the Vladivostok
earthquake, we skipped the first two cycles after the first arrival where small incident angles were
encountered (around 28°). From Eq. (10), it is recognized that small incident angles result in large
azimuth errors, so we selected a time window of 1.15 s later but well before the next arrival, which
was recognized as converted waves. The resulting azimuth is very close to the real one (see Table
1). That for the East Sea earthquake is also very reliable.
Reflections and refractions take place frequently when the seismic rays from deep sources
penetrate through layers with different physical medium features, and converted as well as
reflected waves reach the station in a mixed way. In polarization analyses, it seems that the more
complicate the waveforms are, the more unstable and unreliable the results are. It has been
recognized that azimuth residuals are due to the characteristics of the medium both at the epicenter
and at the station, and consequently depend on both the azimuth and epicentral distance (Kedrov
and Ovtchinnikov, 1990).
@seismicisolation
@seismicisolation
128 Chapter Six
(a) (b)
Fig. 3. The same as in Fig. 1, but for (a) the Vladivostok earthquake (7/22/1994) using the first arrival P
phase and (b) the East Sea earthquake (7/22/1994) using the P phase. Both seismograms in (a) and (b) are
1-3 Hz bandpass filtered at order 4.
From Table 1,it can be seen that Pn azimuths for the four earthquakes can be regarded as accurate.
From the epicenter distribution of the events in this study (see Fig. 4), it can be seen that they
almost cover an azimuth range of 360°. From this study, it seems that azimuth residuals are not
azimuth dependent though our data is not sufficient to prove it.
Nowadays, more and more attention is being paid to explosions due to seismological as well as
public interest. We here applied the polarization method to two series of explosions. The first
series of explosions (including 3 events) took place in January 1988, recorded by a short-period
digitized seismometer with all sampling rates of 20 Hz. The seismometer had a response curve flat
from 1 to 10 Hz. These 3 explosions are listed in Table 2, together with their polarization results.
Their magnitudes range from M b 1.8 to 3.0.
True azimuths, which are from local reports, are also listed in Table 2, and can be compared with
the azimuths determined by polarization analyses. From Table 2, again we can see that azimuths
from P are reliable. The P azimuths for the LKSRS09 and LKSRS21 explosions can be regarded as
accurate for their small residuals.
@seismicisolation
@seismicisolation
Fig. 4. Epicenter distribution map of events listed in Tables 1-3. The number of the event corresponds that
.
in the tables. All epicenters are plotted using parameters determined by the S-P method in this study.
Arrows 1, 2, and 3 point to the epicenters.
Polarization analysis gives us a clue to discriminating multievents with close onset times through a
combination of azimuth and motion pattern monitoring. We have applied the polarization method
to two explosions, each of which includes two or three events. Their results are shown in Table 3.
For the LKSRS 11 explosion, there were three events with near onset times (about 3-s intervals).
We found that the three independent phases were all of the P-wave type corresponding to the three
events (see motion patterns in Fig. 5). Azimuths obtained from polarization analyses of these
phases were close to the true azimuths and residuals were very small (less than 3°, see Table 3).
The S waves of the subevents were well developed (including SV and SH components). So, from
the study, it is recognized that we can estimate azimuths and incident angles of explosions reliably
using P waves. Similar phenomenon could be seen in the LKSRS 23 explosion, which was
composed of two events (for details see Table 3).
@seismicisolation
@seismicisolation
130 Chapter Six
Table 2. Polarization results for the first series of explosions.
*Number of events correspond to those of Fig. 4
(a) (b) (c)
Fig. 5. The same as in Fig. 1, but for (a) event 1 in the LKSRS 11 explosion (1/25/1988) using the P phase;
(b) event 2 in this explosion using the P phase; and (c) event 3 in this explosion using the P phase. All
seismograms are 1-3 Hz bandpass filtered at order 4.
Polarization analysis can find different azimuths, and if the corresponding motion patterns are both
of typical P-wave type, the most convincing conclusion is that the two signals belong to different
events. This is specially useful in explosion discrimination with close onset times.
From above study, it is recognized that the polarization technique provides us information not only
on the source parameters of seismic events such as azimuth and incident angle, but also on the
@seismicisolation
@seismicisolation
polarization characteristics of different wave types. It can be used to distinguish some ambiguous
signals like decoupled-explosions and/or multievents as well as to locate event sources.
Table 3. Polarization results for multievents.
* Numbers of events correspond to those in Fig. 4.
Fig. 6. Distribution map of explosion sites around Bongdam station (126.9°E,37.2°N).
The second series of explosions is composed of 13 events totally (see Table 4), 10 of which (BT
1-10 in Table 4) were recorded by a MARK RAND three-component digitized seismometer and
the rest (BT 11-13 in Table 4) recorded by a S6000/DR2000 digitized seismometer with all
sampling rates of 100 Hz at Bongdam station (126.9°E, 37.2°N), which has a “quiet” circumstance.
Most events recorded there are from construction or mining blasts nearby. The 13 events, together
with their polarization analysis results, are listed in Table 4.
From Table 4, we can see blasts 2, 3, and 4 were from the same explosion site by the close
azimuths (255° or so, see Table 4) and epicentral distances (S-P §1.2 s). The site corresponds to an
industrial* mining point about 10 km away from the station (Changgok site in Fig. 6). For these
three events, S waves had larger UD and EW components but smaller NS component, which
@seismicisolation
@seismicisolation
132 Chapter Six
shows that the S-wave movements were mainly in the UE plane (see Fig. 7). So the S phases in
these explosions were of SY type, which may be converted from P phases due to a discontinuity
plane under the explosion site or generated by severe heterogeneous medium around the source.
The other explosions took place at several other mining or industrial construction sites.
Fig. 7. The same as in Fig. 1,but for (a) BT 2 in the second series of explosions using the P phase; (B) BT 3 in the
second series of explosions using the P phase; and (c) BT 4 in the second series of explosions using the P phase. The
three-component waveforms in (a) and (b) are butterworth filtered in a band of 1-4 Hz at order 4 while those in (c) are
in a band of 1-5 Hz at order 4.
From the above analyses, we made a distribution map of explosion sites, which is plotted in Fig. 6.
BT 1 and BT 6 may correspond to two accidental blasts because no explosion sites are reported to
the northeast of the station. So their epicenters aren’t plotted in Fig. 6. The sites plotted in Fig. 6
coincide well with those from field investigation. Explosions take place frequently in these sites
due to the industrial activities there. From the epicenter results listed in Table 4, we can see that
BT 5 and BT 11 are from the Kangwon explosion site; BT 8 and BT 9 are from the Samchang site;
and BT 10 and BT 13 are from the Dongchul site (see Fig. 6). These sites are very near the station
(< 20 km), which is the reason why we have to preserve two digits after the decimal point in
longitude and latitude results.
From the above study, it can be seen that, in most cases, azimuths for explosions can be computed
quite reliably using P phases and, in some cases, accurate results are obtained. Compared with
those for earthquakes, the P-wave azimuths for explosions have smaller systematic deviation
because seismic waves of explosions may propagate in more “simple” media than those for
@seismicisolation
@seismicisolation
earthquakes discussed in this paper.
Table 4. Polarization results for the second series of explosions (blasts).
In this paper, a technique for computing particle-motion information using three-component

seismograms from a single sensor has been outlined. We have processed data from different kinds
of regional earthquakes, such as deep-focus and shallow-focus earthquakes, and two series of
explosions (27 events totally, including 2 multievents), all recorded in the Korean Peninsula with
different characteristics. The following conclusions can be drawn from this study:
1) The azimuth and incident angle of earthquakes can be estimated quite reliably through body
waves such as Pn. For local explosions, more accurate results are obtained using P waves in most
cases.
@seismicisolation
@seismicisolation
134 Chapter Six
2) As for earthquakes, azimuths and incident angles can be computed reliably with a
three-component single station in the case of shallow-focus events, such as the Kobe earthquake,
etc. The azimuths and incident angles of deep-focus earthquakes can also be estimated reliably
despite complicated tectonic features around the sources and along their propagation paths as long
as the proper phases are utilized.
3) In the case of noise-contaminated signals, we applied a butterworth bandpass filter with a

central frequency similar to that of the signals in order to obtain stable results.
4) Multievents with close onset times can be distinguished effectively through azimuth
determination and motion pattern monitoring.
ACKNOWLEDGMENTS
This research was funded by the Ministry of Education (BSRI-95-5420), Korea and Hanyang
University. We are grateful to Michael Baumbach for personal communication during the study.
We appreciate Z. L, Wu and S. K. Lee for their comments and criticism at various stages of this
investigation.
REFERENCES
Flinn, W. A., Signal analysis using rectilinearity and direction of particle motion, Proc. IEEE,53,
1874-1876, 1965.
Jurkevics, A., Polarization analysis of three-component array data, Bull. Seismol. Soc. Am., 78,
1725-1743, 1988.
Kanasewich, E., Seismic Noise Attenuation, Pergamon Press, New York, 1990.
Kedrov, K. and V. M. Ovtchinnikov, An on-line analysis system for three-component seismic data:
method and preliminary results, Bull. Seismol. Soc. Am., 80, 2053-2071, 1990.
Kim, S. G. and S. K. Lee, Determination of the hypocentral parameter outside the seismic array
using a single station of three-component, J. Eng, GeoL, 5(1), 59-74, 1994.
Krüger, F., Sediment structure at GRF from polarization analysis of P waves of nuclear explosions,
Bull. Seismol Soc. Am., 84, 149-170, 1994.
Roberts, R. G. and A. Christoffersson, Decomposition of complex single-station three component
seismograms, Geophys. J. Int., 103, 55-74, 1990.
Roberts, R. G., A. Christoffersson, and F. Cassidy, Real-time event detection, phase identification
and source location estimation using single station three-component seismic data, Geophys. J.,
97, 471-480, 1989.
@seismicisolation
@seismicisolation
CHAPTER SEVEN
SPECTRAL DISCRIMINATION ANALYSIS OF EURASIAN NUCLEAR TESTS

AND EARTHQUAKES RECORDED BY THE ISRAEL SEISMIC NETWORK
AND THE NORESS ARRAY 1
YEFIM GITTERMAN, VLADIMIR PINSKY AND AVI SHAPIRA
Fragment of a granite stone from the epicenter of the first underground nuclear explosion in the
Degelen mountain on November 10, 1961 (the Museum of Semipalatinsk Test Site history in the
city of Kurchatov, by Y. Gitterman).
ABSTRACT
The energy spectral ratio and the innovative spectral semblance discriminants, successfully
performed previously on local Israeli events, were verified on teleseismic short-period recordings.
The events tested include 29 nuclear explosions and 41 earthquakes (mb =5.2–6.5), mainly from
China and Kazakhstan, recorded by the Israel Seismic Network (ISN) and the NORESS array. A
15-s window comprising P- and P-coda waves was selected for the analysis. The ‘semblance’
statistic commonly used in seismic prospecting for phase correlation in the time domain was
modified and utilized as a measure of coherency of the smoothed spectra across the network/array.
The semblance and the average spectral ratio of low-to-high frequency energy were evaluated,
1
Physics of the Earth and Planetary Interiors (1999), 113, 111–129
@seismicisolation
@seismicisolation
136 Chapter Seven
using a subset of 7–10 stations for a given event. Semblance and spectral ratio values, calculated
from ISN seismograms, were found to be higher for earthquakes, where the analyzed waves are
richer in low-frequency energy and have more coherent spectral shapes than explosions. These
observations are contrary to those observed for local events. The best performance is provided in
the frequency bands (0.6–1 Hz.)/(1–3 Hz) for the ratio and (0.6–2 Hz) _for the semblance. Joint
application of the two discriminants showed almost full separation (95%) between the two
populations. Some explosions exhibited pronounced minima (nulls) near 1 Hz which could be
interpreted in terms of interference of P- and pP-waves from a source at the depth of several
hundreds of meters. Nevertheless, this feature could not be utilized as a discriminant: many
explosions revealed strong variability of this minimum across the ISN network and some
earthquakes also distinctly exhibited this feature. The ISN and NORESS discrimination
performances were compared. The latter records showed the same (as the ISN) relation between
spectral ratio values for earthquakes and explosions, whereas the character of semblance was
reversed. The ratios in the frequency bands (0.6–0.8 Hz)/(0.6–3 Hz) yield full separation of the
two populations. In general, the spectral ratio performance appears to be better for NORESS
recordings, whereas the spectral semblance is better for the ISN data. © 1999 Elsevier Science
B.V. All rights reserved.
Keywords: Teleseismic discrimination; Semblance statistics; Spectra coherency; Low-to-high
frequency energy ratio
INTRODUCTION
Seismological means for detecting and identifying nuclear tests are of critical importance for
CTBT monitoring, but still need further improvement. The recent trend in classification studies
focuses mainly on relatively weak events at local and regional distances. Physical factors
affecting regional discrimination of man-made and natural seismic events have been investigated
by many researchers and presented in a detailed review by Blandford (1995). The dominant ones
are generation of remarkable low-frequency surface waves by quarry blasts, associated with sub-
surface sediments; differences in spectral content and spatial coherency of spectral shapes. These
physical phenomena observed at a regional dense seismic network provide the basis for the
development of efficient classification algorithms, optimally accounting for the multi-station
registration system. Re-cently, appropriate spectral discriminants (energy spectral ratio and
spectral semblance. have been developed in the framework of the CTBT program and successfully
applied to local quarry blasts and earthquakes recorded by the Israel Seismic Network (ISN)
(Shapira et al., 1996; Gitterman et al., 1998).
Occasionally, regional classification is not feasible either, because close stations are either too
sparse for reliable discrimination or data are simply unavailable. In such cases, teleseismic
discrimination is the only alternative. Moreover, seismic waves recorded at teleseismic distances
travel through more homogeneous media and exhibit less distorted source information than the
regional ones.
The classic teleseismic discriminant is the mb:MS ratio (e.g., see Richards and Zavales, 1995) using
long-period observations of surface waves (MS) from strong events. For smaller events, MS is not
observed, hence the interest is using spectral discriminants based on short-period (SP) recordings
only. The shallow focal depth of underground nuclear explosions and source spherical symmetry
lead to strong destructive interference between the down-going P-wave energy and the pP-wave
reflected from the Earth’s surface. The phenomenon results in pro-nounced minima (nulls) in the
spectra of the initial part of the teleseismic records of the Nevada nuclear explosions (Kulhanek,
1971), which, if generally observable, can be used as an identification feature.
@seismicisolation
@seismicisolation
Spectral Discrimination Analysis of Eurasian Nuclear Tests and Earthquakes Recorded by the Israel Seismic 137
Network and the NORESS Array
Table 1: Underground nuclear tests used in the study
No. Location Date hour/minute/ Latitude Longitude Distance Azimuth mb
Second (deg.} (deg.) (deg.) (deg.)
a NT1 Kazakhstan 850615 005700.7 49.889 78.881 37 47 6.0

a NT2 Kazakhstan 850720 005314.5 49.951 78.829 37 47 5.9
a NT3 Kazakhstan 870312 015717.3 49.932 78.785 36 47 5.4
a NT4 Kazakhstan 870417 010304.8 49.886 78.691 36 47 6.0
a NT5 Kazakhstan 870506 040205.5 49.777 78.089 36 47 5.5
a NT6 China 870605 045958.3 41.584 88.737 43 61 6.2
NT7 Kazakhstan 870606 023706.9 49.865 78.143 36 47 5.4
a NT8 Kazakhstan 870802 005806.8 49.880 78.917 37 47 5.9
a NT9 Kazakhstan 871115 033106.7 49.871 78.791 36 47 6.0
NT10 Kazakhstan 871213 032104.8 49.989 78.844 37 47 6.1
NT11 Kazakhstan 871227 030504.7 49.864 78.758 36 47 6.1
NT12 Kazakhstan 880213 030505.9 49.954 78.910 37 47 6.1
NT13 Kazakhstan 880403 013305.8 49.917 78.945 37 47 6.1
NT14 Kazakhstan 880504 005706.8 49.928 78.769 36 47 6.1
NT15 W. Siberia 880822 161958.2 66.316 78.548 43 23 5.3
NT16 Kazakhstan 880914 035957.4 49.833 78.808 36 47 6.1
NT17 Kazakhstan 881112 033003.7 50.078 78.988 37 47 5.3
NT18 Novaya 881204 051953.0 73.387 54.998 42 8 5.9
Zemlya
NT19 Kazakhstan 881217 041806.9 49.886 78.926 37 47 5.9
NT20 Kazakhstan 890122 035706.5 49.915 78.857 37 47 6.0
NT21 Kazakhstan 890212 041506.7 49.895 78.758 36 47 5.9
NT22 Novaya 901024 145758.1 73.361 54.707 42 8 5.7
Zemlya
NT23 China 920521 045957.5 41.604 88.813 43.6 61 6.5
NT24 China 931005 015956.5 41.647 88.681 43 61 5.9
NT25 China 940610 062557.8 41.527 88.710 43.5 61 5.8
NT26 China 941007 032558.1 41.662 88.753 43.5 61 6.0
NT27 China 950515 040558.0 41.665 88.821 43.6 61 6.1
NT28 China 950817 010004.8 41.600 88.700 43.5 61 6.0
a NT29 China 960608 025558.0 41.600 88.600 43.4 61 6.0
a Only ISN data are presented.
b Only NORESS data are presented.
@seismicisolation
@seismicisolation
138 Chapter Seven
Table 2: Selected Eurasian earthquakes
No. Location Date hour/minute/ Latitude Longitude Depth Distance Azimuth mb

second (deg.) (deg.) (km) (deg.) (deg.)
a QT1 Uzbekistan 840319 202838.2 40.320 63.350 15 24 61 6.5
a QT2 Tajikistan 841026 202221.8 39.155 71.328 33 30 65 6.0
QT3 China 850823 124156.1 39.431 75.224 7 33 65 6.4
a QT4 China 850911 204549.5 39.356 75.407 15 33 65 5.8
QT5 Tajikistan 851013 155951.2 40.301 69.823 16 29 63 5.8
b QT6 Uzbekistan 860325 234934.4 40.336 63.658 33 24 61 5.2
a QT7 Afghanistan 860426 141507.6 36.495 71.114 187 29 71 5.6
a QT8 Afghanistan 860821 013417.3 36.472 71.083 235 29 71 5.4
a QT9 Afghanistan 860915 214229.2 36.714 71.092 89 29 70 5.8
QT10 Afghanistan 860917 120809.4 37.290 71.730 120 30 69 5.5
a QT11 China 870225 195636.4 38.029 91.144 33 45 66 5.7
a QT12 China 870430 051737.1 39.770 74.590 9 32 64 5.7
a QT13 China 870918 215836.6 47.276 89.674 33 43 53 5.3
a QT14 Afghanistan 871003 110005.2 36.454 71.437 95 30 71 5.9
a QT15 Tajikistan 880326 225842.8 38.309 73.234 121 31 67 5.7
QT16 Tajikistan 880720 062051.4 37.028 72.914 41 31 69 5.5
QT17 Afghanistan 880806 090321.9 36.461 71.043 195 29 71 6.1
QT18 Afghanistan 880909 211236.1 36.483 71.381 188 30 71 5.4
QT19 China 880923 044640.6 39.570 74.506 33 32 65 5.3
QT20 Afghanistan 880926 071700.2 36.294 71.374 107 30 71 5.6
QT21 Afghanistan 900205 051645.1 37.047 71.250 110 30 69 6.1
QT22 China 900417 015933.4 39.436 74.900 33 32 65 6.0
a QT23 Kazakhstan– 900803 091539.6 47.949 84.958 19 40 51 6.1
China
QT24 China 901024 233815.1 44.117 83.856 20 39 57 5.2
QT25 China 901024 234657.6 44.119 83.876 22 39 57 5.3
a QT26 China 910819 060553.0 46.954 85.333 44 40 53 5.7
b QT27 China 930202 160510.7 42.212 86.126 10 41 60 5.7
QT28 Afghanistan 930918 050227.2 36.369 71.600 117 30 71 6.1
QT29 China 931002 084232.8 38.141 88.638 16 43 66 6.3
QT30 Tibet 860110 034629.9 28.648 86.527 55 44 80 5.4
QT31 Tibet 860620 171246.9 31.240 86.847 33 43 76 5.9
QT32 China 860826 094300.3 37.724 101.496 8 53 64 6.2
QT33 China 870810 121217.8 38.118 106.357 10 57 62 5.4
QT34 N. China 880103 213225.5 38.111 106.336 14 57 62 5.5
QT35 China 881105 021430.3 34.354 91.880 8 47 70 5.9
QT36 Kirgiz 881221 082103.7 41.220 72.298 33 31 61 5.4
QT37 China 890922 022553.5 31.545 102.464 33 56 71 6.0
QT38 China 891102 072239.2 36.206 106.346 10 57 64 5.0
QT39 China 900114 030319.2 37.819 91.971 12 46 66 6.1
QT40 Tibet 900602 003235.0 32.432 92.743 13 48 73 5.6
QT41 Kazakhstan– 900927 211232.5 47.903 84.961 33 40 51 5.0
China
See legend of Table 1.
There is a number of SP observations of Soviet and Chinese nuclear tests at different sites,
confirming more high-frequency energy for the explosions than for the shallow crustal earthquakes
(e.g., Taylor and Marshall, 1991). This can be explained by a smaller spatial area of energy release
for nuclear tests (the finite elastic radius) and shorter duration of the source function (the time
rise), as compared to earthquakes (Savage, 1972), which produces a shorter wave length of P-
waves. Nevertheless, it was noticed that Nevada test site explosions have, in contrast, more low-
frequency content than earthquakes in the western United States (Taylor and Denny, 1991). These
observations may be due to differences in the dynamic responses of the near-source geology or
propagation path peculiarities.
@seismicisolation
@seismicisolation
A few papers have been devoted to teleseismic SP discrimination of China and Kazakhstan events
using spectral parameters such as spectral ratio and third moment of frequency for six Finnish
stations (Tiira and Tarvainen, 1994), or peak frequencies and ratio of P/P-coda spectral maximums
for the NORESS array (Tsvang et al., 1993). In this study, we applied spectral discriminants which
have performed well on local events in various areas to Eurasian nuclear tests and earthquakes
from adjacent areas recorded by ISN. For comparison, the NORESS records of the same events
were included in the analysis. Instead of single station or several stations and a voting method, we
used a subset of stations to estimate a single discrimination factor, which is a simple average (for
energy spectral ratio), or a combination (for spectral semblance) of parameters from different
stations.
DATABASE
The database consists of 29 nuclear explosions from the test sites in China (Lop Nor), Kazakhstan
(Semipalatinsk) and Russia (W. Siberia, Novaya Zemlya), 12 deep (H > 50 km), mostly
Afghanistan earthquakes and 29 shallow earthquakes from China and Tajikistan. The magnitude
range of the events was mb = 5.2–6.5. A total of 68 events recorded by the vertical SP stations of
the Israel Seismic Network (50 Hz sampling rate and a recording frequency range of 0.2–12.5 Hz
due to bandpass filtering) were selected for analysis. In parallel, we collected and processed
NORESS seismograms of 20 explosions and 14 earthquakes, also contained in the ISN data base
plus two additional earthquakes. We personally examined all records and eliminated those which
appeared to be affected by glitches, spikes, noise bursts, or other type of system malfunctions.
Fig. 1. Locations of nuclear test sites (triangles) and teleseismic events (circle) used in the study. Deep (Afghan)
earthquakes are indicated by darker circles. Israel Seismic Network is indicated by the station BGIO (Bar-Giyyora).
placed in the middle of ISN.
@seismicisolation
@seismicisolation
140 Chapter Seven
Fig. 2. Several Chinese tests recorded at the ISN station PRNI (a). Instrument-corrected spectra of time windows about
7 s, marked by arrows, show distinct spectral nulls (b), almost identical for all events. After enhancement of spectra
resolution by expanding the analyzed signal length by 0 to 28 s, and reducing ¨ f from 0.1 Hz to 0.02 Hz, the spectral
nulls became more pronounced (c).
@seismicisolation
@seismicisolation
Fig. 3. Records and spectra of the Chinese tests at the BB station BGIO (vertical component). demonstrating similarity
to ISN short-period data. The analyzed signal in time windows about 7 s, marked by arrows, was expanded by 0 to 28
s, for better spectra resolution.
Detailed source information is presented in Tables 1 and 2; locations of selected events and
recording stations are shown in Fig. 1. Location, origin time, magnitude mb and focal depth are
taken from PDE (NEIC) bulletins; distance and azimuth are calculated relative to the station BGIO
(Bar-Giyyora) of the ISN. The selected ISN records (2–2.5 min in length) contain only P-wave and
P-coda waves. Time windows of 13–15 s, comprising the P-waves, were used for spectral analysis
of both ISN and NORESS seismograms. All amplitude spectra were smoothed in 0.2 Hz window,
while the ISN spectra were also instrument-corrected. For several Chinese nuclear explosions, we
also analyzed recordings of the Israel broadband (BB) station BGIO.
@seismicisolation
@seismicisolation
142 Chapter Seven
Fig. 4. The nuclear test NT15, demonstrating distinct spectral nulls and modulation and high coherency of spectral
shapes at eight different stations.
DISCRIMINANTS USED IN THE STUDY

Spectral nulls: failed discriminant
Contrary to earthquakes, underground nuclear tests are characterized by shallow focal depths
(several hundreds of meters at most) and by the spherical symmetry of the initial wave source
radiation. These features lead to strong destructive interference between the down-going P-wave
energy and the pP wave reflected from the Earth’s surface. This interference, however, is
complicated by non-linear surface effects on the pP reflection (Lay, 1991) The interference
produces modulation minima in the spectra of teleseisms at frequencies:
f § nv » 2h, n = 1, 2 @seismicisolation
@seismicisolation
, ... (1)
where v is the compressional wave velocity of the medium above the source and h is the depth.
For many nuclear tests conducted at depths h = 500–700 m and for v = 1.1–1.5 km/s, Eq. (1) gives
fm §1 Hz. For deeper sources the velocity increases and the frequency value does not change a
great deal. Spectra of the initial part (4–7 s) of the teleseismic records of nuclear explosions
detonated at the Nevada test site demonstrated pronounced minima (nulls) near 1 Hz, which may
be interpreted in terms of interference of P- and pP-wave (Kulhanek, 1971). Similar minima at 1.5
Hz were observed for Semipalatinsk tests (Kulhanek, 1973).
Being a source-like effect, the spectral null is apparent at about the same frequency in the spectra
of different stations of a network and can be used to discriminate nuclear tests from deep
earthquakes. This phenomenon was also observed for ISN (SP seismometers) (Fig. 2), and the BB
station BGIO (Fig. 3) records of several Chinese tests, with almost identical deep spectral minima,
which suggests that the locations and depths of these explosions are very close. The visible
identity of SP and BB seismograms and spectra confirms the absence of significant distortions of
the low-frequencies of teleseismic P-waves recorded by the SP instruments.
We tried to check the azimuthal independence inherent to this source-like effect and compare
spectra of the P-wave group at different ISN stations. Some explosions (e.g., NT15 on Fig. 4)
demonstrate a distinct spectral modulation and high coherency of spectral shapes at many stations
with minima at about 1 Hz (several multiple minima can be observed at 2, 3 and 4 Hz).
Nevertheless, many explosions revealed strong variability of the null frequency (Fig. 5) which can
hardly be identified for some stations. On the other hand, several earthquakes (e.g., QT26) also
show the same spectral null phenomenon, possibly caused by interference of rupture sub-events
(see Fig. 6). We conclude that this spectral feature is not sufficiently stationary to be utilized here
as a reliable discriminant.
Energy spectral ratio
It was observed that ISN seismograms of quarry blasts exhibit more low-frequency content of
seismic energy than local earthquakes (Gitterman and van Eck, 1993). This phenomenon is due
mainly to dominant surface waves generated by quarry blasts in combination with the regional
crustal structure of widespread unconsolidated, low velocity subsurface sediments, trapping low
frequency energy and absorbing high frequencies along propagation paths. This effect was
successfully utilized in the energy spectral ratio RE of seismic energy in the low-frequency [f1, f2]
and the high-frequency range [f3, f4] (Gitterman and Shapira, 1993):
f2 f4
2 2
RE ³ S ( f ) df ³ S ( f ) df
f1 f3
(2)
where |S(f)| is the smoothed spectrum of ground velocity in the selected time window (including
the whole wave train for local events). This ratio, implying physical meaning of seismic energy
frequency distribution, is similar to the ratios of average spectral amplitude (Bennett and Murphy,
1986; Walter et al., 1995), or power spectrum (Pulli, 1995) calculated for specific regional
phases. For an event, RE values at different stations showed strong variation; therefore, averaging
over a subnet is crucial for separation between earthquakes and explosions (Shapira et al., 1996).
The converse effect is observed for Eurasian teleseismic events recorded by the ISN: the nuclear
tests radiate more high frequency energy in P- and P-coda waves than earthquakes from adjacent
areas (see Fig. 7). The structural variations in the upper crust near the ISN stations, where
thicknesses are changing from north to south, may cause azimuth dependent variations in the
frequency content of an incoming signal, therefore, averaging of the energy spectral ratio
enhances the discriminant performance.
@seismicisolation
@seismicisolation
144 Chapter Seven
Fig. 5. The nuclear explosion NT27 showing strong variability of the null frequency and spectral shapes.
@seismicisolation
@seismicisolation
Fig. 6. The earthquake QT26, revealing explosion-like nulls about 1 Hz, modulation and high coherency in spectra for
seven stations. Seismograms of four stations only are presented for better observation of characteristic interference
features in the P-wave.
Fig. 7. Examples of typical teleseismic records from the DSI station of the ISN: deep earthquake, nuclear test and
shallow earthquake. @seismicisolation
@seismicisolation
146 Chapter Seven
Newly developed spectral semblance discriminant

Semblance and cross-correlation statistics are commonly used in seismic prospecting for phase
correlation of seismic traces in the time domain (e.g., Neidell and Taner, 1971). The first
application of these statistics for discrimination of local events (Shapira et al., 1996) was based on
differences in the coherency of spectral shapes at ISN stations. Distinc- tive features of the quarry
blasts and underwater explosions records are the clear spectral modulation patterns presented by
coherent minima (or nulls) and maxima observed at frequencies 1–10 Hz. This azimuth-
independent modulation, indicative of source- type, is caused by interference of linearly
superimposed waveforms from spaced charges exploded at millisecond delays (‘ripple-firing’), or
from oscillations of gaseous bubbles in water.
In contrast to a ripple-fired blast, the spectral pattern for an earthquake is azimuth-dependent, ow-
ing partly to the directivity effect (e.g., Bakun et al., 1978). Therefore, for stations with broad
azimuthal coverage (as in the case of local network observations), an irregular character of
spectral shapes and minima for different azimuths is observed. These statistics were modified to
assess quantitatively the coherency of the spectral shapes and applied successfully to
discrimination of local events (Shapira et al., 1996; Gitterman et al., 1998). As expected, a clear
equivalence between the two characteristics was observed, so we considered only spectral sem-
blance Sf , calculated in a frequency band [ F1, F2]:
(3)
where Ski = log10Sk (fi) is the log spectral amplitude at the k th station; ǅk is the average spectral
level and N is the number of used stations.
By analogy with the case of local events, we, as a working hypothesis, supposed that spectral sem-
blance will be different for the two classes of teleseismic events. Likewise, it would be expected
that the nuclear tests will show higher Sf values than earthquakes owing to the abovementioned
source spherical symmetry and the spectral nulls modulation. Surprisingly, the situation turned
out to be precisely the opposite, as detailed below.
ANALYSIS OF DISCRIMINANT PERFORMANCE
ISN recordings
The spectral ratio and semblance procedures were applied to 68 Eurasian teleseismic events, and
naturally, in comparison with local events, introducing a shift of spectral bands to lower
frequencies: 0.6–3 Hz, according to observed spectra (Figs. 2–6). The statistics - average ratios of
low-to-high frequency energy (see Eq. (2)) and semblance of spectral shapes (Eq. (3)) -were
computed from smoothed spectra for a subnet of 7–10 stations selected for a given event.
The semblance and ratio values for earthquakes turned out to be higher than those for the
explosions (Fig. 8), in contrast to local events observations (Shapira et al., 1997) (see Fig. 9). This
can be explained by the fact that the analyzed waves (P- and P-coda) from teleseismic earthquakes
have relatively more low-frequency energy as predicted from source physics phenomena (e.g.,
Savage, 1972) and unex- pectedly more coherent spectral shapes than nuclear explosions (Figs. 4–
6). Azimuths for a teleseismic event to the ISN stations are almost identical (compared to a local
event), thus significantly increasing spectra coherency for both types of seismic events.
Nevertheless, short-period explosion signals (including scattered waves in the P-coda), are more
influenced by inhomogeneities in the upper crust beneath the ISN stations spaced several tens of
@seismicisolation
@seismicisolation
kilometers apart, thus leading to a decrease of spectra coherency and associated semblance values.
The best discrimination performance was provided in the frequency bands (0.6-1 Hz)/(1-3 Hz) for
averaged spectral ratios and in the range (0.6-2 Hz) for the semblance parameter. As expected,
removing instrument response from the spectra improved significantly the spectral semblance
performance. Joint application of the two discriminants provided good separation between the
earthquake and explosion populations with only few events overlapping. Outliers include an
explosion from Novaya Zemlya (NT18) with anomalous high Sf .= 0.84 and RE =3..2 and a
relatively deep Tibet earthquake (QT30) with Sf = 0.72. Al- though some deep Afghan events
exhibit spectral content and wave forms very similar to those of explosions (see Fig. 7),
nevertheless they were successfully discriminated.
Fig. 8. Discrimination results (spectral semblance vs. spectral ratio at optimal frequency ranges) for teleseismic
Eurasian earthquakes and nuclear explosions, based on ISN recordings (with instrument response removed).
Fig. 9. Discrimination results (spectral semblance vs. spectral ratio) for local events at different frequency ranges. The
semblance range (1–8 Hz) provided the best separation of the events from Galilee and Gilad regions (Northern Israel)
(from Shapira et al., 1997).
@seismicisolation
@seismicisolation
148 Chapter Seven
We tried to track the possible dependence of the two statistics on depth (for earthquakes only) and
magnitude (Fig. 10). The only weak correlation is that between the spectral ratio and the
magnitude. The trend seen in Fig. 10a is probably due to the shift in corner frequency with
increasing magnitude, and an enlargement of low-frequency component of radiated energy. A
visible saturation of this dependence for mb ‫ ޓ‬6 can be explained by the shift of a fraction of low-
frequency energy beyond the high pass cut-off (0.2 Hz), or the analysis band (0.6–3 Hz). The
dependence of earthquake depth on the discriminants was not observed.
Most of the selected seismograms appear to have a high signal-to-noise ratio. However, for the
earthquake QT26, quite strong high-frequency noise is apparent (Fig. 6). Nevertheless, it did not
affect the discriminants used: relatively high values of Sf = 0.81 and RE =3.7 were obtained, which
are characteristics for the earthquake population (Fig. 8).
NORESS data
The average energy spectral ratios and spectral semblance were computed from smoothed spectra,
which were not corrected for instrument response due to a virtually flat short-period transfer
function in the analyzed frequency band 0.6–3 Hz (Mykkeltveit et al., 1990). A subset of 8–10
vertical SP stations (Fig. 11c). was selected; as a rule we used the central station A0, one station
each from circles A, B and C and four to six stations from the outer- most D-ring.
@seismicisolation
@seismicisolation
Fig. 10. Spectral energy ratio (a). and spectral semblance (b) vs. magnitude for ISN records of teleseismic events. A
weak visible trend is marked by dotted lines.
@seismicisolation
@seismicisolation
150 Chapter Seven
Fig. 11. Typical seismograms (a). and spectra (b). of a nuclear test (NT14). recorded at the NORESS (c).
We found a similar high-frequency character of the P-arrivals for explosions relative to

earthquakes (see Figs. 11 and 12) consistent with the results of Tsvang et al. (1993). Therefore,
spectral ratios showed the same (as for ISN) interrelation between the two populations, whereas
the character of the semblance changed was reversed, i.e., very high (Sf ‫ޓ‬., concentrated
values for explosions and lower, in-average, dispersed values for earthquakes (Fig. 13). In this
case, owing to the small dimensions of the array (about 3 km diameter), local inhomo-geneities in
the upper crust and azimuth differences are negligible. We therefore suggest that the main factor
affecting station spectra coherency is a waveform character in the 0- to 15-s window: i.e., a
single predominant P-pulse for explosions and several equal-amplitude pulses of different shapes
for earthquakes.
@seismicisolation
@seismicisolation
Joint application of the two statistics defined for identical frequency bands (0.6–1 Hz)/(1–3 Hz)
provided reliable discrimination between the two types of seismic events, with the only outlier
being the deep Afghanistan earthquake QT10 (Fig. 13a). By applying different frequency bands
(0.6 – 0.8 Hz)/ (0.6–3 Hz), which is actually the normalized low-frequency energy in a more
narrow band, we obtained complete separation of the earthquake and explosion populations (Fig.
13b).
Fig. 12. Typical seismograms (a). and spectra (b) of an earthquake (QT27) recorded at NORESS (analyzed time
window t 13 s).
@seismicisolation
@seismicisolation
152 Chapter Seven
Fig. 13. Discrimination results äVSHFWUDO semblance vs. spectral ratio. for teleseismic earthquakes äA and nuclear tests
äY ., recorded by NORESS. Different frequency bands are used: ä–0.8 +]Uä–3 Hz. äD and ä–0.8 +]Uä–3
Hz. äE

We have demonstrated that the use of multi-station SP discriminants based on spectral features of
seismic waves represents an attractive approach to effective discrimination. Its usefulness is not re-
stricted to small-magnitude regional events, but it is also applicable also to teleseismic events,
recorded by a regional network or by small array. Rather surprisingly, application of the energy
ratio and spectral semblance statistics to Eurasian nuclear explosions and adjacent earthquakes,
demonstrated their significant event classification power. Using NORESS data, full separation of
the two populations of teleseismic events was obtained; only a few outliers were found for ISN
recordings. In general, the spectral ratio performance appears to be better for array recordings,
whereas semblance is more successful in the case of network data.
As compared to small local (Israel) events, for which use of the spectral ratio was based mainly on
path propagation effects and local geological settings, the spectral semblance statistic performed
well mainly due to specific source features of quarry blasts and underwater explosions. However,
in the teleseismic case of large magnitude events, the situation was reversed: the ratio differences
were due to source effects and the semblance estimates of spectral coherency were caused by wave
propagation peculiarities in geological structures near recording stations.
@seismicisolation
@seismicisolation
ACKNOWLEDGEMENTS
We are grateful to Dr. A. Hofstetter (GII) and Dr. J. Fyen (NORSAR, Norway) for their significant
help in the collection of ISN and NORESS recordings of selected teleseismic events. This study
was supported by the U.S. Department of Energy under Contract No. F19628-95-K-0006.
REFERENCES
Bakun, W.H., Stewart, R.M., Bufe, C.G., 1978. Directivity in the high-frequency radiation of
small earthquakes. Bull. Seism. Soc. Am. 68, 1253–1263.
Bennett, T.J., Murphy, J.R., 1986. Analysis of seismic discrimination capabilities using regional
data from Western United States events. Bull. Seism. Soc. Am. 76, 1069–1086.
Blandford, R.R., 1995. Regional seismic event discrimination. In: Husebye, E.S., Dainty, A.M.
(Eds.), Monitoring a Comprehen- sive Test Ban Treaty. NATO ASI Series, Series E: Applied
Sciences. Kluwer Publ., pp. 689–719.
Gitterman, Y., Shapira, A., 1993. Spectral discrimination of underwater explosions. Isr. J. Earth
Sci. 42, 37–44.
Gitterman, Y., van Eck, T., 1993. Spectra of quarry blasts and microearthquakes recorded at local
distances in Israel. Bull. Seism. Soc. Am. 83, 1799–1812.
Gitterman, Y., Pinsky, V., Shapira, A., 1998. Spectral classification methods in monitoring small
local events by the Israel Seismic Network. J. Seismology, Kluwer Publ. 2, 237–256.
Kulhanek, O., 1971. P wave amplitude spectra of Nevada underground nuclear explosions. Pure
Appl. Geophys. 88, 121–136.
Kulhanek, O., 1973. Source parameters of some presumed Semi- palatinsk underground nuclear
explosions. Pure Appl. Geophys. 102, 51–66.
Lay, T., 1991. Teleseismic manifestations of pP: problems and paradoxes. In: Taylor, S.R., Patton,
H.J., Richards, P.G. (Eds.), Explosion Source Phenomenology. Geophysical Monograph 65,
109–125.
Mykkeltveit, S., Ringdal, F., Kvaerna, T., Alewine, R.W., 1990. Application of regional array in
seismic verification. Bull. Seism. Soc. Am. 80, 1777–1800.
Neidell, N.S., Taner, M.T., 1971. Semblance and other coherency measures for multichannel data.
Geophysics 36, 482–497.
Pulli, J.J., 1995. Extracting and processing signal parameters for regional seismic event
identification. In: Husebye, E.S., Dainty, A.S. (Eds.), Monitoring a CTBT. NATO ASI Series,
Series E.
Kluwer Publ., pp. 741–754.
Richards, P.G., Zavales, J., 1995. Seismological methods for monitoring a CTBT: the technical
issues arising in early negotiations. In: Husebye, E.S., Dainty, A.S. (Eds.), Monitor- ing a
CTBT. NATO ASI Series, Series E. Kluwer Publ., pp. 53–81.
Savage, J.C., 1972. Relation of corner frequency to fault dimensions. J. Geophys. Res. 77, 577–
592.
Shapira, A., Gitterman, Y., Pinsky, V., 1996. Discrimination of seismic sources using the Israel
Seismic Network. Proceedings of 18th Seismic Research Symposium on Monitoring a CTBT,
September 1996, Annapolis, pp. 612–621.
Shapira, A., Pinsky, V., Gitterman, Y., 1997. Discrimination of seismic sources using Israel
Seismic Network. Final Report, PL-TR-97-2207. Phillips Lab., MA, 78 pp.
Taylor, S.R., Denny, M.D., 1991. An analysis of spectral differences between Nevada test site
and Shagan river nuclear explosions. J. Geophys. Res. 96, 6237–6245.
Taylor, S.R., Marshall, P.D., 1991. Spectral discrimination between
Soviet explosions and earthquakes using short-period array data. Geophys. J. Int. 106, 265–273.
Tiira, T., Tarvainen, M., 1994. Discrimination of teleseismic events in Central Asia with a local
network of short period stations. Ann. di Geof. 37, 433–449.
@seismicisolation
@seismicisolation
154 Chapter Seven
Tsvang, S.L., Pinsky, V.I., Husebye, E.S., 1993. Enhanced seismic source discrimination using
NORESS recordings from Eurasian events. Geophys. J. Int. 112, 1–14.
Walter, W.R., Mayeda, K.M., Patton, H.J., 1995. Phase and spectral ratio discrimination between
NTS earthquakes and explosions: Part I. Empirical observations. Bull. Seism. Soc. Am. 85,
1050–1067.
@seismicisolation
@seismicisolation
CHAPTER EIGHT
SIGNAL PROCESSING FOR INDIAN AND PAKISTAN NUCLEAR TESTS

RECORDED AT IMS STATIONS LOCATED IN ISRAEL1
YEFIM GITTERMAN, VLADIMIR PINSKY AND RAMI HOFSTETTER
IMS seismic station EIL, Israel, located inside a N-S oriented 100- m long tunnel drilled into Mt
Amram (composed of granite rocks), 15 km north of the city of Eilat. Several tight metal doors
shield the vault from the outside (photo by Y. Gitterman).
ABSTRACT
In compliance with the Comprehensive Nuclear-Test-Ban-Treaty (CTBT) the International
Monitoring System (IMS) was designed for detection and location of the clandestine Nuclear Tests
(NT). Two auxiliary IMS seismic stations MRNI and EIL, deployed recently, were subjected to
detectability, travel-time calibration and discrimination analysis. The study is based on the three
recent 1998 underground nuclear explosions: one of India and two of Pakistan, which provided a
ground-truth test of the existing IMS. These events, attaining magnitudes of 5.2, 4.8 and 4.6
correspondingly, were registered by many IMS and other seismic stations.
The MRNI and EIL broadband (BB) stations are located in Israel at teleseismic distances (from the
explosions) of 3600, 2800 and 2700 km, respectively, where the signals from the tests are already
weak. The Indian and the second Pakistan NT were not detected by the short-period Israel Seismic
Network (ISN), using standard STA/LTA triggering. Therefore, for the chosen IMS stations we
compare the STA/LTA response to the results of the more sensitive Murdock-Hutt (MH) and the
Adaptive Statistically Optimal Detector (OD) that showed triggering for these three events. The
second Pakistan NT signal arrived at the ISN and the IMS stations in the coda of a strong
Afghanistan earthquake and was further disturbed by a preceding signal from a local earthquake.
However, the NT signal was successfully extracted at EIL and MRNI stations using MH and OD
procedures. For comparison we provide the signal analysis of the cooperating BB station JER,
with considerably worse noise conditions than EIL and MRNI, and show that OD can detect
1
Pure and Applied Geophysics (2002), 159, 779-801
@seismicisolation
@seismicisolation
156 Chapter Eight
events when the other algorithms fail. Using the most quiet EIL station, the most sensitive OD and
different bandpass filters we tried in addition to detect the small Kazakh chemical 100- ton
calibration explosion of 1998, with magnitude 3.7 at a distance approaching 4000 km. The detector
response curve showed uprising in the expected signal time interval, but yet was low for a reliable
decision.
After an NT is detected it should be recognized. Spectra were calculated in a 15-sec window
including P and P-coda waves. The spectra for the first Pakistan NT showed a pronounced spectral
null at 1.7 Hz for all three components of the EIL station. The effect was confirmed by observation
of the same spectral null at the vertical component of the ISN stations. For this ground-truth
explosion with a reported shallow source depth, the phenomenon can be explained in terms of the
interference of P and pP phases. However, the spectral null feature, considered separately, cannot
serve as a reliable identification characteristic of nuclear explosions, because not all the tests
provide the nulls, whereas some earthquakes show this feature. Therefore, the multichannel
spectral discrimination analysis, based on a spectral ratio of low-to-high frequency energy (in the
0.6- 1 Hz and 1-3 Hz bands), and a semblance of spectral curves (in the 0.6-2 Hz band), was
conducted. Both statistics were calculated for the vertical component of the ISN stations as well
for the three components of the EIL station. The statistics provided a reliable discrimination
between the recent NT and several nearby earthquakes, and showed compliance with the former
analysis of Soviet and Chinese NT, where nuclear tests demonstrated lower values of energy ratio
and spectral semblance than earthquakes.
Accurate location of NT requires calibration of travel time for IMS stations. Using known source
locations, IASPEI91 travel-time tables and NEIC origin times we calculated expected arrival time
for the P waves to the EIL and MRNI stations and showed that the measured arrival time has a
delay of about 4 sec. Similar results were obtained for the nearby Pakistan earthquakes. The
analysis was complimented by the P travel-time measurements for the set of Semipalatinsk NT,
which showed delays of about 3.7 sec to the short-period MBH station which is a surrogate station
for EIL. Similar delays at different stations evidence a path- rather than site-effect. The results can
be used for calibration of the IMS stations EIL and MRNI regarding Asian seismic events.
Key words: GT0, CTBT, Quanterra, NEIC, STA/LTA
INTRODUCTION
The recent nuclear test activity in southern Asia (Fig. 1), comprising the announced tests in India
(May 11, 1998) and Pakistan (May 28 and 30, 1998), provided new recordings for verification of
teleseismic discriminants, and for testing new detection and location procedures, as well as an
opportunity for travel-time correction for new stations of the International Monitoring System
(IMS). These discriminants and location procedures can serve as a tool for improving the
monitoring executed by the Comprehensive Test-Ban Treaty. The explosion recordings were
analyzed mainly based on the closest station, Nilore (NIL), at near-regional distance range of
about 7°-9° (e.g., WALTER et al., 1998). In August 1998, a 100-ton chemical explosion was
conducted at the former Semipalatinsk Nuclear Test Site (STS), Kazakhstan (Fig. 1), as a
calibration experiment for the CTBT verification, and was recorded by many stations worldwide
(CTBTO, 1998b).
The number of active IMS stations in the Middle East is rather limited, and thus we stress the need
to assess, as precisely as possible, the monitoring capabilities of each station. The IMS stations
EIL and MRNI were installed in Israel in 1996 and 1998, respectively, when nuclear testing was
nearing termination, therefore every chance to analyze ground-truth explosion data at these
stations should be used. We analyze and interpret here the recordings of the two IMS stations, the
cooperating BB station JER and short-period stations of the Israel Seismic Network (ISN; Fig. 2),
of the nuclear and chemical explosions, located at a distance range of 25°- 38°. We show that these
@seismicisolation
@seismicisolation
Signal Processing for Indian and Pakistan Nuclear Tests Recorded at IMS Stations Located in Israel 157
measurements can be a useful tool for monitoring the CTBT and the calibration of IMS stations in
the region. The case of the second Pakistan test is of special interest, because it simulates in some
way the scenario of a clandestine test masked by other two events: a strong earthquake in
Afghanistan (06:22, MS = 6.9), and a local earthquake (06:57, ML = 2.5) 20 km south of EIL
station.
DATA
There is no ground-truth information GT0 for the recent Indian and Pakistan nuclear tests.
BARKER et al. (1998) presented location and origin time for these tests, based on PIDC REB,
Joint Epicenter Determination and Satellite Imagery. Owing to additionally using in our study
several nearby Pakistan earthquakes and old Semipalatinsk nuclear tests, we address here source
parameters taken from the PDE (USGS) Bulletin (see Table 1), in order to keep homogeneity of
the data.
Figure 1: Relative location of the Israel network and the selected teleseismic events.
We analyzed seismograms observed by the short-period stations (1 Hz seismometer, L4C Mark

Product) and broadband stations EIL, MRNI and JER (STS-2 seismometer and Quanterra data
logger) of the ISN. Only the Pakistan test on May 28, 1998 triggered the short-period stations (Fig.
3a). Two tests on May 11 and 28, 1998, triggered the 80 Hz channel of the broadband station EIL.
The second Pakistan nuclear test on May 30, 1998, occurred about 38 minutes after a major MS =
6.9 earthquake, and was revealed in the continuous 20-Hz channel of EIL using a higher frequency
band (Fig. 3c). In order to provide a discrimination analysis we added several recordings of
earthquakes occurring near the explosion sites in Pakistan (Figs. 1 and 3b). We also analyzed EIL
recordings of the chemical calibration explosion of 100 tons at the STS, on August 22, 1998
(shown below on Fig. 9).
@seismicisolation
@seismicisolation
158 Chapter Eight
Figure 2: Location of short-period and broadband stations of the ISN used in the analysis.
TRAVEL TIMES AND PEAK AMPLITUDES

The same set of teleseismic events from southern Asia was used to estimate station corrections
based on P travel times and NEIC source data. The P arrivals to the IMS station EIL showed
regular delays with an average of about 4 seconds relative to the IASPEI91 model (see Table 2,
Fig. 3c). Analogous delays of about 3.7 seconds on average were found in the analysis of former
nuclear explosions at the Semipalatinsk Test site (Table 2). The explosions were recorded at the
short-period station MBH situated within 14 km North of EIL station (Fig. 2), which was not yet
installed. Figure 3c also presents single cases of similar lags 3.5-4 sec for the IMS station MRNI
and BB station JER. Similar delays at different stations evidence a path- rather than site-effect.
Table 1: Source parameters of nuclear tests and earthquakes used in the study (location, origin
time and magnitude are taken from PDE)
# Date Origin time Lat. N Lon. E Depth Mag. Dist. to Event, region
km mb EIL, deg.
New events
1 1997/12/04 10:17:01.33 29.09 64.11 33 5.0 25.4 Earthquake,
Pakistan
2 1998/01/05 16:58:35.29 29.01 64.35 18 4.9 25.6 Earthquake,
Pakistan
3 1998/05/11 10:13:41.78 27.10 71.80 0 5.2 32.4 Nucl. expl.,
India
4 1998/05/28 10:16:15.23 28.90 64.79 0 4.8 26.0 Nucl. expl.,
Pakistan
5 1998/05/28 20:32:46.51 26.58 62.23 47 4.7 24.2 Earthquake,
Pakistan
6 1998/05/30 06:54:57.1 28.50 63.74 0 4.6 25.2 Nucl. expl.,
Pakistan
@seismicisolation
@seismicisolation
7 1998/08/22 05:00:18.90 49.77 77.99 0 3.8 38.06 Chem. expl.

3.7* 100 ton, STS
Old STS nuclear tests
1 1987/03/12 01:57:17.3 49.93 78.79 0 5.4 38.5
2 1987/04/17 01:03:04.8 49.89 78.69 0 6.0 38.5
3 1987/05/06 04:02:05.5 49.78 78.09 0 5.5 38.1
4 1988/04/03 01:33:05.8 49.92 78.95 0 6.1 38.6
5 1988/11/12 03:30:03.7 50.08 78.99 0 5.3 38.7 Kazakhstan
6 1988/12/17 04:18:06.9 49.89 78.93 0 5.9 38.6
7 1989/01/22 03:57:06.5 49.92 78.86 0 6.0 38.6
8 1989/02/12 04:15:06.7 49.90 78.76 0 5.9 38.5
* JHD solution by WALLACE (1998).
Based on the corrected arrival time we tried to identify the P arrival of the STS 100-ton chemical
explosion using a band pass filtered seismogram at EIL. We estimated a possible P amplitude for
this explosion using the dependence of measured P-wave peak amplitudes for a series of old STS
nuclear explosions (see Table 2) with magnitudes in the mb range 5.3-6.1. The upper limit
magnitude-yield relationship for explosions in hard rock (KHALTURIN et al., 1998) for the 100-
ton explosion gives mb = 3.9, NEIC reports mb = 3.8, and pIDC REB presents mb = 3.7. Rough
extrapolation to lower magnitudes of the magnitude-amplitude dependence observed at MBH
station shows that for mb = 3.7-3.9 one should expect at EIL station P amplitudes much lower than
0.01 micron/sec, whereas presignal noise level is about 0.01-0.02 micron/sec (in the band 1-3 Hz).
A weak phase can be observed on the filtered (2.5-3.5 Hz) vertical component seismogram about 2
sec after the calculated P-arrival time, corrected for the 3.7 sec average delay (an arrow on Fig. 9).
Nevertheless the phase association with P waves from this explosion is not reliable.
Figure 3: Sample seismograms of different types of seismic events: short-period vertical records of the first Pakistan
test (28.5.98) (a), and the same day nearby earthquake in Pakistan (b); filtered broadband vertical records of the three
nuclear tests (c). Vertical axis units here (and at all BB seismogram plots) are counts (1 count = 1.67 10-9 m/s).
@seismicisolation
@seismicisolation
160 Chapter Eight
WAVEFORM ANALYSIS AND TRIGGERING

Station EIL, located inside a 100-m long tunnel, is a quiet station. The broadband and short-period
noise velocity spectral density are presented in Figure 4a, along with selected noise velocity values
of a quiet station CAD in the Pyrenees (VILA, 1998), located in hard rock tunnel (490-m long, 80-
m deep). In the same frequency range the noise level at MRNI and JER is approximately 10 to 20
times larger, respectively (see filtered seismograms in Figs. 7 and 8). However, for the EIL station
the signals of the explosions under consideration appear considerably more attenuated than for
more remote GRF, NORES and ARCES (CTBTO, 1998a).
Figure 3b
We checked the detectability of the EIL BB station using recordings of the Indian and Pakistan
nuclear tests. The 80-Hz channel of the station is supplied with the procedure of MURDOCK and
HUTT (1983) for automatic detection (MH), including the following main stages:
1. The input time series is filtered. For simulation of the detector performance we used a
specific ARMA (AutoRegressive-Moving Average) filter, applied to the 20-Hz channel of
the BB station, instead of the built-in WWSSN-SP standard short-period filter.
2. Relative maxima and minima (peaks and troughs) of the filtered series are found, and
successive peaks and troughs are differenced. These differenced values, together with their
associated times, are named "P-T " time series.
3. Three thresholds (Th1, Th2 and Th3) are calculated. Detection is declared if P-T amplitude
exceeds Th1 and two other P-T values exceed Th2 in a fixed time interval of usually four
seconds. Th3 is used to search for onsets of detected events.
@seismicisolation
@seismicisolation
Figure 3c
Table 2: Travel times (IASPEI91) and P-wave peak amplitudes for the selected events
@seismicisolation
@seismicisolation
162 Chapter Eight
Figure 4: Velocity spectral density of noise at the EIL station compared to: (a) CAD BB station, (b) the Indian test
signal+noise on a 2-sec interval.
The values of the thresholds: Th1 = 4 s, and Th2 = 3 s and Th3 = 2 s are recommended as typical
and provide a high rate of true detections and a low rate of false alarms, where s is one standard
deviation of the P-T series for noise.
Figure 5 illustrates the P-T curves extracted from the EIL observed seismograms for the three
nuclear explosions. The first two nuclear tests do pass the stated above detection criterion, but the
second Pakistan nuclear test (Fig. 5c) does not pass. It would fit with lower Th1 and Th2, but at
the expense of a higher risk of false alarms. Therefore, two additionally different detectors were
tested using the broadband recordings of the events: the first is the STA/LTA detector (based on
JOHNSON, 1979); and the other is adaptive statistically Optimal Detector (OD) by KUSHNIR et
al. (1990).
To extract waveforms from background noise, we manually bandpass filtered all the seismograms
of the broadband stations. Spectral analysis for the vertical channel of EIL (Fig. 4b) shows that the
signal of the Indian nuclear test dominates noise in the frequency band 0.5-2 Hz. To provide less
distortion of the signal we used the filter 0.5-5 Hz, yielding peak amplitudes Am ~ 0.08 micron/sec
(see Fig. 6). We did not manage to extract the signal for MRNI and JER for this explosion.
Though of smaller magnitude but somewhat closer, the Pakistan nuclear test on May 28, 1998 was
well observed at short-period ISN stations (Fig. 3a), and broad band station EIL (Figs. 3c, 7a),
with peak amplitude Am ~ 0.4 micron/sec. A weak signal at JER was observed after a narrow band-
pass filtering 0.5-1.5 Hz (Fig. 3c). We could not reveal the signal at MRNI due to a burst of local
noise, or possible malfunction of the station.
The case of the second Pakistan nuclear test on May 30, 1998 is particularly complicated due to
preceding arrivals from the other two events mentioned above: the strong earthquake in
Afghanistan, and the local earthquake. Thus the body waves of the explosion are partially masked
by the coda waves of these two events (see Fig. 8). However, the P waves of the nuclear test can
be identified on the EIL seismogram in the 2-5 Hz frequency range with amplitude Am ~ 0.04
micron/sec. A strong enough signal with Am ~ 0.1 micron/sec was obtained at MRNI in the narrow
frequency range 1.5-3 Hz (Fig. 3c), while at JER this signal was fully masked by noise.
The seismograms, bandpassed filtered in the chosen frequency ranges, were used for testing by the
above mentioned STA/LTA detector. The output of this algorithm is set to be negative, and
detection is declared once the output is greater than zero. The STA/LTA trigger may fail or be
successful, depending on the selected frequency band and the background noise.
@seismicisolation
@seismicisolation
The Optimal Detector (OD) algorithm is based on autoregressive representation of seismic noise,
preceding the signal, inverse filtering, and producing "whitened" noise. The OD demonstrates
more robust behavior and enhanced sensitivity at relatively wider frequency range than the
STA/LTA algorithm. Detection is declared once the autocovariance of the "whitened" noise
exceeds a given threshold.
Figure 5: Application of the Murdock-Hutt detector algorithm to the BB EIL seismograms for the three nuclear
explosions. P-T curves (1) are obtained using output (2) of the WWSSN_SP filter.
@seismicisolation
@seismicisolation
164 Chapter Eight
Figure 6: Detection of the Indian test at EIL station: (a) vertical unfiltered 20 Hz channel record; (b) BP filtered data;
(c) Optimal Detector output; (d) STA/LTA output.
Comparative performance of the OD and STA/LTA, applied to the filtered (0.5-5 Hz) BB records
(20 Hz) of the considered explosions, is presented in Figures 6-8. Both detectors were successful at
EIL records of the Indian and the first Pakistan tests (Figs. 6, 7a). Detection of the second Pakistan
NT at EIL was complicated by the nearby local earthquake, which blurs NT triggering, though the
OD demonstrates better performance. Records of the Pakistan explosions at JER (Fig. 7b) and
MRNI (Fig. 8b) show a poor SNR, resulting in a failure of the STA/LTA, whereas the OD
provides reliable detections.
@seismicisolation
@seismicisolation
Figure 7: Detection of the first Pakistan test (28.5.98) at EIL (a) and JER (b) BB stations.
Using the records of the most quiet station EIL, the sensitive OD algorithm and different band-
pass filters, we tried in addition to detect the Kazakh chemical explosion. The detector output
showed some uprising in the expected signal time interval, however it was still too low for reliable
decision (Fig. 9).
SPECTRAL NULL FEATURE
We calculated P-wave spectra from seismograms (after removing instrument response) of the test
on May 28, 1998; the only one that triggered short-period ISN stations (see Figs. 2 and 3a). All
spectra were computed using a ~15 sec time window at vertical recordings, including P and P-
coda waves, and smoothed in the 0.5 Hz triangular window (Fig. 10a). The spectra at all stations
showed pronounced coherent spectral minima (nulls) at about 1.7 Hz, which may be interpreted as
interference of P and pP phases, possibly complicated by nonlinear surface effects such as spall
(LAY, 1991). KULHANEK (1971) reported similar observations of 1 Hz spectral nulls for Nevada
nuclear tests, treated as the interference effect. The relatively high null frequency 1.7 Hz possibly
corresponds to the reported shallow depth of the test placed in a horizontal shaft of a steep
mountain (WALLACE, 1998), compared with deeper Nevada tests. The same 1.7-Hz coherent
spectral nulls were also observed at all three components of the BB EIL station for the May 28
explosion, but not for the Indian test.
@seismicisolation
@seismicisolation
166 Chapter Eight
Figure 7b.
Figure 8: Detection of the second Pakistan test (30.5.98) at EIL (a) and MRNI (b) BB stations.
The same spectral analysis was done for a nearby earthquake in Pakistan with similar magnitude
and waveforms (Fig. 3b). The earthquake spectra (Fig. 10b) obtained at the same short-period
stations do not show coherent nulls (also as spectra at BB EIL), thus rejecting the site-effect
version. WALLACE (1998) notes decidedly more complex waveforms for the May 28 Pakistan
test, than for the Indian explosion (see also Fig. 3c), explained possibly by wave scattering from
the complex topography in the test region.
@seismicisolation
@seismicisolation
Nevertheless, GITTERMAN et al. (1999), based on ISN observations of Semipal- atinsk and
Chinese nuclear tests and nearby earthquakes, showed that this spectral null feature, considered
separately, cannot serve as a reliable identification characteristic of nuclear explosions due to its
non-stability over the entire nuclear test population and its presence in the recordings of some
earthquakes.
SPECTRAL DISCRIMINATION
TAYLOR and MARSHALL (1991) applied spectral ratio for discrimination between Soviet (STS)
nuclear tests and earthquakes recorded at the United Kingdom short- period teleseismic arrays.
Using 5-sec windows (including P waves), the spectral ratios from an array beam in the 0.5-1 Hz
and 2-3 Hz frequency bands provided the best discrimination performance. It was concluded that
the explosions are characterized by more high-frequency energy (resulting in lower spectral ratios)
than the shallow Central Asia earthquakes.
Figure 8b
@seismicisolation
@seismicisolation
168 Chapter Eight
Figure 9: Detection trial of the Kazakh chemical explosion at EIL BB station.
Similar results were obtained by GITTERMAN et al. (1999) using slightly different processing
parameters, with higher ratios (low-to-high frequency energy) for teleseismic earthquakes,
compared to nuclear explosions. The database included tests from STS, Novaya Zemlya and Lop
Nor (China), recorded by the short-period Israel Seismic Network. A 15-sec window, comprising
P- and P-coda waves, was selected for computation of spectral ratios of energy in low (0.6-1 Hz)
and high (1-3 Hz) frequency ranges, providing the best performance. Additionally, this study
included the application to the same spectra of a recently proposed spectral semblance
discriminant (GITTERMAN et al., 1998). The semblance statistic, commonly used in seismic
prospecting for phase correlation of seismic traces in the time domain (e.g., NEIDELL and
TANER, 1971), was modified to assess the coherency Sf of smoothed spectral shapes for different
stations (channels) in a frequency band [F1, F2]:
2
1 F2 ª N º F2 N
Sf ¦ ¦ ki k »¼
( S S ) ¦¦ (S Sk )2
N F1 «¬ k 1
ki
F1 k 1
(1)
where Ski = log10 Sk(fi) - log10(spectral amplitude at the k-th station), S k is the average spectral level
and N is the number of used stations (channels). The network-based spectral semblance estimates
showed higher coherency for teleseismic earthquakes than for nuclear explosions (GITTERMAN
et al., 1999).
For the selected recent events (Table 1), the two discrimination statistics were estimated for a
subset of nine short-period vertical ISN stations (shown in Fig. 2), and also for three components
(channels) of the EIL broadband station. The ratios were then averaged for a given event over the
stations, or over the three BB components. The second Pakistan nuclear test and the STS chemical
explosion did not trigger the short-period and broadband stations. Both signals on the 20 Hz
continuous recording are weak and comparable with the noise level (see Figs. 3c and 9), therefore
the spectral ratio and semblance were not calculated.
Both multi- and single-station estimates correspond well to the successful discrimination results
obtained for the former Semipalatinsk and Chinese nuclear tests (GITTERMAN et al., 1999). The
new nuclear tests also demonstrate lower values of energy ratio and spectral semblance than
nearby earthquakes (Table 3, Fig. 11). The analyzed waves (P and P coda) of teleseismic
@seismicisolation
@seismicisolation
earthquakes have relatively more low-frequency energy, as predicted from source physics
phenomena and more coherent spectral shapes than nuclear explosions. The short-period explosion
signals (including scattered waves in the P coda) are more influenced by inhomogeneities in the
upper crust beneath the ISN stations, with varying site structure, thus leading to a decrease of
spectra coherency and associated semblance values. The results show that both types of
discriminant estimates, using the short-period (Fig. 11a) or broadband records (Fig. 11b), being
rather different for a given event (especially the energy ratios), nevertheless provide a reliable
separation of nuclear explosions and earthquakes.
Figure 10: Spectra of the first Pakistan test, observed at short-period stations, showing a spectral null source feature at
1.7 Hz (a), compared with the same day nearby Pakistan earthquake of a similar magnitude (b).
@seismicisolation
@seismicisolation
170 Chapter Eight
Figure 10b

We investigated detectability at the two IMS auxiliary stations EIL and MRNI and cooperating
station JER using 20-Hz recordings of the Indian (mb = 5.2) and the two Pakistan nuclear tests (mb
= 4.6, 4.8). Additionally the detection of the Kazakh chemical test (mb = 3.7) was tested. Three
detection procedures were compared: STA/LTA, Murdock-Hutt and Adaptive Statistically
Optimal Detector. All nuclear tests were detected at EIL, while at MRNI we succeeded only with
the first Pakistan NT, and at JER with the second one. A station detectability is determined by the
maximum of frequency-dependent SNR, therefore the OD proved to be the most efficient
procedure, due to its more accurate adaptation to ambient noise spectra.
Unfortunately, we cannot precisely state detection of the Kazakh explosion at EIL, whereas it was
clearly observed at IMS stations NOA (37.7°) and ILAR (60.6°).
@seismicisolation
@seismicisolation
Table 3: Average spectral discrimination results for Indian and Pakistan tests based on ISN and
BB EIL station records
No. Date Energy ratio Semblance Region, event
(0.6-1)/(1-3 Hz) (0.6-2 Hz)
ISN EIL ISN EIL
1 97/12/04 1.98 0.903 0.840 0.942 Pakistan
earthquake
2 98/01/05 5.39 1.890 0.876 0.879 Pakistan
earthquake
3 98/05/11 - 0.532 - 0.847 India nuclear test
4 98/05/28 1.248 0.588 0.737 0.849 Pakistan nuclear
test
5 98/05/28 1.591 0.666 0.866 0.946 Pakistan
earthquake
Therefore, due to unfavorable propagation conditions for the Asian events, the EIL detectability
may be approximated by the magnitude threshold mb = 4.0-4.5, within a 25°-38° distance range.
Evidently the magnitude threshold for MRNI and JER should be higher for the same type of
events.
The case of the second Pakistan test simulates in a manner the scenario of a clandestine test
masked by two other events: a strong earthquake in Afghanistan and a local earthquake near EIL.
In spite of a poor SNR, the explosion signal was detected by the OD at both IMS stations EIL and
MRNI.
The spectra of the first Pakistan NT exhibited a pronounced spectral null at 1.7 Hz for all three
components of the broadband EIL station, as well as at the vertical component of the short-period
ISN stations. For this ground-truth explosion with a reported shallow source depth (WALLACE,
1998), the interference of P and pP phases seems the most reasonable explanation for this
phenomenon. Other possible reasons, the spallation process (LAY, 1991) and wave scattering from
the complex topography in the source zone, also relate to surface source effects from very shallow
seismic events, i.e. explosions.
However, the spectral null feature, considered separately, cannot serve as a reliable identification
characteristic of nuclear explosions, because not all the tests produce the nulls, whereas some
earthquakes display this effect. The multi-channel spectral discrimination analysis, based on the
spectral ratio of low-to-high frequency energy and semblance of spectral curves, provided a
reliable discrimination between the recent NT and several nearby Pakistan earthquakes. The results
showed compliance with the former analysis of Soviet and Chinese NT recorded at seismic
stations in Israel, where nuclear tests demonstrated lower values of energy ratio and spectral
semblance than earthquakes (GITTERMAN et al., 1999).
@seismicisolation
@seismicisolation
172 Chapter Eight
Figure 11: Application of the semblance and energy ratio discriminants to the recent nuclear tests and nearby
earthquakes (see Table 3): (a) multi-station ISN estimates (large symbols), compared with the results for old Eurasian
events (small symbols) (GITTERMAN et al., 1999); (b) single-station, three-component estimates for the EIL BB
station
It should be noted that both spectral discriminants originally were successfully applied to small
local quarry blasts, underwater explosions, and earthquakes (GITTERMAN et al., 1998), where
the semblance (in the 1-12 Hz band) and the ratio (for 1-3 Hz and 6-8 Hz bands) showed higher
values for explosions, contrasting to the teleseismic case. The higher ratio was explained by path
effects due to local geological settings (thick soft sediments). The higher semblance was related to
specific source features of quarry blasts (ripple-firing) and underwater explosions (bubbling),
producing the azimuth-independent spectral modulation, opposites earthquakes having less spectra
similarity (at local distances) due to radiation pattern and directivity effects.
In some way the spectral energy ratio is similar to the Ms: mb discriminant used by the IDC for
event screening, however it can be applied (as well as the spectral semblance) to small local events
where Ms: mb technique is not relevant, whereas for teleseismic events the spectral discriminants
can provide an additional independent check.
For the recent tests and nearby earthquakes, we observed at the EIL stations consistent delays of P
arrivals of about 4 sec relative to the IASPEI91 model. Slightly shorter delays of about 3.7 seconds
in average were found for old STS explosions recorded at the ISN station MBH situated within a
short distance of EIL. Similar residuals for MBH imply that only a few tenths of this effect are site
related. P-arrival lags 3.5-4 sec were also observed for the IMS station MRNI and BB station JER.
Similar delays at different stations evidence a path- rather than site-effect.
The obtained results will contribute to calibration of the IMS auxiliary stations EIL and MRNI
regarding Asian seismic events. Possibly, events from Africa, Europe and other areas have
different station corrections relative to IASPEI91, and this should be checked when ground-truth
@seismicisolation
@seismicisolation
events from these areas become available at Israel stations.

ACKNOWLEDGEMENTS
We are grateful to William R. Walter for constructive comments resulting in a significant
improvement of the manuscript. The study was supported by the Defense Threat Reduction
Agency of the U.S. Department of Defense under Contract No. DSWA01-97-C-0151, and the
Ministry of the National Infrastructures, Israel.
REFERENCES
BARKER, B., CLARK, M., DAVIS, P., FISK, M., HEDLIN, M., ISRAELSSON, H.,
KHALTURIN, V., KIM, W. Y., MCLAUGHLIN, K., MEADE, C., MURPHY, J., NORTH, R.,
ORCUTT, J., POWELL, C., RICHARDS, P. G., STEAD, R., STEVENS, J., VERNON, F.,
AND WALLACE, T. (1998), Monitoring Nuclear Tests, Science 281, 1967-1968.
CTBTO (1998a), Preliminary IDC reviewed event solution for the announced Pakistan
underground nuclear test 30-May-1998, Preparatory Commission for the CTBTO, Provisional
Technical Secreteriat, Vienna. CTBTO (1998b), Preliminary IDC reviewed event solution for
the announced Kazakhstan Calibration Explosion 22-August-1998, Preparatory Commission
for the CTBTO, Provisional Technical Secreteriat, Vienna.
GITTERMAN, Y., PINSKY, V., and SHAPIRA, A. (1998), Spectral Classification Methods in
Monitoring Small Local Events by the Israel Seismic Network, J. of Seismol. 2, 237-256.
GITTERMAN, Y., PINSKY, V., and SHAPIRA, A. (1999), Spectral Discrimination Analysis of
Eurasian Nuclear Tests and Earthquakes Recorded by the Israel Seismic Network and the
NORESS Array, Physics Earth and Planet. Inter. 113, 111-129.
JOHNSON, C. E. (1979), CEDAR - An Approach to the Computer Automation of Short-period
Local Seismic Networks, Ph.D. Dissertation, California Institute of Technology, Pasadena, CA.
KHALTURIN, V., RAUTIAN, T., and RICHARDS, P. (1998), The Seismic Signal Strength of
Chemical Explosions, Bull. Seismol. Soc. Am. 88, 1511-1524.
KULHANEK, O., (1971), P-wave Amplitude Spectra of Nevada Undergroud Nuclear Explosions,
Pure appl. geophys. 88, 121-136.
KUSHNIR, A. F., LAPSHIN, V. M., PINSKY, V. I., and FYEN, J. (1990), Statistically Optimal
Event Detection Using Small Array Data, Bull. Seismol. Soc. Am. 80, 1934-1947.
LAY, T. (1991), Teleseismic manifestations of pP: Problems and paradoxes. In Explosion Source
Phenomenology, (S.R. Taylor, H.J. Patton, and P.G. Richards, eds.), Geophysical Monograph
65, 109-125.
MURDOCK, N. J and HUTT, C. R. (1983), A New Event Detector Designed for the Seismic
Research Observatories, USGS Open-File Report 83-785, October 1983.
NEIDELL, N. S. and TANER, M. T. (1971), Semblance and Other Coherency Measures for
Multichannel data, Geophysics 36, 482-497.
TAYLOR, S. R. and MARSHALL, P. D. (1991), Spectral Discrimination between Soviet
Explosions and Earthquakes Using Short-period Array Data, Geophys. J. Int. 106, 265-273.
VILA J. (1998), The Broadband Seismic Station CAD (Tunel del Cadi, Eastern Pyrenees): Site
Characteristics and Background Noise, Bull. Seismol. Soc. Am. 88, 297-303.
WALLACE, C. T. (1998), The May 1998 Indian and Pakistan Nuclear Tests, Seism. Res. Lett. 69,
386-393.
WALTER, W. R., RODGERS, A. J., MAYEDA, K., MYERS, S. C., PASYANOS M., and
DENNY, M. (1998), Preliminary Regional Seismic Analysis of Nuclear Explosions and
Earthquakes in Southwest Asia, Proc. 20th Symposium on Monitoring a Comprehensive Test-
Ban Treaty, Santa Fe, 442-451.
@seismicisolation
@seismicisolation
CHAPTER NINE
OVERVIEW AND REVIEW OF MOMENT TENSOR INVERSION FOR NORTH

KOREAN UNDERGROUND NUCLEAR EXPLOSIONS
SO GU KIM
DPRK’s Nuclear Explosion on 03-SEP-2017 recorded on RB8BB ¨ = 92.7°), Raspberry Shake

Seismograph at Saint Louis University, St. Louis, USA by courtesy of Bob Herrmann.
ABSTRACT
Various researchers have studied source mechanisms for the North Korean nuclear tests using
focal mechanisms via moment tensor inversion with various methods of analysis and interpretation.
In this paper several studies utilizing moment tensor inversion are highlighted and additional
relevant studies published in the Geophysical Research Letters are listed in the Appendix
(Vavrycuk and Kim, 2012).
CASE STUDY 1
Seismic waves are radiated by seismic sources and modified along their travel path through the
Earth’s structure. Accurate analysis of a seismic source is reliant on the detailed knowledge of the
Earth’s crustal structure and vice versa. Long-period seismic waveform data (long wavelengths)
are affected by the largescale structure of the earth. Knowledge of these structures makes it
possible to calculate synthetic long-period waveforms. This can determine the explosive character
of a complex seismic source utilizing the whole waveform signal. The moment tensor is a second
order tensor that can pinpoint the seismic source as a point source mathematically. Moment tensor
inversion is a standard tool used for the determination of earthquake source mechanisms on
@seismicisolation
@seismicisolation
Overview and Review of Moment Tensor Inversion for North Korean Underground Nuclear Explosions 175
different scales by minimizing the discrepancy between observed and synthetic waveforms
calculated generally for a 1D Earth model. Barth (2014) used a model3D and found both tests
correlate well with the observed orientation of fault systems in the Northn Korean test region,
which are predominantly NE-SW to NNE-SSW.Moreover provided normal faulting mechanisms
for the 2009 and 2013 North Korean nuclear tests were observed which deviate from many other
studies (see Fig. 1).
The existence of shear energy radiated by nuclear explosions was previously observed. Well-
contained underground explosions which have a high degree of coupling to the surrounding
material may radiate shear waves due to block motions, spall, damage or tectonic triggering (Baker
et al. 2010). Block motion movements on pre-existing faults allow discrete(near-vertical)
displacements along existing joints and faults. In contrast, tectonic triggering may occur due to
pre-stressing of the affected volume possibly by previous nuclear explosions. Small explosions in
water-filled cavities may also radiate shear waves without any damage to the surrounding medium.
In this case the generation of these waves occurs near the explosion source most likely due to
oscillations of an asymmetrical shaped cavity. Patton and Taylor (2011) found that a non-isotropic
radiation pattern for long period waves might be a contribution from material damage in the source
medium. Observed normal faulting mechanisms might be a consequence of slapdown of spalled
layers. The two North Korean nuclear tests in 2009 and 2013 occurred in close vicinity to each
other and were probably well contained and therefore coupled to the surrounding medium since
both of them did not show any significant release of radioisotopes. This study showed that the2013
nuclear explosion released a major shear energy while the 2009 test did not. The obtained strike of
the potential fault planes of the shear energy released for both tests correlateed well with the
observed orientation of the fault systems in the region (predominantly NE-SW to NNE-SSW).
However, the regional stress regime is rather strike-slip with orientations of the maximum
horizontal stress around 70° to 74° east in eastern China as well as in South Korea and up to 90°
East in NE South Korea and is not consistent with the normal faulting mechanism observed for
both nuclear tests. West of this strike-slip regime, a region of normal faulting is apparent in the
Yellow Sea, and this is interpreted as a collision belt between the North and South China blocks,
Kim et al., (2004) explained the normal faulting mechanisms in the northern part of the Yellow
Sea and NW Korea in terms of the counter-clock wise movement of the Amurian Plate in NE Asia.
Another local anomaly is present offshore of eastern South Korea, where a reverse faulting
structure is adjoined, suggesting reverse activation. Pure tectonic triggering which would reflect
both regime and fault orientation can be ruled out. This is supported by the fact that only the 2013
test and not the 2009test released a considerable amount of shear energy as would be expected by a
tectonic release according to Patton (1991).
Cho (2016), Dreger (2018), Liu et al. (2018) and Vavrycuk and Kim (2014) found most of the
North Korean nuclear tests to be reverse faulting mechanisms as illustrated in the following
examples.
It is very unusual to find normal faulting mechanisms in underground nuclear explosions including
the high portion of double couple of 49.7 % for the 2013 test. The detailed mechanism for the
2013 test is explained using moment tensor inversion and waveform modeling at a depth of 2 km
in the Appendix of this chapter.
@seismicisolation
@seismicisolation
176 Chapter Nine
Fig. 1. Source mechanisms by moment tensor inversion for the 2009 and 2013 nuclear tests by Barth (2014). Both
mechanisms represent normal faulting mechanisms (Barth, 2014).
a; North Korea Nuclear Test 09. Date: 25.5.9 Time: 54:43.3, Lat: 41.30 Long: 129.00, Depth: 1.0
Plane (strike/dip/slip): NP1: 36/36/-104, NP2: 233/55/-79 Frequency range (Hz): 0.0200-0.360
Variance= 0.377, Mw = 4.8, Double Couple =27.3 %, Isotropic = 27.3 %, CLVD=21.2 %
b; North Korea Nuclear test 13. Date: 12. 2 13 Time: 2:57:51.4, Lat: 41.30 Long: 129.10, Depth 1.0 Plane
(strike/dip/slip): NP1: 37/43/-88, NP2: 215/47/-91 Frequency range (Hz): 0.0200-0.360
Variance=0.402, Mw=5.0, Double Couple=49.7%, Isotropic=50.2 %, CLVD=0.1 % (?)
CASE STUDY 2
Shin et al. (2010) carried out focal mechanism by moment tensor inversion studies for the 2009
nuclear test. Full moment tensor inversion of full waveform data shows that the 2009 test had a
very large isotropic component. Pure isotropic moment tensor inversion also resulted in good
recovery of observed waveforms with clear indication that the 2009 test was explosive in origin
(Fig. 2).
@seismicisolation
@seismicisolation
Fig. 2. Comparison of full moment tensor inversion (a) and isotropic moment tensor inversion (b) for the may 25,
2009 nuclear test of North Korea. Solid and dashed lines represent observed and theoretical waveforms, respectively.
Seismograms show negligible tangential components. The isotropic component of (a) is significant and variance
reduction is almost the same for both cases. Strike, dip and rake do not have physical meanings due to the nominal
double-couple component in the explosion.
Although the focal depth of an earthquake can be estimated by choosing the solution based on the
goodness of ¿t in the range of possible depths, the focal depth of an explosion cannot be resolved
accurately via moment tensor inversion (Ford et al. 2009). Moment tensor inversions for the 2009
test with source depths at every 0.2 km from 0.2 km to 4 km. However, the inversions with source
depths between 0.2 and 3.4 km did not give notable variations and source depth was estimated at
1.0 km. Figure. 2 shows the results of full moment tensor and isotropic moment tensor inversions
at the source depth of 1.0 km. Theoretical wave forms from both inversions closely correlate
observed waveforms suggesting that the elements of the moment tensor were recovered well. The
average VR from full moment tensor inversion is larger than that from isotropic moment tensor
inversion. This is because a small amount of the non-isotropic component of the 2009 test can be
@seismicisolation
@seismicisolation
178 Chapter Nine
utilized for the full moment tensor inversion (Figure 2). Full moment tensor inversion of full
waveform data shows that the 2009 test had a very large isotropic component.
CASE STUDY 3
The results of moment tensor inversion for the events were acquired by a modified TDMT (Cho,
2016) with a frequency band of 0.04~0.08Hz (see Fig. 3). The moment magnitude of the first,
second, the third, fourth and fifth in the same depth of 0.7 km is about 4.0, 4.4, 4.6, 4.6 and 4.9
respectively. The fifth event has higher amplitude of the vertical component than the others. The
data also showed the tangential component. It may be caused by the tectonic stress, topography
effect, inhomogeneous structure near source or anomalous chemical effect including configuration
of source. The last nuclear explosion of the sixth event will be discussed in details in Case Study 4
of this chapter. The moment inversion provides evidence of artificial explosions of the nuclear
tests because of the high isotropic component. The deeper the depth, the higher the isotropic (ISO)
and the CLVD components are, but the lower the double-couple (DC) is. Moment tensor inversion
was performed up to 2 km of depth and it does not significantly affect the parameters of the MT
solution (Fig. 4). The double component gives a reverse faulting due to tectonic stress. The psedo-
horizontal direction of the double couple component correlates with the horizontal tectonic stress
direction (NNE-SSW) of the Korean Peninsula.
a
@seismicisolation
@seismicisolation
@seismicisolation
@seismicisolation
180 Chapter Nine
Fig. 3. Focal mechanisms by moment tensor inversion for the 2006 (a), 2009 (b), 2013 (c), 2016J (d), 2016S (e) by
Cho (2016). Only the 2006 test represents a reverse faulting mechanism which correlates well with the spectral hole of
the fundamental-mode Rayleigh wave (Rg wave) amplitude spectra for both synthetics and observations (2018a, Kim
et al., ; 2018b, Kim et al.) and the rest present an oblique reverse type
@seismicisolation
@seismicisolation
@seismicisolation
@seismicisolation
182 Chapter Nine
Fig. 4. Moment tensor inversion for the 2006, 2009 and 2013 North Korean nuclear tests at a depth of 2.3 km (Cho,
2015). No significant discrepancy from the moment tensor inversion at a depth of 0.7 km.
The frequency band used for moment tensor inversion was used with 0.04-0.08 Hz. However,
Vavrycuk and Kim (2013) presented an oblique-reverse faulting mechanism from waveform
modeling for the 2013 test and observed that the mechanism by waveform modeling with the
synthetics at a depth of 2.0 km rather than at a depth of 1.0 km using particle motions of surface
waves (see Appendix).
CASE STUDY 4
Moment tensor inversions for nuclear explosions cannot be accurately determined from surface
wave records because the wavelength is much larger than the source depth and the major
mechanism is activated without faulting in motion characteristics. Thus, various moment tensor
inversions were observed for the North Korean nuclear tests from multiple researchers (Barth,
2014; Cho et al.,2016; Dreger, 2018). The 2017 event moment tennsor inversion was obtained
from IRIS (Fig. 5)
@seismicisolation
@seismicisolation
Fig. 5. The moment tensor inversion for the 6th nuclear test of DPRK on September 3, 2017 from IRIS
dreger_dprk09032017_moment_tensor. We can see the moment magnitude for total moment (MW 5.2) is much larger
than that of the pure explosion (MW 4.95). One can find more details in https://ds.iris.edu/.../2017/09/03/2017-north-
korean-nuclear-test. The moment tensor inversion was analyzed assuming that the depth is about 1.0 km, resulting in
an oblique reverse faulting mechanism which is similar to that of the 2013 test by 9DYU\þXN & Kim (2014) (see
Appendix).
Unlike earthquake inversions, full moment tensor inversions of an explosive source cannot
constrain the source depth by comparing fits at different depths at the frequencies examined. Event
locations put the source at less than 1 km so the results discussed above assume a source depth of
600 m. However, the isotropic component and total moment are dependent on the assumed depth.
Ford et al. (2009b) showed that though the source-type is robust, the total moment trades-off with
the assumed depth and the moment goes into the vertical elements of the moment tensor as the
depth decreases. Since the Green’s functions for these elements diminish at the limit as the depth
of the source decreases due to vanishing traction at the free surface.
@seismicisolation
@seismicisolation
184 Chapter Nine
A series of source inversions were performed for the 25 May 2009 (Memorial Day) North Korean
seismic event using intermediate period (10–50 s) complete waveform modeling (Ford et al..
2009b). An earthquake source is inconsistent with the data and the best-fit full seismic moment
tensor is predominantly explosive (~60%) with a moment magnitude (MW) of 4.5. A pure
explosion solution yields a scalar seismic moment of 1.8 x 1022 dyne-cm (MW4.1) and fits the data
almost as well as the full solution. The difference between the full and explosion solutions is the
predicted fit to observed tangential displacement which requires some type of non-isotropic (non-
explosive) radiation. Possible causes of the tangential displacement are additional tectonic sources,
tensile failure at depth and anisotropic wave propagation. Similar displacements may be hidden in
the noise of the 2006 test. Modeling of intermediate period and regional distance waveforms
identifies the Memorial Day event in Kimchaek, North Korea as decidedly non-tectonic with the
best-fit model dominated by an explosion source. There are Love waves observed at several
stations indicating that the source may have some non-isotropic component.
CASE STUDY 5
Five tests were conducted at North Korea’s Punggye-ri test site in 2006, 2009, 2013, and 2016
(twice). Their body wave magnitude (mb), reported by the United States Geological Survey
(USGS), increased from 4.3 in 2006 to 5.3 in 2016. Relative methods, including waveform
interferometry (Murphy et al., 2013; Wen & Long, 2010; Zhang & Wen, 2013; Zhao et al., 2017)
located the 2006–2016 events within an ~5 × 5-km region. Seismic estimates of burial depths
(<800 m) were supported by satellite images of the entrance tunnels under Mt. Mantap (Murphy et
al., 2013). The explosion yield estimate which depends strongly on depth remained challenging to
accurately assess (Koper et al., 2008). The full moment tensor (MT) inversion of the 2009 test
revealed a non-double-couple source (e.g., Ford et al., 2009b). Nonisotropic radiation, including
shear motion, was observed in the 2013 test (Barth, 2014; 9DYU\þXN & Kim, 2014). The 2016 tests
helped to delimit a VLJQL¿FDQW uncertainty in the source parameters (Cesca et al., 2017). The
uncertainty is due to poor resolution of the depth and certain moment-tensor components when a
shallow nondouble-couple source is studied at long wavelengths (e.g., Bukchin et al., 2010; Henry
et al., 2002). It yields VLJQL¿FDQW trade-offs between moment magnitude Mw and the moment-
tensor parts (double couple [DC], compensated linear vector dipole [CLVD], and isotropic [ISO]).
Fortunately, resolvability of the source type (explosion, implosion, and crack) appears to be less
affected, mainly if combining waveform data with ¿rst-motion polarities (Chiang et al., 2016). In
this paper the resolution is studied with a Bayesian approach (Hallo et al., 2017; 9DFNiĜ et al.,
2017), thus, providing probability density functions (PDFs) of the source parameters.
Many nuclear tests in the past were accompanied by an underground collapse (e.g., Springer et al.,
2002) but only a few studies of related seismic events were published, mostly before the 1980s.
For example, a sequence of weak events (mb < 3.8) following the explosion at Amchitka were
observed eventually terminating with a large complex shock (mb 4.9, 38 hr after the ~5 Mt nuclear
test CANNIKIN) (Engdahl, 1972). The sequence was due to an explosion generated, progressively
deteriorating cavity, while the largest shock was attributed to a major, complete collapse,
concurrent with surface subsidence. We study a similar event, as large as ML 4.1 but occurring
shortly (8.5 min) after the 2017 test in North Korea.
On 3 September 2017, at 03:30:01 UTC, the thus-far strongest North Korean test of mb 6.3 (USGS)
was detonated near the previous explosions. At 03:38:31, an event of ML4.1 (USGS) was reported
at the test site. Source processes of the initial explosion (mainshock) and its early ML4.1
aftershock hereafter referred to as Events 1 and 2, respectively. Focus predominantly on Event 2
with its crack-closing nature without any shear-slip signature. A deep insight into Event 2 is
possible thanks to unique data from nearby broadband seismic stations. At more distant stations,
@seismicisolation
@seismicisolation
Event 2 is obscured by noise and seismic coda of Event 1. There was also a tectonic aftershock
(Event 3) that occurred 20 days after the 2017 test (Fig. 6).
Fig. 6a. Uncertainty of nodal lines for mainshock (a) and non-tectonic aftershock (b). The uncertainty source depth
0.5-2.5 km is considered. Physical meaning of these DC-nodal lines is highly limited due to the very small DC part of
the moment tensor. Symbols T and P denote the mean tension and Pressure axes implying reverse and normal faulting
mechanisms respectively.
Fig.6b. (a) The mainshock (star) is forward simulated at 10 stations in Jpan (triangles). (b) The observed and synthetic
displacement data are shown in black and red, respectively; Focal mechanism by moment tensor inversion using
seismic stations for DPRK’s 6th nuclear test on September 3, 2017 (Liu et al., 2018). From S3, the mainshock of the
nuclear test is a reverse faulting mechanism whereas the aftershock of the nontectonic event is a normal faulting
mechanism.
As you see in Fig.6, the focal mechanisms of moment tensor inversion for the nuclear test and the
second nontectonic event are found to be a reverse faulting mechanism and a normal faulting
mechanism, respectively. The 2017 source mechanism is similar to the tests of 2016 (MW 5.1 and
MW 5.2), but it has a much higher magnitude of MW 5.8. In contrast, the DPRK nuclear test of
2013 (MW 5.0) showed a VLJQL¿FDQWO\ larger DC component in the source mechanism. In particular,
most of researchers have confused to identify whether or not the second event is an artificial or
natural event when it occurred shortly (8.5@seismicisolation
min) after the 2017 test in North Korea at that time.
@seismicisolation
186 Chapter Nine
Moment magnitude Mw also trades off with depth, decreasing from 6 to 5.6 over the examined
depth range of 0.5–2.5 km. At an estimated 1.5-km depth, which might perhaps be preferred, in
analogy with the seismic array study of the 2016 test (Cesca et al., 2017), the Mw 5.7 is obtained;
the full MT is given as ISO part (ISO ~55–60%), together with CLVD (CLVD ~30–45%) and
clearly dominates DC part (DC ~0–10%) as a small portion.
CASE STUDY 6
Wang et al. (2018) carried out an extensive study titled “Accurate source location from waves
scattered by surface topography.” The solution remains relatively robust despite a wide range of
random noise in data, unmodeled random velocity heterogeneities, and uncertainties in moment
tensors. This method can be applied to locate pairs of sources in close proximity by utilizing
differential waveforms using source-receiver reciprocity further reducing errors caused by
unmodeled velocity structures. Using synthetic experiments, Wang et al. (2018) demonstrated a
novel approach of using waves scattered by the surface topography to locate shallow seismic
sources. Different from traditional methods using travel time information based on ray theory for
earthquake location, the waveform information is used to determine hypocenter locations. For
computational HI¿FLHQF\ precalculation of SGTs from the receivers and apply the grid search
method using source-receiver reciprocity. The best solution is offset from the real source location
especially in the z direction when only P wave data are used but its accuracy is improved and
uncertainty is reduced when P coda or both P and P coda waves are considered. This synthetic
study will be followed by the application of this new method to real seismic events (e.g., nuclear
explosions and mining activities) in places with VXI¿FLHQW topographic variations. While the
application of this method requires well-constrained topographic data, these data are usually
available at the scales of the wavelengths of regional seismic waves. Nevertheless, care should be
taken to ensure that topography resolution is adequate for the frequencies of waves used (i.e., a
coarse/ smoothened topography will not be adequate for accurate simulation of high-frequency
waves). Care should also be taken in the selection of the coda wave window as the topography-
scattered waves usually dominate immediately after the direct arrival, while the heterogeneity-
scattered waves are strong in the later part of the coda. Since scattering by surface topography
strengthens at higher frequencies (Rodgers et al., 2010), it may be necessary to use ¿ner grids.
This method can be applied to sources other than explosions, such as earthquakes as long as there
are separate moment tensor solutions from the inversion of longer period body and/or surface
waves. In this study, the effect of scattering by surface topography is dominant over those of
random noise or unmodeled velocity heterogeneities. In application, this will depend on the
strength of the signal, topographic variations, depth of the source, and magnitude of velocity
heterogeneity (Fig. 7).
As the topography Àattens, the source occurs deeper, the velocity heterogeneity becomes larger,
and the dominance of surface scattering weakens. Therefore, the application of scattered waves to
real situations must be evaluated on a case-to-case basis. However, it was found that azimuth-
averaged spectra for flat and various topographic models are almost the same (Rodgers et al., 2010;
Lay, et al., 2011). Consequently, depth estimates (Kim et al., 2018a and Kim et al., 2018b) using
depth phases (pP and sP/pPn and sPn) from azimuth averaged spectra can be used to adopt the
appropriate approach to estimate the depth of a nonlinear source in the nonlinear topographic
effects in the presence of rough surface topography around Mt. Mantap for the North Korean
nuclear tests.
@seismicisolation
@seismicisolation
Fig. 7. Full-wave form modeling can be also used to determine epicenter locations for 2016J, 2016S and 2017 nuclear
tests of North Korea in addition to focal mechanism study.

It is generally not easy to obtain accurate focal mechanism estimates via moment tensor inversion
for underground nuclear explosions. There are multiple interpretations and findings utilizing
moment tensor inversion for the North Korean nuclear tests. In moment tensor inversion for
nuclear tests, one can see the isotropic corresponding to changes in volume (ISO), a sudden change
in the shear modulus, a compensated linear vector dipole (CLVD) and finally double couple (DC).
CLVD plus DC corresponds to a shear fracture resulting in deviatoric stress which cannot be
neglected in an underground nuclear explosion due to Earth structure and source parameters even
if it is supposed to be ideally isotropic. Most of researchers assumed that the source depth would
be at around 700-1000 m for the moment tensor inversion analysis, however, 9DYU\þXN and Kim
(2013) used a source depth of about 2 km which led to closer resolution in terms of polarization of
surface waves for the 2013 test. Nevertheless, utilizing the azimuth averaged spectra of depth
phases was a pragmatic way to calculate the depth of a nonlinear source in a nonlinear topographic
@seismicisolation
@seismicisolation
188 Chapter Nine
region such as Mt. Mantap where nuclear tests were conducted. The methodology for the azimuth
averaged spectra for depth phases are explained in detail in previous studies (Kim et al., 2018 and
Kim et al,. 2019) including further research by BGR showing body wave magnitudes for 6 North
North Korean nuclear tests (Fig. 8).
.
Fig. 8. Body wave magnitudes of 6 DPRK’s nuclear tests determined using GERES Array by Federal Institute for
Geosciences and Natural Resources (BGR), Hannover, Germany.
REFERENCES
Baker, G. E., Xu, H., & Stevens, J. L. (2010) Generation of shear waves from explosions in water-
filled cavities. Bulletin of the Seismological Society of America. 100, 1196–1210.
Barth, A. (2014). 6LJQL¿FDQW release of shear energy of the North Korean nuclear test on February
12, 2013. Journal of Seismology, 18(3), 605–615. https://doi.org/10.1007/s10950-014-9431-6
Bukchin, B., Clévédé, E., & Mostinskiy, A. (2010). Uncertainty of moment tensor determination
from surface wave analysis for shallow earthquakes. Journal of Seismology, 14(3), 601–614.
https://doi.org/10.1007/s10950-009-9185-8
Cesca, S., Heimann, S., Kriegerowski, M., Saul, J., & Dahm, T. (2017). Moment tensor inversion
for nuclear explosions: What can we learn from the 6 January and 9 September 2016 nuclear
tests, North Korea? Seismological Research Letters, 88(2A), 300–310. https://doi.org/10.1785/
0220160139.
Chiang, A., Dreger, D. S., Ford, S. R., Walter, W. R., & Yoo, S. (2016). Moment tensor analysis
of very shallow sources. Bulletin of the Seismological Society of America, 106(6), 2436–2449.
https://doi.org/10.1785/0120150233.
Cho, C. S (2015). Moment tensor inversion application for the DPRK’s 2006, 2009 and 2013
nuclear tests at a depth of 2.3 km. Unpublished publication and personal communication.
@seismicisolation
@seismicisolation
Cho, C, Shin, J. S., Kim, G. (2016). Comparison of results of relative location methods and
moment tensor inversion for the nuclear explosions experimented in North Korea, S31A-2722,
2016AGU Fall Meeting, San Francisco, pp. 10–12.
Coutant, O. (1990). Program of numerical simulation AXITRA, Lab. Géophysique Interne
Tectonophys. Report, Univ. Joseph Fourie, Paris.
Day, S. M., & McLaughlin, K. L. (1991). Seismic source representations for spall. Bulletin of the
Seismological Society of America, 81(1), 191–201.
Engdahl, E. R. (1972). Seismic effects of the MILROW and CANNIKIN nuclear explosions.
Bulletin of the Seismological Society of America, 62(6), 1411–1423.
Ford, S. R., Dreger, D. S., & Walter, W. R. (2009a). Identifying isotropic events using a regional
moment tensor inversion. Journal of Geophysical Research, 114, B01306.
https://doi.org/10.1029/2008JB005743
Ford, S. R., Dreger, D. S., & Walter, W. R. (2009b). Source analysis of the Memorial Day
explosion, Kimchaek, North Korea. Geophysical Research Letters, 36, L21304.
https://doi.org/10.1029/2009GL040003
Henry, C., Woodhouse, J. H., & Das, S. (2002). Stability of earthquake moment tensor inversions:
Effect of the double-couple constraint. Tectonophysics, 356(1–3), 115–124.
https://doi.org/10.1016/S0040-1951(02)00379-7
Julian, B. R., Miller, A. D., & Foulger, G. R. (1998). Non-double earthquakes, 1. Theory. Reviews
of Geophysics, 36(4), 525–549. https://doi.org/ 10.1029/98RG00716.
Kim, S. G. (2018). Surveys in Solid – Earth Geophysics, Munundang, pp479.
Kim, S. G., Lkhasuren, D. & Park, P. H. (2004). The low seismic activity of the Korean Peninsula
surrounded by high earthquake countries, Journal of Seismology, 8, 91-103.
Kim, S. G., Gitterman, Y. & Lee. S. (2018a). Depth estimate of the DPRK's 2006-10-09, 2009-05-
25 and 2013-02-12 underground nuclear tests using local and teleseismic arrays, Journal of
Asian Earth Sciences, 163, 249-263.
Kim, S. G., Gitterman, Y. & Lee. S. (2018b). Depth calculation for the January 06, 2016, the
September 09, 2016 and the September 03, 2017 nuclear tests of North Korea from detailed
depth phases using regional and teleseismic arrays, Pure and Applied Geophysics,
https://doi.org/10.1007/s00024-018-1958-y.
Koper, K. D., Herrmann, R. B., & Benz, H. M. (2008). Overview of open seismic data from the
North Korean event of 9 October 2006. Seismological Research Letters, 79(2), 178–185.
https://doi.org/10.1785/gssrl.79.2.178.
Lay, T., Avants, M., Xie, X-B., & Rodgers, A. J. (2011). Effects of 3D surface topography on
regional and teleseismic signals from underground explosions. 2011 Monitoring Research
Review: Ground-Based Nuclear Explosion Monitoring Technologies.
Liu, J. Li, L., Zahradník, J., Sokos, E., Liu, C., & Tian, X. (2018). North Korea’s 2017 test and its
nontectonic aftershock, Geophysical Research Letters, 10.1002/2018GL077095.
Masse, R. P. (1981). Review seismic source models for undergrouns nuclear explosions, Bulletin
of the Seismological Society of America, 71, 1249-1268.
Murphy, J. R., Stevens, J. L., Kohl, B. C., & Bennett, T. J. (2013). Advanced seismic analyses of
the source characteristics of the 2006 and 2009 North Korean nuclear tests. Bulletin of the
Seismological Society of America, 103(3), 1640–1661. https://doi.org/10.1785/0120120194
Patton, H. J., & Taylor, S. R. (2008). Effects of shock-induced tensile failure on mb-Ms
discrimination: Contrasts between historic nuclear explosions and the North Korean test of 9
October 2006. Geophysical Research Letters, 35, L14301. https://doi.org/10.1029/
2008GL034211
Patton, H. J., & Taylor, S. R. (2011). The apparent explosion moment: Inferences of volumetric
moment due to source medium damage by underground nuclear explosions. Journal of
Geophysical Research, 116, B03310. https://doi.org/10.1029/2010JB007937.
@seismicisolation
@seismicisolation
190 Chapter Nine
Rodgers, A. J., Petersson, N. A., & Sjogreen, B. (2010), Simulation of topographic effects on
seismic waves from shallow explosion snear the North Korean nuclear test site with emphasis
on shear wave generation, J. Geophys. Res., 115, B11309, doi:10.1029/2010JB007707.
Shin, J. S., Sheen, D. & Kim, G. (2010). Regional observations of the second North Korean
nuclear test on 2009 May 25. Geophys. J. Int., 180, 243-250.
Toksoez, M. N., Thomson, K. C., & Ahrens, T. J. (1971). Generation of seismic waves by
explosions in prestressed media. Bulletin of the Seismological Society of America, 61(6),
1589–1623.
Tompson, A. F. B., Bruton, C. J., Pawloski, G. A., Smith, D. K., Bourcier, W. L., Shumaker, D. E.,
Kersting, A., Carle, S. & Maxwell, R. (2002). On the evaluation of groundwater contamination
from underground nuclear tests. Environmental Geology, 42(2–3), 235–247.
https://doi.org/10.1007/s00254-001-0493-8.
9DFNiĜ J., Burjánek, J., *DOORYLþ F., Zahradník, J., & Clinton, J. (2017). Bayesian ISOLA: New
tool for automated centroid moment tensor inversion. Geophysical Journal International, 210(2),
693–705. https://doi.org/10.1093/gji/ggx158
9DYU\þXN V. (2011). Tensile earthquakes: Theory, modeling, and inversion. Journal of
Geophysical Research, 116, B12320. https://doi.org/ 10.1029/2011JB008770
9DYU\þXN V., & Kim, S. G. (2014). Nonisotropic radiation of the 2013 North Korean nuclear
explosion. Geophysical Research Letters, 41 (20), 7048–7056.
https://doi.org/10.1002/2014GL061265.
Wang, Shen, Flinders, & Zhang (2018). Accurate source location from seismic waves scattered by
surface topography, J. Geophys. Res. Solid Earth, 125, 4538-4554, doi: 10.1002/2016JB012814.
Wen, L., & Long, H. (2010). High-precision location of North Korea’s 2009 nuclear test.
Seismological Research Letters, 81(1), 26–29. https://doi. org/10.1785/gssrl.81.1.26
Willis, D. E. (1963). Comparison of seismic waves generated by different types of source. Bulletin
of the Seismological Society of America, 53(5), 965–978.T
Zhang, M., & Wen, L. (2013). High-precision location and yield of North Korea’s 2013 nuclear
test. Geophysical Research Letters, 40, 2941–2946. https://doi.org/10.1002/grl.50607
Zhao, L., Xie, X., Wang, W., Fan, N., Zhao, X., & Yao, Z. (2017). The 9 September 2016 North
Korean underground nuclear test. Bulletin of the Seismological Society of America, 107(6),
3044–3051. https://doi.org/10.1785/0120160355
@seismicisolation
@seismicisolation
Appendix
NONISOTROPIC RADIATION OF THE 2013 NORTH KOREAN NUCLEAR
EXPLOSION
Václav 9DYU\þXN and So Gu Kim
Geophys. Res. Lett. (2014). 41 (20), 7048-7064
@seismicisolation
@seismicisolation
192 Chapter Nine
@seismicisolation
@seismicisolation
@seismicisolation
@seismicisolation
194 Chapter Nine
@seismicisolation
@seismicisolation
@seismicisolation
@seismicisolation
196 Chapter Nine
@seismicisolation
@seismicisolation
@seismicisolation
@seismicisolation
198 Chapter Nine
@seismicisolation
@seismicisolation
@seismicisolation
@seismicisolation
200 Chapter Nine
Auxiliary Material for

Non-isotropic radiation of the 2013 North Korean nuclear explosion
9iFODY9DYU\þXN1, So Gu Kim2
1
1nstitute of Geophysics, Academy of Sciences, Prague, Czech Republic,
2
Korea Seismological Institute, Goyang 410734, Republic of Korea
Geophysical Research Letters, 2014
Introduction
The auxiliary material contains two tables and three figures stored in the following files:
* ‘Table S1.pdf’ with the continental crustal velocity model

* ‘Table S2.pdf’ with the oceanic crustal velocity model
* ‘Figure S1a.pdf’ showing the fit between waveforms and synthetics in the moment tensor
inversion for the explosion depth of 2 km
* ‘Figure S1b.pdf’ showing the fit between waveforms and synthetics in the moment tensor
inversion for the explosion depth of 2 km for the rest of stations
* ‘Figure S2a.pdf’ showing the fit between waveforms and synthetics in the moment tensor
inversion for the explosion depth of 1 km
* ‘Figure S2b.pdf’ showing the fit between waveforms and synthetics in the moment tensor
inversion for the explosion depth of 1 km for the rest of stations
* ‘Figure S3.pdf’ showing the focal mechanisms obtained by the waveform inversion and the fit
between the observed and synthetic ratios of the maximum amplitudes of the transverse (T) and
radial (R) components of the surface waves.
Description of tables:
1. ‘Table S1.pdf’ Velocity model with the continental crust
1.1 Column “Depth”, km, depth of the top of the layer.
1.2 Column “Vp”, km/s, P-wave velocity.
1.3 Column “Vs”, km/s, S-wave velocity.
1.4 Column “Density”, g/cm3, density of the medium.
1.5 Column Qp’, quality factor of the P waves.
1.6 Column Qs’, quality factor of the S waves.
2. ‘Table S2.pdf’ Velocity model with the oceanic crust

2.1 Column “Depth”, km, depth of the top of the layer.
2.2 Column “Vp”, km/s, P-wave velocity.
2.3 Column “Vs”, km/s, S-wave velocity.
2.4 Column “Density”, g/cm3, density of the medium.
2.5 Column ‘Qp’, quality factor of the P waves.
2.6 Column ‘Qs’, quality factor of the S waves.
Table S1. Velocity model with the continental crust

@seismicisolation
@seismicisolation
Depth [km] Vp [km/s] Vs [km/s] Density [g/cm3] Qp Qs

0.0 4.50 2.56 2.10 200 100
0.5 5.30 3.12 2.44 300 150
1.3 5.70 3.29 2.64 300 150
6.0 6.01 3.41 2.79 400 200
18.0 6.32 3.59 2.86 600 300
24.0 6.73 3.82 3.04 1000 500
32.0 8.04 4.63 3.24 1000 500
40.0 8.15 4.71 3.24 1450 725
80.0 8.30 4.77 3.14 1450 725
Qp and Qs are the P- and S-wave quality factors.
Table S1. Velocity model with the continental crust

Depth [km] Vp [km/s] Vs [km/s] Density [g/cm3] Qp Qs
0.0 4.50 2.56 2.10 200 100
0.5 5.30 3.12 2.44 300 150
1.3 5.70 3.29 2.64 300 150
6.0 6.01 3.41 2.79 400 200
18.0 6.32 3.59 2.86 600 300
24.0 6.73 3.82 3.04 1000 500
32.0 8.04 4.63 3.24 1000 500
40.0 8.15 4.71 3.24 1450 725
80.0 8.30 4.77 3.14 1450 725
Qp and Qs are the P- and S-wave quality factors.
@seismicisolation
@seismicisolation
202 Chapter Nine
@seismicisolation
@seismicisolation
@seismicisolation
@seismicisolation
204 Chapter Nine
@seismicisolation
@seismicisolation
@seismicisolation
@seismicisolation
206 Chapter Nine
@seismicisolation
@seismicisolation
CHAPTER TEN
GT0 EXPLOSION SOURCES FOR IMS INFRASOUND CALIBRATION:

CHARGE DESIGN AND YIELD ESTIMATION FROM NEAR-SOURCE
OBSERVATIONS 1
Y. GITTERMAN AND R. HOFSTETTER
On-surface calibration explosion of 82 tons TNT-like explosives at Sayarim Military Range on 26

August 2009; a snapshot (after 0.9 s) from a speed video-recording at distance 755 m (video by the
Experiment Division of the Israel Defense Forces).
ABSTRACT
Three large-scale on-surface explosions were conducted by the Geophysical Institute of Israel
(GII) at the Sayarim Military Range, Negev desert, Israel: about 82 tons of strong high explosives
in August 2009, and two explosions of about 10 and 100 tons of ANFO explosives in January
2011. It was a collaborative effort between Israel, CTBTO, USA and several European countries,
with the main goal to provide fully controlled ground truth (GT0) infrasound sources, monitored
by extensive observations, for calibration of International Monitoring System (IMS) infrasound
stations in Europe, Middle East and Asia. In all shots, the explosives were assembled like a
pyramid/hemisphere on dry desert alluvium, with a complicated explosion design, different from
the ideal homogenous hemisphere used in similar experiments in the past. Strong boosters and an
upward charge detonation scheme were applied to provide more energy radiated to the
atmosphere. Under these conditions the evaluation of the actual explosion yield, an important
source parameter, is crucial for the GT0 calibration experiment. Audio-visual, air-shock and
1
Pure and Applied Geophysics (2014), 171, 599–619. First Online: 06 September 2012,
doi.org/10.1007/s00024-012-0575-4. @seismicisolation
@seismicisolation
208 Chapter Ten
acoustic records were utilized for interpretation of observed unique blast effects, and for
determination of blast wave parameters suited for yield estimation and the associated relationships.
High-pressure gauges were deployed at 100–600 m to record air-blast properties, evaluate the
efficiency of the charge design and energy generation, and provide a reliable estimation of the
charge yield. The yield estimators, based on empirical scaled relations for well-known basic air-
blast parameters—the peak pressure, impulse and positive phase duration, as well as on the crater
dimensions and seismic magnitudes, were analyzed. A novel empirical scaled relationship for the
little- known secondary shock delay was developed, consistent for broad ranges of ANFO charges
and distances, which facilitates using this stable and reliable air-blast parameter as a new potential
yield estimator. The delay data of the 2009 shot with IMI explosives, characterized by much
higher detonation velocity, are clearly separated from ANFO data, thus indicating a dependence on
explosive type. This unique dual Sayarim explosion experiment (August 2009/January 2011), with
the strongest GT0 sources since the establishment of the IMS network, clearly demonstrated the
most favorable westward/eastward infrasound propagation up to 3,400/6,250 km according to
appropriate summer/winter weather pattern and stratospheric wind directions, respectively, and
thus verified empirically common models of infrasound propagation in the atmosphere.
Key words: Controlled surface chemical explosion, IMS infrasound station calibration, yield
estimation, airblast secondary shock delay, infrasound propagation.
INTRODUCTION
The International Monitoring System (IMS) comprises an infrasound network (currently of 60
stations) for nuclear tests monitoring via recording of low-frequency acoustic waves emitted from
remote sources placed on the Earth surface or in the atmosphere (e.g., CHRISTIE et al., 2001). To
improve monitoring of explosion sources, i.e., detection, identification, location and yield
estimation, the infrasound stations should be calibrated. The best procedure includes fully
controlled large-scale on- surface explosions producing strong acoustic signals that can be
observed at large distances. A well-established calibration experiment should include
measurements of ground truth zero (GT0) information: detonation time, accurate GPS coordinates,
precise TNT equivalent yield, and extensive seismic and acoustic observations at near-source zone
and also at local/regional distances.
A number of large nuclear and chemical (up to 8 kT) on-surface and near-surface tests were con-
ducted in the 1970–1990’s, many of them at White Sands Military Range (WSMR) (Website
http:// GlobalSecurity.org). Unfortunately, all of them were conducted prior to the establishment of
IMS infrasound stations, starting in 1996. In the last decade, a very few controlled explosions
were conducted providing well-documented observations of infrasound waves at large distances
beyond 1,500 km (CHRISTIE, 2005), which were recorded at IMS stations. The remarkable
Watusi experiment was conducted recently at the Nevada Test Site (BHATTACHARYYA et al.,
2003), where the high explosives (HE) charge (19 tons of TNT equivalent) was detonated in a
cylindrical container that was partially above the ground level. The acoustic signals of this
relatively small explosion were observed at several IMS infrasound stations up to 2,165 km. Two
controlled ground-level explosions (5 and 27 tons of ammunition) were conducted at the Woomera
test facility in Australia, where clear infrasound signals were recorded at 470–1,260 km at
different azimuths (BROWN et al., 2003).
Strong surface explosions in NW Russia (presumably due to demolition of old ammunition) were
reported in a recent NORSAR study (RINGDAL, 2005). Clear acoustic signals were observed at
both infrasound and seismic sensors of Apatity and ARCES arrays placed at ~ 250 km from the
source. Combined seismic and infrasound signal processing showed that location estimations
based on infrasonic detections match closely the standard seismic data locations, but a joint
seismic/infrasound location procedure implies assigning (to the distinct infrasound arrivals dataset)
of reliable weights which cannot be chosen without GT0 events. Bolide explosions were used to
@seismicisolation
@seismicisolation
GT0 Explosion Sources for IMS Infrasound Calibration 209
study the capabilities and limitations of source location procedures based on infrasound travel
times and azimuth deviations derived from ray tracing formulations (GARCES et al., 2002).
Seismic and infrasound data from construction explosions near Seoul were jointly used to produce
a combined event location (STUMP et al., 2002). However, the location accuracy of used
procedures in all these cases cannot be reliably estimated without GT0 explosions.
The propagation features of infrasonic waves from atmospheric explosions with yields <1 kT in
the distance range of more than 1,500 km are still poorly understood. Such experiments may
provide considerable insight into the detection capability. Until recently no GT0 explosions,
relevant for IMS infrasound calibration, and no IMS stations recorded signals from GT0 sources
were presented anywhere in the Eastern Mediterranean/Middle East (EM–ME) region.
To improve infrasound monitoring in the EM–ME region, a number of large-scale surface
controlled explosions were conducted in August 2009 and January 2011 by the Geophysical
Institute of Israel (GII) in different weather and wind conditions, which were supported by the US
Army Space and Missile Defence Command (SMDC) and PTS CTBTO (GITTERMAN, 2010,
2011). The collaborative efforts of several organizations provided this creative, cost- efficient
venture, improving monitoring and verification of the Comprehensive Test Ban Treaty (CTBT) in
the region. The explosions were realized at Sayarim Military Range (SMR), Negev desert, in the
site of the regular demolition of outdated ammunition by Israel Defense Forces (IDF).
The main goals of the explosion experiments - recording infrasound signals at several IMS
stations, and experimental demonstration of seasonal (summer/winter) variation in infrasound
propagation at far-regional distances - were reached. Dense seismo-acoustic network of portable
and permanent stations provided good datasets at near-field and local ranges. We analyze here air-
shock and acoustic records aiming to understand observed unique blast effects, determining wave
parameters that are better suited for explosive yield estimation, and deriving empirical relations
between the yield and the relevant wave parameters.
This paper covers analysis and interpretation of the explosion parameters and near field recordings
and provides a reference to other papers devoted to Sayarim experimental explosions, in-
particular: long distance propagation and modeling of the infrasound signals (FEE, 2012), and
seismo-acoustic energy partitioning at near-source and local distances (BONNER, 2012).
CHARGE DESIGN
Major design conceptions, elaborated by GII, were maximal concentration of explosives in the
charge assembled from numerous similar units and increased energy release to the atmosphere - in
order to provide more distant infrasound observations (Fig. 1). Special attention was given to
details of charge design and configuration to ensure the understanding of observed explosion
effects and energy generation. A similar design was used in both experiments, however main agent
explosives were rather different (Table 1).
@seismicisolation
@seismicisolation
210 Chapter Ten
Figure 1. Charge design of main Sayarim shots in August 2009 (a), and in January 2011 (b)
Table 1. Parameters of explosives used in large-scale Sayarim surface explosions
Event Date Main Density Velocity of Charge unit Charge
agent (g/cm3) detonation, (kg) booster
VOD (m/s)
Ex1 26/08/09 Cast IMI 1.54–1.67 7,130–7,980 Barrel 315 Mines M-15
Ex2 24/01/11 Bulk ANFO 0.80–0.81 2,400 Big bag 870 (each one 10 kg
composition B)
Ex3 26/01/11
The large-scale surface explosion Ex1 in August 2009 was characterized by (Fig. 1a): (a) big
barrels (~ 315 kg each) filled up to the top with strong cast HE, provided by Israel Military
Industries Ltd. (IMI); (b) minimal air voids between the barrels by filling the voids with plastic
bags containing HE bulk explosives, and between six charge layers; (c) nearly compact pyramidal
shape (measured dimensions: base 5.8 x 5.9 m and height 5.5 m); (d) a strong booster composed of
several mines M-15, placed on the ground upside down, to provide the upward detonation and
additional upward cumulative effect, and 63 mines on the central platform of the first layer; (e)
multiple-point initiation scheme to ensure reliable detonation (GITTERMAN, 2010). The IMI
explosives consist of a mixture of different recuperated HE: TNT and gun powder as major agents,
smaller amounts of Composition B and RDX, and some other components.
Two explosions, Ex2 and Ex3, conducted in January 2011 have slightly different features (Fig.
1b): (a) primary agent ANFO mixture (94 % ammonium nitrate and 6.0 % fuel oil), placed in
waterproof big bags, supplied by Explosives Manufacturing Industries (1997) Ltd., Israel (EMI);
(b) nearly hemispherical shape (measured actual dimensions: base 8 9 8 m and height 4 m); (c)
single mines M-15 beneath each bag (double mines for the upper layer) for reinforcement of the
@seismicisolation
@seismicisolation
detonation wave front propagating upward; (d) additional mines placed in air-voids in four corners
of the first layer (GITTERMAN, 2011).
CONDUCTING OF THE EXPLOSIONS AND GT0
Parameters
All charges in summer (Ex1) and winter (Ex2, Ex3) experiments were placed on the soft sediment
surface at the same site (Fig. 2). In the explosion area the subsurface layer of ~ 0.5–1 m consists of
soft and loose sediments, with consolidated sediments below. Geologically the subsurface media is
presented by Quaternary alluvial conglomerates, underlined by consolidated limestone, chalk and
chert rocks. The available shallow subsurface velocity model is shown in Table 2. The model was
obtained from a seismic refraction survey at Sayarim Valley, where GII conducted a Seismic
Calibration Explosion of 32.5 tons in boreholes (GITTERMAN et al., 2005), not far from the
Infrasound Calibration Explosion site (~16 km to the South).
In January 2011, per IDF safety requirements, the territory within 6 km of the large explosion was
surveyed by helicopter about 2 h before the detonation to ensure that no people or animals were in
the area (Fig. 2). In addition, there was GII installed on- line video broadcasting for both
explosions using digital network cameras for enhancing the safety surveillance (see the layout map
in Fig. 3). The real time picture was simultaneously displayed on monitors at SMR Command
Post, PTS CTBTO office in Vienna, and GII.
Figure 2. Helicopter view of the large explosion site (morning of 26 January 2011). The previous
82-ton shot place is shown, with distance between the charge centers ~ 30 m
Table 2. Velocity model of the shallow subsurface at Sayarim Valley
Layer no. Depth interval Interval velocities, Averaged layer Average
(m) Vp (m/s) thickness (m) Vp (m/s)
1 0 down to 5–7 1,130–1,300 0–6 1,215
2 5–7 down to 17–20 1,650–1,720 6–18.5 1,685
3 17–20 down to 85–100 1,930–2,020 18.5–92.5 1,975
4 Deeper than 85–100 3,750–4,120 >92.5 3,935
@seismicisolation
@seismicisolation
212 Chapter Ten
Figure 3. The Ex2 and Ex3 experiment layout (winter 2011): location of the explosions (and the
2009 shot Ex1) and near-source measurement systems: pressure gauges (G1–G6), accelerometers
(A1, A2), home and speed video-cameras, and monitoring safety cameras that were broadcasting
the real time video through the internet (a); view of the dust and gases column from the Command
Post at 6 km for Ex3 after 6 min (b)
Table 3 presents origin (detonation) time (OT), GPS coordinates (accuracy 4–5 m), total weight of
all explosives, altitude of the charge and near-surface air temperature during the detonation. The
OT values were based on an electric circuit attached to the detonator with an appropriate PC-based
system that includes GPS time. In addition, we checked the timing via analysis of raw data files
from the GII data acquisition system and verification by near-source accelerometric records. The
day and time (~0930 hours local time) for the 2009 shot Ex1 were chosen due to favorable wind
conditions for infrasound propagation (GARCES et al., 2009; BOWMAN et al., 2009). In 2011
the smaller 10-ton shot Ex2 was conducted as late as possible in the afternoon ~15:18 hours local
time) given the short daytime in January and the large 100-ton shot Ex3 was conducted as early as
possible in the morning, to check the day variability of weather conditions.
NEAR-SOURCE OBSERVATIONS
GII deployed various measuring and observation systems at close distances including pressure
gauges, accelerometers, regular and speed video-cameras. The experiment layout of the explosion
and GII recording systems, similar for both experiments, is shown on Fig. 3.
The maximal height of the dust and gases column was estimated using recordings of a video
camera: ~1,250 m for Ex2 (after ~4 min), and ~2,750 m for Ex3 (after ~ 6 min) (Rafael Ltd.,
personal communication) (see Fig. 3b).
Table 3. Parameters of large-scale surface explosions, at Sayarim Military Range
# Date Nominal Detonation Altitude Air Latitude Longitude
charge time (m) temperatur (°N) (°E)
weight (kg) (GMT) e
(°C)
Ex1 26/08/09 81,664 06:31:54.00 558 35 30.00057 34.81351
Ex2 24/01/11 10,240 13:17:53.80 546 24 29.99555 34.81668
Ex3 26/01/11 102,080 07:17:42.44 558 13 30.00064 34.81324
@seismicisolation
@seismicisolation
Figure 4. Sample records for Ex2, gauge G1 (filtered) (a), and Ex3, gauge G3 (b). Exponential
fitting curves (red) and intervals (green) are also shown.
Note a significant difference of the maximal recorded amplitude and the peak fitting overpressure
value, for Ex3
HIGH-PRESSURE MEASUREMENTS
High-pressure sensors are required to evaluate the efficiency of the charge design and energy
generation of ANFO explosives and estimate the energy released by the explosions, as the TNT
equivalent yield of the charge. Furthermore, the records of air-shock waves are used to analyze and
explain some interesting blast phenomena. The IDF team installed six pressure gauges XTL-190-
5G/50A/100A along a line in the distance range 100–600 m (Fig. 3a). The gauges with disc type
baffles were mounted on steel rods ~1.5 m above the surface providing side-on free-field over-
pressure measurements with sampling rate of 2 MHz.
We measured accurately basic parameters of recorded air-blast waves:
• Time of arrival TAms, peak pressure Pmms, positive phase impulse I+, positive phase duration
τ+ - for the main shock;
• parameters (Tss, Pss) - for the secondary shock (see below in details).
Samples of the air-shock overpressure records with explanations of the measured parameters are
presented in Fig. 4, together with calculated shock-wave impulse (blue curves).
The records of the small shot Ex2 are contaminated by high-frequency noise (at frequency about
107.5 kHz), that hampered accurate measurement of the parameters. Therefore, we applied the
33.33 kHz low pass filter (Fig. 5), thus improving the measurement accuracy.
Due to some irregularities in the high-pressure time history, high-frequency noise and spikes, the
waveform was approximated by 2–3 exponent fit curves, for more reliable estimation of (Pm, τ+)
values for the main shock and (TA, Pm) values for the secondary shock (Fig. 4). Then we calculated
the secondary shock delay ∆t = TAss - TAms. Sample measured values for the explosion Ex2 are
presented in Table 4.
@seismicisolation
@seismicisolation
214 Chapter Ten
Figure 5. Application of low pass 33.33 kHz filtering to a sample record for Ex2
Table 4. Accurately measured basic air-blast wave parameters for Ex2 and expected (standard)
values (in brackets) as obtained by BECv4 procedure for the charge 10.3 ton ANFO, actual
altitude and air temperature
The air-blast measured parameters were found enlarged (Pm) or reduced (τ+, I+), for all gauges and
all three explosions, compared to expected (standard) values obtained by the DDESB Blast Effects
Computer, Version 4.0 (BECv4), an Excel template (SWISDAK, 2000) (see Table 4). This
procedure provides easy computation of a wide variety of free-field air-blast parameters (in
English or Metric units), for different explosives and charges (including on-surface sources), in
broad pressure, charge weight and distance ranges. The parameters are calculated by reliable
empirical scaling relationships based on numerous fully-controlled experimental explosions of
hemispherical charges (e.g., SADWIN and SWISDAK, 1970). Air-blast estimations can be
obtained also by commonly used ConWep computation procedure (CONWEP, 1997), but only for
sea level conditions, whereas BECv4 takes into account also an altitude (or atmospheric pressure)
and the temperature (in some cases the difference is significant).
ESTIMATION OF TNT EQUIVALENT YIELD FROM AIR-BLAST WAVE DATA
The TNT equivalent estimation is based on the mentioned BECv4 procedure, using the actual
altitude and air temperature at the explosion site (see Table 3). The accurately measured values of
peak over-pressure, positive phase duration and impulse for the main shock were utilized for
calculation of an appropriate yield for each gauge, and then averaged over several gauges, for each
specific air-shock wave parameter. The gauge G1 was excluded from the yield estimation for Ex3
due to anomalously low pressure amplitudes (distance 103 m), supposedly because the air-shock
propagation may have been affected by the nearby IDF bunker (see Fig. 2). The results are
@seismicisolation
@seismicisolation
presented in Table 5.
Increased Pm and decreased τ+ values resulted in appropriately overestimated and underestimated

yields. The positive phase impulse I+ is the integral and stable characteristic of the air-blast wave
(unlike the one-point peak amplitude), and we consider the impulse-based estimation as the most
reliable, accepted as GT0 parameter of Sayarim calibration explosions (shown bold in Table 5),
and used in the following developing of charge-scaled relationships. Note that the impulse-based
yield estimation is close to the average of estimations by both parameters (Pm and τ+), because the
impulse is about proportional to their multiplication.
The same blast wave parameters and a similar multi-station procedure were applied to charge esti-
mation for track bomb explosions, concluding that the most reliable and accurate acoustic wave
property (for yield estimation) is the impulse of the airblast (KOPER et al., 2002).
The obtained TNT yield estimation for explosion Ex1 is about 20 % more than the nominal weight
of the charge. There are several possible factors for the enhanced air-blast energy and the
appropriate enlarged TNT yield: (1) strong IMI explosives with high detonation velocity ~7,500
m/s on the average for the whole charge, ~10 % higher than for TNT (6,900 m/s); (2) high
concentration of explosives, when most air-voids between the charge units were filled with HE; (3)
very strong booster and multiple initiation scheme; (4) upward detonation of the charge.
Note that the ratio of yield estimations based on peak over-pressure and maximal impulse is very
similar for the two large shots: 152/96 = 1.58 for Ex1 with IMI explosives stronger than TNT, and
119.6/76.8 = 1.56 for Ex3 with ANFO explosives, which is weaker than TNT.
Table 5. Peak pressure, positive phase duration and impulse for the main shock and appropriate
estimations of the TNT equivalent yield for the Sayarim explosions
Event Gauge Distance, r (m) Peak overpressure, Pm Positive phase duration, τ+ Positive phase impulse, I+
Measured Yield (tons) Measured (ms)Yield (tons) Measured (Pa s)Yield (tons)
(kPa)
Ex1 G2 197 74.4 164 129.9 33 2,963 101
G3 294 34.9 157 163.9 52 1,953 93
G4 394 22.3 163 185.6 59 1,508 95
G5 509 15.0 155 199.6 57 1,202 98
G6 611 10.6 123 219.5 67 966 92
Average 152.4 53.6 96.0
Ex2 G3 351 7.14 9.20 103.60 5.1 331.8 7.7
G2 452 5.40 10.1 107.51 4.5 252.4 7.4
G1 552 4.31 10.8 119.32 5.6 203.6 7.2
Average 10.0 5.1 7.4
Ex3 G2 203 53.5 106.8 127.4 24.7 2,424.4 72.1
G3 303 30.16 132.0 156.36 34.4 1,731.5 76.0
G4 405 18.7 124.0 179.9 42.2 1,314.7 75.8
G5 513 13.3 122.0 200.54 48.8 1,109.85 82.7
G6 580 10.9 113.0 206.03 46.5 943.3 77.3
Average 119.6 39.3 76.8
Based on the known standard ANFO to TNT equivalence of about 0.82,the expected equivalent
TNT yield values are 8.5 ton for Ex2, and 84.5 ton for Ex3. The obtained yield estimates are a
little smaller than the expected values (~13 % for Ex2, and ~9% for Ex3). We note several
possible factors for lower air-blast energy and appropriate reduced equivalent TNT yield: (1) not
quite hemispherical shapes (especially for Ex2); (2) inhomogeneous charges (bags and boxes for
Ex2, placement of numerous mines with stronger explosives in the ANFO charge body for Ex3);
(3) numerous air-voids between the charge units (plastic boxes and big bags) that were not filled
by explosives (as was done for Ex1). All these factors caused a non-uniform charge that resulted in
observed blast anomalies: jetting, asymmetrical blast fronts, multiple shock-wave phases,
@seismicisolation
@seismicisolation
216 Chapter Ten
turbulence, collisions of wave fronts, and, consequently, in some energy losses.

SECONDARY SHOCK EFFECT
In the 2011 ANFO explosions Ex2 and Ex3, a distinct secondary shock (SS) wave was observed at
all gauges during the negative phase of the pressure– time curves, showing negative or close to
zero peak pressures (Figs. 4, 7a, b). It is similar to SS waves observed for surface 20 and 100 ton
ANFO shots in Alberta, Canada (SADWIN and SWISDAK, 1970), and opposite to the case of
Sayarim 2009 Ex1, where positive SS peak pressures were observed (Fig. 7c) (GITTERMAN,
2010). The explosion Ex1 comprised cast IMI explosives with higher density and velocity of
detonation than ANFO that was used in 2011 (see Table 1). Evidently these strong explosives
caused smaller time delays between main and secondary shocks, resulting in positive SS peak
pressures.
Clear SS waves were observed also at two Kinemetrics K2 accelerometers (A1 and A2 in Fig. 3a),
placed on the surface and subjected to the strong impact of the air shock wave. For the 102 ton
shot Ex3, a vertical acceleration ~4 g was measured at the closest station, corresponding to the
main air-shock MS arrival (Fig. 6).
In this known, but rarely reported phenomenon, the air-blast wave for any finite chemical
explosion source can exhibit numerous repeated shocks of small amplitudes at various times,
caused by successive implosion of rarefaction waves from the contact surface between explosion
products and the air (BAKER, 1973). A higher pressure shock front propagates faster, therefore
the time delay between the main shock (MS) and SS phases increases with distance (see Fig. 7a,
b), as well as with the charge yield.
Using the charge cubic root scaling law we developed relationships for the scaled delay Dt and the
scaled distance R (for estimated TNT equivalent charge W, presented in Table 5):
Dt = Dt/W1/3 (s/kg1/3), (1)
R=r/W1/3 (m/kg1/3)
Some differences in air temperature and pressure (altitude) for different explosions are considered
to cause minor changes in air-blast parameters, comparable with measurement errors of distance
and wave parameters, and are not applied here in the distance and time delay scaling.
We extended the SS dataset, utilizing a broader observation range for more complete analysis of
this unique air-blast feature. In the Sayarim 2011 experiment (Ex2 and Ex3) a number of acoustic
sensors were deployed by Weston Geophysical Corporation (WGC), the University of Mississippi
(UM), and the University of Firenze (UF), at a distance range 1–37 km (GITTERMAN, 2011),
providing good records of SS waves that we included in the analysis.
A linear RMS fit regression curve of the scaled delay Dt versus logarithmic scaled distance R was
obtained for two Sayarim ANFO explosions, with a high correlation parameter C2 (Fig. 8):
Dt(s/kg1/3) = 0.0057565 x log(R) + 0.0032,
C2 = 0.985, 2˂R˂1,000 m/kg1/3 (2)
In addition, we tried to extend also the source range by checking some acoustic records of the
WSMR large-scale ANFO explosions Distant Image (2,214 metric tons ANFO, equivalent to
1,815 tons TNT), and Minor Uncle (2,472 metric tons ANFO, equivalent to 2,027 tons TNT), at
distances 28–60 km, presented in NORRIS (2007). We identified SS phases, measured the delays
and scaled for TNT equivalent yield, and found a good agreement with Sayarim ANFO explosions
and the fit curve (Fig. 8).
@seismicisolation
@seismicisolation
Figure 6. Accelerogram of Ex3 at A1 (~300 m), showing arrivals for P-waves, and air-blast main
(MS) and secondary (SS) shocks
Figure 7. Samples of shock-waves records for Ex3 at different distances (a, b). Gauge G5 showed
SS delay Δt ~0.42 s (b), compared to a smaller Δt ~0.24 s, for Ex1, at a similar distance (c)
Thus Eq. (2) describes the data over a broad range of charges (10 – 2,725 tons ANFO) and
distances (0.1–60 km).
However the data for the Sayarim 2009 explosion Ex1 with IMI explosives, using records of five
high- pressure gauges and also three accelerometers at a distance range 200–610 m, are clearly
separated from the ANFO data, and, though showing a linear relationship between the scaled delay
and the log-scaled distance, but significantly lower than the fit curve (Eq. 2) (Fig. 8). The IMI
explosives are much stronger than ANFO and, as we estimated, more energetic than TNT, having
a higher detonation velocity (see Table 1). Supposedly, this factor explains the smaller SS delays.
The obtained results show that there is an option to use the stable and reliable air-blast parameter,
SS delay, as a new potential yield estimator. For ANFO shots on dry soft sediment surface, if the
SS phase is clearly identified on acoustic or seismic records, then the charge weight can be
accurately estimated.
EMPIRICAL SCALING RELATIONSHIP FOR BASIC AIR-BLAST PARAMETERS
FROM SAYARIM EXPLOSIONS
When analyzing the dependence of the main shock peak pressure in air-shock waves on the scaled
distance, in a broad distance range (0.1–10.5 km) we found that the data of all three Sayarim
explosions, although of very different ANFO and IMI explosives, show a unified linear
relationship in the double logscale, indicating a power law curve, with a low spreading (Fig. 9).
simple power law with a high correlation parameter C2:
The RMS regression procedure provided a@seismicisolation
@seismicisolation
218 Chapter Ten
Pm(Pa) = 410, 801 x R-1.3461, C2 = 0.98,

2˂R˂500 m/kg1/3 (3)
In the peak pressure regression analysis, we used the same near-source high-pressure and close
acoustic records as for the SS delay regression for Sayarim shots, including WGC, UM and UF
sensors. The distance range was restricted to 10.5 km, in order to stay in the field of strong air-
shock waves (Pm > 100 Pa, approximately corresponds to the threshold of window breakage), and
to avoid atmospheric/wind effects on propagation of weak (elastic) infrasound waves. Equation (3)
is similar to one of the standard reference empirical relations for the open air detonation
(PERKINS, 1964): Pm(Pa) = 360,070 x R-1.38, which fits well the Sayarim data (Fig. 9).
For comparison, we present in Fig. 9 also an empiric - analytical relationship derived from the
Figure 8. Scaled time delay between MS and SS phases versus scaled distance (calculated for TNT
equivalent charge weights)
Landau analytical form, that was developed for weak air-shock waves at large distances
(LANDAU, 1945), using measured near-source peak pressures at 200–600 m from Sayarim 2009
shot Ex1 (V. PERGAMENT, personal communication):
∆P/P0 = 1/{R[log(R/π)1/2]1/2} (4)
where P0 is normal atmosphere pressure, 105 Pa.
Equation (4), which is valid for R ˃ 3.4 m/kg1/3, demonstrates a good correspondence to Sayarim
data, especially at close distances.
We also developed empirical scaling relationships for the main shock basic parameters: peak over-
pressure Pm, positive phase duration τ+ and impulse I+, using only high-pressure gauges data (i.e.,
in a narrow distance range 100–600 m). In addition to the charge cubic root scaling (only τ+ and I+,
using the estimated TNT equivalent charge W, Table 5), we scaled these measured air-blast
parameters to the reference atmospheric conditions at the sea level (altitude H = 0 m, temperature
T = 15°C), following Sachs’ scaling procedures (SACHS, 1944; SADWIN and SWISDAK, 1970).
The results are shown in Fig. 10, the following RMS fit expressions were obtained:
@seismicisolation
@seismicisolation
Figure 9. Peak over-pressure versus scaled distance (calculated for TNT equivalent charge
weights) for Sayarim explosions
Pm(Pa) = 662,764 x R-1.5451 (5)
τ+(ms/kg1/3) = 1.4063 x R0.46639 (6)
I+(Pa ms/kg1/3) = 268,343 x R-0.94907 (7)
The reference (TNT standard) values, obtained by BECv4 procedure, are also presented (Fig. 10).
The scaled observed data demonstrate a high consistency for all three shots with different types of
explosives and TNT yield, and correspond well to power fit curves (5–7), except for τ+ values
(Fig. 10b) that seems to fit much better to a polynomial curve (Fig. 10d). There is a discrepancy of
the scaled Sayarim data with the reference (by BECv4) values (as appropriate to the analysis of the
TNT yield estimation presented before): measured peak pressures are a little higher (Fig. 10a),
positive phase durations are lower (Fig. 10b, d), whereas positive impulses show almost the same
values as the TNT standard curve (Fig. 10c). This result corresponds, evidently, to impulse-based
estimation of yield values, which are crucial for scaling relations.
Note an expected stronger attenuation with distance for high near-source peak pressures, compared
with weaker Pm values in a broad local distance range (Eq. 3).
Obtained small spreading and high uniformity for all three shots of the scaled data, especially of
the positive impulse, verify the yield evaluation procedure and estimated TNT yields.
AUDIO-VISUAL OBSERVATIONS OF BLAST EFFECTS
Two home video-cameras, three special safety monitoring cameras (radio and cellular transmitted),
and a speed Phantom-3 camera (3,100 frames/s) were placed at different distances and azimuths in
the two explosions (Fig. 3a), for recording the unique blast phenomena. Snapshots from home
video of the 102-ton shot show an expanding, evidently following the shock wave, short-term
(~0.5 s) white spherical condensation cap, which is clearly visible due to specific air (humidity 61
%, temperature 13 °C) and lighting conditions (Fig. 11). Snapshots of speed video-record for the
ten ton shot show non-spherical segments and multiple phases in the air-shock wave front at short
times and distances (Fig. 11). Apparently, the same multiple shock-front phases were also
observed in the large Ex3 and correspond to multiple peaks in the positive (compression) phase of
the air-shock wave recorded by close pressure gauges (Fig. 7a). Supposedly, these front
phases/peaks are due to non uniformity of the charges, and cannot be attributed to any specific
@seismicisolation
@seismicisolation
220 Chapter Ten
charge elements. Away from the explosion, the peaks are merged, and after ~ 70–100 m, a stable
uniform main shock phase is formed.
Figure 10. Air-blast data measured for three Sayarim large explosions, scaled to the TNT yield
charge and the sea level conditions, show a high uniformity: peak pressure (a), positive phase
duration (b, d) and positive phase impulse (c). Reference (TNT standard) values, obtained by
BECv4 procedure, are also shown
Many observers of the large Ex3 placed at ~9 km reported hearing two clear ‘‘bang’’ sounds with
a delay of less than 1 s, which were interpreted as two separate explosions. Multiple pronounced
‘‘bang’’ sounds can be clearly identified at recordings of the monitoring video-cameras (at several
hundreds meters) for both shots Ex2 and Ex3. Initially it was suggested that these calibration shots
were not simultaneous as planned, however, detailed analysis of the data showed that these audio-
phenomena were caused by secondary shocks in the air-blast wave observed at all records of high-
pressure gauges and accelerometers (Figs. 4, 6, 7). These records, along with the yield estimates
using the main-shock peak pressures (Table 5) confirm that all explosive materials were fully
detonated in the initial 2–3 ms.
CRATER OBSERVATIONS AND SEISMIC MAGNITUDES
A regular hemispherical crater was created by the 82-ton IMI explosion in 2009 (GITTERMAN,
2010). The Survey of Israel conducted accurate GPS crater measurements just after the shot, and
provided 2D– 3D crater images, and accurate diameter D and depth H estimations (Fig. 12; Table
6).
@seismicisolation
@seismicisolation
Figure 11. Snapshots from speed video-recording for Ex2 at 360 m (left) and from usual video
recording at 6 km for Ex3 (right), show unique blast effects
Unlike the 2009 explosion, complex irregular- shaped craters were found for the two January 2011
ANFO shots, with step-like walls and a cone of crashed rocks in the center (Fig. 13).
For comparison, we used empirical equations for craters of large-scale explosions (in Russia) on
soft soils surface (ADUSHKIN and KHRISTOFOROV, 2004):
D(m) = 2 x 3.36 W0.336; H(m) = 1.78 W0.316 (8)
and another diameter equation based on 200 large surface shots (KINNEY and GRAHAM, 1985):
D(m) = 8 x W1/3, (9)
where W is TNT equivalent yield in tons.
The crater parameters calculated using Eqs. (8) and (9) are presented in Table 6, jointly with mea-
sured values, demonstrating significantly smaller sizes.
A hemispherical surface explosion of 90.7 ton TNT, on alluvium, in the US Army Waterways
Experiment (1962), was similar to Sayarim shot Ex1, but created a bigger crater with radius D =
42.6 m, H = 6.3 m (CRATERDATABASE_V1.3 2004).
@seismicisolation
@seismicisolation
222 Chapter Ten
The ConWep procedure (CONWEP, 1997) provided also much larger crater size estimations for
Ex3 (102 tons ANFO on dry sandy clay): D = 41 m, H = 13 m.
Figure 12. Crater created by the 2009 explosion Ex1 (top), and 2D and 3D images provided by the
Survey of Israel (bottom)
Table 6. Measured crater dimensions, compared with calculated values and local seismic
magnitudes Md
The local duration magnitude Md estimated from the Israel Seismic Network (ISN) records showed
two interesting important features (Table 6). Firstly, the magnitude for Ex1 with higher TNT yield
is smaller that for Ex3 (though the crater is larger). Evidently, the main reason is the much higher
VOD of stronger IMI explosives compared to ANFO. A similar effect was found for small
borehole shots NEDE in the USA in 2008, where larger amplitude Rg and Love waves were
generated from shots using black powder (VOD 530 m/s) and ANFO emulsion (VOD 5,260 m/s),
than shots with Composition B (VOD 8,100 m/s) (STROUJKOVA, 2012). Secondly, all
magnitudes seem smaller than could be expected, based on extensive GII experience in monitoring
of numerous detonations of outdated ammunition at SMR. For example, an ammunition shot at
SMR in June 2008 (consisted mainly of M-15 mines) with TNT yield about 10 tons, similar to
Ex2, but with the downward detonation (the booster mines were on the charge top, as on Fig. 14),
produced a higher Md = 2.5 (GITTERMAN and HOFSTETTER, 2008; GITTERMAN, 2009).
@seismicisolation
@seismicisolation
Figure 13. Crater general view for 2011 Sayarim explosions Ex2 (a) and Ex3 (b), and 2D (c) and
3D (d) crater images for Ex3, provided by the IDF Engineer Corps team
The results show that crater parameters, the diameter and especially the depth, and seismic
magnitudes, for all Sayarim explosions were smaller than the expected values, based on data for
previous similar experiments, thus indicating a decreased coupling of explosive energy with the
ground, supposedly due to the specific charge design and especially the upward detonation.
A striking evidence for importance of the detonation direction for cratering was found for test
explosions at SMR with equal charges of 1 ton of TNT and Composition B, with different
detonation direction (GITTERMAN, 2009). For two explosions (separately TNT and Composition
B), conducted in June 2008, the boosters (mines) were placed on the top of the charge (Fig. 14a),
providing downward detonation. Both explosions produced large deep (~0.5–0.7 m) craters (a little
deeper for the explosion with more powerful Composition B) (Fig. 14b). In distinction, the upward
detonation was applied to 1 ton (mix of TNT, Composition B and RDX) shot in December 2008,
resulting in a very small shallow crater (Fig. 14c, d; Table 7). Evidently a significant difference in
the crater size prevail over non-identical charge shape and placement conditions. The smaller
crater for the upward detonation evidences that relatively large and small portions of explosive
energy were released to the atmosphere, and penetrated the ground as seismic energy, respectively,
compared with downward detonation. It is also confirmed by lower amplitudes of seismic waves
from the December 2008 explosion recorded at ISN stations (Fig. 15), and by an appropriately
smaller magnitude, compared to June 2008 shots (Table 7). Consequently, cratering effects
observed for surface explosions can be considered as an indicator for explosion energy partitioning
between seismic waves in the ground and acoustic waves in the atmosphere.
@seismicisolation
@seismicisolation
224 Chapter Ten
Figure 14. Cratering effects for two SMR test shots: downward detonation (red arrows) of 1 ton
TNT in June 2008 (a), that created a large deep crater (b), and upward detonation of the same size
charge in December 2008 (c), that produced a very small shallow crater (d) (from GITTERMAN,
2009)
Table 7. Observed crater sizes for 2008 test shots at SMR
Date Charge (kg) Detonation Md Crater size
explosives direction
Diameter (m) Depth (m)
2008/06/24 1,020 TNT Down 1.8 2.5–2.6 0.5–0.6
2008/06/25 1,020 CompB Down 2.0 2.4–2.5 0.6–0.7
2008/12/02 1,040 TNT, Up 1.6 1.8–2.0 0.3–0.4
CompB&RDX
DISCUSSION
Based on an empirical scaling relationship for the dominant period of infrasound waves, another
yield estimation for the 2009 explosion was obtained. The physical basis for this relationship is the
increased acoustic transit time of the blast radius with increased yield (REVELLE, 1998). Besides,
the dominant period is not influenced much by propagation like peak amplitudes. Utilizing a well-
recorded dataset of Israel and Cyprus infrasound stations at a distance range of 150–570 km,
where single clear main arrivals were observed, the yield of 128 tons of TNT was roughly
estimated based on the average dominant period (GITTERMAN, 2010). This estimation is
relatively close to the reliable yield of 96 tons from air-blast measurements (Table 5).
@seismicisolation
@seismicisolation
Figure 15. Calculated energy (in counts) of the whole seismic signal (vertical component) from
2008 test shots recorded at five close ISN stations
Similar to the near-source pressure gauges, anomalous enhanced peak pressure amplitudes were
found at local infrasound stations for the 2009 explosion, indicating a possible upward directivity
effect and asymmetric energy radiation to the atmosphere (GITTERMAN, 2010). Possibly, this
effect caused an overestimated yield of 0.3–0.5 kT, obtained at two far-regional IMS infrasound
stations I26DE and I48TN of the International Data Center (IDC), using wind-corrected
amplitudes, as applied to several yield-range-amplitude attenuation laws, including LANL2003
(BROWN, 2009). This large misfit in yield estimation denotes an obvious need for refinement and
tuning of the monitoring procedures, using GT0 infrasound sources, to obtain reliable yield values.
However, using a new-developed Parabolic Equation- based semi-empirical relation, a reasonable
value ~0.1 kT at the dominant frequency ~0.5 Hz was estimated from I48TN data for the 2009
explosion (LE PICHON, personal communication, 2010).
As mentioned before, the air-blast parameter SS delay could be used as a yield estimator if the
explosive type is known, based on the developed scaled relationships. On the other hand, the
explosive type can be identified and the detonation velocity can be roughly estimated, if the yield
is known and SS delays are measured. It seems that atmospheric or surface nuclear explosions do
not produce air-blast SS phases (e.g., BRODE, 1956), because a nuclear test provides an
instantaneous and point-like source, where parameters of chemical shots (crucial for the SS effect)
- charge size, detonation velocity, expanding detonation products (gases) - are not applicable.
Then, by a logical extension, it is guessed that the SS phase could be used as a discriminant
between a small nuclear and large chemical explosion. Theoretically it is possible under specific
conditions, but it can be practical only at close distances where the main shock and especially
secondary shock are still observable, and not transformed into numerous phases of
acoustic/infrasound wave.
CONCLUSIONS
Two large-scale shots of about 100 tons, with a pyramid/hemisphere charge design, were
successfully conducted by GII in summer 2009 (Ex1) and winter 2011 (Ex2, Ex3). The explosion
design was rather complicated and different from the ideal homogenous hemisphere used in
similar past experiments. Strong boosters and an upward charge detonation scheme were applied
to ensure that more energy is radiated to the atmosphere. Enhanced peak pressures at all distances
and smaller craters and seismic magnitude indicated that the developed charge design provided a
strong explosion energy generation and necessary energy partition: large portions of energy to the
atmosphere and less to the ground as seismic energy. Under these conditions evaluation of the
actual explosion yield, as one of the important source parameters, is crucial for the GT0 calibration
experiment. The accurate yield estimations were obtained, based on measured positive phase
@seismicisolation
@seismicisolation
226 Chapter Ten
impulse in air-blast waves at near-source distances, and considered as ground truth parameter.
Compared to the nominal explosives weight, the estimated TNT equivalent yield was enlarged for
the 2009 shot and reduced for the 2011 shots.
Dense near-field observation systems provided valuable abundant data sets of audio-visual, air-
shock and acoustic records that were utilized for interpretation of observed unique blast effects and
determination of blast wave parameters, suited for yield estimation. Empirical functional
relationships between basic blast-wave parameters, the yield and the distance were developed,
including a novel secondary shock delay relation, that provided a base for a new yield estimator.
Figure 16. Modeling of stratospheric winds for the pair of large Sayarim shots (Ex1 and Ex3),
based on atmospheric specifications from the ECMWF, provided 6-hourly models of 91 layers on
a global 0.58 3 0.58 grid, the figure is derived from these models at 50 km altitude (courtesy of L.
Evers, KNMI). The explosion site and recorded IMS (triangle) and some regional portable
(inverted triangle) infrasound stations are shown
The main goal of this dual Sayarim calibration experiment was reached: fully controlled
infrasound sources, the strongest since the establishment of the IMS network, were observed at
several IMS infrasound stations in Europe, Middle East and Asia, thus establishing the first GT0
infrasound dataset for this region.
In the 2009 experiment, ten portable infrasound arrays were deployed by collaborating institutions
from eight countries at regional distances in Europe and EM–ME countries, coordinated by the
University of Hawaii and CTBTO teams. In the 2011 experiment, institutions from 20 countries
collaborated to set up a dense infrasound network for near-regional observations. All together 20
portable arrays were deployed in 13 countries throughout the EM–ME region by local institutions,
coordinated by the CTBTO and UM teams, that provided recording equipment for the far field
observations in 2011 (GITTERMAN, 2011).
The infrasound signals were observed at numerous regional and IMS stations up to 3,400 km to the
west/ north-west for the summer 2009 shot Ex1, and up to 6,250 km to the east for the winter 2011
shot Ex3 - further than expected. The very clear westward/eastward infrasound propagation
according to appropriate summer/winter stratospheric wind directions, respectively, was
demonstrated (Fig. 16), thus verifying empirically common models of infrasound propagation in
the atmosphere.
The complementary smaller Ex2 shot (winter 2011) of 10 tons was conducted during the afternoon
2 days prior to the Ex3 (main 102-ton explosion detonated in the morning), providing additional
valuable data for the analysis of charge scaling and infrasound propagation features affected by
various atmospheric (wind) conditions.
The unique collected database provides an important contribution in the modeling of long-range
@seismicisolation
@seismicisolation
infrasound propagation in the atmosphere, detection analysis at different stations, and refining and
tuning of IDC evaluation procedures of infrasound monitoring and yield estimation.
ACKNOWLEDGMENTS
Many organizations and persons participated in the preparation of Sayarim calibration explosions,
measurements and data processing. High-quality explosives in convenient packages were supplied
by IMI Ltd. (I. Veksler) for the 2009 experiment, and by EMI Ltd. (Dr. D. Hershkovich) for the
2011 experiment. Elita Security Ltd. (S. Kobi) assembled the 2009 IMI charge with maximal
concentration of explosives. The IDF Experiment Division (E. Stempler, Y. Hamshidyan)
provided appropriate territory, logistics and near-source measurements, and assembled the 2011
ANFO charges with the optimal initiation/detonation scheme. The IDF Engineer Corps (Z. Savir)
contributed to preparation of infra-structure for close-in measurements, and conducted crater
parameters measuring and processing. GII personnel helped in logistics procedures, preparation
and deployment of numerous near-source local observation systems, and in data processing and
graphic presentation of air-blast data (U. Peled, N. Perelman). Thanks to our collaborators Dr. J.
Bonner, Weston Geophysical Corporation, Dr. R. Waxler, the University of Mississippi and Dr. E.
Marchetti, the University of Firenze, for supplementing data of close seismo-acoustic observations
in 2011 experiment. Thanks to V. Pergament of Magnitogorsk State Technical University for
assisting in the analysis of the pressure-distance relationship, to Dr. L. Evers, the Royal
Netherlands Meteorological Institute (KNMI), for supplementing stratospheric winds modeling.
Sayarim experiments were supported by the US Army SMDC in 2009 (M. Pickens), and PTS
CTBTO in 2011 (Dr. L. Zerbo and Dr. J. Given). Research work of one of the authors (Y.G.) was
supported by the Israel Ministry of Immigrant Absorption. We are thankful to anonymous
reviewers for valuable comments, especially relating to additional aspects of the air-blast
secondary shock delay application, and the analysis of seismic magnitudes.
REFERENCES
ADUSHKIN, V. and B. KHRISTOFOROV (2004). Craters of Large-Scale Surface Explosions,
Combustion, Explosion, and Shock Waves, Vol. 40, No. 6, pp. 674–678.
BAKER, W. E. (1973). Explosions in Air, University of Texas Press, Austin and London, 268p.
BHATTACHARYYA, J., H.E. BASS, D.P. DROB, R.W. WHITAKER, D.O.
REVELLE and T.D. SANDOVAL (2003). Description and analysis of infrasound and seismic
signals recorded from the Watusi explosive experiment, September 2002, Proceedings of the
25th SRR, Tucson, Arizona, September 2003.
BONNER, J., R. WAXLER, Y. GITTERMAN and R. HOFSTETTER (2012). Seismo-acoustic
energy partitioning at near-source and local distances from the 2011 Sayarim explosions in the
Negev desert, Israel. Bull. Seis. Soc. Am. (in press).
BOWMAN J.R., H. ISRAELSSON, G. SHIELDS, M. O’BRIEN and Y. GITTERMAN (2009).
Detection and Characterization of Infrasound Signals at IMS Stations from Explosions at the
Sayarim Military Range, Israel: 2005-2009. Presented at the Infrasound Technology Workshop,
Brasilia, Brazil, November 2-6, 2009.
BRODE, H. L. Point Source Explosions in Air; The Rand Corp., Research Memo RM-1824-AEC,
1956.
BROWN, D., C. COLLINS and B. KENNET (2003). The Woomera infrasound and seismic
experiment, September-October 2002, in Proceedings of Infrasound Technology Workshop, La
Jolla, California.
BROWN, D., I. KITOV, N. BRACHET, P. MIALLE and R. LE BRAS (2009). Enhancements to
the CTBTO infrasound processing system, Presented at the Infrasound Technology Workshop,
Brasilia, Brazil, November 2-6, 2009.
CHRISTIE, D., B. KENNET and Ch. TARLOWSKI (2005). Detection of regional and distant
atmospheric explosions at IMS infrasound stations, Proceedings of the 27th SRR, Palm Springs,
CA, Sep- tember 2005. @seismicisolation
@seismicisolation
228 Chapter Ten
CHRISTIE, D., J. VIVAS VELOSO, P. CAMPUS, M. BELL, T. HOFFMANN, A. LANGLOIS,

P. MARTYSEVICH, E. DEMIROVIC, and J. CARVALHO (2001), Detection of atmospheric
nuclear explosions: The infrasound component of the International Monitoring System,
Kerntechnik, 66, 96–101.
CONWEP (1997), a collection of conventional weapons effects calculations from the equations
and curves of TM 5-855-1 ‘‘Design and Analysis of Hardened Structures to Conventional
Weapons Effects’’. USAE Engineer Research and Development Centre, Vicksburg, MS.
CRATERDATABASE_V1.3 (2004). http://keith.aa.washington.edu/ craterdata/.
FEE, D., R. WAXLER, Y. GITTERMAN, J. GIVEN, J. COYNE, P. MAILLE, P. GRENARD, M.
GARCES, J. ASSINK, D. DROB, D. KLEINART, H. BUCHANAN, and L. ZERBO (2012).
Overview of the 2009 and 2011 Sayarim Infrasound Calibration Experiments (in preparation).
GARCES M., D. FEE, R. WAXLER, C. HETZER, J. ASSINK, D. DROB, A. LE PICHON, Y.
GITTERMAN, and R. HOFSTETTER (2009). The Sayarim Calibration Experiment: Theory
and Observations. Presented at the Infrasound Technology Workshop, Brasilia, Brazil,
November 2-6, 2009.
GARCES, M., C. HETZER, K. LINDQUIST and D. DROB (2002). Source location algorithm for
infrasonic monitoring, Proceedings of the 24th SRR, Ponte Vedra Beach, Fl, September 2002.
GITTERMAN, Y., V. PINSKY, A.-Q. AMRAT, D. JASER, O. MAYYAS, K. NAKANISHI, and
R. HOFSTETTER (2005). Source features, scaling and location of calibration explosions in
Israel and Jordan for CTBT monitoring, Isr. J. Earth Sci., 54: 199-217.
GITTERMAN Y. and R. HOFSTETTER (2008). Infrasound Calibration Experiment in Israel:
Preparations and Test Shots. Proceedings of the 30th Monitoring Research Review, LA-UR-08-
05261.
GITTERMAN, Y., M. GARCES, R. BOWMAN, D. FEE, H. ISRAELSSON, R. HOFSTETTER,
V. PINSKY (2009). Near-source and far-regional observations for Sayarim test explosions.
Proceedings of the 2009 Monitoring Research Review: Ground-Based Nuclear Explosion
Monitoring Technologies, LA-UR-09-05276, pp. 724- 734.
GITTERMAN, Y. (2010). Sayarim Infrasound Calibration Explosion: near-source and local
observations and yield estimation, in Proceedings of the 2010 Monitoring Research Review:
Ground- Based Nuclear Explosion Monitoring Technologies, LA-UR-10- 05578, Vol. II, pp.
708–719.
GITTERMAN, Y., J.W. GIVEN, J. COYNE, R. WAXLER, J. L. BONNER, L. ZERBO, R.
HOFSTETTER (2011). Large-scale controlled surface explosions at Sayarim, Israel, at
different weather patterns for Infrasound Calibration of the International Monitoring System, in
Proceedings of the 2011 Monitoring Research Review: Ground-Based Nuclear Explosion
Monitoring Technologies, LA- UR-11-04823.
GLOBALSECURITY.org, http://www.globalsecurity.org/wmd/ops/testing- effects.htm.
KINNEY, G. and K. GRAHAM (1985). Explosive shocks in Air, Springer-Verlag, New York, 9-
10.
KOPER, K., T. WALLACE, R. REINKE and J. LEVERETTE (2002). Empirical Scaling Laws for
Track Bomb Explosions based on Seismic and Acoustic Data, Bull. Seis. Soc. Am., Vol. 92, No.
2, 527-542.
LANDAU, L. (1945). On shock waves at large distances from the place of their origin, Applied
Mathematics and Mechanics, V. 9, No. 4, pp.96-103 (in Russian).
PERKINS, B. and W. JACKSON (1964). Handbook for prediction for air-blast focusing, BRL
report no. 1240.
NORRIS, D., J. BHATTACHARYYA and R. WHITAKER, 2007. Development of advanced
propagation models and application to the study of impulsive infrasonic events. Proceedings of
the 29th Monitoring Research Review, September 2007.
REVELLE, D., R. WHITAKER, W. ARMSTRONG (1998). Infrasound from the El Paso super-
bolide of October 9, 1997. LANL Report LA-UR-98-2983.
@seismicisolation
@seismicisolation
RINGDAL, F. and J. SCHWEITZER (2005). Combined seismic/infrasound processing: A case

study of explosions in NW Russia, NORSAR Sci. Rep. 2-2005, Semiannual Technical
Summary, Kjeller, August 2005.
SACHS, R.G. (1944). The dependence of Blast and Ambient Pressure and Temperature. BRL
report 466, May 1944.
SADWIN L. and SWISDAK, M. (1970). Blast Characteristics of 20 and 100 Ton Hemispherical
ANFO Charges. NOL Data Report, NOL TR 70-32, 17 Mar 1970.
STROUJKOVA, A., J. BONNER, M. LEIDIG, R. MARTIN, and P. BOYD (2012). Seismic
studies of explosions with different velocities of detonation in Barre granite, In press, BSSA.
STUMP, B. W., S. M. MCKENNA, C. HAYWARD and T-S. KIM (2002). Seismic and
Infrasound data and models at near-regional distances, Proceedings of the 24th SRR, Ponte
Vedra Beach, Fl, September 2002.
SWISDAK M. (2000). DDESB Blast Effects Computer, Version 4.0.
@seismicisolation
@seismicisolation
CHAPTER ELEVEN
SECONDARY SHOCK FEATURES FOR LARGE SURFACE EXPLOSIONS:

RESULTS FROM THE SAYARIM MILITARY RANGE, ISRAEL
AND OTHER EXPERIMENTS 1
YEFIM GITTERMAN
On-surface calibration explosion of 82 tons TNT-like explosives at Sayarim Military Range on 26

August 2009; a snapshot (after 0.9 s) from a speed video-recording at distance 755 m (video by the
Experiment Division of the Israel Defense Forces).
ABSTRACT
A series of surface explosions was designed and conducted by the Geophysical Institute of Israel at
the Sayarim Military Range in the Negev desert, including two large-scale explosions: approx. 82
tons of high explosives in 2009, and approx. 100 tons of low-grade ANFO explosives in 2011. The
main goal of the explosions was to provide large controlled sources for calibration of global
infrasound stations designated for monitoring nuclear tests; however, the geophysical experiment
also provided valuable observations for shock wave research. High-pressure gauges were deployed
at distances between 100 and 600 m to record air blast properties and to provide reliable estimation
of the true charge yield compared to the design value. Secondary shock phenomena were clearly
1 Shock Waves - An International Journal on Shock Waves, Detonations and Explosions (2014), Volume 24, Issue 3, 267–282. This
paper was based on work that was presented at the 22nd Symposium on the Military Aspects of Blast and Shock, Bourges, France,
November 4–9, 2012. @seismicisolation
@seismicisolation
Secondary Shock Features for Large Surface Explosions 231
observed at all near-source gauges as characteristic shock wave shapes. Secondary shocks were
also observed at numerous seismic and acoustic sensors deployed in the range 0.3–20 km as
acoustic phases. Empirical relationships for standard air blast parameters (peak pressure and
impulse) and for a new parameter called secondary shock time delay, as a function of distance,
were established and analyzed. The standard parameters, scaled by the cubic root of the estimated
TNT yield, were found to be consistent for all analyzed explosions. However, the scaled secondary
shock delays were clearly separated for the 2009 and 2011 explosions, thus demonstrating
dependence on the explosive type. Additionally, air blast records from other experiments were
used to extend the charge and distance ranges for the secondary shock observation, and showed
consistency with the Sayarim data. Analysis and interpretation of observed features of the
secondary shock phenomenon are proposed and a new empirical relationship of scaled secondary
shock delay versus scaled distance is established. The results suggest that the secondary shock
delay can be used as a new additional waveform feature for simple and cost-effective explosive
yield estimation.
Keywords Surface explosion · Air blast wave · Yield estimation · Secondary shock delay · Delay
dependence on explosives · Delay as new yield estimator
INTRODUCTION
The Geophysical Institute of Israel (GII) conducted several large-scale on-surface explosions at the
Sayarim Military Range (SMR) in the Negev desert of Southern Israel. The tests included an 82-
ton surface explosion in August 2009 and a 102 ton explosion in January 2011. The main goal of
these explosions was to provide fully controlled explosion sources for calibration of the
International Monitoring System (IMS) infrasound stations in Europe, the Mediterranean, the
Middle East and Asia, and to verify models of infrasound propagation in the atmosphere, for
summer and winter wind patterns. The IMS stations are designed for global nuclear test moni-
toring using recordings of low-frequency infrasound waves emitted from manmade and natural
sources on the Earth surface or in the atmosphere. To improve explosion monitoring capability
(detection, identification, location and yield estimation), the stations must be calibrated. The best
procedure for calibration includes fully controlled, large-scale surface explosions that produce
strong acoustic signals that can be observed at large distances. A well-established calibration
experiment should include accurate measurements of the detonation time, the GPS coordinates of
ground zero, as well as an estimate of the TNT equivalent yield, which may be different from the
design value depending on explosive efficiency or whether full detonation is obtained. The main
geophysical results of this unique dual explosion experiment are presented in [5,12,21].
Both tests, especially the 2011 one, were densely instrumented at near-source (0.1–0.6 km) and
local (1–37 km) distances with high-pressure, seismic and acoustic sensors that provided valuable
observations for shock wave research. The focus of this paper is to present some particular blast
effects and to examine characteristics of the generated air blast waves. Special attention is devoted
to numerous observations of the rarely reported secondary shock phase at different sensor types. A
new parameter is analyzed—the time delay between secondary shock and main shock arrivals—
that is dependent on charge and distance, as are standard air blast parameters, but also on the type
of explosive (e.g., its velocity of detonation, VOD). The obtained data and records from other
explosion experiments provide evidence that the easily measured secondary shock time delay can
be used as a new simple and cost-effective yield estimator.
SECONDARY SHOCK PHENOMENON
Numerous on-surface chemical explosions demonstrate clearly a remarkable air blast feature that
often arrives during the negative phase of the primary shock. A description of this known, but
rarely reported phenomenon, is presented by Baker [4]: “For any finite explosion source… blast
wave can also exhibit numerous repeated shocks of small amplitudes at various times… caused by
@seismicisolation
@seismicisolation
232 Chapter Eleven
successive implosion of rarefaction waves from the contact surface between explosion products
and air. Secondary and tertiary shocks of this nature have indeed been observed”. A more detailed
explanation is given by Needham [28], showing that in conventional explosions, after formation of
the primary shock, the high momentum of the expanding detonation products causes an
overexpansion and rarefaction of the interior, resulting in formation of a backwards-moving shock
that converges on the center and reflects, thus creating the secondary shock. Similar theoretical
descriptions and diagrams of the secondary and tertiary shocks for expanding spherical blast waves
are presented by Dewey [11].
In some reference works on shock waves, the secondary shock wave phenomenon is either not
mentioned (e.g. [15]), or not illustrated by overpressure–time empirical records (e.g. [11]). In some
papers, the secondary shock waves are not identified on included air blast plots [1,3,15,29,31].
This may be explained by less interest in this phenomenon due to its much smaller amplitudes
compared to those of the primary main shock and its corresponding weaker impact on structures
and/or humans.
Very few papers can be found devoted especially to an analysis of the secondary shock effect. It is
shown for small atmospheric spherical TNT charges 1–8 lb (0.45–3.63 kg) that secondary shock
delays depend on distance and charge- booster configuration [13]. The secondary shock waves are
identified for ANFO hemispherical charges of 20 and 100 tons, and distance-dependent relations
for secondary shock arrival times and peak pressures are presented [33]. A theory is developed
where the formation and motion of the secondary shock are given by explicit formulae, but only up
to returning back to the charge center [14]. Baker [4] mentions a note by Rudlin which indicates
that scaled secondary shock arrival times depend on the type of explosion source and presence of a
ground reflecting surface.
Theoretical overpressure time histories are presented in [6], calculated by Brode for two surface
explosions of 106 lb (453.6 tons) TNT-equivalent: chemical—with secondary shocks [8] and
nuclear—without secondary shocks [7]. There is evidence that in the past on-surface or near-
surface nuclear tests, secondary shocks were not found on air blast records, but only for high-
altitude explosions (Needham, Adushkin, pers. comm.).
Some researchers suggest that secondary shocks from gaseous explosions are implausible, or if
observed, are very weak because the density of the expanding gaseous detonation products is only
slightly above ambient air density and, therefore, the overexpansion is not as pronounced as in the
case of high density detonation products from a solid explosive [28]. However, a very clear
secondary shock for detonation of propane-oxygen gas in a small soap bubble is presented on a
pressure–time plot in [31], and other examples are observed in similar experiments (Sochet, pers.
comm.). A recent publication presented also the interpretation of some observed peaks from
bubble gas detonations as secondary shocks [36].
CHARGE DESIGN AND EXPERIMENTAL LAYOUT
Detailed descriptions of charge design are presented in [18– 20]. Only the key features important
for the description of the physical experiment and for the interpretation of the observed air blast
features are included here.
GII designed the sources for maximal concentration of explosives in the charge assembled from
numerous similar units and for increased energy release to the atmosphere to provide more distant
infrasound observations. A similar design was used in both experiments, however, the main agent
explosives, charge units and total charge yields were significantly different (Fig. 1; Table 1). The
explosion charge in August 2009 (Ex1) was composed of large barrels (315 kg each) filled with
strong cast high explosives, provided by Israel Military Industries Ltd. (IMI), deployed in a
compact pyramidal shape (Fig. 1a). Most of the air voids between the barrels were filled by plastic
bags with HE bulk explosives. A strong booster was composed of M-15 land mines (each one
@seismicisolation
@seismicisolation
containing 10 kg of Composition B) in a plastic box in the center of the basal layer, with a
multiple-point initiation scheme. The IMI cast explosives consisted of a mixture of different
recuperated high explosives: TNT and Gun powder as major agents, smaller amounts of
Composition B and RDX, and some other components providing a VOD higher than that of TNT
(Table 1).
Fig. 1 Desert view of the charge pyramid in August 2009 (a), final assembling stage of the 102 ton charge used in the
experiment on January 26, 2011 (b)
Table 1 Data of explosives and boosters used in the Sayarim explosions
The two explosions of 10 (Ex2) and 102 tons (Ex3) were conducted in January 2011 (on different
days). Unlike the 2009 test, the primary agent was an ANFO mixture (94 % Ammonium Nitrate
and 6 % Fuel Oil), supplied by Explosives Manufacturing Industries (1997) (EMI), Israel, placed
in large bags, assembled in an approximate hemispherical shape. The M-15 mines were again used
in the strong central booster and additional mines were placed beneath each bag and in air voids in
the four corners of the basal layer of the 102-ton explosion.
The VOD and density for the 2011 ANFO charges were measured by the supplier (EMI) as 2,400
m/s and 0.80–0.81 g/cm3, respectively [19,20]. ANFO has a broad range of VOD according to
numerous references. Higher values of 4,250–4,600 m/s were reported for similar 20 and 100 ton
ANFO surface explosions in Canada; however, these were not measured directly, but inferred from
an ionization probe and time of arrival data [34]. A slightly lower value of 3980 m/s was assumed
(Needham, pers. comm.), based on results of computer modeling of the 102-ton explosion by the
SHAMRC code [40]. It is supposed that the difference is caused by the fact that the supplier
conducted tests for small samples of ANFO only, whereas for very large quantities of such
explosives, increased VOD can be expected, due to essential self-tamping, resulting in a VOD
increase to 4,000–4,500 m/s. The important fact is that the VOD of the IMI explosives in 2009 is
much higher than for the ANFO tests in 2011. The parameters of different explosives are relevant
for the following analysis of the secondary shock delay variance (Sect. 10).
It is known that the energy released by a surface explosion varies greatly with the soil conditions
(e.g. [22]). The charges in both the 2009 and 2011 experiments were placed on the same soft
sediment surface with less than 30 m separation (Fig. 2). In the explosion area, the subsurface
layer of approx. 0.5–1 m consisted of dry soft and loose sediments, with consolidated sediments
@seismicisolation
@seismicisolation
234 Chapter Eleven
below. Geologically, the subsurface media are presented by Quaternary alluvial conglomerates,
underlain by consolidated limestone, chalk, and chert rocks [20].
Various measurement and observation sensors and systems were deployed at close distances
including high- pressure gauges, accelerometers, normal-speed and high- speed video cameras.
The experiment layout of the explosions and recording systems, similar for both experiments, is
shown in Fig. 2.
Fig. 2 Location of the explosions and near-source measurement systems (a): high-pressure gauges (filled bullet),
accelerometers (filled square), and high-speed ( ) video cameras, (b) helicopter view of the explosion site in 2011
Table 2 shows the accurately measured detonation time (DT), GPS coordinates (accuracy 4–5 m),
the total weight of all explosives and local seismic magnitude Md of the explosions. The DT values
were recorded by an electric circuit attached to the detonator with a PC-based system that includes
GPS time. The magnitude was estimated from signals recorded on the Israel Seismic Network
[21].
@seismicisolation
@seismicisolation
In-general, cost-effective compact charges, assembled in only 1–2 days, were used in this
geophysical experiment. They were significantly less expensive than several large- scale uniform
hemispherical charges detonated in the US and Canada [27,33] designed for studying nuclear
weapon effects in the near-source zone. The Sayarim charges were complex, non-homogeneous,
with air voids between units, far from the ideal hemispherical shape, resulting in nonlinear blast
effects and anomalies at close distances (see Sects. 6, 7). However, these effects were not
significant for far-regional infrasound observations [5,12]. Strong boosters and the upward
charge detonation scheme, utilized in the Sayarim surface explosions, provided a reduced energy
release to the ground and an enlarged energy radiation to the atmosphere, thus providing enhanced
opportunities for observation of infrasound signals at great distances. The smaller than expected
crater sizes and local seismic magnitudes confirmed the required explosives energy partition [21].
FREE-FIELD HIGH-PRESSURE MEASUREMENTS
High-pressure sensors were deployed in the near-source zone to (a) evaluate the efficiency of the
charge design and energy generation of explosives and (b) estimate the energy released by the
explosions, as the TNT equivalent yield of the charge. Furthermore, the records of air blast waves
were utilized to analyze and explain the explosion phenomena from these complex charges. Six
high-pressure gauges XTL- 190-5G/50A/100A were installed along a line in the distance range
100–600 m (Fig. 2a). The gauges with disc type baffles were mounted on steel rods approx. 1.5 m
above the surface providing side-on free-field overpressure measurements with a sampling rate of
2 MHz.
Obtained high-quality records demonstrated the expected positive pressure pulse of the main
shock, but interestingly there are secondary shocks observed, and at some records, even Tertiary
Shocks (TS) can be interpreted (Fig. 3) (details are given in the next sections). Basic (standard)
parameters of recorded air blast waves were accurately measured, including time of arrival TA,
peak pressure Pm, positive phase impulse I+, positive phase duration Ĳ+—for the main shock; and
parameters (TA, Pm)—for the secondary shock. From these values, a new air blast parameter—the
secondary shock delay ¨t T ss T ms—was calculated.
Because of the ringing of a fast acting transducer, some irregularities in the pressure time history
were observed, and the true peak pressure was difficult to measure directly. It was best determined
by fitting an exponent equation to a major part of the time history and extrapolating this to the time
of arrival of the primary shock. Then, more reliable estimates of ( Pm, Ĳ+) values for the main
shock and (TA, Pm) values for the secondary shock were obtained [17,20].
Table 2 Parameters of the SMR explosions (from [21])
a Pressure is calculated based on the altitude and the temperature.
@seismicisolation
@seismicisolation
236 Chapter Eleven
Fig. 3 Samples of the air blast overpressure records at different distances for Ex1 of ANFO (a), and Ex3 of IMI (b).
Secondary (SS) and tertiary shocks (TS) are also shown
@seismicisolation
@seismicisolation
Fig. 4 3-component accelerometer records of Ex1 at about 300 m (a) and Ex3 at 2 km (b). Arrivals of seismic P- and
S- waves, air blast main (MS) and secondary shock (SS) are shown (GMT time is presented in the headings)
Fig. 5 a Positions of seismic and acoustic sensors during the 2011 explosion Ex.3. b Sample record of seismic and
acoustic (air blast) waves at a WGC station at 2.2 km showing acoustic section (right). The station included co-located
3-component 1-Hz Sercel L4 velocity seismometer and Validyne overpressure sensors. Arrivals of seismic P- waves,
and air blast main, secondary and supposed tertiary shocks are shown
@seismicisolation
@seismicisolation
238 Chapter Eleven
AIR BLAST PHASES AT ACOUSTIC AND SEISMIC SENSORS

Two accelerometers were deployed by GII on the surface in the near-source zone and thus were
subjected to the impact of the air blast wave. For test Ex1, vertical accelerations up to 4 g,
corresponding to the main air shock (Fig. 4a), were measured at the closest sensor ACC1 (A1 on
Fig. 2).
For the 2011 explosions, additional acoustic and seismic sensors were deployed by GII partners:
the US Weston Geo- physical Corporation (WGC), the University of Mississippi (UM), and
University of Firenze (UF), Italy, at a distance range 1 to 37 km [5,19,20] (Fig. 5a) that also
provided good records of air blast waves (Fig. 5b). These records showed clear and strong main
shock peaks and smaller amplitude secondary peaks to distances of 20 km, with easily measured
arrival times.
Acoustic and seismic records obtained at these large distances from the explosions, far beyond the
near-source zone, are included here to demonstrate clear phases, interpreted as secondary shock
wave arrivals. Similar measurements and interpretations have not been presented before in
previous open publications. These arrival time measurements were utilized in the development of
the secondary shock delay scaling relationship (Sect. 10).
AUDIO-VISUAL BLAST OBSERVATIONS
Several home and safety monitoring video cameras (30 frame/s), and a high-speed Phantom-3
camera (3,100 frames/s) were placed at different distances and azimuths (Fig. 2), and recorded
distinctive blast phenomena. A snapshot from a video of Ex3 (102 tons), taken from a distance of
approx. 6 km, demonstrated an expanding, evidently following the shock wave, short-term white
spherical condensation cap, which is clearly visible for approx. 0.5 s due to specific air (humidity
61 %, temperature 13 ƕ& and lighting conditions (Fig. 6a). An analogous “milk-white hemi-
spherical dome” was reported for a large-scale surface explosion in Kazakhstan (the MASSA
experiment), conducted on 28.11.1981 at 8:30 local time in similar weather conditions [2]. This
phenomenon is common in more humid (e.g. ocean) environments, but is rarely seen for
explosions in arid (desert) test sites. A similar condensation effect is often observed around a jet
fighter at transonic flight in humid air, when an arising expansion wave rapidly cools a supersonic
flow and causes condensation of the ambient water vapor e.g. [25].
A snapshot of the high-speed video record for Ex2 (10 tons) showed non-spherical segments and
multiple air shock wave fronts at short times and distances (Fig. 6b). The same multiple shock
front phases were also observed in Ex3 (102 tons) and correspond to multiple peaks in the posi-
tive phase of the air shock wave recorded by close pressure gauges (Fig. 6c). Evidently, these front
phases/peaks are related to non-uniformity of the charges (Fig. 1), and cannot be attributed to any
specific charge elements. As distance increases to 100 m from the explosion, the different shocks
are merged resulting in a stable uniform shock wave (as in Fig. 3).
The secondary shock generation was uniquely revealed in the high-speed video record of Ex3,
demonstrating rapid processes in the epicenter area during the first 0.2 s: the record shows the
initial fast outward propagation of the primary shock in the air and on the ground surface, then a
cloud of gases and dust, which gradually decelerated up to a full stop (at approx. 0.12 s), and then
started moving inward back to the center of the explosion [20,21]. This observation is similar to
the above-mentioned theoretical descriptions from [4,11,28].
In the 2011 experiment, numerous observers, placed at 9 km and as far away as 37 km (Bonner,
pers. comm.) from the explosion center, heard two clear “bangs” with a delay of less than 1 s.
Observers interpreted these bangs as two separate explosions, due to the lack of reflecting walls or
hills in the area. Multiple “bangs” were also clearly registered at recordings of the safety
monitoring regular-speed video cameras (at several hundred meters) for both Ex2 and Ex3 [20].
@seismicisolation
@seismicisolation
Fig. 6 Snapshots from a home video camera at 6 km for Ex3 (a) and from a high-speed video recording for Ex2 at 360
m, 14 ms after the detonation (b), air blast overpressure record of Ex3 at a close gauge at 49 m (installed by Rafael
corp.) (c)
Initially, some researchers suggested that these calibration explosions were not instantaneous as
planned, but occurred in multiple steps (and that this was the reason for the observed double
“bang”), which resulted in numerous questions about the charge design and the initiation of
detonation being asked to GII. However, detailed analysis of the data showed that the peak
overpressures and impulses were consistent with the designed explosive yield, and that the audio-
phenomena reported by the observers and cameras were instead caused by air blast secondary
shocks, observed at records of high- pressure gauges and accelerometers (Figs. 3–5). The double
crack audio-effect was clearly heard in a video clip of Ex3 taken at the observation point. The
audio-channel was extracted from the video and visualized, providing accurate estimate of time
@seismicisolation
@seismicisolation
240 Chapter Eleven
interval between the two “bangs” (main and secondary shocks) as 0.75 s (Fig. 7a), which
corresponds to the secondary shock delay in air blast recording at a nearby UM acoustic sensor
(Fig. 7b). These observations, along with the yield estimates based on main-shock parameters (see
next section, Table 3), confirmed that both 2011 explosions were instantaneous and all explosives
were fully detonated in the initial 2–3 ms.
Fig. 7 Visualization of the audio-channel from a video clip, taken during Ex3 at the observation point at 9 km,
provided an accurate estimate of the time interval between the two “bangs” (main and secondary shocks) as 0.75 s (a),
the air blast wave record by a nearby NCPA acoustic sensor (deployed by UM) showed about the same secondary
shock delay (b)
Table 3 Estimates of the TNT equivalent yield for the SMR explosions, obtained from high-
pressure profile records by the BECv4 procedure, for the actual altitude and air
temperature.
Though the secondary shock peak pressure peak was about 3 times smaller than that of the main
shock (Fig. 3), some observers claimed that the two booms were of similar strength. The inner ear
is insensitive to overpressure jumps, but hears induced high (sound) frequencies. It is clearly
demonstrated at the video clip, where the microphone recorded two high-frequency audible
phases of similar amplitude, corresponding to observers hearing of main and secondary shocks
(Fig. 7a). It is interesting and important that during the 2009 explosion, observers at the same point
did not report any “double bang” effect, while clear secondary shock phases were also recorded
(see Fig. 3a). The reason is that secondary shock delay was much shorter for the IMI strong
explosives (see Sect. 10), and supposedly the two “bangs” were essentially merged and felt by the
ear as one long sound.
This audio-effect was also reported in some publications on large surface explosions. The “two
shocks” were heard distinctly, with an interval of about 0.5 s, at 2 km from the MASSA explosion,
but not explained and not correlated with the secondary shock effect [2]. Two distinctly audible
bangs of approximately the same loudness were heard within a short interval for charges of 3,000
kg and more, at 13 km from the firing point depending on the environmental conditions; the cause
of the double bang was not found, because there were no mountains or structures to reflect the
sound [30]. @seismicisolation
@seismicisolation
ESTIMATING TNT EQUIVALENT

Accurate and reliable TNT yield is an important Ground Truth (GT) parameter of a calibration
explosion, and it is crucial for development and testing of scaling relations of air blast parameters.
The reasons and importance of using TNT as equivalent for comparison of blast effects of differ-
ent explosions and types of explosives are discussed in many publications (e.g. [10,38]).
The estimation method used here is based on the DDESB Blast Effects Computer, Version 4.0
(BECv4), an Excel template [39]. This procedure provides computation of a wide variety of free-
field air blast parameters, for different explosives and charges (including on-surface sources), in
broad pressure, charge weight and distance ranges. The parameters are calculated by well-
established empirical scaling relationships based on numerous fully controlled experimental
explosions of hemispherical charges (e.g. [33]), calibrated to the observed altitude and air
temperature at the explosion ground zero.
The accurately measured values of peak overpressure, positive phase duration and impulse for the
main shock of the explosions were utilized for calculation of an appropriate yield for each high-
pressure gauge that provided reliable records, and then averaged for each parameter over several
gauges [20,21]. The resulting yield estimates and the appropriate standard deviations are presented
in Table 3. The TNT equivalences of an explosive charge based on these different air blast
parameters should not be expected to be the same. Observed increased Pm and decreased Ĳ+ values,
compared to expected values obtained by the BECv4 procedure, resulted in significantly
overestimated and underestimated yields, respectively. The positive phase impulse I+ is the
integral and stable characteristic of the air blast wave, unlike the one-point peak pressure
amplitude, and provides the best yield estimates with minimal standard deviations. Therefore, the
impulse-based estimate was considered as the GT parameter for the Sayarim calibration explosions
[5,12,21]. These values were used in the following evaluation of scaled-yield relationships for
basic air blast parameters, discussed in the next section. Note that the impulse-based yield estimate
is close to the average of estimates by both parameters (Pm and Ĳ+), because the impulse is
proportional to their product. The same blast wave parameters and a similar multi-station
procedure were applied to charge estimation for track bomb explosions, concluding that the most
reliable and accurate acoustic wave property for yield estimation is the air blast impulse [24].
The obtained TNT yield estimation for the explosion Ex1 is about 20 % more than the total weight
of the charge. There are several possible factors that can explain the enhanced air blast energy and
the increased TNT yield, including strong IMI explosives with high VOD (approx. 7,500 m/s for
the whole charge), approx. 10 % higher than for TNT (6,900 m/s); high concentration of
explosives, when most air voids between the charge units were filled with HE; very strong booster;
multiple initiation scheme; upward detonation of the charge.
Based on the known standard ANFO to TNT equivalence of 0.82, the expected equivalent TNT
yield values should be 8.5 tons for Ex2, and 84.5 tons for Ex3, whereas the obtained yield
estimates are lower by approx. 13 % for Ex2, and approx. 9 % for Ex3. Possible factors for lower
air blast energy and appropriate reduced equivalent TNT yield are not quite hemispherical shapes,
inhomogeneous charges and numerous air voids between the charge units. All these factors caused
a non-uniform charge that resulted in the observed blast anomalies, such as jetting, asymmetrical
and multiple air blast fronts, turbulence, collisions of wave fronts, and consequently in some
energy losses.
The obtained TNT yield estimations were validated by computer modeling of these explosions
using a hydrocode numerical approach [23, Kivity, pers. comm.]. For this modeling, the effective
internal energy of the charge was determined that provided the best fit to near-source high-
pressure measurements: 4.9 MJ/kg for 2009 (82 tons) and 3.05 MJ/kg for 2011 (102 tons), when
for TNT this value is the commonly used average 4.2 MJ/kg (e.g. [26]). Using these data, the TNT
@seismicisolation
@seismicisolation
242 Chapter Eleven
equivalent yields can be calculated as 95.7 and 74.1 tons for 2009 and 2011 explosions,
respectively, very close to the GII estimates in Table 3.
EMPIRICAL SCALING RELATIONSHIP FOR BASIC AIR BLAST PARAMETERS
Scaling relationships for the main shock peak overpressure Pm, positive phase duration Ĳ+ and
impulse I+ were calculated, using high-pressure gauges at near-source distances (0.1–0.6 km) for
the three explosions. The charge cubic root scaling law W1/3 was applied to distance, r, and I+,
using the estimated TNT equivalent charge, W (Table 3). In addition, available data for altitude H
and temperature T (Table 2) facilitated scaling of the measured air blast parameters to the
reference atmospheric conditions at sea level (H = 0 m, T = 15 ƕ& following the Sachs scaling
procedures [32,33]. The obtained RMS power fit curves for Pm and I+ are presented in Fig. 8.
Positive phase duration Ĳ+ values provide an improved fit to a polynomial curve [21]. The
reference (TNT standard) values, obtained by the BECv4 procedure, are also shown.
The scaled observed data demonstrate a high consistency for all 3 tests of different type of
explosives and TNT yield. There is a discrepancy of the scaled data with the reference (by BECv4)
values (as appropriate to the analysis of the TNT yield estimation presented before): measured
peak pressures are slightly larger (Fig. 8a), whereas positive impulses show almost the same
values as the TNT standard curve (Fig. 8b). This result corresponds to impulse-based yield
estimates. Low variance and high uniformity of the scaled data, especially of the positive impulse,
for all 3 shots, verify the yield evaluation procedure and estimated TNT yields that are crucial for
scaling relations.
Fig. 8 Air blast data measured in 3 large tests at SMR (with different type of explosives) versus distance, scaled to the
TNT yield charge and the sea level conditions, show a high uniformity for peak pressure (a) and positive phase
impulse (b). Reference (TNT standard) values, obtained by the BECv4 procedure, are also shown
SECONDARY SHOCK OBSERVATIONS IN THE SAYARIM EXPERIMENT

Distinct secondary shock waves were observed during the negative phase of the pressure–time
curves for all three surface explosions at high-pressure gauges placed at distances >100 m. For
both ANFO explosions (Ex2 and Ex3), the secondary shock peak overpressures were found to be
negative or close to zero (Fig. 3b), similar to secondary shock wave observations for ANFO trials
in Canada [33]. The secondary shock arrival times correspond to the statement that secondary
shock phase is “usually located at or near the minimum of the negative phase” [35]. However, this
is valid only for ANFO explosions. The cast IMI explosives used in the 2009 test (with increased
density and VOD, compared to ANFO, Table 2) produced smaller secondary shock delays than for
ANFO explosives (see next section). The secondary shock wave arrives much earlier than the
minimum of the negative phase, and superposition of the main and secondary shocks at an
observation point results in positive secondary shock peak overpressures, as was recorded for the
@seismicisolation
@seismicisolation
82-ton Ex1 in 2009 (Fig. 3a).

The secondary shock wave is formed and starts propagation at very close distances [4,28],
approximately next to the charge boundary. However, in these tests, it is not observed at high-
pressure sensors closer than 100 m to the explosion center (e.g. see Fig. 6c). A higher pressure air
shock front propagates faster; therefore, the time delay !Jt between the main and secondary shocks
increases with distance and larger secondary shock delays are observed for larger distances, as
well as for larger charge sizes [13,33]. Presumably, at short distances, when the secondary shock
delay is still small, the secondary shock wave arrives during the main shock positive phase, at the
time when pressures are significantly higher than the secondary shock peak pressure.
Consequently, superposition of the main and secondary shocks would not result in a clear
manifestation of a secondary shock arrival (e.g., a jump) (see Fig. 11a), as in the case of late
secondary shock arrivals at larger distances, where they occur during the main shock negative
phase with low pressures, comparable to the secondary shock peak pressure.
The secondary shock phases were observed on all portable seismic and acoustic sensors at
medium-range distances from 0.3 to 20 km. Tertiary shocks can be also tentatively identified at
some high-pressure gauges, although more clearly for the 2009 test (Fig. 3), as well as at some
acoustic and seismic sensors (Fig. 5).
Numerical two- and three-dimensional simulations of the 102-ton ANFO detonation (Ex3) were
conducted by Applied Research Associates, Inc. (USA), using the hydrocode SHAMRC, and
yielded overpressure and impulse waveforms that agreed well with the experiment, including the
capture of the observed secondary shocks [40].
SECONDARY SHOCK DELAYS: SCALING RELATIONSHIPS AND DEPENDENCE ON
EXPLOSIVE TYPE
For more complete analysis of this unique air blast feature, the secondary shock delay data set of
available high-pressure gauges at distances 0.1–0.6 km was extended, utilizing a broader
observation range. Accelerometer observations of secondary shock phases at distances from 0.3 to
2 km were added. The 2011 ANFO explosions (Ex2 and Ex3) provided observable secondary
shock arrivals at seismic and acoustic sensors up to 20 km.
Fig. 9 Scaled secondary shock delay versus scaled distance for the Sayarim tests with different type of explosives
When applying the charge cubic root scaling to the secondary shock delay !Jt and distance r (1),
using the TNT equivalent charge W (from Table 3), a remarkable correlation was found between
the scaled delay Dt and scaled distance R, for the two 2011 ANFO tests (Fig. 9).
Dt = ¨t/W1/3( s/kg1/3), R = r/W1/3 (m/kg1/3) (1)
@seismicisolation
@seismicisolation
244 Chapter Eleven
Small differences in the air temperature and altitude for different explosions, causing minor
variances in air blast parameters, comparable with measurement errors of distance and wave
parameters, were assumed to be negligible.
A new empirical scaling relationship was established, based on observations of the 2011 Sayarim
tests. A linear RMS fit regression curve of the scaled delay Dt versus logarithmic scaled distance R
was obtained, with a high correlation parameter C 2, in the distance range 0.1–20 km, for surface
hemispherical ANFO charges, placed on dry soft sediment surface [(2), Fig. 9]:
Dt (s/kg1/3) = 0.0057565 ‫ כ‬log (R) + 0.0033,
C 2 = 0.987, 2 < R < 1,000 m/kg1/3 (2)
However, the secondary shock delay data for the 2009 Ex1 with IMI explosives, comprising all
available records of high- pressure gauges and accelerometers at a distance range 0.2–
0.6 km, were found clearly separated from the 2011 ANFO data. The IMI regression also showed a
linear relationship between the scaled delay and the log-scaled distance, but with significantly
lower values than the ANFO fit curve [(3), Fig. 9]:
Dt (s/kg1/3)= 0.0039144 ‫ כ‬log (R) + 0.0017,
C 2 = 0.997, 4 < R < 14 m/kg1/3 (3)
(Note the higher correlation for IMI data, although in a much smaller distance range).
The revealed distinction indicates a dependence in the secondary shock delay on the charge
explosives type. The VOD for the IMI explosives is much higher (approx. 2 times) than that of
ANFO (Table 1). The IMI explosives are much stronger than ANFO and more energetic than TNT
(Table 3). Contrary to the secondary shock results, the standard air blast parameters ( Pm and I+),
corrected for the charge cubic root scaling, showed a high uniformity for all 3 tests with different
explosives (Fig. 8). This suggests that there is explosion product discrimination information
contained in the secondary shock delay.
APPLICATION OF SECONDARY SHOCK ANALYSIS TO OTHER EXPLOSIONS
The source yield and observation distance ranges were extended by compiling acoustic records of
additional ANFO surface explosions obtained from a literature search. These records include:
1. detonation of propane-oxygen gas in a (hemispherical) soap bubble of radius 0.06 m,
recorded at 0.15 m, with TNT equivalence approx. 76 % and VOD = 2,400 m/s [37], when
the estimated TNT equivalent was only 0.0013 kg [31] (Fig. 10a);
2. large-scale ANFO hemispherical explosions at White Sands Military Range (WSMR)
including Distant Image (2,214 metric tons), and Minor Uncle (2,472 metric tons) (which
was a particularly uniform charge) [27]; raw data recorded at distances 28–60 km were
utilized, obtained from the USA [Bonner, pers. comm.] (Fig. 10b);
3. the earlier mentioned MASSA explosion in Kazakhstan of 251 tons of ANFO-like
explosives, records at distances 1.4, 2.0 and 8.6 km [3] (Fig. 10c);
4. ANFO trials of 20 and 100 tons in Alberta, Canada, where secondary shocks arrival times
were measured [33] (Fig. 10d).
Secondary shock phases were identified and the secondary shock delays were measured and scaled
for the TNT equivalent yield. Adequate agreement between the additional data and the 2011
ANFO tests was found (Fig. 12). Thus, the established relationship (2) sufficiently describes the
secondary shock delay data over a broad range of ANFO (and ANFO-like) charges from 1gram up
to 2,000 tons (TNT equivalent) and distance range 0.15 m–60 km.
@seismicisolation
@seismicisolation
The secondary shock from the gaseous detonation (Fig. 10a) is weak for the reason mentioned
before (see Sect. 2); however, it is clear and easily measured. More examples of gaseous
detonations should be collected (e.g. from [36]) for detailed analysis and verification of the
secondary shock effect in a separate research.
Fig. 10 Examples of ANFO and ANFO-like explosions where secondary shock phases were found including: (a) gas
bubble detonation (reprinted from [31] with permission from Sochet), (b) WSMR explosion Distant Image (20 June
1991) of 2,214 tons ANFO, recorded at 28 km by co-located pressure sensor and 2 seismic channels, (c) MASSA
explosion of 251 tons of ANFO-like explosives (reprinted from [3] with permission from Kulichkov), (d) ANFO trial
of 18.2 tons in Alberta, Canada, at 235 and 408 m (reprinted from [33] with permission from Sadwin)
@seismicisolation
@seismicisolation
246 Chapter Eleven
Data of several surface explosions with different explosives more powerful than ANFO (with
higher VOD approx. 6,900–7,800 m/s) were also examined for comparison to the secondary shock
delay observations for the 2009 test Ex1 with similar explosives parameters. These explosions
include:
1. small TNT charge of 0.734 kg, recorded at 5.2 m [15, Fig. 3.39];
2. cast TNT hemispherical charge of 830 kg, detonated at SMR in January 2008, and recorded
by high-pressure gauges at 18 and 19 m; the charge of radius r0 0.5m was placed on a
wooden plate of approx. 0.1 m height; the scaled height is approx. 0.01 m/kg1/3, the
configuration can thus be considered as a surface burst, and the observations were made at a
distance r >> r0 [16, Fig. 8];
3. Approximately hemispherical charges of recuperated TNT, RDX and Composition B (pieces
in big bags) with estimated TNT equivalent of 1 and 5.3 tons, conducted by GII at SMR on
2–3 Dec 2008 (tests prior to the large 82-ton test Ex1 in 2009), and recorded by high-
pressure
4. explosion of 1,000 tons TNT in Russia, 25.08.1987 [1] (Fig. 11b).
Good agreement was found with the 2009 test data (Fig. 12), thus forming a consistent set of data
for high explosives with VOD approx. 6,900–7,800 m/s, clearly separated from ANFO (and
ANFO-like) data. A clear dependence of the secondary shock delay on the type of explosives with
different velocity of detonation is evident.
Fig. 11 Examples of non-ANFO explosions where secondary shock (SS) phases were observed and measured: a SMR
test 1 ton high explosives, recorded at 149 m (from [17]), note an early secondary shock arrival during the positive
main shock phase resulting in relatively weak appearance (as mentioned in the previous section), b air blast wave
profile for 1,000-ton explosion, recorded at 70 and 997 m (reprinted from [1] with permission from Adushkin)
@seismicisolation
@seismicisolation
Fig. 12 Scaled secondary shock delay versus scaled distance for different explosions and explosives
SECONDARY SHOCK DELAY AS A NEW YIELD ESTIMATOR

Numerous observations of the surface SMR explosions and previous experiments with variable
explosives and design, in broad charge and distance ranges, demonstrate a special character of the
secondary shock delay. It is a stable air blast feature that is different from other basic parameters
(e.g., Pm, Ĳ+ and I+) that must be measured by expensive and well-calibrated high-pressure sensors
that are protected from air blast impact in the near-source zone. By contrast, the secondary shock
delay is:
1. easily measured, by any simple low-cost acoustic (microphone) or seismic (accelerometer)
sensor; it can also be measured by the audio-channel of any home video camera (as shown
in Sect. 5, Fig. 7) that may be the only available close record in the case of an unexpected
blast accident; the sensor does not have to be calibrated;
2. a sensor that can be deployed at remote locations, in the low overpressure range, and the
station, cables and recorder need not be protected from the blast impact like the expensive
high-pressure gauge system (e.g. used in the SMR experiments);
3. a differential parameter ¨W = ࢀ࢙࢙ ࢓࢙
࡭ í ࢀ࡭ ), therefore gauges at distances 138–244 m [17] (Fig.
11a); records of free-field or reflected (e.g. as in Fig. 10a) air blast wave can be utilized;
4. it is not necessary to know (or measure) the exact detonation time (e.g., in the case of using
the main shock arrival time for the yield estimation), this is critical in the case of a blast
accident;
5. stable to atmospheric conditions—speed and direction of the wind, that affect air
blast/acoustic wave amplitudes (at local distances), but evidently change equally arrival
times of the main and secondary shocks; and it depends on the explosives properties, thus
serving in some cases as a blast product discriminator.
The obtained results show that the air blast parameter secondary shock delay can be used as a new
potential yield estimator, based on the established scaled relationship. For surface tests of ANFO
explosives placed compactly on dry soft sediment surface, if secondary shock phases are identi-
fied on acoustic and seismic records, then the charge weight can be accurately estimated.
On the other hand, it is possible that the explosive type can be identified and the detonation
velocity can be roughly estimated, in the case the charge weight is known and secondary shock
delays are measured. Moreover, a delay variance may provide valuable information about the
quality of the detonation (as pointed out in@seismicisolation
[13]), if the explosive and VOD are well-known.
@seismicisolation
248 Chapter Eleven
DISCUSSION
It is presumed that an increase of the secondary shock delay with charge size and a decrease with
detonation velocity correlate with the characteristic time of complete detonation of a finite
hemispherical charge. A second possible explanation is that secondary shock delays for IMI
explosives (and for TNT, see Figs. 11, 12) are decreased due to the TNT afterburning effect which
results in approaching the secondary shock from the minimum overpressure (in the negative
phase), very close to the end of the positive phase ([9], [Dewey, pers. comm.]). As mentioned, the
main agent of the IMI explosives was recuperated TNT (pieces).
Evidently, the VOD is not the only explosive parameter that may affect the delay between the two
shocks, but clearly provided strong differences between the 2009 and 2011 tests. Possibly, the
effective internal energy or acoustic impedance of the explosive material (density multiplied by
sound speed) is important. Ambient media factors can also affect the position of the second shock,
but this analysis needs additional research.
Possibly, the new method of TNT yield estimation based on secondary shock time delay
measurements is less accurate than traditional procedures, based on near-source high- pressure
measurements, providing the best characterization of air blast waves and yield estimation.
However, it is significantly less expensive (valuable for low budget experiments), and in the case
of an accidental explosion it may be the only method available. The new estimator can be also
relevant for some field experiments with complicated logistics, when it is necessary to conduct a
series of explosions at different locations.
The proposed technique has some limitations, as it is sensitive to the explosive type and currently
limited to ANFO and ANFO-like explosives. It cannot be used directly to compare the charge
masses of different explosives; however, it should work well when determining the relative size of
two charges of the same explosive [Dewey, pers. comm.].
As mentioned before (Sect. 2), surface nuclear tests, unlike chemical explosions, may not produce
a secondary shock. Supposedly, this distinction, if it exists, could be explained by a different
source phenomenology, because a nuclear test provides an instantaneous and point-like source, for
which the mass of the charge is extremely small relative to the energy release. In this case, the
features of chemical tests, crucial for the secondary shock effect—charge size, detonation velocity,
expanding detonation products, overexpansion and rarefaction of the interior—are not applicable.
Nevertheless, some cases can be imagined, when the secondary shock feature appearance can be
used for identification of the source: nuclear or chemical explosion.
CONCLUSIONS
Large-scale surface explosions were conducted by the Geo- physical Institute of Israel at the
Sayarim Military Range in Israel for calibration of infrasound stations in Europe, Middle East and
Asia. The explosions generated the rarely reported air blast feature—the secondary shock
phenomenon that initiated this study. Secondary shock waves were observed at all near-source
high-pressure gauges as characteristic shock wave shapes, but also at numerous seismic and
acoustic sensors, deployed in the 0.3–20 km distance range as arrivals of high-frequency acoustic
phases—such measurements and interpretations have not been presented in previous studies.
The observations have demonstrated a special characteristic of the secondary shock delay,
different from other commonly used basic air blast parameters such as peak pressure, positive
duration and impulse. Using the charge cubic root scaling law (for the estimated TNT equivalent
charge), standard relationships for the basic parameters versus distance were calculated, showing a
high uniformity for all analyzed Sayarim tests with different explosives type.
A clear correlation between the scaled delay and the logarithmic scaled distance was revealed, and
@seismicisolation
@seismicisolation
a new empirical relationship was established. Noticeable separation between tests of ANFO and
stronger and faster VOD explosives was found, thus demonstrating clearly dependence on
explosives properties and discrimination capability for the explosion type. This means that two
tests with different explosives but the same TNT equivalent yield, recorded at the same distance,
will show approximately the same peak pressure, impulse, positive phase duration, but different
secondary shock delays that are smaller for explosives with higher detonation velocity (one of the
factors affecting the delay).
Analysis of secondary shock features helped to explain the reason and conditions of an audio
“double bang effect”, reported for the 2011 explosion Ex3 and some other large- scale explosions.
The results of the secondary shock delay observations for the SMR tests were verified using data
of previous explosion experiments with different design and explosives, for a broad range of
charges (1g–2,000 tons) and distances (0.15 m– 60 km).
The results provide evidence that the secondary shock delay, which is a stable, reliable and easily
measured air blast parameter, can be considered as a new yield estimator, providing a cost-
effective estimation procedure requiring simple measurement systems. This new method for yield
estimation has not been considered before, and future studies should take the derived empirical
relationships in this paper to estimate yields of additional surface explosions.
ACKNOWLEDGMENTS
The Sayarim experiments were supported by the US Army Space and Missile Defense Command
(in 2009) and the Comprehensive Test Ban Treaty Organization, Vienna (in 2011). Thanks to
collaborators from the WGC (J. Bonner), the UM (R. Waxler) and the UF (E. Marchetti) for
supplementing data of close seismo-acoustic observations in 2011 experiment. The research group
of the Protective Technologies R&D Center, Ben-Gurion University (G. Ben-Dor, O. Sadot and A.
Britan) provided fruitful discussions of observed air blast wave effects and secondary shock
features, and supplement of useful references. Thanks to C. Needham, J. Dewey, L. Sadwin and to
anonymous reviewers for valuable comments to the research and the manuscript. I appreciate kind
help of J. Bonner, R. Waxler and R. Hofstetter (GII) for essential improving the English and
writing style of the manuscript. Research work of the author was supported by the Israel Ministry
of Immigrant Absorption.
REFERENCES
1. Adushkin, V.V., Khristoforov, B.D.: Action of the coastal 1000-ton surface explosion on the
environment. Combust. Explos. Shock Waves, 40(6), 686–693(8) (2004)
2. Alperovich, L.S., Gokhberg, M.B., Drobzhev, V.I., Troitskaya, V.A., Fedorovich, G.V.:
Project MASSA: an investigation of magnetosphere-atmosphere relationships during
seismoacoustic events. Izvestiya Earth Phys. 21(11), (1985)
3. Alperovich, L.S., Afraymovich, E.L., Vugmeister, B.O., Gokhberg, M.B., Drobzhev, V.I.,
Yerushchenkov, A.I., Ivanov, E.A., Kalikhman, A.D., Kudryavtsev, V.P., Kulichkov, S.N.,
Krasnov, V.M., Mordukhovich, M.I., Matveyev, A.K., Nagorskiy, P.N., Ponomarev, E.A.,
Salikhov, N.M., Tarashchuk, Yu.E., Troitskaya, V.A., Fedorovich, G.V.: The acoustic wave
of an explosion. Izvestiya Earth Phys. 21(11), (1985)
4. Baker, W.E.: Explosions in Air. University of Texas Press, Austin (1973)
5. Bonner, J., Waxler, R., Gitterman, Y., Hofstetter, R.: Seismo- Acoustic Energy Partitioning
at Near-Source and Local Distances from the 2011 Sayarim Explosions in the Negev Desert.
Israel. Bull. Seismol. Soc. Am. (2012). doi:10.1785/0120120181
6. Bracco, F.V.: Air Shock Parameters and Design Criteria for Rocket Explosions. Wyle
Laboratories, Report WR 65-21 (1965)
7. Brode, H.L.: Point Source Explosions in Air. The Rand Corp., Research Memo RM-1824-
AEC (1956)
@seismicisolation
@seismicisolation
250 Chapter Eleven
8. Brode, H.L.: A Calculation of the Blast Wave From A Spherical Charge of TNT, AD
144302 (1957)
9. Dewey, J.M.: The air velocity in blast waves from t.n.t. explosions. Proc. R. Soc. A 279,
366–385 (1964)
10. Dewey, J.M.: The TNT equivalence of an optimum propane-oxygen mixture. J. Phys. D
Appl. Phys. 38, 4245–4251 (2005). doi:10. 1088/0022-3727/38/23/017
11. Dewey, J.M.: Spherical shock waves: chapter 13.1 expanding spherical shocks (Blast
Waves). In: Ben-Dor, G., Igra, O., Elperin, T. (eds.) Handbook of Shock Waves. ISBN: 978-
0-12-086430-0. Elsevier (2001)he 18th
12. Fee, D., Waxler, R., Assink, J., Gitterman, Y., Given, J., Coyne, J., Mialle, P., Garces, M.,
Drob, D., Kleinert, D., Hofstetter, R., Grenard, P.: Overview of the 2009 and 2011 Sayarim
infrasound calibration experiments. JGR Atmos. 118, 1–22 (2012). doi:10. 1002/jgrd.50398
13. Fisher, E.M., Pittman, J.F.: Air Blast Resulting from the Detonation of small TNT Charges,
NAVORD, Report 2890 (1953)
14. Friedman, M.P.: A simplified analysis of spherical and cylindrical blast waves. J. Fluid
Mech. 11, 1–15 (1961)
15. Gelfand, B., Silnikov, M. (2006) Barothermic Effect of Explosions. S. Petersburg (in
Russian)
16. Gitterman, Y., Hofstetter, R.: Infrasound calibration experiment in Israel: preparation and
test shots. In: The 30th Monitoring Research Review on Ground-Based Nuclear Explosion
Monitoring Technologies, Portsmouth, Virginia, pp. 882–893 (2008)
17. Gitterman, Y., Garces, M., Bowman, R., Fee, D., Israelsson, H., Hofstetter, R., Pinsky, V.:
Near-source and far-regional infrasound observations for Sayarim test explosions. In: The
31th Monitoring Research Review on Ground-Based Nuclear Explosion Monitoring
Technologies, Tucson, Arizona, pp. 724–734 (2009)
18. Gitterman, Y.: Sayarim infrasound calibration explosion: near-source and local observations
and yield estimation. In: The 32th Monitoring Research Review on Ground-Based Nuclear
Explosion Monitoring Technologies, Orlando, Florida, pp. 708–719 (2010)
19. Gitterman, Y., Given, J., Coyne, J., Waxler, R., Bonner, J., Zerbo, L., Hofstetter, R.: Large-
scale controlled surface explosions at Sayarim, Israel, at different weather patterns, for
infrasound calibration of the International Monitoring System. In: The 33th Mon- itoring
Research Review on Ground Based Nuclear Explosion Monitoring Technologies, Tuscon,
Arizona. LA-UR-11-04823, pp. 766–777 (2011)
20. Gitterman, Y., Hofstetter, R.: GT0 Explosion Sources for IMS Infrasound Calibration, Final
Report on Contract No. 2010–1037 of CTBTO, GII, Report no. 547/631/11, 85 p. (2011)
21. Gitterman, Y., Hofstetter, R.: GT0 explosion sources for IMS infrasound calibration:
charge design and yield estimation from near- source observations. Pure Appl. Geophys.
(2012). doi:10.1007/ s00024-012-0575-4 (Springer Basel AG)
22. Hlady, S. L. Effect of soil parameters on land mine blast. In: The 18th International
Symposium on Military Aspects of Blast and Shock, Bad Reichenhall, Germany (2004).
http://www.gichd.org/ fileadmin/pdf/LIMA/MABS18_HLADY.pdf
23. Kivity, Y., Savir, Z., Edri, I., Ben-Dor, G.: The blast wave resulting from an 82 ton
explosion: test and analytical model. In: Shapiro, J, (ed.) Proceedings of the 14th
International Symposium on Interac- tion of the Effects of Munitions with Structures
(ISIEMS), 19–23 Sep 2011, Seattle, USA
24. Koper, K., Wallace, T., Reinke, R., Leverette, J.: Empirical scaling laws for track bomb
explosions based on seismic and acoustic data. Bull. Seismol. Soc. Am. 92(2), 527–542
(2002)
25. Krehl, P.O.K.: History of Shock Waves, Explosions and Impact. Springer, Berlin (2009)
26. Lee, J.H., Guirao, C.M., Chiu, K.W., Bach, G.G.: Blast effects from vapor cloud explosions.
Loss Prev. 2, 59–70 (1977)
27. Leverette, J.A., McCrory, R.A., Reinke, R.E.: Chapter 4: Far-Field Ground Motion and
@seismicisolation
@seismicisolation
Airblast Measurements. Report of Experimental Shock Physics Office, Field Command

Defense Nuclear Agency, 27 p. (1993)
28. Needham, C.E.: Blast Waves. Springer, Berlin (2010)
29. Pachman, J., Šelešovský, J., Künzel, M., Matyáš, R.: Blast wave parameters and TNT
equivalency of improvised explosives: Urea Nitrate. In: The 22th International Symposium
on Military Aspects of Blast and Shock, paper number P140 (2012)
30. Roach, P.K., Verster, W.F.: Discussing the detonation effects of large charges. In: SABO
Conference, 15–17 August 2006, 22p. (2006). http://hdl.handle.net/10204/1075
31. Trelat, S., Sochet, I., Autrusson, B., Loiseau, O., Cheval, K.: Strong explosion near a
parallelepipedic structure. Shock Waves 16(4–5) (2007)
32. Sachs, R.G.: The Dependence of Blast and Ambient Pressure and Temperature. BRL report
466 (1944)
33. Sadwin, L.D., Swisdak, M.M.: Blast Characteristics of 20 and 100 Ton Hemispherical
AN/FO Charges. NOL Data Report, NOL TR 70–32, 77 p. (1970)
34. Sadwin, L.D., Swisdak, M.M.: AN/FO charge preparation for large scale tests. NOL Data
Report, NOL TR 70–205, 62 p. (1970)
35. Sadwin, L.D., Swisdak, M.M.: The airblast and sonic boom second shock revelation. In: The
22th International Symposium on Military Aspects of Blast and Shock, paper number P6
(2012)
36. Sauvan, P.E., Sochet, I., Trélat, S.: Analysis of reflected blast wave pressure profiles in a
confined room. Shock Waves 22, 253–264 (2012). doi:10.1007/s00193-012-0363-1
37. Schultz, E.: Detonation Diffraction Through an Abrupt Area Expansion, PhD Thesis,
California Institute of Technology, 322 p. (2000).
http://www2.galcit.caltech.edu/EDL/publications/ reprints/schultz_thesis.pdf
38. Sochet, I.: Blast effects of external explosions. In: The 8th International Symposium on
Hazards, Prevention, and Mitigation of Industrial Explosions, Yokohama, Japan (2010).
http://hal. archives-ouvertes.fr/hal-00629253/
39. Swisdak, M.M.: DDESB Blast Effects Computer, Version 4.0 (2000).
http://orise.orau.gov/emi/hazards-assessment/files/resources/BlastEffectsComputer4.0
Description.pdf
40. Weber P., Millage, K., Crepeau, J., Happ, H., Needham, C., Git- terman, Y.: Numerical
simulation of a 100-ton ANFO detonation. In: The American Society of Mechanical
Engineers Verification and Validation Symposium (2012).
http://asmeconferences.org/VVS2012/pdfs/FinalProgram.pdf. #page=15
@seismicisolation
@seismicisolation
CHAPTER TWELVE
SPECTRAL MODULATION EFFECT IN TELESEISMIC P-WAVES FROM NORTH

KOREAN NUCLEAR TESTS RECORDED IN BROAD AZIMUTHAL RANGE
AND POSSIBLE SOURCE DEPTH ESTIMATION 1
YEFIM GITTERMAN, SO GU KIM AND RAMI HOFSTETTER
Seismic Arrays used in this study for North Korean Nuclear Explosions.
ABSTRACT
Three underground nuclear explosions, conducted by North Korea in 2006, 2009 and 2013, are
analyzed. The last two tests were recorded by the Israel Seismic Network. Pronounced coherent
minima (spectral nulls) at 1.2–1.3 Hz were revealed in the spectra of teleseismic P-waves. For a
ground-truth explosion with a shallow source depth, this phenomenon can be interpreted in terms
of the interference between the down-going P-wave and the pP phase reflected from the Earth’s
surface. This effect was also observed at ISN stations for a Pakistan nuclear explosion at a
different frequency 1.7 Hz and the PNE Rubin-2 in West Siberia at 1 Hz, indicating a source-effect
and not a site-effect. Similar spectral minima having essentially the same frequency, as at ISN,
were observed in teleseismic P-waves for all the three North Korean explosions recorded at
networks and arrays in Kazakhstan (KURK), Norway (NNSN), Australia (ASAR, WRA) and
1
First published in Pure Appl. Geophys. 173 (2016), 1157–1174.
@seismicisolation
@seismicisolation
Spectral Modulation Effect in Teleseismic P-waves from North Korean Nuclear Tests Recorded in Broad 253
Azimuthal Range and Possible Source Depth Estimation
Canada (YKA), covering a broad azimuthal range. Data of 2009 and 2013 tests at WRA and
KURK arrays showed harmonic spectral modulation with three multiple minima frequencies,
evidencing the clear interference effect. These observations support the above- mentioned
interpretation. Based on the null frequency dependency on the near-surface acoustic velocity and
the source depth, the depth of the North Korean tests was estimated about 2.0–2.1 km. It was
shown that the observed null frequencies and the obtained source depth estimates correspond to P-
pP interference phenomena in both cases of a vertical shaft or a horizontal drift in a mountain.
This unusual depth estimation needs additional validation based on more stations and verification
by other methods.
Key words: North Korean tests, teleseismic P-wave observations, spectral nulls, source-depth
estimation.
INTRODUCTION
The source depth is a key parameter for the assessment of the characteristics of an underground
nuclear test. An accurate estimate of the source depth can provide information about the technical
features of the test, as well as providing a more reliable TNT yield estimate. The yield is usually
calculated from the seismic magnitude, but the magnitude-yield relation is also known to be
dependent on the depth of the explosion (e.g., GITTERMAN 2009; MURPHY et al. 2013).
Location procedures based on propagation times of P and S-waves observed in regional and
teleseismic records normally result with large errors (several km) for these shallow seismic
events, and, therefore, a nuclear explosion depth is defined usually as zero (see for example USGS
and CTBTO bulletins).
The presence of characteristic minima in amplitude spectra (spectral nulls) of teleseismic P-waves
from nuclear tests has been considered in several publications, in relation to more accurate source-
depth evaluation. Spectral nulls were observed for three US tests and the obtained depth
estimations were compared with actual values, showing a relatively small error ~0.2 km (COHEN
1970). Teleseismic P-waves for two large Amchitka explosions were analyzed from records at 8
US stations in azimuthal range 44°–86°, and surface reflection time delays pP-P were estimated
(BAKUN and JOHNSON 1973). Pronounced spectral minima about 1 Hz were found in
teleseismic records of the Swedish seismic network for Nevada tests, with azimuths from 16° to
25° (KULHANEK 1971), and about 1.5–1.8 Hz for Semipalatinsk tests, with azimuths from 304°
to 324° (KULHANEK 1973). The minima, observed in rather narrow azimuthal ranges, were
interpreted in terms of interference of P- and pP-waves, with the subsequent depth estimation
analysis.
The main goal of this research was to study these characteristic spectral features from the three
underground nuclear explosions in North Korea, based on records at seismic stations and arrays, in
a very broad azimuthal range, and to show that it is not a site-effect but rather a source one, that
facilitates the test depth estimation.
SPECTRAL NULL PHENOMENON
Underground nuclear explosions are characterized by shallow focal depth, usually several
hundreds of meters, and more than 2 km only in a few tests (MIKHAILOV 1999), and by spherical
symmetry of the initial wave radiation. These features result in some cases in the destructive
interference effect between the down-going direct P-wave and pP-wave reflected from the Earth’s
surface (COHEN 1970; KULHANEK 1971; BAKUN and JOHNSON 1973), and this effect is
usually observed at teleseismic records.
The interference produces spectral modulation (scalloping), presented by minima (or spectral
nulls) at the fundamental (first) and multiple frequencies (e.g., KULHANEK 1971):
@seismicisolation
@seismicisolation
254 Chapter Twelve
f01 = Vp /(2 * h), f0n = n * f01, n = 1, 2, .. .; (1)

where Vp(m/s) is the average compressional (P-wave) velocity of the medium between the source
and the surface (near-surface velocity), and h(m) is the source depth.
Equation 1 is valid for teleseismic distances, where the incidence angle at the surface for pP is
small (neglecting take-off angles), and the fundamental period T01 equals the pP-wave delay
relative to the direct P-wave.
The destructive interference effect can be complicated by non-linear surface effects on the pP
reflection, like spallation and disrupted medium for the downgoing pP reflection that can obscure
spectral nulls (LAY 1991). Rough site topography will reduce the ideal free-surface reflection
coefficient (k = -1) and smear the spectral modulation.
Being the source-depth effect for the point-like nuclear explosion, the spectral null appears at
about the same frequency in the spectra of different network stations or array elements, placed at
different azimuths. Nevertheless, local subsurface geological features at the receiver can shift
individual null frequencies causing weakening of the spectral shape coherency.
SPECTRAL NULLS OBSERVATIONS AT ISRAEL SEISMIC NETWORK FROM
PREVIOUS TESTS
Spectral analysis of the Israel Seismic Network (ISN) records for some previous nuclear
explosions has revealed pronounced coherent minima in P-waves at about 1 Hz (and multiples at 2
and 3 Hz) from the Peaceful Nuclear Explosions (PNE) Rubin-2 test in West Siberia, Tyumen
region, in 1988 (GITTERMAN et al. 1999) and about 1.7 Hz for the first Pakistan test in 1998
(GITTERMAN et al. 2002), which were interpreted as P-pP interference (Fig. 1).
The Rubin-2 test, aimed for deep seismic sounding, was placed in underground shaft, where the
charge elevation beneath the sea level was -830 m. Considering the ground surface altitude in this
region about 40 m above the sea level (provided by GoogleTM Earth), the source depth h was about
0.87 km. Assuming an average near-surface P-wave velocity of about 1.8–2 km/s, for the 1-km
thick sedimentary structure in the West-Siberian Platform (MUSATOV 1998), Eq. 1 gives a
fundamental null frequency of 1 Hz as observed. The relatively high null frequency for the
Pakistan test possibly corresponds to the reported shallow depth of the test which was placed in a
horizontal drift of the mountain (WALLACE 1998).
PROCESSING AND ANALYSIS PROCEDURES
For better identification of coherent spectral shapes, indicating source-effect (and not site-effect),
our processing procedures included calculation and joint plotting of FFT amplitude spectra of
network vertical seismograms recorded at different azimuths. In order to enhance the spectral
modulation features related to the source-depth effect, the spectra were smoothed, depending on
the window length, using the convolution triangle weights filter, and then the average of all
analyzed station spectra was calculated. The joint log–log spectral plots provided better
observation of coherent minima and maxima at multiple frequencies (harmonics) with very
different amplitudes for network stations, and more accurate evaluation of spectral minima (nulls)
frequencies in the low-frequency range. In order to separate signal and noise spectral features, in
some cases the spectra of the pre-signal noise window were also calculated using the same window
length.
@seismicisolation
@seismicisolation
Figure 1. Vertical seismograms and P-wave spectra at Israel Seismic Network stations
demonstrating distinct spectral nulls and modulation and high spectral shape coherency: a PNE
test Rubin-2 in W. Siberia on Aug. 22, 1988 (reprinted from GITTERMAN et al. 1999); b the first
Pakistan nuclear test on May 28, 1998 (reprinted from GITTERMAN et al. 2002)
Initially this technique was developed and applied for revealing and enhancing spectral modulation
in the seismic waveforms, caused by interference phenomena for ripple-fired quarry blasts
(GITTERMAN and VAN ECK 1993), and for underwater explosions due to the bubble-effect
(e.g., GITTERMAN and SHAPIRA 1994; GITTERMAN 2002; KIM and GITTERMAN 2012). It
was also used in the analysis of ISN teleseismic records of previous nuclear tests (GITTERMAN
et al. 1999, GIT- TERMAN et al. 2002; GITTERMAN 2002).
Table 1. Compilation of basic parameters of North Korean tests as reported by the agencies
USGS, CTBTO and NORSAR
Date Origin time (UTC) Magnitude Yield (kT) Depth (km)
Oct 9, 2006 01:35:28 4.1–4.3 0.5–1 0

May 25, 2009 00:54:43 4.5–4.7 2.5–7 0
Feb 12, 2013 02:57:51 4.9–5.1 6–10 1
@seismicisolation
@seismicisolation
256 Chapter Twelve
Figure 2. Observation of North Korean tests at the Israel Seismic Network and two additional
stations: CSS, Cyprus, and GHAJ, Jordan (the ISN+ network). Only P-waves were observed after a
narrow bandpass filtering
In the case of low signal-to-noise ratio, different narrow band-pass filters were applied to reveal
the first P-arrivals, within the common range of 0.5–10 Hz, containing all analyzed null
frequencies (including multiples). As mentioned above, this range is placed within the flat part of
the instrument response of all used, short-period and broadband seismometers at each network
stations and arrays; therefore, the instrument correction of the amplitude spectra was not
necessary.
FFT spectra were calculated for vertical recordings in time windows of 7–30 s beginning from the
first arrival comprising teleseismic P- and P-coda signals (or in similar windows of pre-signal
noise). Actually both direct P and free surface reflection pP phases are contained in the first 3 s of
teleseismic P-waves from nuclear tests. The 5-s time window was used by BAKUN and
JOHNSON (1973); however, much larger windows ~50 s were used by COHEN (1970): ‘‘… to
include in the analysis all daughter phases of the parent P and pP phases’’. Evidently, it will
accumulate and enhance interference source features over non-correlated background noise along
the analyzed large window. In a window larger than 3 s, for any later arrival, originated from P (as
a reflected, refracted, or other phase), there is a paired arrival, originated from pP, and shifted at
the same time delay T01 as P and pP, thus providing the same interference spectral modulation.
Therefore, in each case we selected empirically the optimal time window length, providing the
best manifestation of the coherent minima (spectral nulls) at different network stations or array
elements. A sensitivity test for the window length demonstrated more pronounced spectral null and
modulation for the large window of 30 s than for a smaller window 10 s (see below Fig. 8). For a
fixed explosion, the same window (highlighted on the seismogram plots) was used within all
elements of an array or within all stations of a network.
@seismicisolation
@seismicisolation
Figure 3. Vertical seismograms and P-wave spectra at ISN+ stations (distance ~8000 km, azimuth
~295): a 2009 test, records were band-pass (BP) filtered at 0.7–3 Hz; b 2013 test, records were BP
filtered at 0.7–2 Hz; spectra of pre-signal noise are also shown (left side). The average spectrum is
shown as a dotted line, the arrows show the spectral minimum position, and the smoothing value
denotes a number of spectral amplitudes included in the convolution triangle weights filter (also
applicable in the next figures)
@seismicisolation
@seismicisolation
258 Chapter Twelve
Figure 4. Map of seismic networks and arrays used for spectral analysis of teleseismic P-waves
from North Korean explosions
The conducted analysis utilizes only spectral shapes and positions of spectral minima (nulls), but
not absolute amplitude values, and, therefore, we used waveform amplitudes in counts, and the
spectral amplitudes are presented in count  second. All procedures of data view, filtering, FFT
calculation, smoothing, averaging and plotting are performed by jSTAR software developed in the
Geophysical Institute of Israel (POLOZOV and PINSKY 2005).
OBSERVATIONS OF NORTH KOREAN NUCLEAR TESTS AT ISN STATIONS
This study was initiated by spectral analysis of P-waves from North Korean nuclear explosions
(Table 1) recorded at ISN stations (distance ~8,000 km, azimuth ~295°). The 2006 test was too
small with poor teleseismic observations, and reliable P-signals could not be found in seismograms
even after narrow band-pass filtering. The next tests (in 2009 and 2013) were observed at several
ISN stations and two additional stations: CSS, Cyprus, and GHAJ, Jordan, separated by several
dozen km (denoted here as ISN? network) (Fig. 2).
The spectra demonstrate clear spectral scalloping and coherency of spectral shapes at most stations
with minima at 1.2–1.3 Hz (Fig. 3). The minima can be hardly identified for some stations, and the
null frequency variability at different stations is observed. Evidently this spectral feature, which
was not found in spectra of pre-signal noise (Fig. 3b), has resulted from P-pP interference.
The region around the test site consists of low- porosity, dense, intrusive and extrusive igneous
rocks, including granites (considered as the most likely source rock), basalts, and rhyolites, which
are Mesozoic or younger, assuming an average near- surface elastic P-wave velocity of 5 km/s
(MURPHY et al. 2013). The Korean seismic model of HERRMANN et al. (2005) suggests P-
wave velocity in the mountainous region around the test site of ~5 km/s, and a model of BONNER
et al. (2008) uses a value of Vp = 5.1 km/s. Based on these data and the estimated null frequency,
the depths of the 2009 and 2013 tests were estimated by Eq. (1) at h ≥ 2 km.
@seismicisolation
@seismicisolation
Figure 5. Vertical seismograms (BP filtered 0.7–5 Hz, P-arrivals aligned) and P-wave spectra at
NNSN and NORSAR stations, Norway (distance about 7400 km, azimuthal range 314–332°): a -
2009 test, no filter; b - 2013 test, no filter
TELESEISMIC OBSERVATIONS OF NORTH KOREAN TESTS AT WORLD
STATIONS IN A BROAD AZIMUTHAL RANGE
We tried to check the broad-range azimuthal independence inherent to this source-depth effect and
compared spectra of the P-wave group from the North Korean tests recorded at different seismic
stations and arrays placed at teleseismic distances of 4000–7500 km in other parts of the world
(Fig. 4). The same (as for ISN data) FFT calculation and smoothing procedures were utilized for
time windows of 7–30 s of vertical seismograms. The results are presented on Figs. 5, 6, 7, 8, 9.
Similar to the ISN observations, records of the 2009 and the 2013 tests at NNSN (Norwegian
National Seismic Network) and NORSAR stations, Norway, demonstrate clearly the spectral
@seismicisolation
@seismicisolation
260 Chapter Twelve
minima. The null frequencies display variability in the range of 1.1–1.2 Hz at different stations
that are located far one from each other, with the averages of 1.11 Hz for the 2009 test, and 1.18
Hz for the 2013 test. How- ever, a rather high semblance of spectral shapes around the first null
frequency is found thus evidencing the source-depth effect (Fig. 5). Reliable P-signals were not
found for the 2006 test.
We analyzed also data from several other arrays. All three tests were recorded at Kurchatov
(KURK) array, Kazakhstan (IMS station AS058), which is the closest one among all analyzed
arrays (about 4000 km). For the 2006 test we observed weak P-signals with visible minima at
about 1.2 Hz, whereas the next two explosions demonstrated very distinct coherent spectral nulls
at all array elements at frequency 1.23 Hz, which is not observed in the spectra of pre-signal noise,
and as well at two multiple frequencies of ~2.39 Hz (n = 2) and ~3.66 Hz (n = 3) (Fig. 6).
@seismicisolation
@seismicisolation
Figure 6. Vertical seismograms and P-wave spectra at KURK array (distance 3960 km, azimuth
303°): a 2006 test, BP filter 0.7–5 Hz; b 2009 test, BP filter 0.5–10 Hz (arrivals aligned); c 2013
test, no filter, spectra of pre-signal noise are also shown
Figure 7 shows data of the YKA array, Canada (IMS station), which is located further than KURK
(about 7200 km). The P-signals for the small 2006 test cannot be observed in the seismograms.
How- ever, the two next tests demonstrated very distinct spectra minima at ~1.3 Hz with rather
small variability across the array elements.
We analyzed data from two Australian arrays, Warramunga (WRA) and Alice Springs (ASAR),
located at about opposite azimuths than the above- mentioned networks and arrays (Fig. 4). ASAR
array (IMS station PS03), at a distance of 7200 km away from the test site, did not record visible
P-signals from the 2006 test. We observed a considerable variability of spectral minima shapes for
the 2009 test, and an outstanding coherency of deep spectral minima for the 2013 test, but the null
frequency was shifted to 1.1 Hz for both tests (Fig. 8). However, data of a slightly closer WRA
array (IMS station PS02) placed at the same azimuth, with about the same propagation path,
demonstrated again the pronounced 1.25–1.35 Hz minima for all three tests. Moreover, two
multiple frequencies were revealed, especially clear for the largest 2013 test (Fig. 9), around 2.4
Hz (n = 2) and 3.3 Hz (n = 3), with the minima not well pronounced, but still observed in all
analyzed spectra. The similarity in the frequencies of the spectral minima in these arrays and those
in KURK, NORSAR, YKA and ISN rules out the possibility of a propagation effect.
According to the obtained null frequency estimates the pP-P delay value should be about 0.8 s. We
could identify pP phase arrivals corresponding to this delay only at a few elements of WRA array
for the largest 2013 test (Fig. 9c).
ACCURACY OF THE DEPTH DETERMINATION
The correctness of the depth determination is limited by the accuracy of the null frequency f01
estimation (or the time delay between P- and pP waves) demonstrating a variance for different
stations (due to local subsurface inhomogeneity) and reliable knowledge of the average
compressional near-surface velocity Vp. Table 2 presents the fundamental spectral null
(minimum) frequency f01 measured from average spectra at analyzed network stations or array
elements. The overall mean estimate for all stations and the standard deviation were then
calculated for each test.
@seismicisolation
@seismicisolation
262 Chapter Twelve
Figure 7. Vertical seismograms (no filter) and P-wave spectra at YKA array, Canada (distance
7200 km, azimuth 48°): a 2009 test, no filter; b 2013 test, no filter
Based on the average near-surface compressional velocity of 5.1 km/s (BONNER et al. 2007), Eq.
1 provides the following close depth estimations for the 3 tests: h1 = 2.01 ± 0.12 km, h2 = 2.13 ±
0.12 km, and h3 = 2.11 ± 0.11 km. Utilizing a little bit smaller near-surface velocity of 5.0 km/s
(HER-- RMANN et al. 2005; MURPHY et al. 2013) will decrease the depth estimation by a few
percent, whereas considering the uncertainties from neglecting take-off angles will increase it by a
few percent.
DISCUSSION
The reciprocal of a given first null frequency provides an estimate of the pP-P delay time τ ~ (2 x
h)/Vp. Due to a shallow source depth of nuclear explosions, the delays are rather short, about 0.5–
1.0 s, and in most cases pP arrivals usually could not be revealed in relatively low-frequency
teleseismic P-signals with a low signal-to-noise ratio (the only exception in the analyzed data is
WRA seismograms for the 2013 test, Fig. 9c). Therefore, numerous special complicated
processing techniques were utilized for direct determination of τ and the evaluation of the source
depth or the near-surface velocity: cepstral and pseudo-autocorrelation analysis (COHEN 1970),
@seismicisolation
@seismicisolation
homomorphic deconvolution into pulse wavelets representing the source (like slapdown phase)
and propagation paths (BAKUN and JOHNSON 1973). However, if estimating only the source
depth, then identification and accurate calculation of null frequencies in amplitude spectra seem to
be the most simple and effective procedure, based on a general assumption that wave interference
features are best pronounced in the spectral domain.
Spherically symmetrical explosive sources cannot produce pronounced spectral minima in
teleseismic P-waves in the frequency range of the dominant seismic energy (KULHANEK 1971).
The existence of a propagation effect, which causes narrow-band filtering in the frequencies
around 1.2–1.3 Hz in all directions, providing similar spectral modulation and null frequencies at
different azimuths is unlikely. Comparable null frequencies at the same station and at different
networks for all 3 tests suggest that the explosion depths are very similar. Almost identical
(normalized) waveforms for 2009 and 2013 tests recorded at a near-regional station MDJ,
presented at IRIS website, indicate about the same source depth (Fig. 10).
The obtained unusual depth values greater than 2 km for relatively small nuclear tests, with
estimated range 0.5–10 kT (Table 1), seem unexpected, when a source depth less than 1 km (200–
300 m) is sufficient to provide fully contained explosions of this size (in hard rock) (GLASSTONE
and DOLAN 1977).
Almost all NTS and STS explosions and underground tests in other countries were placed at
depths less than 1 km (SPRINGER et al. 2002). However, there were more than 20 PNE in the
USA and USSR of a similar size with reported depths of more than 2 and some are deeper than 2.5
km; the deepest test was a 2.5-kiloton nuclear device detonated at the bottom of a vertical shaft
2850 m deep, in Siberia, Russia, on 18 June 1985 (Benzol test), to stimulate oil production
(MIKHAILOV 1999). Larger depths for the North Korean tests can be suspected: (1) to prevent
release of radioactive gases to the atmosphere that can be detected by IMS radionuclide stations,
thus providing sensitive information about the design of a clandestine explosion; (2) to avoid any
changes of the surface in the test epicenter that could be revealed by satellite observations to
determine the exact place of the test.
MURPHY et al. (2013) estimated yields, which depend on source depths, for two North Korean
explosions, and found that broad-band source spectral ratios (2009/2006) for regional South Korea
stations are best modeled (Mueller/Murphy explosion source model) by source depths of about 200
m for the 2006 test (of yield 0.9 kT) and 550 m for the 2009 test (4.6 kT). They claim that ‘‘these
depth estimates were obtained from the inferred pP delay times assuming an average near-surface
elastic P-wave velocity of 5 km/s…’’, whereas any evidence of appropriate null frequencies, being
~5 Hz for the 2009 test and ~12 Hz for the 2006 test (according to Eq. 1), is not presented. Note
that the pP delay times were obtained from regional (several hundreds of km), and not teleseismic
observations, and, moreover, MURPHY et al. (2013) conclude that ‘‘… the narrowband,
teleseismic P-wave spectral data available for these two shallow, low-yield explosions do not
provide the resolving power needed to confidently determine source depths …’’. Our results
evidence a good resolving power of teleseismic P-wave spectra for the source depth determination
even for the minimal- yield 2006 test, observed at WRA array at 6700 km (Fig. 9a).
@seismicisolation
@seismicisolation
264 Chapter Twelve
@seismicisolation
@seismicisolation
Figure 8. Vertical seismograms and P-wave spectra at ASAR array, Australia (distance 7200 km,
azimuth 175°): a 2009 test, P-arrivals aligned, filter 0.7–5 Hz; b 2013 test, no filter, window 30 s;
c 2013 test, no filter, window 10 s
The obtained estimation of North Korean test source depth was initially presented based on ISN
records only (GITTERMAN et al. 2013), indicating that the observed spectral minima at 1.2–1.3
Hz have resulted from the source-depth effect (P-pP interference), and not from a local site-effect
beneath ISN. Several arguments support this suggestion: (1) the ISN stations are located far from
each other (Fig. 2), in different geological conditions and subsurface structure (HOFSTETTER et
al. 2000) that rules out the possibility of coherent site-effect minima at this frequency; (2) as
mentioned above, about the same ISN stations showed rather different spectral null frequencies for
previous nuclear tests (GITTERMAN et al. 1999, GITTERMAN et al. 2002); (3) about the same
null frequency was observed at numerous seismic stations and arrays, in many countries, in a
broad azimuthal range. Nevertheless, a specific local site-effect can be suspected in the case of
ASAR, producing a spectral maximum that shifted the source-depth effect spectral minima from
1.2–1.3 to 1.1 Hz, for both observed tests (Fig. 8).
We could not identify in the analyzed records the slapdown phase associated with the free surface
spallation near the explosion epicenter, which arrives later than pP phase and was found for very
large explosions MILROW (yield ~1 Mt, depth h ~ 1.22 km) and CANNIKIN (yield ~5 Mt, depth
h ~ 1.79 km), placed in pillow lavas and breccia with near-surface P-velocity ~3.7 km/s (BAKUN
and JOHNSON 1973). As reported, significant surface effects were observed for these megaton
tests (e.g. MERRITT 1971). The smaller North Korean tests were conducted in harder granite
rocks having higher velocity of ~5 km/s and at larger depths, thus significantly exceeding the
explosion inelastic zone radius, and as a result no surface effects were reported in the test area
after explosions; therefore, the non-linear effects on the surface pP reflection are not applicable.
@seismicisolation
@seismicisolation
266 Chapter Twelve
Figure 9. Vertical seismograms and P-wave spectra at WRA array, Australia (distance 6700 km,
azimuth 174°): a 2006 test, BP filter 0.7–5 Hz, P-arrivals aligned; b 2009 test, BP filter 0.5–10 Hz;
c 2013 test, BP filter 0.5–10 Hz, P-arrivals aligned, pP phases can be identified at some channels
@seismicisolation
@seismicisolation
Supposed Horizontal Drift Case
Since the late 1990s, surveillance satellites have detected tunneling activity in this area suspected
to be indicative of North Korea aiming at conducting nuclear tests at this site (BROAD et al.
2005). Satellite imagery of the test site showed horizontal tunnels placed very close to Mantap-san
(Mount Manthap, 2205 m above sea level) (YONHAP news agency 2013). If this case of a
horizontal drift is real, then a simple scheme of hypothetical P-wave reflection from the mountain
summit surface and possible P-pP interference can be proposed (Fig. 11a). This scheme is very
similar to the vertical shaft case of P-pP interference (Fig. 11b), resulting in the same source-depth
effect of spectral minima observations at different azimuths. In this case the h value obtained from
the spectral null frequency corresponds to the vertical distance from the source on the drift level to
the mountain summit, i.e., 2.0–2.1 km, which seems quite reasonable. In reality, perhaps, a more
complicated tunnel case design could exist (a combination of horizontal drifts and vertical shafts),
and the proposed scheme should be corrected if additional data and analysis results are available.
Table 2. Spectral null frequency f01 (Hz) measured from average spectrum minimum at
analyzed seismic stations and the overall mean estimate of the frequency and the source depth
(by Eq. 1)
Test ISN NNSN KURK YKA ASAR WRA Total mean Total mean
and STDEV and STDEV
network network array array array array for f01 (Hz) for depth
(km)
1-2006 – – 1.19 (10) – – 1.35 (13) 1.27 ± 0.08 2.01 ± 0.12
2-2009 1.25 (8) 1.11 (11) 1.23 (10) 1.27 (17) 1.10 (18) 1.25 (20) 1.20 ± 0.07 2.13 ± 0.12
3-2013 1.22 (12) 1.18 (9) 1.23 (9) 1.30 (17) 1.10 (18) 1.25 (10) 1.21 ± 0.06 2.11 ± 0.11
Number of used network stations or array elements is also shown (in brackets)
Figure 10: Vertical seismograms of 2009 and 2013 tests recorded at MDJ station (from IRIS
website: http://www.iris.edu/dms/nodes/dmc/specialevents/ 2013/02/12/north-korea-nuclear-
explosion/)
CONCLUSIONS
Analysis of teleseismic P-waves from North Korean nuclear tests in 2009 and 2013, recorded at
Israel Seismic Network stations, revealed spectral minima at null frequencies 1.2–1.3 Hz,
interpreted as P-pP interference effect, providing source depth estimation about 2 km. A simple
effective spectral technique was utilized for clear observation of spectral modulation (scalloping)
in smoothed coherent amplitude spectra that indicate directly the wave interference phenomena
and easy accurate determination of the null frequencies. Similar spectral minima at about the
same frequencies (1.10–1.35 Hz) were observed, for the first time in a very broad azimuth range,
@seismicisolation
@seismicisolation
268 Chapter Twelve
at network stations and arrays in Israel, Kazakhstan, Norway, Australia and Canada, for the North
Korean tests in 2006, 2009 and 2013, thus confirming that it is the source- and not a site- effect.
These results validate very similar depth estimates about 2.0–2.1 km for all three North Korean
nuclear tests. It was shown that the observed null frequencies and the obtained source depth
estimates are appropriate to the P-pP interference phenomenon for both cases of vertical shaft or
horizontal drift in a mountain. This unusual depth estimation needs an additional validation based
on more stations and verification by other methods.
Figure 11. Two hypothetical design cases for North Korean nuclear tests and schemes of P-wave
reflection and P-pP interference: a horizontal drift in Mt. Manthap; b vertical shaft
ACKNOWLEDGMENTS
We are thankful to Dr. Inna Sokolova from KazNDC for data supplement from KURK array and
to Drs. Rick Benson (IRIS) and David Jepsen (Geoscience Australia) for providing more
teleseismic data. Thanks to anonymous reviewers for valuable comments on the manuscript.
Research work of one of the authors (Y.G.) was supported by the Israel Ministry of Immigrant
Absorption.
REFERENCES
BAKUN, W. H., L. R. JOHNSON (1973). The deconvolution of teleseismic P waves from
explosions MILROW and CANNIKIN, Geophys. J. R. Astr. Soc. 34, 321–342.
BONNER, J., R. B. HERRMANN, D. HARKRIDER and M. PASYANOS (2008). The Surface
Wave Magnitude for the 9 October 2006 North Korean Nuclear Explosion, Bull. Seis. Soc. Am,
V.98, No.5.
BROAD, W., D. JEHL, D. SANGER, and T. SHANKER (2005), North Korea nuclear goals:
Case of mixed signals, NY Times, 25 July.
COHEN, T.J. (1970). Source-Depth Determinations using Spectral, Pseudo-Autocorrelation and
Cepstral Analysis, Geophys. J. R. astr. Soc., 20, 223–231.
GITTERMAN, Y. (2002). Implications of the Dead Sea experiment results for analysis of seismic
recordings of the submarine ‘‘Kursk’’ explosions. Seism. Res. Lett., 73: 14–24.
GITTERMAN, Y. (2009). Source phenomenology experiments with borehole explosions of special
design in Israel, Bull. Seism. Soc. Am., 99, 3, 1892–1905, June 2009,
doi:10.1785/0120080245.
GITTERMAN, Y. (2013). Evaluation of source depth for North Korea nuclear tests from ISN
teleseismic data. In Poster presented at Science and Technology meeting SnT2013, Vienna,
Austria.
GITTERMAN, Y. and VAN ECK, T. (1993). Low frequency spectra of quarry blasts and
microearthquakes recorded at local distances in Israel, Bull. Seism. Soc. Am., 83, 1799–1812.
GITTERMAN, Y. and SHAPIRA, A. (1994). Spectral characteristics of seismic events off the
coast of the Levant, Geophys. J. Int.,116: 485–497.
GITTERMAN, Y., PINSKY, V. and SHAPIRA, A. (1999). Spectral dis-crimination analysis of
Eurasian nuclear tests and earthquakes recorded by the Israel Seismic Network and the
@seismicisolation
@seismicisolation
NORESS array, Physics of the Earth and Planetary Interiors, 113, 111–129.
GITTERMAN, Y., V. PINSKY, R. HOFSTETTER (2002). Signal Processing for Indian and
Pakistan Nuclear Tests Recorded at IMS Stations Located in Israel. Pure Appl. Geophys., 159,
779–801.
GITTERMAN, Y., KIM, S. G. and HOFSTETTER (2013). Estimation source parameters of
large-scale chemical explosives and recent nuclear tests, AGU Fall Meeting, San Francisco.
GLASSTONE, S. and P.J. DOLAN (1977). The Effects of Nuclear Weapons, U.S. Government
Printing Office, Washington.
HERRMANN, R. B., Y. S. JEON, and H. J. YOO (2005). Broadband source inversion using
digital data from Korean seismic networks, in Proc. 4th International Seminar on Seismic
Tomography of Far-East Asia and Related words, 2 December 2005, Korea Institute of
Geoscience and Mineral Resources, Daejeon, Korea.
HOFSTETTER, A., C. DORBATH, M. RYBAKOV, V. GOLDSHMIDT (2000). Crustal and
upper mantle structure across the Dead Sea rift and Israel from teleseismic P-wave
tomography and gravity data. Tectonophysics, 327, 37–59.
KIM, S.G. and Y. GITTERMAN (2012). Underwater Explosion (UWE) Analysis of the ROKS
Cheonan Incident. Pure Appl. Geophys, Springer Basel AG, doi:10.1007/s00024-012-0554-9.
KULHANEK, O. (1971). P wave amplitude spectra of Nevada underground nuclear explosions,
Pure Appl. Geophys., 88, 121–136.
KULHANEK, O. (1973). Source parameters of some presumed Semipalatinsk underground
nuclear explosions, Pure Appl. Geophys., 102, 51–66. IRIS website (http://www.iris.edu/dms/
nodes/dmc/specialevents/2013/02/12/north-korea-nuclear- explosion/).
LAY, T. (1991). Teleseismic manifestations of pP: problems and paradoxes, in Explosion Source
Phenomenology, in S.R. Taylor, H.J. Patton and P.G. Richards (Eds.), Geophysical Monograph
65, 109–125.
MIKHAILOV, V.N. (editor-in-chief) (1999). Catalog of worldwide nuclear testing, by Begell-
Atom, LLC ISBN 1-56700-131-9 (http://www.iss-atom.ru/ksenia/catal_nt/index.htm).
MERRITT, M. (1971). Ground Shock and Water Pressures from Milrow. BioScience (American
Institute of Biological Sciences) 21 (12): 696–700. doi:10.2307/1295751.
MURPHY, J.R., J. L. STEVENS, B. C. KOHL and T. J. BENNETT (2013). Advanced seismic
analyses of the Source characters of the 2006 and 2009 North Korean nuclear tests, Bull.
Seism. Soc. Am. 103 (3), 1640–1661.
MUSATOV E. E. (1998). Cenozoic sedimentary structure and neo- tectonics of the Barents-Kara
shelf from reflection profiling data. Russian Journal of Earth Sciences, Vol 1, No. 2
(http://elpub. wdcb.ru/journals/rjes/v01/tje98007/tje98007.htm#fig09hook).
POLOZOV, A., and V. PINSKY (2005). New software for seismic network and array data
processing and joint seismological database, Report of Geophysical Institute of Israel,
546/185/05, 27 pp.
SPRINGER, D. L., G. A. PAWLOSKI, J. L. RICCA, R. F. ROHRER, and D. K. SMITH (2002).
Seismic source summary for all U.S. below- surface nuclear explosions, Bull. Seism. Soc. Am.,
92, 1806–1840.
WALLACE, C.T. (1998). The May 1998 India and Pakistan Nuclear Tests, Seism. Res. Let., 69,
386–393.
YONHAP news agency (2013).
http://english.yonhapnews.co.kr/
national/2013/02/04/58/0301000000AEN20130204006652315F. HTML.
@seismicisolation
@seismicisolation
CHAPTER THIRTEEN
THE RATIO BETWEEN CORNER FREQUENCIES OF SOURCE SPECTRA

OF P- AND S-WAVES—
A NEW DISCRIMINANT BETWEEN EARTHQUAKES AND QUARRY BLASTS 1
GALINA ATAEVA, YEFIM GITTERMAN AND AVI SHAPIRA
Calibration explosion of 20 tons ANFO in large diameter boreholes at Beit-Alpha basalt quarry on
June 6, 2005 (photo by Y. Gitterman).
ABSTRACT
This study analyzes and compares the P- and S-wave displacement spectra from local earthquakes
and explosions of similar magnitudes. We propose a new approach to discrimination between low-
magnitude shallow earthquakes and explosions by using ratios of P- to S-wave corner frequencies
as a criterion. We have explored 2430 digital records of the Israeli Seismic Network (ISN) from
456 local events (226 earthquakes, 230 quarry blasts, and a few underwater explosions) of
magnitudes Md= 1.4–3.4, which occurred at distances up to 250 km during 2001–2013 years. P-
wave and S-wave displacement spectra were computed for all events following Brune’s source
model of earthquakes (1970, 1971) and applying the distance correction coefficients (Shapira and
Hofstetter, Teconophysics 217:217–226, 1993; Ataeva G, Shapira A, Hofstetter A, J Seismol
19:389-401, 2015). The corner frequencies and moment magnitudes were determined using
1
Originally published in the J of Seismology (2017). 21, 209-220, doi.org/10.1007s10950-016-9598-0.
@seismicisolation
@seismicisolation
The Ratio between Corner Frequencies of Source Spectra of P- and S-waves 271
multiple stations for each event, and then the comparative analysis was performed.
The analysis showed that both P-wave and especially S-wave displacement spectra of quarry blasts
demonstrate the corner frequencies lower than those obtained from earthquakes of similar
magnitudes. A clear separation between earthquake and explosion populations
was obtained for ratios of P- to S-wave corner frequency f0(P)/f0(S). The ratios were computed for
each event with corner frequencies f0 of P- and S-wave, which were obtained from the measured f0I
at individual stations, then corrected for distance and finally averaged. We obtained empirically
the average estimation of f0(P)/ f0(S) = 1.23 for all used earthquakes, and 1.86 for all explosions.
We found that the difference in the ratios can be an effective discrimination parameter which does
not depend on estimated moment magnitude Mw.
The new multi-station Corner Frequency Discriminant (CFD) for earthquakes and explosions in
Israel was developed based on ratios P- to S-wave corner frequencies f0(P)/f0(S), with the empirical
threshold value of the ratio for Israel as 1.48.
Keywords Event discrimination. Spectral displacement. Seismic source parameters. P- and S-
wave corner frequency
INTRODUCTION
The problem of discriminating between man-made explosions and natural earthquakes is an
important issue in seismic monitoring. In spite of many efforts over several decades, the
discrimination of small magnitude events is still a difficult problem. A variety of waveform-based
methods have been studied including ratios between amplitudes of different seismic phases (e.g.,
Bennett and Murphy, 1986; Wuster 1993; Plafcan et al. 1997; McLaughlin et al., 2004, Morozov,
2008, Anderson et al., 2009), analysis of the velocity spectra (e.g., Taylor et al., 1988; Kim et al.
1993, 1994; Walter et al., 1995; Gitterman et al., 1998), analysis of coda waves (e.g., Su et al.,
1991; Hartse et al., 1995) and detection of ripple- firing spectral modulation (e.g., Hedlin et al.,
1990; Gitterman and van Eck, 1993; Smith, 1993; Carr and Garbin, 1998; Arrowsmith et al. 2006,
2007).
One of the main spectral features, proposed for identification of quarry blasts conducted by the
ripple-firing technology, is the spectral modulation, similar at different network stations and not
presented in earthquake velocity spectra (Gitterman and van Eck, 1993). It was also observed in
Israel that seismograms of explosions are richer in low-frequency energy as compared to
earthquakes, either because of higher attenuation of high-frequency seismic waves or the efficient
wave interference during generating surface waves, or both (Gitterman et al., 1998). These effects
were utilized in a multi-station spectral discrimination method based on energy spectral ratio (RE)
between the low-frequency (1–3 Hz) and high-frequency (6–8 Hz) bands, and the semblance of
signal spectral shapes at different stations (Gitterman et al., 1998).
It is known that source spectra of earthquakes and explosions show significant differences, due to
the different source mechanisms. Explosions release their energy in a 3D symmetric volume and
their waveforms are dominated by P-waves. Quarry blasts are commonly fired with a series of
delays (ripple firing) in order to control the movement of rock masses during the blast. In contrast,
earthquakes represent a line source, with a dimension several times larger than explosion sources
and generate much larger S-waves (Wyss, and Hanks, 1971, Chapman, 2008, Shearer and
Allmann, 2007, Allmann et al., 2008, Dahy and Hassib 2010). The classical source model of Brune
(1970, 1971) is a theoretical representation of the source spectrum of earthquakes. Wyss and
Hanks (1971) pointed the differences between P-wave displacement spectra of underground
explosions and shallow earthquakes of similar body- wave magnitude. It was demonstrated that in
southern California, the best earthquake/explosion discriminator is the RMS misfit between P-
wave spectra and the theoretical earthquake source model (Shearer and Allmann, 2007, Allmann et
@seismicisolation
@seismicisolation
272 Chapter Thirteen
al., 2008). Quarry blast spectra show a steeper fall-off at high frequencies for all magnitudes and
do not fit well to the standard source model. The deficiency of high frequency seismic energy from
explosions was explained by ripple-firing and/or
strong near-surface attenuation (Shearer and Allmann, 2007, Allmann et al., 2008). In other
studies, enrichment of low frequency energy content for explosions has been pointed out, e.g.,
Taylor et al. (1988), Bennett and Murphy (1986), Su et al. (1991).
We have investigated earthquake and explosion source parameters from P- and S-wave
displacement spectra, focusing on the corner frequency f0, and demonstrated the effectiveness of
the P- to S-wave corner frequencies ratio f0(P)/f0(S) as a new multi-station discriminant between
small-magnitude shallow earthquakes and explosions in terms of spectral properties.
We propose the following formulation of the Corner Frequency Discriminant (CFD), which
includes (1) estimations of all corner frequencies at individual stations for an event from P- and S-
waves, (2) correction of these frequencies for distance, and then (3) computation of the
discriminant with the averaged multi-station values f0(P) and f0(S):
CFD = f0(P) / f0(S),
where f0 = (f01 , f02….f0n)= ೙ඥ݂ଵ଴ ݂଴ଶ … ݂଴௡ -the geometrical mean of corner frequencies of P- or S-
waves from n stations for an event, f0i = f0I eȖR - corner frequency from an individual station i,
corrected for distance R (Shapira and Hofstetter 1993 and Ataeva et al., 2015), f0I - corner
frequency observed at individual station i, Ȗ - the empirical coefficient.
The CFD, determined as ratio of corner frequencies, provided quite robust results for the data in
Israel.
DATA
In this study, we explored 2430 seismograms of 226 earthquakes and 230 explosions (quarry blasts
and a few of underwater explosions occurred off-coast Israel and Lebanon) of magnitudes Md =
1.4–3.4 with epicentral distances up to 250 km (Fig. 1a). Most of quarry blasts were known
(independently of seismic means) from monthly reports sent by the quarries to the Geophysical
Institute of Israel. In some cases, quarry blasts are identified by microphone sensors presented in
stations of Israel Seismic Network (ISN) that register specific explosion acoustic phases at close
distances. We have tried to keep up a balance through the same magnitude ranges for earthquakes
and explosions for a reliable statistical analysis (Fig. 1b, c). We used only earthquakes shallower
than 20 km. The data consist of the vertical components of broadband and short period ISN
stations.
The waveforms (stations) were selected by the condition that the seismic event was recorded by at
least three calibrated seismic stations located at different epicentral distances and the signal-to-
noise ratio exceeds 2. The data processing program JSTAR (Polozov and Pinsky, 2007) was used
in the analysis. Available parametric data: local duration magnitude Md, epicentral distance, and
focal depth, were collected from the seismological bulletin of the Geophysical Institute of Israel.
Figure 2 demonstrates a typical example of waveforms at different stations from an earthquake of
2010/02/06 and from a quarry blast of 2013/02/24, both of the same magnitude, Md = 2.0. The
distance between the epicenters of these events is about 18 km (see Fig. 1, position 1). It is obvious
that the waveforms in this displayed example do not reveal any clear difference between the
earthquake and the quarry blast, although sometimes identification of quarry blast seismograms
is evident due to longer period surface waves or sonic wave arrivals thus indicating the event
blasting character. However, the objective of this study is the comparison of the seismic source
effect features of earthquakes and explosions in the frequency domain.
@seismicisolation
@seismicisolation
Fig. 1 a Location map of explosions, earthquakes, and seismic stations used in this study; the events and stations SLTI
and MMAOB used as examples are denoted (positions 1, 2). b, c Histograms of explosion (blue) and earthquake (red)
duration magnitudes
@seismicisolation
@seismicisolation
Fig. 2 Sample vertical component seismograms (bandpass filtered 0.5–15 Hz) from an earthquake of 2010/02/06, Md
= 2.0 and from a quarry blast of 2013/02/24, Md = 2.0 (position 1 at Fig. 1). Traces are sorted according to epicentral
distance—from 16 to 240 km for the earthquake and from 18 to 238 km for the explosion. Also the P-wave and S-
wave phases are shown
METHOD
Spectral displacement analysis
We tried to estimate the similarity of the source dis-placement spectrum of P- and S-waves of an
earthquake and an explosion to the theoretical earthquake model. The source parameters of all
events were analyzed in accordance with Brune’s seismic source model (1970, 1971), using
computations of the P- and S-waves displacement Fourier spectra from the vertical component of
seismograms. The calculation windows start at the arrival times of P- and S-wave. The P-wave
window includes the first group of P phase, increasing with distance from 1.5 to 3 s. The S-wave
time window is usually 10–20 s, depending on the distances, and includes the most significant part
of the signal energy, as generally accepted for estimation of seismic moment (Brune 1970, 1971,
Hanks and Wyss 1972, Baumbach and Borman 2012). The quarry blast S-wave time win-dows at
short distances include also surface Rg waves (actually S + Rg windows). The long S-wave
window provided optimal results in separation of earthquakes and explosions.
Then, the corner frequencies, f0I(P), f0I(S), and the low-frequency spectral amplitudes ȍ(P), ȍ (S)
of P- and S-waves are manually measured from calculated displacement spectra for each station
record, uncorrected for attenuation, through approximating the spectrum by two lines which refer
to constant low-frequency level and high-frequency decay of f -2. The corner frequency is defined
as the intersection of low- and high-frequency asymptotes (Brune 1970; Baumbach and Borman
2012; Kiratzi and Louvari 2001; Ashkpour Motolagh and Mostafazadeh 2008; Havskov and
Ottemoeller 2010). Records of an earthquake and a quarry blast at station SLTI, located to the
south from these events (Fig. 1, position 1), and their P-wave and S-wave displacement
@seismicisolation
@seismicisolation
spectra, not corrected prior for attenuation of the seismic waves, are shown in Fig. 3. It is obvious
that the dis-placement spectra of the quarry blast from station SLTI show the lack of high-
frequency energy in P-wave and especially in S-wave, and therefore the measured corner
frequencies f0I(P) and especially f0I(S) are lower than those obtained from the earthquake of the
same magnitude at the same station.
In addition, we checked a few underwater earthquakes and explosions and found that the
displacement spectra of P- and S-wave of the underwater explosions also demonstrate the similar
anomalous behavior as distinct from the underwater earthquakes of the same magnitudes (Fig. 4).
Note that spectral modulations were found both in S-wave and in P-wave displacement spectra, for
some quarry blasts, due to the ripple fired effect, and for some underwater explosions due to the
bubbling effect. These modulations were previously observed in the velocity spectra (Gitterman
and van Eck, 1993, Gitterman et al., 1998, Gitterman and Shapira, 1994). We also found in some
explosions a noticeable scatter of f0I values at different stations.
Source parameter estimation
The values of observed corner frequency f0I and low- IUHTXHQF\VSHFWUDODPSOLWXGHȍDWGLVWDQFH
R, are used to determine the corner frequency f0i and seismic moment M0i at the source from each
station i (Brune, 1970, 1971), considering attenuation effect due to geometrical spreading and
inelastic attenuation. The principal task is to evaluate the spectra high-frequency decay as the
source effect only, excluding the attenuation effect. It was done by utilizing empirical distance
correction coefficients for calculating of the corner frequency f0 and seismic moment M0 for ISN,
following Shapira and Hofstetter (1993) and Ataeva et al. (2015). In this approach, the seismic
source parameters, M0(Nm), f0 (Hz) and the distance corrections, in general, are described by
equations
M 0 = ȍʌȡDV 3/CF, D = RĮeįR, f0 = f0I eȖR,
with the empirical coefficients:
Į į S-ZDYH Į į
P-ZDYH Ȗ P- and S-wave),
where ȍ - observed low-frequency spectral amplitude (m*s), ȡ - density, 2700 kg/m3; V - P-wave
(6200 m/s) and S-wave (3600 m/s) velocity; C—free surface amplification, equals 2; F—radiation-
pattern correction factor, Fs = 0.18, FP = 0.64; R—hypocentral distance (m), f0I - corner frequency
(Hz) at a station at distance R (Shapira and Hofstetter 1993, Ataeva et al., 2015).
The moment magnitude Mw (Kanamori 1977; Hanks and Kanamori 1979) is calculated as:
Mw = ORJM0 í 9)
The corner frequencies f0(P) and f0(S) at the source for each event are calculated as the geometrical
average of corner frequencies f0i from individual stations i. The f0i values are obtained after the
distance correction to the manually measured frequencies f0I. The averaging reduces the
interstation scatter due to possible errors from the manual measuring and the possible local site
effect at some stations. For earthquakes, we determined corner frequencies with the standard
deviation (in logarithmic units) in the range 0.005–0.116 (average 0.038) for P- wave, and in the
range 0.006–0.084 (average 0.035) for S-wave. The standard deviations of explosion corner
frequency are estimated between 0.002 and 0.109, the average value ı = 0.040 (P-wave), and
between 0.001 and 0.083, the average value ı = 0.036 (S-wave). Al- though the number of used
stations in each event is different (3–15 for earthquakes, in most cases 6–8 stations, and 3–8 for
explosions, in most cases 4–6 stations), the average VWDQGDUG GHYLDWLRQV YDOXHV ı RI FRUQHU
frequency estimations, for earthquakes and explosions are acceptable and practically identical as
@seismicisolation
@seismicisolation
required for the confidence of the comparative analysis

Seismic moment M0 and moment magnitude Mw from P- and S-wave for all studied events were
also calculated by geometrical averaging of values estimated from the individual stations i that
recorded the event, thus providing more reliable parameter evaluation. The averaging of source
parameter estimates over stations reduces the scatter caused by different station azimuths,
especially for P-wave.
RESULTS
Spectral characteristics of quarry blasts in different areas
We have analyzed the spectra of quarry blasts in accordance with quarries location in Israel and
adjacent countries and with their duration magnitudes determined by the Israeli Seismic Network.
We defined three areas, different by the corner frequency f0(P) and f0(S) and the duration
magnitude Md range, possibly due to diversity in quarry technology and near- source geology (Fig.
5). For two areas (A, B) prevailing values f0 are about 2.0 Hz for S-wave and about 4.0–4.5 Hz for
P-wave spectra, when the maximum explosion magnitudes are different. The third area (large
quarries in Jordan) is visibly characterized by higher values of the corner frequencies in the
displacement spectra: about 3 Hz for S-wave and about 5 Hz for P-wave and by the maximum
magnitude up to 3 (Fig. 5c). We found that spectral displacement corner frequencies of explosions
do not depend on the duration magnitude, i.e., the quarry blasts of different magnitudes in Israel
have roughly similar corner frequencies. It is important result for practical application of the
discrimination procedure. For comparison, the corner frequencies f0(P) and f0(S), for studied
earthquakes (Fig. 5d) show a clear dependence of f0 on duration magnitude Md.
@seismicisolation
@seismicisolation
Fig. 3 An earthquake of 2010/02/06, Md = 2.0 and a quarry blast of 2013/02/24, Md = 2.0, (position 1 at Fig. 1). a
Seismograms (bandpass filtered 0.5–15 Hz) recorded by the short period station SLTI at distance 37 and 29 km,
respectively. The analyzed time windows of P- and S-wave trains are marked by solid blue and red lines. b
Uncorrected P-wave and S-wave displacement spectra (m*s) from seismograms of station SLTI. The green lines
indicate the low-frequency amplitude level and high- frequency decay, the red line - the value of the corner frequency
and the dashed blue lines show f í lines. The measured f0I and ȍ are shown
@seismicisolation
@seismicisolation
Fig. 4 Uncorrected P-wave and S-wave displacement spectra (m*s) calculated from seismograms of broad band station
MMAOB of an earthquake (2008/03/11, Md = 1.7, distance 62 km) and of an explosion (2013/03/11, Md = 1.9, 39
km) in E. Mediterranean Sea (position 2 at Fig. 1). Green lines indicate the low-frequency amplitude level and high-
frequency decay. The red line shows the corner frequency value and dashed blue lines show f í decay. The measured
f0I and ȍ are shown
@seismicisolation
@seismicisolation
Fig. 5 S-wave corner frequency (red circles) and P-wave corner frequency (blue circles) versus the duration
magnitude Md, for explosions in different areas (a–c) and for earthquakes (d); location map of studied earthquakes and
explosions at dedicated areas (e)
Comparison of source spectra parameters for earthquakes and explosion

Average multi-station values M0, Mw, and f0 for earthquakes and explosions were used for
statistical and comparative analysis, separately for P- and S-waves. Seismic moment versus corner
frequency is plotted in Fig. 6 for 226 earthquakes and 230 explosions, with histogram distribution
of the corner frequency for P- and S-waves. The analysis shows significantly lower corner
frequencies f0 of explosions than earthquakes, especially for S-wave displacement spectra. Seismic
moments M0 of earthquakes and explosions, located in the same range, are not separated and
cannot be used as a discriminant (Fig. 6a, b). Presented data suggest also the lower stress-drop for
explosions compared to earthquakes. The stress drop concept has no meaning for explosions and
qualified here as if interpreted for an earthquake. Nonetheless, determination of stress-drop values
@seismicisolation
@seismicisolation
and their analysis as discriminant is beyond the scope of this study. Figure 6 shows a minor
overlap in the corner frequency distributions for earthquakes and explosions, the overlap is larger
for P-wave values. In general, the shift f0(P) relative to f0(S) is obviously higher for explosions
than for earthquakes.
Corner frequency ratio f0(P)/f0(S) and discriminant formulation
In accordance with the earthquake source model and following Hanks and Wyss (1972), Molnar et
al. (1973), later by Watanabe et al. (1996) and by Tusa and Gresta (2008), it was found that f0(P)>
f0(S). In Israel for earthquakes of magnitude Mw = 2.7–5.6 an average empirical ratio was
obtained: f0(P)/f0(S) = 1.24 (Ataeva et al., 2015). Based on the conducted analysis presented in Fig.
6, the ratio between f0(P) and f0(S) is larger for explosions than for earthquakes. This result
assumes an ability for discrimination. Figure 7a demonstrates the comparison between corner
frequencies estimated from P- and S-waves for studied earthquakes and explosions. The average
ratios are obtained: f0(P)/f0(S) = 1.23 for 226 earthquakes with standard deviation of 0.18,
coefficient of determination of 0.98 and f0(P)/f0(S) = 1.86 for 230 explosions with standard
deviation of 0.27 and coefficient of determination of 0.97. The ratios f0(P)/f0(S) do not depend on
estimated moment magnitude Mw, and distributed for earthquakes in the range 1.05–1.48, which is
clearly different from the range 1.49–2.85 for explosions, with a suggested threshold value about
1.48 (Fig. 7b). The empirical threshold value was derived as the average of the largest earthquake
and the smallest explosion ratios f0(P)/f0(S). The observed full separation of earthquakes and
explosions was obtained for the selected dataset; no event was excluded as an outlier. Some
earthquakes and explosions show very close ratios, within measurement errors. Some overlap for
another event selection can be expected.
Fig. 6 Seismic moment M0 versus corner frequency f0 for studied earthquakes (red circles) and explosions (blue
circles) obtained from P-wave (a) and S-wave (b). The solid lines indicate the constant stress drop in accordance with
the Brune’s source model. Histograms of the corner frequency for all studied earthquakes (red) and explosions (blue
cross) were obtained from P-wave (c) and S-wave (d) spectra
@seismicisolation
@seismicisolation
The presented analysis demonstrates feasibility of the f0(P)/f0(S) ratio as an effective discriminant,
separating earthquakes and explosions. The significant scatter in the explosion corner frequency
ratios (Fig. 7b) might be explained by different quarry technology feature (variable ripple firing
delays, number of delays) and by local near-source geology. The effect of near-source
(explosion) material properties on the spectral characteristics (spectral slope and corner
frequency) of P-wave was demonstrated by Ford et al. (2011) and Walter et al. (1995). It should be
noted that most of small f0(P)/f0(S) are obtained from displacement spectra of the quarry blasts in
Jordan, which are characterized by higher corner frequency values (Fig. 5c).
The Corner Frequency Discriminant (CFD), which we propose, includes estimations of all corner
frequencies from the individual stations of the event from P- and S-waves, then corrected for
distance, and finally computation of the discriminant with the averaged multi-station values f0(P)
and f0(S):
CFD = f0 P) /f0 S)
The P- wave and S-wave displacement spectra from local earthquakes and explosions with similar
magnitudes were analyzed and compared in this study. We explored 2430 digital records of the
Israeli Seismic Network (ISN) from 456 events (226 earthquakes and 230 explosions) of
magnitudes Md = 1.4–3.4. The short period and broad band vertical component seismograms were
used in the analysis.
This study concluded that the low-magnitude shallow earthquakes and the explosions in Israel
demonstrate significant differences in their spectral properties.
Fig. 7 a Corner frequency f0(P), determined from P-wave analysis, versus corner frequency f0(S), determined from S-
wave analysis. The green solid lines are the least-squares fits to the data. B Ratio between the corner frequencies
versus moment magnitude Mw estimated from P- and S-wave. The black line represents the empirical discrimination
threshold value ‫׽‬1.48 for Israel
We focused on the corner frequency only, assuming the standard model of fí 2 spectral fall (Brune.
1970, 1971), and did not analyzed the change in slope of the high-frequency fall-off because of the
complexity of explosion and earthquake spectral shapes. The corner frequency from displacement
spectra is the most important information about the seismic source type that can be used for
discrimination. To obtain the source corner frequency, we applied the empirical attenuation param-
eter (Ataeva et al., 2015) that characterizes the propagation effect in the medium and provides
reliable results similar to those of Brune’s method. We used for measurements all available
stations (3–15 for earthquakes and 3–8 for explosions), in some cases with one recorded phase.
Although uncertainties in manual measuring of the corner frequencies can be quite large, yet, the
@seismicisolation
@seismicisolation
obtained relatively small standard deviations of geometrical average of corner frequencies for
earthquakes and explosions suggest that our approach is valid.
The quarry blast S-wave windows, used in our study, at short distances may include Rg phase. The
amplitude of this fundamental mode Rayleigh wave is strongly depth dependent, it is often
observed with large amplitudes in near surface events such as quarry blasts (e.g., McLaughlin et al.
2004) and not usually observable in earthquakes with depths greater than a few km. We found that
including Rg large amplitudes in the calculating of S- wave displacement spectra (actually S + Rg
waves) has 15 % lowering effect on the observed corner frequencies, staying within obtained
standard deviation (logarithmic units) of averaged corner frequencies for each event. Since we use
available stations from the different distances, the corner frequencies for quarry blasts at the larger
distances more often obtained from S-wave.
Our results show that the corner frequencies of P-wave and especially S-wave source displacement
spectra of quarry blasts are lower than obtained from earthquakes of the same magnitudes, due to
losses of the high frequency seismic energy, caused by wave propagation in near-surface soft
sediment layers. Many researchers observed this effect in seismograms of quarry blasts (Taylor et
al., 1988, Bennett and Murphy, 1986, Su et al., 1991, Gitterman, et al., 1998) and in P-wave
displacement spectra (Shearer and Allmann, 2007, Allmann, et al., 2008). However, we observed
in the corner frequency distributions of earthquakes and explosions an overlap (larger for P-waves
than for S-waves), which does not allow completely separate explosions from earthquakes. We
demonstrated that prevailing values of corner frequency of all selected explosions are about 2.0–
2.5 Hz for S-wave and about 4.0–4.5 Hz for P-wave spectra, i.e., the quarry blasts of different
magnitudes have roughly the same corner frequencies.
We found that the ratio between f0(P) and f0(S) is higher for explosions than for earthquakes.
Ratios f0(P)/f0(S) between the corner frequencies of P-waves and S-waves are distributed in the
range 1.05–1.48 (average 1.23) for earthquakes, and in the range 1.49–2.85 (average 1.86) for
explosions. We used the difference in these ratios as an effective discrimination parameter which
is not dependent on moment magnitude Mw. The presented new discriminant is more closely
related to source than to depth type, because is based on the ratio of P to S corner frequencies, but
not on the S + Rg window corner frequency alone.
However, we can suggest that scatter in the explosion corner frequencies ratios might be caused
not only by the different technologies of quarries and local near- source geology but also by the
ambiguity of S or S + Rg corner frequency determination.
The proposed discrimination parameter, CFD = f0(P)/f0(S) is formed from corner frequencies
estimated for individual stations, then corrected for distance, finally the ratio is calculated from
the averaged values f0(P) and f0(S). This is a new approach providing effective multi-station
discrimination between low-magnitude shallow earthquakes and explosions in Israel. The same
method is likely to be applicable in other regions with similar quarry technology. We may expect a
different discriminant value for different detonation conditions.
ACKNOWLEDGMENTS
This study was supported by The Ministry of Energy and Water Resources of Israel. Special
thanks to Dr. A. Hofstetter for constructive comments. We thank the reviewers and Editor Mariano
Garcia-Fernandez for useful and constructive comments that significantly improved the
manuscript.
@seismicisolation
@seismicisolation
REFERENCES
Allmann BP, Shearer PM, Hauksson E (2008) Spectral discrimination between quarry blasts and
earthquakes in southern California. Bull Seis Soc Am 98:2073–2079
Anderson DN, Walre WR, Fagan DK, Mercier TM, Taylor SR (2009) Regional multistation
discriminants: magnitude, distance, and amplitude corrections, and sources of error. Bull Seis
Soc Am 99(2A):794–808
Arrowsmith SJ, Arrowsmith MD, Hedlin MAH, Stump B (2006) Discrimination of delay-fired
mine blasts in Wyoming using an automatic time-frequency discriminant. Bull Seis Soc Am
96(6):2368–2382
Arrowsmith SJ, Hedlin MAH, Arrowsmith MD, Stump B (2007) Identification of delay-fired
mining explosions using seismic arrays: application to the PDAR array in Wyoming, USA. Bull
Seis Soc Am 97(3):989–1001
Ashkpour Motolagh S, Mostafazadeh M (2008) Source parameters of the Mw 5.8 Fin (south of
Iran) earthquake of March 25, 2006. World Appl Sci J 4:104–115
Ataeva G, Shapira A, Hofstetter A (2015) Determination of source parameters for local and
regional earthquakes in Israel. J Seismol 19(2):389–401
Baumbach M, Borman P (2012) Determination of source parameters from seismic spectra. New
Manual of Seismological Observatory practice 2 (NMSOP-2) (pp. 1–7). Deutsches Geo
Forschugs Zentrum GFZ, Potsdam
Bennett T, Murphy J (1986) Analysis of seismic discrimination capabilities using regional data
from western United States events. Bull Seis Soc Am 76(4):1069–1086
Brune J (1970) Tectonic stress and spectra of seismic shear waves from earthquakes. J Geophys
Res 75:4997–5009
Brune JN (1971) Correction. J Geophys Res 76:5002
Carr DB, Garbin HD (1998) Discriminating ripple-fired explosions with high-frequency (>16 Hz)
data. Bull Seis Soc Am 88(4):963–972
Chapman M (2008) Seismological discrimination of blasts and natural earthquakes. Workshop on
Blasting, West Virginia
Dahy SA, Hassib GH (2010) Spectral discrimination between quarry blasts and microearthquakes
in Southern Egypt. Res JEarth-Sci 2:01–07
Ford SR, Walter WR, Ruppert SD, Matzel EM, Hauk TF, Gok R (2011) Toward an empirically-
based parametric explosion spectral model in Proceedings of the 2011 Monitoring Research
Review: Ground-Based Nuclear Explosion Monitoring Technologies, DE-AC52-
07NA27344/LL08.
Gitterman Y, Shapira A (1994) Spectral characteristics of seismic events off the coast of the
Levant. Geophys J Int 116:485–497 Gitterman Y, van Eck T (1993) Spectra of quarry blasts
and microearthquakes recorded at local distances in Israel. Bull Seis Soc Am 83:1799–1812
Gitterman Y, Pinsky V, Shapira A (1998) Spectral classification methods in monitoring small local
events by the Israel seismic network. J Seismol 2(3):237–256
Hanks T, Kanamori H (1979) A moment magnitude scale. J Geophys Res 84:2348–2350
Hanks T, Wyss M (1972) The use of body-wave spectra in the determination of seismic-source
parameters. Bull Seis Soc Am 62(2):561–589
Hartse HE, Phillips WS, Fehler MC, House LS (1995) Single station spectral discrimination using
coda waves. Bull Seis Soc Am 85(5):1464–1474
Havskov J, Ottemoeller L (2010) Routine data processing in earthquake seismology. Springer,
Dordrecht, p 347. doi:10.1007/978-90-481-8697-6
Hedlin MA, Minster JB, Orcutt JA (1990) An automatic means to discriminate between
earthquakes and quarry blasts. Bull Seis Soc Am 80(6B):2143–2160
Kanamori H (1977) The energy release in great earthquakes. J Geophys Res 82:2981–2987
Kim WY, Simpson DW, Richards PG, Smith AT (1993) Discrimination of explosions from
simultaneous mining blasts. Bull Seis Soc Am 83(1):160–179
@seismicisolation
@seismicisolation
Kim WY, Simpson DW, Richards PG (1994) High-frequency spectra of regional phases from
earthquakes and chemical explosions. Bull Seis Soc Am 84(5):1365–1386
Kiratzi A, Louvari E (2001) Source parameters of the Izmit-Bolu 1999 (Turkey) earthquake
sequences from teleseismic data. Ann Geofis 44:33–47
McLaughlin KL, Bonner JL, Barker T (2004) Seismic source mechanisms for quarry blasts:
modeling observed Rayleigh and Love wave radiation patterns from a Texas quarry. Geoph J
Int 156(1):79–93
Molnar P, Tucker B, Brune J (1973) Corner frequencies of P and S waves models of earthquake
sources. Bull Seism Soc Amer 63:2091–2104
Morozov AN (2008) Method of identification of explosive seismicity on territories of the
Arkhangelsk region. Institute of Ecological Problem in the North of UB RAS, Arkhangelsk
Plafcan D, Sandvol E, Seber D, Barazangi M, Ibenbrahim A, Cherkaoui T (1997) Regional
discrimination of chemical explosions and earthquakes: a case study in Morocco. Bull Seis Soc
Am 87:1126–1139
Polozov A, Pinsky V (2007) New software for seismic network and array data processing and joint
seismological data base. GII report 546/307/07
Shapira A, Hofstetter A (1993). Source parameters and scaling relationship of earthquakes in
Israel. Teconophysics 217:217– 226
Shearer PM, Allmann BP (2007) Spectral studies of shallow earthquakes and explosions in
Southern California, in Proceedings of the 29th Monitoring Research Review: Ground-Based
Nuclear Explosion Monitoring Technologies, p 656-662
Smith AT (1993) Discrimination of explosions from simultaneous mining blasts. Bull Seis Soc
Am 83:160–179
Su F, Aki K, Biswas NN (1991) Discriminating quarry blasts from earthquakes using coda waves.
Bull Seis Soc Am 81(1):162– 178
Taylor SR, Sherman NW, Denny MD (1988) Spectral discrimination between NTS explosions and
western United States earthquakes at regional distances. Bull Seis Soc Am 78(4): 1563–1579
Tusa G, Gresta S (2008) Frequency-dependent attenuation of P waves and estimation of
earthquake source parameters in southeastern Sicily, Italy. Bul Seis Soc Amer 98(6):2772–
2794
Walter WR, Mayeda KM, Patton HJ (1995) Phase and spectral ratio discrimination between NTS
earthquakes and explosions; part I, empirical observation. Bull Seis Soc Am 85(4):1050–1067
Watanabe K et al (1996) Source characteristics of small to moderate earthquakes in the Kanto
Region, Japan: application of a new definition of the S-wave time window length. Bul Seis Soc
Amer 86(5):1284–1291
Wuster J (1993) Discrimination of chemical explosions and earthquakes in central Europe - a case
study. Bul Seis Soc Amer 83:1184–1212
Wyss M, Hanks T (1971) Comparison of P-wave spectra of underground explosions and
earthquakes. J Geoph Res 76: 2716–2729
@seismicisolation
@seismicisolation
CHAPTER FOURTEEN
OVERVIEW AND REVIEW ON ANALYSIS FOR HYDROACOUSTIC

STUDIES IN UNDERWATER EXPLOSIONS
SO GU KIM
Seismograms and accelerograms showing evidences that the ROKS Cheonan sinking were due to
an underwater explosion showing the characteristics of the positive higher amplitude of the first
P-wave arrival and bubble pulses on the vertical component and bubble jets on the horizontal
components.
SUMMARY
The dynamics of bubble pulses, or “cavitation” have been studied extensively in hydroacoustics
since World War II. After the initial shock wave is produced by the explosion, there follow a series
of positive pressure pulses generating by the expanding and contracting gas sphere (bubble) that
rises to the surface. This paper showcases the fundamental studies of the underwater explosions
showing study examples such as Cole (1948), Weinstein (1968), Koper et al. (2001), Gitterman
(2002), Raymond et al. (2003), and Kim and Gitterman (2013), etc. The imortant research aims on
hydroaouustcis are to investiagte bubble effects and reverberation effects for underwater
@seismicisolation
@seismicisolation
286 Chapter Fourteen
explosions resulting in detonation depth and explosive charge weight, but it is not appropriate to
determine a size of un undrwater explosion in terms of seismic magnitude (especially Richter
scale). This study attepmts to provide some comments and clear understandings on results by
several researchers including some reviews if any. In general, there are no shear waves in
underwater explosions,because of no fracturing or spalling like land explosions. However, we can
see some shear waves produced by mode conversion of the acoustic waves in the water to S waves
in the sedimentary layers, on the sea bottom, and other discontinuities deeper in the crust.
CASE STUDY EXAMPLE 1
The Chase program of large underwater explosions has provided an excellent basis for examining
these spectral characteristics (Weinstein, 1968). Details of the Chase 2, 3, and 4 experiments are
given in Table 2. Magnetic tape recordings of the hydroacoustic signals generated by the
explosions were obtained at Bermuda by the Columbia University Geophysical Field Station.
[Recordings of the seismic signals were obtained at permanent and mobile seismological
observatories established by the U.S. Technical Applications Center as a task under the Advanced
[Research Project Agency's project Vela Uniform. Copies of the hydroacoustic and seismic
recordings have been made available to us for spectral analysis.
In the general case three different spectral series may be present in the seismic signal. They are the
previously cited odd harmonic series with fundamental fH = c/4H, an odd harmonic series with
fundamental fS= c/4d, where d is the detonation depth, and fS is the fundamental frequency from
the free surface, also the complete harmonic series with fundamental frequency of fb = l/T, where
T is the first bubble pulse period. Using a bubble pulse frequency 2.75 Hz, the detonation depth is
estimated at 797 m for detonation depth using at Bermuda recording data and at a depth of 1786 m
in the ocean using the vertical short-period seismometer at Winner, South Dakota at a range of
2319 km from Fig. 1 and Table 2. This information provides the unknown parameters (or some
incorrect estimates) of detonation depth and the ocean depth for the underwater explosions for the
Chase Program.
@seismicisolation
@seismicisolation
Overview and Review on Analysis for Hydroacoustic Studies in Underwater Explosions 287
Fig. 1. Bubble pulse frequency (upper), using 2.75 Hz, 5.50 Hz, 8.25 Hz… using 2.75 Hz and 2 kt TNT, we can
estimated detonation depth, i. e. 2.75=(d+10)5/6/2.1*20000001/3, d=797.3 m and bottom reflection frequency (lower),
0.21 Hz, 0.63 Hz, 1.05 Hz…using 0.21 Hz, Ocean depth H=1500/4*0.21=1785.7 m (~1800 m). (Weinstein, 1968).
@seismicisolation
@seismicisolation
Fig. 2a. Bubble pulse frequency, 0.83 Hz, 1.66 Hz, ~2.49 Hz….using 0.83 Hz and 0.3 kt, detonation depth is estimated
from 0.83 = (d+10)5/6/2.1*300001/3, d=292 m (Weinstein, 1968).
Table 2. Source details for the Chase Program

Experiment Chase 2 Chase 3 Chase 4
Epicenter latitude 38⁰49’00’’N 37⁰11’48’’N 37⁰11’34’’N
Epicenter longitude 72⁰14’30’’W 74⁰21’06’’W 74⁰26’34’’W
Date 1964-09-17 1965-07-15 1965-09-16
Origin time (UTC) 22:07:45.0 14:16:08.1 19:51:10.2
Yield (kt) 2.0 (nominal) 0.70 0.30
Detonation depth (m)
Nominal 300 300
Measured at sea 275 275
Measured in this 1800 ?? (797) 300 275 (292/289)
paper
Water depth 2150 (1786) 1525 1550
The brackets refer to the present study.
@seismicisolation
@seismicisolation
Fig. 2b. We can see the fundamental bubble pulse frequency, 0.83 Hz and the surface reflection frequency, 1.3 Hz and
the second higher-mode bubble pulse frequency 1.66 H from recording on the vertical short-period seismometer at
Wichita Mountain. Using surface reflection frequency, 1.3 Hz, we can estimated the detonation depth at d=1500/4*1.3
= 288.5 m which is closely agreeable with the measured values (Weinstein, 1968).
The results presented here, however, when combined with the previously cited publications,
demonstrate that under appropriate conditions spectral maxima will appear in the seismic signals at
frequencies corresponding to the odd harmonic series related to detonation depth and water depth
and the full harmonic series generated by the first bubble pulse. By contrast, only the harmonic
series generated by the bubble pulse will appear in the hydroacoustic signals observed at long
range due to the large charge weight. However, one can hardly see the reverberation frequency
from the seabed except for Chase 2 because of the rough variation of the seafloor.
Reymond et al. (2003) studied two presumed underwater explosions,detonatedon13 April 2000
(approximate times 00:19 and 23:29 coordinated universal time), at a site located approximately
215 km southwest of Oahu, Hawaii, and detected from a combination of T phases recorded at
shore-based seismic stations and acoustic waves recorded by hydrophones. The explosions were
initially detected by the Polynesian Seismic Network, and a preliminary location obtained in the
vicinity of Kauai. the variation of spectral amplitude with frequency,the observation of a strong
frequency dispersion in the spectrograms, and the identification of a bubble period (0.45 sec) in the
cepstra of the signals, which translates into a yield of 275 kg of equivalent TNT (ML=2.27) (Kim
and Gitterman, 2013) at a depth of 50 m. In the context of monitoring the Comprehensive nuclear
Test Ban Treaty, these two explosions provide a perfect opportunity to assess the capabilities of T-
phase stations and hydrophones for detection, location, identification, and quantification of these
sources. Our study, conducted in the absence of any ground-truth information, stresses the
@seismicisolation
@seismicisolation
possibility of a powerful synergy between these two types of recording facilities, but also points to
several limitations in the performance of certain shore-based seismic stations.
The quefrency Q of the spike is “period”of pulsation of the bubble generated by the explosion, and
has been related to its depth h and yield Y, through the empirical formula proposed by Chapman
(1985).
Q = 2.1Y1/3/ (h+ 10.1)5/6 (1)
where Q is in seconds, Y in kilograms of equivalent TNT, and h in meters. In principle, the
question of the trade-off between Y and h can be addressed through a comparison of
amplitudesrecordedonPolynesianatollsforthe2000events and explosions with published source
parameters (Nava et al., 1988; Weigel, 1990; Talandier and Okal, 2001), taking into account an
appropriate distance correction. We conclude that a possible scenario would involve a yield of 275
kg fired at a depth of 49.3 m, from Equation (1) and Fig. 4.
Fig. 3. Map of the epicenter of event I (solid dot) and of the stations used in this study (seismic, upward triangles;
hydrophones, inverted triangles). The inset details the position of the shot with respect to the Hawaiian Islands. Note
that the paths to BNB and H25 are unobstructed by the islands. Even on the scale of the inset, event II would not be
distinguishable from event I.
This article studies two presumed underwater explosions, detonated on 13 April 2000,
approximately 215 km southwest of Oahu, Hawaii, and detected from a combination of T phases at
shore-based seismic stations and acoustic waves recorded by hydrophones (Fig. 3). In the context
of monitoring the Comprehensive Nuclear Test Ban Treaty (CTBT), these sources provide a
perfect opportunity to assess the capabilities of T-phase stations and hydrophones for detection,
location, identification, and quantification of marine events. They illustrate the power of synergy
between the two types of recording facilities, but also point to limitations in the performance of
certain shore-based seismic stations. The two explosions, hereafter events I and II, were detected
during routine seismic processing at the Central Laboratory of the Polynesian Seismic Network
(hereafter, RSP) in Papeete, Tahiti, on Thursday, 13 April 2000 local time (event I) and Friday, 14
April 2000 (event II). This network has been described in previous publications (Talandier and
Kuster, 1976; Okal et al., 1980) and it includes short-period stations equipped with special band-
@seismicisolation
@seismicisolation
pass filters in the frequency range 2–10 Hz. Such instrumentation, initially developed in the 1970s,
has helped define the concept of so called “T-phase stations,” later mandated by the CTBT (Okal,
2001).
We clearly see a clear negative peak at around 0.13 s which may be related to the surface
reverberation reflection and we can estimate the detonation depth using this reverberation
frequency. The detonation depth is: d=1500/4*7.69 = 48.9 m where 0.13 s= 7. 69 Hz. but there
may not be the reverberation frequency from seafloor due to the deep Pacific Ocean.
Table 1. Seismic and Acoustic Stations Used in this study (Reymond et al., 2003).
Code Name Network Latitude Longitude SPS (Hz) Distance (m)
PMO Poma RSP -15.017 -147.906 50 4073
PPT Papeete RSP -17.569 -149.574 50 4293
TBI Tubuai RSP -23.349 -149.461 50 4916
XMAS Christmas IRIS 2.045 -157.445 20 1996
KIP Oahu IRIS 21.420 -158.020 20 225
HON Honolulu PTWC 21.322 -158.008 100 219
KAA Hawaii HVO 19.266 -155.871 100 392
RNB Barry CNSN 52.576 -131.752 100 4339
WK30 Wake Is. IDC 19.410 165.856 240 3623
WK31 Wake Is. IDC 17.927 167.499 240 3474
H25 East Pac. EEPHA -7.992 -109.937 250 6251
The second maximum in the cepstrum, at about 0.82 sec (Fig. 4), emerges marginally from the
background noise at WK31 and H25. However, it may be the second bubble pulse signal by 1.8 x
0.46=0. 819 assuming that the second bubble period is about 1.8 times of the first bubble due to
the bubble-shape deformation increasing a bubble pulse period (Baumgardt and Der, 1998). (b)
Same as (a) for event II. Probably we may find the fundamental frequencies reflected form the
seafloor since the seismic energy may not reach the bottom of the Pacific Ocean floor owing to its
deep depth compared to a small size of yield and the irregular seafloor.
We illustrate this limitation on Fig. 5, which presents the cepstra of the land-based records at KAA
and PMO, the former using a sampling rate of 100 Hz, the latter 50 Hz. At KAA, the spike of
event I is extracted at a quefrency of 0.45 sec, in full agreement with the hydrophone results; at
PMO, the maximum of the cepstrum is indeed found at Quefrency= 0.44 sec, but the sampling is
too coarse to give it the character of an undisputed spike. In event II, the spike is tentatively
identified in the KAA cepstrum, but does not emerge from background noise at PMO.
We conclude that, although land-based T-phase stations have the potential to resolve bubble
periods, a high sampling rate, on the order of 100 Hz, is an absolute necessity. In this respect, the
traditional sampling rate of 20 Hz used in the broadband IRIS stations was found to be insufficient
to allow an adequate study of the decay attenuation parameter, let alone to resolve the bubble
periods. Even the improved sampling of 40 Hz, now available at many IRIS sites, may be too
coarse for the latter.
@seismicisolation
@seismicisolation
Fig. 4. Investigation of the bubble pulse in the records of event I at the hydrophone stations WK30, WK31, and H25.
Top, Individual time series; the vertical scales are arbitrary digital units. Bottom, Cepstrum of the windowed time
series at each location. Note the consistent spike at the quefrency 0.455(±0.001) sec.
@seismicisolation
@seismicisolation
Fig. 5. Extension of cepstral investigation to land-based stations KAA (left) and PMO (right). Note the generally lower
signal-to-noise ratios at PMO. The bubble spike in the cepstrum is well extracted only for event I at KAA.
Another good example for cepstral study is found from the Russian Journal of Acoustical Physics.
Gomashava and Zakharov (2002) expressed the period of the first fluctuation T1 (bubble period) in
a different form with more detailed numerical constants and estimate the explosion depth by
T1 = 2.08W1/3/(Z + 10.07)5/6 (2)
where Z is the explosion depth , W is the mass of the explosive charge in kilograms, and T1 is the
period of the first fluctutation in seconds.case, from the autocorrelation.
Russian scientists (Khristoforov, 1996; Gromashava and Zakharov, 2002) have studied some
special underwater explosions using a very small seismic source and a very large on showing
characteristics of bubble pulses in cepstrum and waveform shown as below:
@seismicisolation
@seismicisolation
Fig. 6a. Cepstrum of the bottom reverberation caused by an almost simultaneous explosion of two 0.21-kg TNT
charges fired at the depths 198.5 and 206 m (Gromashava and Zakharov, 2002).
This method provides a high resolution in depth, which is confirmed by Fig. 6a showing the
cepstrum of the reverberation and bubble pulse caused by an simultaneous explosion of two
charges fired at depths of 198.5 and 206 m. One can see that the cepstral peaks corresponding to
different depths are well resolved by the autocorrelation function. We can estimate the bubble
pulse periods at 0.0140 s and 0.0144 s using T=2.08 (0.21)1/3/(206 + 10.07)5/6 = 0.0140 s and
T=2.08 (0.21)1/3/(198.5 + 10.07)5/6 = 0.0144 s, respectively. We cannot see the reflections from the
surface and the seabed from reverberation effects because they may be in the range of about 0.5 s -
8.0 s which is beyond the scale. However, the reverberation phenomena may have not appeared
because of the too small explosive charge weight compared to the depth of detonation and ocean.
Fig. 6b. The underwater nuclear explosion on October 23, 1961 was conducted at the Bay Chernaya
of Novaya Zemlya Island at a depth of about 101 m with 1.56 ktTNT (recalculated from M 5.10).
P, S, LR, BP, RF and SR indicate P- and S- wave arrivals, Rayleigh wave, bubble pulse and
reverberation frequencies reflected from the sea floor and sea surface at 3.61 s after P-wave onset
which were recorded on the broadband (0.2-30 sec) at town Belushya 120 km away from the
epicenter.
A principal feature of underwater explosions is a distinct rarefaction impulse (negative) arrival (BP)
with amplitude 2 times larger than that of the first P-wave arrival. The first bubble pulse period is
4.82 sec (0.207 Hz) and the second one is about 2.53 sec (0.395 Hz). Using bubble period 4.82 s
and the modified Rayleigh-Willis Equation [T=2.1 W1/3/(d + 10)5/6], we calculated the detonation
@seismicisolation
@seismicisolation
(actual) depth at about 101 m even if the reported depth was 20-50 m (Khristoforov, 1996). Using
the seafloor reverberation frequency (1.93 Hz) in the record, the sea depth is estimated at 194 m
(see Fig. 6b). It should be also noted that one can see a negative impact reflected from the sea
surface at 3.61 s from the P-wave arrival onset.
The most definitive seismic evidence that the main Kursk event was dominated by an explosion
source is the observation of a "bubble pulse." Explosions that occur underwater generate a bubble
of hot gases that quickly rises to the surface. This gas bubble oscillates in response to the confining
hydrostatic pressure, and this oscillation has a dominant frequency that is related to the type of
explosive, the yield of the explosive, and the depth of detonation. The oscillation of the bubble is
called the bubble pulse, and it produces a distinctive signature on seismic records. In particular,
waveforms from underwater explosions have a scalloped amplitude spectrum, similar to that
produced by ripple fired mining blasts. We illustrate stacked amplitude spectra for the main Kursk
event in Fig. 7. A scalloped spectral pattern is consistently observed, with adjacent troughs (peaks)
separated by -1.45 Hz. The fact that the same pattern is observed at stations located at varying
distances and azimuths implies that the scalloping is a source effect and is unrelated to propagation
effects. Furthermore, the same spectral pattern is observed whether the time window is chosen to
single out a specific phase or entire waveform is selected. Spectra of comparably sized time
windows of seismic noise do not demonstrate the pattern of peaks and troughs illustrated in Fig. 6.
A second phenomenon associated with underwater explosions that causes distinct peaks and
troughs in the amplitude spectra is the reverberation of seismic energy in the water column. The
reverberations come about because the ocean-atmosphere interface is an excellent reflector of
acoustic energy Observations of these "water multiples" are very common in seismic exploration
experiments and even for shallow earthquakes. The broad spectral peak centered at 9 Hz is
probably due to water multiples. Assuming that the wave speed in the ocean is -1500 m/s, this
frequency gives a water depth of -85 m, which is similar to the reports from the sinking site of
-100 m. Note that the water depth is much too shallow for reverberations to cause spectral peaks at
the lower frequencies where the bubble pulse peaks are prominent. A relationship between bubble
pulse frequency and explosive yield, and detonation depth has been developed and verified using a
large population of chemical explosions. The tradeoff between detonation depth and explosion size
makes it impossible to determine a unique yield estimate. However, if we assume a detonation
depth of 85-100 m, the bubble pulse frequency of 1.45 Hz results in yields of 3100-4500 kg
equivalent TNT. However, if we assume a detonation depth of 61 m, the bubble pulse frequency
of 1.45 Hz (0.69 s) results in yields of 1500 kg equivalent TNT resulting in magnitude
(ML=2.83) (Gitterman, 2002; Kim and Gitterman, 2013) which is very acceptable in our
study. The spectral scalloping can be used to address another important question about the Kursk
explosion: Was the main shock due to a single large explosion or to several small explosions
detonated sequentially? The stacked spectra are very clean and only the fundamental scalloping at
1.4-1.5 Hz stands out sharply.
@seismicisolation
@seismicisolation
Fig. 7. Bubble pulse frequency at 1.45 Hz and sea floor reflection frequency at 9 Hz for the Kursk main shock (Koper
et al., 2001).

Gitterman (2002) carried out detailed spectral analyses for the Kursk underwater explosion and
found out the detonation depth and the explosive charge weight in terms of TNT. It has been some
time since the tragic sinking of the Russian submarine Kursk in the Barents Sea, on 12 August
2000. However, uncertainty regarding possible causes and scenarios of the accident still remain.
The below figure is one of spectral analyzing showing bubble pulse frequencies from the Kursk
under water explosion.
An analysis conducted by NORSAR indicated that there were two explosions, the larger of which
occurred at 07:30:42 GMT. This explosion had a magnitude of 3.5 on the Richter scale,
corresponding to about 1-2 tons of explosive in water. A smaller explosion with a magnitude of
1.5 was recorded at the same location 135 seconds earlier. However, the NORSAR bulletin, placed
on the same Web page and based on recordings of the closest station APA (Apatity, Russia) and of
Scandinavian stations ARCES, FINES, SPITS, and NORES, shows the estimated average local net
magnitude 2.8 which is the same result that is calculated from 1500 kg TNT and magnitude-charge
weight relation by Kim and Gitterman (2013). Nevertheless, there is known to be a significant
discrepancy between the local magnitudes for seismic events in this region and the magnitude
based on Richter scale (Fig. 7).
. For underwater explosive sources this may lead to significant differences due to the uncertainties
of high Q factor with low attenuation in sea waters, optimum detonation, etc., a magnitude
correction of 0.7 units was introduced by NORSAR, Richter magnitude of 3.5 in the press release.
@seismicisolation
@seismicisolation
Fig. 8a. A: Seismograms of the large explosion recorded on the ARC (ARCES) array (457 km);B: joint analysis of the
closest seismic stations APA and ARC, showing smoothed spectra of different channels in the 30 sec window,
including P waves; C: the averaged spectrum , stability in the spectral modulation pattern and maximum frequencies is
observed for both stations. The bubble pulse frequencies from spectral analyses for the Kursk Submarine explosion in
the Barents Sea (Gitterman, 2002). The series of 1.46, 2.91 and 4.08 Hz corresponds to bubble pulses which are close
to 1.45 Hz, 2.90 Hz and 4.35 Hz etc.
@seismicisolation
@seismicisolation
Fig. 8b. From the ARCES array C; smoothed signal spectra for all plotted channels D: the average spectrum.
Maximum frequencies are also shown. All spectra are calculated for the window 60 sec. The bold and thin arrows
indicate the reverberation frequencies (3.13 Hz and ~9 Hz) from the sea bottom and from free surface (6.1 Hz) which
result in detonation at a depth 61 m in the ocean depth of 120 m (Gitterman, 2002).
In contrast to the complete harmonic series caused by the bubbling effect, water reverberations
produce two odd harmonic series with fundamentals fH and fd (Weinstein, 1968):
fH = V/4H; fnH = (2n-1)fH, n = 1, 2, ….n (3)
fd = V/4d; fnd = (2n - 1)fd, n = 1, 2 .... n (4)
where V is the sound velocity (m/sec), H is the water depth (m), and d is the detonation depth (m).
For an underwater explosion, the resonant frequencies constitute a complete harmonic series with
a fundamental frequency fb, which is characterized by the detonation depth d (m) and the explosive
yield W (kg of TNT) (e.g., Willis, 1963):
fb = (d + 10)5/6/2.1*W1/3 (5)
The 3.13 Hz peak is most likely the fundamental reverberation frequency reflected from the sea
floor whereas the strong peak at 6.1 Hz in Fig. 8bcahe 3.13 Hz peak represents the the
reverberation frequency fH and the peak 8.7 Hz (~9 Hz) corresponds to 3fH amplified by the first
reverberation harmonic 3 fH . (See Fig. 8b)
@seismicisolation
@seismicisolation
Fig. 9. Velocity and acceleration records with 20 and 100 sps by an UWE of ROKS Cheonan which prove an
underwater explosion for the cause of the ROKS Cheonan sinking showing the characteristics of the positive higher
amplitude of the first P-wave motion with bubble pulse (BP) on the vertical component (HHZ) and bubble jets (BJ) on
the horizontal components (HHN and HHE).
The above seismograms and accelerograms (Fig. 9) are the most important evidences that the
ROKS Cheonan sinking was due to an underwater explosion, i.e. the seismological record could be
the smoking gun in investigating the cause of the ROKS Cheonan Sinking. P, S, BP , PP, LR and
LQ indicate the first of P- and S-wave arrivals, gas Bubble, peak pressure, Rayleigh wave and
Love wave. This photo must have approved that the ROKS Cheonan sinking was due to the non-
contact underwater explosion because one could see all compressions (upward, +) of P-wave
arrivals in the 3-components (N-S, E-W and Z). Using the direction of the resultant of N-S and E-
W components and the first motion compression of Z-component, the epicenter of an explosion is
SW (226⁰). The maximum acceleration and velocity of 0.084 gal and 8.4 µm/sec are recorded on
Z-component about 10 km away from the epicenter. Those values are negligible at the faraway
distance even if the detonation affected the fatal damage to the ship (Fig. 9)
@seismicisolation
@seismicisolation
Fig. 10. Seismograms and spectra of 3-components at the Baengnyeong Island seismic station showing predominant
Rayleigh waves (LR) on HHZ and Love waves (LQ) on HHE and HHZ as well as Bubble Pulse (BP), Bubble Jet (BJ)
and Toroidal Bubble Deformation (TB) on the spectra. This figure clearly proves that the ROKS Cheonan sinking was
due to an underwater explosion.
The first and second bubble pulse frequencies (BP arrow marks) were determined at 1.012 Hz
(0.988 sec) and at 1.723 Hz (0.5804 sec), respectively (Fig. 10). The small arrow mark at HHZ
may correlate with the bubble pulse arrival which is difficult to find a rarefaction motion
(downward impact phase) in the time domain because the detonation depth is too shallow to keep
ascending up as a gas bubble, but we can see the evident high amplitude at 1.012 Hz in the
frequency domain. The higher amplitude spectra at 1.65 - 1.85 Hz (TB) on HHN may correlate
with toroidal bubble deformation whereas those at 1.077 Hz (BJ) on HHN and HHE must be the
bubble jet impact (vortex) breaking a hole straight through the ship hull resulting in the splitting
the ship into two parts. The new findings of bubble pulse frequencies (1.012 Hz), bubble jet (1.077
Hz) and toroidal bubble (1.65-1.85 Hz) are the most up-to-date research results in this study.
Some people suggest that there must have been a water columnar plume (a pillar of water) if the
ROKS Cheonan sinking was to do an underwater explosion. However, nobody had seen the water
plume at that time except for watching a flash by two watch sailors on board and a watch soldier
on land. The underwater explosion could not produce a water column because the shallow
detonation depth could not have enough water volume to build a water column.
@seismicisolation
@seismicisolation
Fig. 11 Seismograms and spectrograms using 3-components recorded at the Baengnyeong Island seismic station.
BAR3C_spectrogram_time window 5sec_logscale_smooth-spline_total 50 sec _40Hz _frequency range. The strong
energy appears at bubble pulse and reverberation frequency from the seabed.
Utilizing T=2.08 (W)1/3/(d + 10.07)5/6, 1.012 = (d + 10.07)5/6 /2.08 (136)1/3, d = 7.36 m.

Utilizing Fb = ( d + 10)5/6/2.1W1/3, 1.012 = (d + 10)5/6/2. 1W1/3, d = 7.63 m.
Using reverberation 45 Hz of the surface reflection (Figs, 11, 12 and 13), the detonation depth d =
1450/4*45 = 8.06 m. The bubble pulse in cepstral analysis is not very clear due to the very shallow
detonation and a contamination with shock waves via a sonic boom at the almost same time.
Raypaths refract away from the sea surface and are going up to the surface due to the increase of
sound velocity with depth in the maritime environment of the cold season (Kim et al., 2014). The
raypaths clearly show the surface reverberation from the free surface as well as the seabed
reverberation reflected from the sea bottom because the detonation depth and size are well
adjusted to generate raypaths (seismic waves) to fully reach the sea surface as well as the seabed.
@seismicisolation
@seismicisolation
Fig. 12. BP represents a bubble pulse which may be immediately followed by a bubble jet peak that penetrates the gas
bubble pulse and a gas recompression on the vertical component even if it normally appears on the horizontal
components (N-S and E-W). RH and Rd are related to reverberation from the bottom and the surface immediately
followed by the first P-wave onset.
Fig.13. Ray-tracing based on hydroacoustic wave speed in the early spring season with a UWE source detonated at a
depth of about 8 m offshore the Baengnyeong Island in the Yellow Sea. We can see two kinds of channel waves in the
shallow depth at the detonation depth and from the sea bottom.
@seismicisolation
@seismicisolation
The ocean depth is about 45 m and the distance between the explosion site and land is about 2,5
km. The ray-tracing modeling presents the significant reverberation effects as well as the bubble
effects when the source (explosion) occurs at a depth of about 8 m in case of the ROKS Cheonan
Sinking because there are significant raypaths reflected from the free surface as well as from the
seabed (Fig. 13).
There are some good examples showing the guide-wave channels in the underwater studies (Pilger
and Ceranna, 2017). Pilger and Ceranna (2017) showed the excellent guide-wave channels through
the open deep sea in the below Fig. 14. In order to observe SOFAR, there should be the lower
velocity layer around 1000 m in the Atlantic Ocean in Fig 14, but in Fig. 15 which does not
include the low velocity zone of SOFAR, it is impossible to see the wave-guide channel..
Fig.14. Underwater sound wave simulated by ray-tracing method. The upper frame shows ray tracing propagation
from an equator source to the polar regions along 30° W longitude. The lower Frame zooms into a 3°(¬ 330 km)
segment highlighting single rays within the SOFAR channel (Pilger and Ceranna, 2017).
@seismicisolation
@seismicisolation
Fug. 15. The underwater sound speed profiles showing attenuation of the sound pressure in dB in color-coded. There
are two channels as a SOFAR channel (lower frame) and no SOFAR (upper frame) (Pilger and Ceranna, 2017).

Relationship between seismic magnitude and explosive charge weight for underwater explosions
First of all, the preliminary reported USGS magnitude 3.7-3.8 of US Navy shock trial underwater
explosions (Heyburn et al., 2018) was based on Richter scale, not body wave magnitude (mb). It
may not be reliable to estimate such small size events outside the seismic network of the United
States. Furthermore, the USGS magnitude is estimated based on the land formula from land events,
not those from underwater explosions resulting in larger amplitude due to high quality factor Q in
the sea waters (Shapira, 1981).
The averaged magnitude from the REB magnitudes determined by IDC/CTBTO is more
appropriate than USGS preliminary reported magnitude (PDE) to use the relationship of
magnitude vs charge weight for US Navy shock trial underwater explosions (Table 1).
Heyburn et al. (2018) attempted to use the Dead Sea formula which has a high salinity (33.7%)
compared to the normal seawater (0.35%) resulting in a larger magnitude due to the higher
acoustic impedance, low attenuation of seismic waves. The USGS magnitude 3.7 is a preliminary
reported Richter magnitude right after an underwater explosion was detonated. In general, it is
most difficult to make an appropriate relationship between an explosive charge weight and a
magnitude for underwater explosions. There are various researchers (Jacob and Neilson, 1977;
Gitterman, 1998; Gitterman and Shapira, 2001; Booth, 2009; Kim and Gitterman, 2013) who have
intensively studied underwater explosions in relation to explosive charge weight and magnitude
which is substantially a body wave magnitude (mb) for a modified local magnitude. I would like to
refer authors (Heyburn et al., 2018) to use one of formulas for magnitude – charge weight relation
for an underwater explosion (Kim and Gitterman, 2013) as follows:
ML = 0.753 log (W) + 0.436 (6)
@seismicisolation
@seismicisolation
where ML is the modified local magnitude equivalent to mb. The recorded local magnitude for the
Dead Sea explosions agreed well with values predicted by an empirical magnitude charge weight
relation derived using a series of small-scale Dead Sea explosions detonated in 1993 (Gitterman,
1998).
ML = 0.285 + log (W) (7)
However, the higher salinity and hence its acoustic impedance is significantly higher than normal
sea waters in regions such as the coastal waters of Florida, which leads to a larger magnitude
estimate as the energy flux density scales with the reciprocal of the acoustic impedance. I am sure
that a relationship between charge weight and ML from previous studies (Kim and Gitterman, 2013)
of explosions in the normal seawater is shown to provide reasonable estimates of the charge
weights of the seawater near Florida.
Table 3. Ground-Truth (GT) Parameters and Parameters Published in the Reviewed Event
Bulletin (REB) for Explosion Shock Trails Conducted East of Florida in 2001, 2008, and
2016 (Heyburn et al., 2018).
Ground-truth Estimate (GT) Revised Event Bulletin (REB)
Date OT Lat (N) Lon (W) OT Lat (N) Lon (W) Mag
2001/05/24 17:16 30.307 79.91 N/A N/A N/A N/A
2001/06/03 14:58 30.243 79.851 14:58:12.25 29.890 79.480 3.3
2001/06/03 18:28 30.165 79.908 18:27:54.43 30.199 79.681 3.4
2008/08/16 19:15 29.887 79.725 N/A N/A N/A N/A
2008/08/26 21:01 29.824 79.583 N/A N/A N/A N/A
2008/09/13 17:05 29.811 79.569 N/A N/A N/A N/A
2016/06/10 17:12 29.941 79.575 17:10:48.97 30.084 79.633 3.6
2016/06/23 17:20 29.948 79.479 N/A N/A N/A N/A
2016/07/16 20:01 29.676 79.573 20:00:12.28 29.442 79.687 3.2
2016/09/04 N/A N/A N/A 18:29:31.68 30.362 79.537 3.2
2016/09/21 N/A N/A N/A 16:30:54.83 30.173 79.568 3.2
Mean 3.32±0.15
Authors should also find the corrected plot star in the Figure using this average magnitude. The
authors will see the plotting of the Florida shock trial explosions exactly stay along the line of
Equation 1. Furthermore authors should ensure that a magnitude in the Dead Sea is always larger
than that in the normal seawater, so Gitterman’s formula is perfectly correct in the Dead Sea and it
must be modified in order to use in the normal seawater. Therefore using Table 1 and W=6759 kg,
ML=0.753 log (W) + 0.436, authors will obtain ML= 3.32 whereas using Equation 2, and W=6759
kg, ML=0.285 + log (W)=4.11 in the Figure 11. This estimate also complies with many other study
methods (Jacob, 1975; Jacob and Neilson, 1977; Booth, 2009) as well as the plot in the Figure
(Kim and Gitterman, 2013). Equation 1 trend was well presented in Figure 5 in authors’ paper with
other studies such as Gitterman and Shapira (2001), Kim and Gitterman (2013) including land
explosions of Khalturin et al. (1998) and Brocher (2003) as well as main plots of UK studies off
@seismicisolation
@seismicisolation
the UK coast and North Sea (Jacob, 1975; Jacob and Neilson, 1977; Booth, 2009). The similar
figure as below was already presented at SnT2017 as T2.3-P13 _Heyburn posters at SnT2017,
CTBTO, 20-24 June, 2017, Hofburg Palace, Vienna, Austria
Fig. 16. Reported magnitude for the underwater and land detonations versus the charge size. empirical relations between
local magnitude and underwater explosions from this study (Eq. 1) (solid line), by Gitterman and Shapira (2001) for
Dead Sea underwater explosions (Eq. 2) (dashed line), by Khalturin et al. (1998) the empirical upper limit magnitude of
chemical land explosions in the hard rock, fitting nuclear explosions (Eq. 3) (dotted and dashed lines) and coda
magnitude of land explosions in California and Nevada, including most of coda duration magnitudes (Brocher, 2003)
(Eq. 4) (dotted line). The magnitude-charge weight relation for underwater and land explosions stressing on the US
Navy shock trials off coast Florida. Open star, diamond and triangle symbols indicate relationships between magnitude
and explosive charge weight for US Navy shock trials using Equations 1 (ML=3.3 from 6759 kg TNT), this paper (7045
kg TNT VS. ML=3.8) and Equation 2 (ML=4.1 from 6759 kg TNT) (Heyburn et al., 2018). The red star indicates an
underwater explosion of the ROKS Chenan Sinking.
Equation 1 (Kim and Gitterman, 2013) is calculated from data of UK off Sea and North Sea
whereas Equation 2 (Gitterman, 1998) is obtained from data of the Dead Sea. This suggests that
provided an appropriate correction is made for the effect of the lower acoustic impedance in
seawater with normal levels of salinity, the magnitude–charge weight relationship in Equation 1
can provide reasonable estimates of the charge weight for underwater explosions at similar
detonation. The plot of diamond is away from major distribution of magnitude-charge weight
distribution in Fig. 16. I would suggest that the Equation 1 may be an easy and simplified formula
to estimate the magnitude-weight relation for the underwater explosions for normal sea waters not
the Dead Sea because of overestimated magnitude such as 2.0 (ML)using 24 kg TNT in the Dead
Sea (Gitterman et al., 1998). In study of Heyburn et al. (2018), the authors tried to apply the
formula obtained from the Dead Sea with high salinity instead of using normal sea waters resulting
in inappropriate interpretation between the magnitude and charge weight for the underwater
explosions of the US Navy shock trials (Heyburn et al., 2018). In Fig. 12, it is the best way to
present the relationship between the seismic yield (explosive charge weight) and bubble pulse
period in terms of detonation depth, charge weight and babble pulse period in the empirical
monograph (Swisdak, 1978). It is confused for interpretation of Figure 8. 21 in the classical and
@seismicisolation
@seismicisolation
exclusive underwater explosion book (Cole, 1948) sometimes unless one understands that the plot
is done by period versus (d + 33)5/6 in feet which is similar to the thick red line in Fig. 17.
Fig. 17. Nomograph of the first bubble pulse period, maximum bubble radius and detonation depth for a gas bubble
from an underwater explosion (Swisdak, 1978). The detonation yield in terms of TNT can be calculated by estimating
bubble pulse period and detonation depth from the spectral analysis. The thin blue and red lines represent the
detonation of 136kg at a depth of 8 m and 7.32 m (Kim and Gitterman, 2013), whereas the dashed red and thin red
lines indicate the detonation of 250 kg TNT at a depth of 15 m with a bubble radius of 8 m and the detonation of 250
kg TNT at a depth 10 m with a bubble pulse 1.1 s (from Joint Investigation Report of South Korean Defense Ministry).
The thick black line indicates the detonation of 100 kg TNT at a depth of 500 m with the bubble pulse period of 0.054s
and the maximum bubble radius of 2.033 m (Swisdak, 1978). Orange, green and thin black lines indicate underwater
explosions of US navy shock trials at a depth of 60 m with 6759 kg TNT (Heyburn et al., 2018), at 50 m with 275 kg
TNT (Reymond et al., 2003), and at 61 m with 1500 kg TNT for the Kursk underwater explosion (Gitterman, 2002),
respectively.

In general, it is important to use forensic seismology in order to find physical parameters such as
detonation charge weight, detonation depth including water depth as a smoking gun for a
clandestine underwater explosion. The best analyzing technologies for underwater explosions are
to use spectral and cepstral analyses. There are two basic characteristics for underwater explosions
are bubble and reverberation effects which can be found through the spectral analysis. We estimate
detonation depth, detonation weight, and ocean depth, but it is not appropriate to apply a
magnitude based on Richter scale for a size of underwater explosions instead of explosive charge
weight (kg) equivalent TNT because a source size in water is not the same as in land due to a high
Q factor with little attenuation, salinity and pressure including the optimum depth in which an
experiment conducted using depth charges, it was found that signal amplitudes recorded for
@seismicisolation
@seismicisolation
explosions detonated away from the optimum depth could produce amplitudes that are only 40%
that of an explosion at the optimum depth (the depth at which the first surface reflection and
bubble pulse are in phase). It is always possible to estimate a bubble pulse period, but
reverberation reflections from surface and bottom depend upon a depth of detonation and ocean,
including seabed and a charge weight. The uncertainty associated with yield estimation for
underwater explosions, however, comparable with similar uncertainties observed for underground
nuclear explosions. The method was observed to consistently underestimate depth and yield and
this was speculated to be a result of bubble-shape deformation causing increased period of the
second oscillation. As a result, the ROKS Cheonan sinking was due to an underwater explosion
which occurred at a depth around 8 m, approximately 5 m portside of the hull centerline with 136
kg-TNT net explosive weight equivalent to 2.04 of seismic magnitude.
REFERENCES
Baumgardt, D. R. and Z. Der (1998). Identification of presumed shallow underwater chemical
explosions using land-based regional arrays. Bull. Seism. Soc. Am., 88 (2), 581-595.
Booth, D. C. (2009). The relation between seismic local magnitude ML and charge weight for UK
Explosions, Keyworth, Nottingham, BGS, UK, Brithsh Geological Survey, Earth Hazards
Programme, OR/09/062, Open Report pp. 25.
Brocher, T. M. (2003). Detonation charge size versus coda magnitude relations in California and
Nevada. Bull. Seism. Soc. Am., 93 (5), 2089-2105.
Chapman, N. R. (1985). Measurement of waveform parameters of shallow explosive charges, J.
Acoust. Soc. Am. 78, 672–681.
Cole, R. H. (1948). Underwater Explosions, Princeton University Press, Princeton, New Jersey, pp.
464.
Gitterman, Yefim, Z. Ben-Avraham and A. Ginzburg (1998). Spectral analysis of underwater
explosions in the Dead Sea, Geophys. J. Int., 134, 460–472
Gitterman, Y. (2002). Implications of the Dead Sea Experiment Results for Analysis of Seismic
Recordings of the Submarine Kursk Explosions, Seismological Research Letters 73(1), 14-24.
Gromasheva, O. S. and V. A. Zakharov (2002). Estimation of the underwater explosion depth from
the modified cepstral analysis of sea reverberation. Acoustical Physics, 48 (3), 273–278.
Heyburn, Ross, Stuart E. J., Nippress and David Bowers (2018). Seismic and Hydroacoustic
Observations from Underwater Explosions off the East Coast of Florida, Bull. Seism. Soc.
Am.108 (6), doi: 10.1785/0120180105.
Jacob, A. W. B., and G. Neilson (1977). Magnitude determination on OWNET .Edinburgh, UK,
Institute of Geological Sciences, GSU, Report, 86.
Jacob, A. W. B., (1975). Dispersed shots at optimum depth-an efficient seismic source for
lithospheric studies. J. Geophysics, 41, 63-71.
Khalturin, V. I., T. G. Rautian and P. G. Richards (1998). The seismic signal strength of chemical
explosions. Bull. Seism. Soc. Am.88 (6), 1511-1524.
Khristofrov, B. (1996). About the Control of the Underwater and Above Water Nuclear
Explosions by the Hydroacoustic Methods. Final Report for the Project SPC-95-4019, Russian
Academy of Sciences, Institute of Dynamics and Geospheres, 91pp.
Kim, S. G. and Y. Gitterman (2013). Underwater explosion (UWE) analysis of the ROKS
Cheonan Incident. Pure and Appl. Geophys. 170 (4), 547-560.
Kim, So Gu (2013). Forensic Seismology and Boundary Element Method Application vis-à-vis
ROKS Cheonan Underwater Explosion, J. Marine Sci. Appl. 12, 422-433. DOI:
10.1007/s11804-013-1213-y.
Kim, So Gu, Yefim Gitterman and Orlando Camargo Rodriguez (2014). Estimating depth and
explosive charge weight for an extremely shallow underwater explosion of the ROKS Cheonan
sinking in the Yellow Sea, Methods in Oceanography, 11, 29-39.
Koper, Keith D., Terry C Wallace, Steven R. Taylor and Hans E. Hartse (2001). Forensic
Seismology and the Sinking of the Kursk. EOS, AGU, 82 (4), 37-52.
@seismicisolation
@seismicisolation
Pilger, C. and Ceranna, L. (2017). The hydroacoustic network of the international monitoring
system (IMS), 111-122. – Pilger, Ceranna and Bonnoennemann (eds): Monitorig Compliance
with the Comprehensive Nuclear-Test- Ban Treaty (CTBT), Bundesanstalt for
Geowissenschaften und Rohstoffe (BGR), Honnover, Germany, 328pp.
Reymond, D., O. Hyvernaud, J. Talandier, and E. A. Okal (2003). T- wave detection of two
underwater explosions off Hawaii, Bull. Seism. Soc Am. 93(2), 804-816.
Shapira, Avi (1981). T phase from underwater explosions off the coast of Israel. Bull. Seism. Soc.
Am.71 (4), 1049-1059.
Swisdak, M. M. (1978). Explosion effects and properties: part III-explosion effects in water.
Dahlgren, VA, USA, NSWC/WOL, TR 76-116, 109.
Talandier, J., and G. T Kuster (1976). Seismicity and submarine volcanic activity in French
Polynesia, J. Geophys. Res. 81, 936–948.
Talandier, J., and E. A. Okal (1998). On the mechanism of conversion of seismic waves to and
from T waves in the vicinity of island shores, Bull. Seism. Soc. Am. 88, 621–632.
Talandier, J., and E. A. Okal (2001). Identification criteria for sources of T waves recorded in
French Polynesia, Pure Appl. Geophys. 158, 567– 603.
Weinstein, M. S. (1968). Spectra of acoustic and seismic signals generated by underwater
explosions during Chase experiment, J. Geophys. Res., 73 (16), 5473-5476.
Willis, D. E. (1963). Seismic measurements of large underwater shots, Bull. Seism. Soc. Am. 53,
789-809.
@seismicisolation
@seismicisolation
APPENDIX
Spectral analysis of underwater explosions in the Dead Sea1

Yefim Gitterman, Zvi Ben-Avraham, Avihu Ginzburg
SUMMARY
The present study utilizes the Israel Seismic Network (ISN) as a spatially distributed multichannel
system for the discrimination of low-magnitude events (ML<2.5), namely earthquakes and
underwater explosions in the Dead Sea. In order to achieve this, we began with the application of
conventional single-station methods, such as spectral short-period ratios. We then applied a newly
developed, network-oriented algorithm based on different spectral features of the seismic radiation
from underwater explosions and earthquakes, i.e. spectral semblance statistics.
Twenty-eight single-shot underwater explosions (UWEs) and 16 earthquakes in the magnitude
range ML=1.6–2.8, within distances of 10–150 km, recorded by the ISN, were selected for the
analysis. The analysis is based on a smoothed (0.5 Hz window) Fourier spectrum of the whole
signal (defined by the signal-to-noise criterion), without picking separate wave phases. It was
found that the classical discriminant of the seismic energy ratio between the relatively low-
frequency (1–6 Hz) and high-frequency (6–11 Hz) bands, averaged over an ISN subnetwork,
showed an overlap between UWEs and earthquakes and cannot itself provide reliable
identification.
We developed and tested a new multistation discriminant based on the low- frequency spectral
modulation (LFSM) method. In our case the LFSM is associated with the bubbling effect in
underwater explosions. The method demonstrates a distinct azimuth-invariant coherency of
spectral shapes in the low-frequency range (1–12 Hz) of short-period seismometer systems. The
coherency of the modulated spectra for different ISN stations was measured by semblance
statistics commonly used in seismic prospecting for phase correlation in the time domain. The
modified statistics provided an almost complete separation between earthquakes and underwater
explosions.
Key words: Dead Sea, explosion seismology, seismic spectra coherency, seismograph network.
INTRODUCTION
A seismicity study of the northern basin of the Dead Sea was conducted by the Dead Sea Research
Center, Tel Aviv University. The experiment comprised the deployment of several Ocean Bottom
Seismometers in the lake and their use for recording microseismic signals over a period of two
months. During the recording period a large number of industrial explosions were set off in the
lake in the course of blasting salt monoliths on the bottom. No accurate record of the exact
locations and times of these explosions was kept; we found it necessary, therefore, to develop and
test an efficient and accurate method for discriminating between the signals of underwater man-
made explosions and those of microseismic activity in the Dead Sea basin.
The primary goal of most regional seismic networks, as in the case of the Israel Seismic Network
(ISN), is to monitor earthquakes. The application of earthquake monitoring to earthquake hazard
assessment also depends upon the reliable identification of man-made seismic events and the
exclusion of those events from earthquake catalogs. In this respect, the results of a regional
discrimination study are of great value to operators of local and regional seismic networks,
particularly in regions of relatively low seismicity.
1
Originally published in the Geophys. J. Int. (1998) 134, 460–472.
@seismicisolation
@seismicisolation
Simple seismic source physics would suggest that earthquakes, being dislocation sources, should
generate more shear-wave energy relative to compressional-wave energy than explosions,
therefore, the P/S amplitude ratio is expected to be greater for explosions than for earthquakes (see
e.g. a review by Pomeroy et al., 1982). In practice, P/S ratios for events from Galilee (Northern
Israel) at various distances showed no distinction between earthquakes and quarry blasts
(Gitterman et al., 1996). This discriminant performed satisfactory for only a limited number of
ISN stations and only for strong events (ML > 2) at distances greater than 100 km (for about 18
events from 69).
A number of known regional discriminants is based on spectral differences of earthquakes and
explosions. Different kinds of spectral ratios are used (see a review by Blandford 1995): ratios of
separate phases, for example P to S, in selected narrow bands; and ratios of peaks, r.m.s. or
averaged amplitudes (or power) in different frequency bands for the same phase, mainly Lg.
Suteau-Henson & Bache (1988) showed that quarry blasts in southwest Norway and offshore
events (supposed underwater explosions) recorded by NORESS, can be separated by the ratios of
the average high-frequency power to the average low-frequency power. A similar discriminant, the
spectral ratio, RE, of total seismic energy of the signal in low-frequency and high-frequency
ranges was proposed by Gitterman & Shapira (1993):
f2 f4
RE 

f1
| S ( f ) | 2 df /

f3
| S ( f ) | 2 df (1)
where |S(f)| is the smoothed spectrum of ground velocity in the selected time window, including
the whole wave train. This statistic, averaged over an ISN sub-network, provided a complete
separation of earthquakes and underwater explosions off the Levant coast.
These short-period discriminants depend heavily on the structure and constitution of the uppermost
crust. Their performance varies from region to region and should be tested on a data base with
ground-truth information in each case.
One of the main spectral features used in the identification of industrial explosions is spectral
modulation (SM) which is caused by ripple firing or the bubbling effect. Most of the studies are
based on time-independent patterns for a single station or array. These patterns are presented in
spectrograms or sonograms of the entire seismogram (for example Hedlin et al., 1989), or in
spectra of different regional phases (e.g. Baumgardt & Zigler, 1988).
A different multi-station approach based on the azimuth-invariant SM caused by bubble
oscillation, considered jointly for ISN stations, was implemented for identification of underwater
explosions in the Mediterranean (Gitterman & Shapira, 1993, 1994). This approach was first
discussed by Gitterman & van Eck (1993) for discrimination of ripple-fired quarry blasts and
earthquakes in Galilee. This network-oriented method utilizes low frequency (1-12 Hz) minima
and demonstrates the distinct coherency of spectral shapes for broad ranges of azimuths and
distances. In the present study the advantages provided by the ISN are utilized in the newly
developed discriminant based on the coherency of spectral shapes for underwater explosions, at
different azimuthally distributed stations, which is not present in earthquake patterns. The
coherency is measured by the “semblance” and “cross-correlation” statistics, used in seismic
reflection prospecting for phase correlation in the time domain (Neidell & Taner, 1971). After
some modification, including taking the logarithms of the spectra and subtraction of the average,
the statistics can be written as (Gitterman et al., 1996b):
@seismicisolation
@seismicisolation
semblance:
2
1 F2  N  F2 N
S f    ( S ki  S k )  (S ki  Sk )2 (2)
N F1  k 1  F1 k 1
cross-correlation:
2
1 F2  N  N
2
F2 N
Sc    ( S ki  S k )   ( S ki  S k ) 
N  1 F1  k 1  k 1 
 (S
F1 k 1
ki  Sk )2 (3)
where Ski = log10 Sk(fi) - log spectral amplitude at k station; S k - the average spectral level in the
analyzed frequency band [F1,F2]; N - the number of used stations. The form of Eqs. (2) and (3)
suggests a strong coupling of the two statistics.
THE DATA
The Israel Seismic Network (ISN) is operated by the Seismology Division of the Geophysical
Institute of Israel (GII, formerly the Institute for Petroleum Research and Geophysics, IPRG).
From the ISN we selected a sub-network of stations as shown in Fig. 1. The network consists of
short period (1 Hz) seismometers. Signals are bandpass filtered (0.2 to 12.5 Hz), amplified and
digitally recorded by the ISDA system (Shapira & Avirav, 1990) with a sampling rate of 50
samples per second. All seismograms used in the study were recorded by vertical seismometers.
During the period of observation, the data were transmitted via FM telemetry to the National
Seismological Center at the GII.
During the course of a refraction study of the crust in the Dead Sea transform fault zone carried
out by the Tel-Aviv University in 1993, a series of 78 single-shot underwater explosions with
charges of W = 16-304 kg was detonated. 75 of these explosions were recorded by the ISN. For
our analysis we selected a representative dataset of 28 explosions of four different charge sizes,
distributed uniformly along a line in the middle of the northern basin of the Dead Sea (Fig. 1). The
rest of the recorded explosions with charges 16 and 24 kg were not included due to their identity
and close proximity (several tens meters) to the selected ones. Ground-truth information, including
accurate coordinates, charge weight, shot depth and local time of explosions is presented in Table
1. The local magnitudes, determined from coda duration measurements (Shapira, 1988), were
evaluated for a few explosions of different yields. Overestimated ML values were observed,
especially for the largest shot (W=304 kg), owing to long wave trains of reverberations in the
water layer.
In the context of reliable monitoring of detailed natural seismicity in the area, we also selected for
comparison 16 earthquakes (mostly nocturnal) with magnitudes ML=1.5-2.8, which occurred in the
Dead Sea basin, along the seismically active main fault system, in close proximity to the locations
of the explosions (Table 2). According to ground truth information being regularly collected by
GII, none of the selected events corresponds to quarries activity in the area.
Simple comparison by eye of the seismograms shown in Figures 2a and 3a show differences that
are due to features of the source. For the “typical” earthquake seismograms the S/Rayleigh arrivals
can usually be identified and have the largest amplitudes for each station. For the “typical”
explosions the largest amplitudes are usually in the P group with S/Rayleigh onsets being difficult
to identify. However, a number of events demonstrate anomalous amplitude characteristics. The
small explosion EX13 shows some earthquake-like wave features, i.e. large amplitudes of the
S/Rayleigh arrivals at each station (Fig. 2b) [this is the only example among the underwater
explosions (UWEs) studied here to show such behaviour]. Explosion-like wave features can be
@seismicisolation
@seismicisolation
observed for an earthquake QS12 (see Fig. 3c): the largest amplitudes in the P-group for the five
closest stations, at distances less than 90-100 km. The same effect was noted also in the
seismograms of the earthquakes QS11, QS14 and QS16 (located very close to QS12, near Mt.
Sdom, with a specific geological setting in upper crust salt diapirs, and possibly related to the
Eastern segment of the Dead Sea transform), and in the very shallow earthquakes QS9 and QS10
(see Table 2).
These observations at close distances show that the P/S discriminant, as in the previously
mentioned case of the Galilee events (Gitterman et al., 1996), cannot provide good results for
earthquakes and underwater explosions in the Dead Sea area. Therefore, for reliable identification
of seismic events in the region we should apply some discriminants, based on spectral features of
seismic waves, which have a clear physical interpretation.
Figure 1. Epicentres of earthquakes, underwater explosions (UWEs) and ISN stations in the Dead Sea region. The
single UWEs used for seismic profiling form a line along the Dead Sea. Some events analysed in the study are
labelled. The two sets of axis labels relate to local coordinates (km) and geographical coordinates (degrees).
Table 1. Single underwater explosions used for seismic profiling (shots 70 m below sea level;
X, Y, local coordinates).
EVENT ORIGIN TIME ML X Y LAT LONG. SIZE

(yr/mo/dy/hr/mn) (km) (km) (N) (E) (kg)
EX1 9301280827 3.1 189.7 82.7 31.33 35.42 304
EX2 9301281043 - 189.6 82.3 31.33 35.42 24
EX3 9301281103 - 189.7 82.8 31.34 35.42 24
EX4 9301281203 - 189.9 83.2 31.34 35.42 24
EX5 9301281237 - 190.4 84.6 31.35 35.43 24
EX6 9301281354 2.0 191.1 86.4 31.37 35.44 24

@seismicisolation
@seismicisolation
EX7 9301281446 - 191.7 88.3 31.39 35.44 24
EX8 9301281541 - 192.5 90.2 31.40 35.45 24
EX9 9301281638 - 193.0 92.1 31.42 35.46 24
EX10 9301290456 - 198.0 126.3 31.73 35.51 24
EX11 9301290636 - 197.4 124.3 31.71 35.50 24
EX12 9301291006 2.5 195.9 122.6 31.69 35.49 192
EX13 9301291155 - 197.1 123.3 31.70 35.50 24
EX14 9301310829 - 193.6 94.1 31.44 35.46 24
EX15 9301310916 - 193.5 95.6 31.45 35.46 24
EX16 9301311138 - 193.6 99.6 31.49 35.46 24
EX17 9301311242 - 193.6 101.6 31.50 35.46 24
EX18 9301311321 - 193.6 103.1 31.52 35.46 24
EX19 9301311449 - 193.6 106.8 31.55 35.46 24
EX20 9301311528 - 193.6 108.7 31.57 35.46 24
EX21 9301311621 - 193.9 110.7 31.59 35.47 16
EX22 9301311729 - 194.4 112.2 31.60 35.47 16
EX23 9301311741 - 194.5 112.7 31.62 35.47 16
EX24 9301311921 - 195.0 114.1 31.62 35.48 24
EX25 9301312030 - 195.5 116.3 31.64 35.48 24
EX26 9301312118 - 195.7 117.8 31.65 35.49 24
EX27 9301312246 - 196.6 121.0 31.68 35.49 24
EX28 9301312301 - 196.3 120.0 31.67 35.49 16
SPECTRAL ANALYSIS
We computed the Fourier spectra of ground motions recorded by the ISN subnet which included
10 stations in the distance range 10-150 km (Fig. 1). Recordings at some stations with noise bursts,
spikes and malfunctions were excluded from the analysis. Computations were made usually for a
20-30 sec time window. The analyzed window for a given event is the same for all stations and
includes the entire signal (up to the end of the coda, defined by the signal-to-noise criterion). By
using this window we avoided the unrealistic task of picking separate seismic phases such as Lg
waves (necessary for the P:Lg discriminant, but which are hardly visible at short distances) and S-
onsets (difficult to identify on vertical channels) and accumulated information about source and
possibly propagation features kept in all wave forms, thus enhancing the resolving power of the
considered discriminants.
In order to provide equivalent spectral resolution for different stations, the spectra were instrument-
corrected and smoothed with a triangular operator (Hanning window) in a 0.5 Hz moving window.
The log-log plotting mode was used for single event spectra for a subnet of azimuthally distributed
stations, enabling clear identification of the coherent low-frequency spectral modulation (LFSM)
extrema for underwater explosions in the higher part of the frequency range.
@seismicisolation
@seismicisolation
Detonations in deep water produce seismic waves bearing spectral features caused by a source
effect (bubble oscillation) and a path propagation effect (reverberations). Gitterman & Shapira
(1994) gave a detailed description of low-frequency azimuth-invariant spectral modulation,
observed in ISN recordings of UWE off the Levant coast, which is similar to that observed in
ripple firing.
Table 2. Earthquakes in the southern Dead Sea basin selected for comparison.
EVENT ORIGIN TIME ML X Y DEPTH LAT. LONG. AREA

(YR/MO/DY/HR/MN) (KM) (KM) (KM) (N) (E)
QS1 8901120345 2.2 194.3 88.5 18 31.39 35.47 Dead Sea
QS2 8901120752 2.4 197.7 87.8 18 31.38 35.51 Dead Sea
QS3 8908110029 2.4 182.2 71.0 18 31.23 35.34 Dead Sea
QS4 9001130116 2.0 181.5 44.0 17 30.99 35.33 Arava
QS5 9001141228 2.7 180.4 34.9 18 30.90 35.32 Arava
QS6 9011161920 2.8 199.1 84.9 18 31.36 35.52 Dead Sea
QS7 9102230059 1.9 183.8 37.2 24 30.93 35.35 Arava
QS8 9106031202 1.6 189.1 76.2 21 31.28 35.41 Dead Sea
QS9 9106232222 2.6 199.1 79.0 0 31.30 35.52 Dead Sea
QS10 9109040811 2.1 181.4 45.9 2 31.00 35.33 Arava
QS11 9109272224 2.6 191.5 54.7 11 31.08 35.44 Dead Sea
QS12 9109301155 2.1 193.4 54.1 10 31.08 35.46 Dead Sea
QS13 9110030128 1.5 184.3 31.5 17 30.87 35.36 Arava
QS14 9110050744 1.9 190.5 58.9 18 31.12 35.43 Dead Sea
QS15 9111170201 2.6 184.8 25.5 21 30.82 35.36 Arava
QS16 9112240106 1.6 192.9 54.7 10 31.08 35.45 Dead Sea
It was shown that this is also an interference source-effect, but with a different physical nature: the
interference derives from gas bubble oscillations which are produced by the interaction of the
expanding volume of explosion-generated gases and the hydraulic pressure.
This bubbling effect produces the complete harmonic series of spectral maxima (unlike the odd
harmonic series due to vertical reverberation in the water layer near the source) with a fundamental
frequency f1b, which is characterized by the detonation depth d (m), and the explosive yield W (kg
of TNT) (e.g. Willis, 1963):
fb = (d+10)5/6 /(2.1W1/3); fnb = nfb, n = 1, 2,... (4)
These effects, which are not observed in the case of offshore earthquakes, can be utilized in the
semblance discrimination procedure. Long wave trains of low frequency surface waves are
observed on seismograms of the Dead Sea explosions due to the bubbling effect and reverberations
(see the example for the largest event, EX1, in Fig. 2a).
The high efficiency of the Dead Sea UWEs should be emphasized here: very small yields of
and local magnitudes of ML2.5, exceeding those
W=200-300 kg provide peak amplitudes@seismicisolation
@seismicisolation
produced by local quarry blasts with 50-60 times bigger charges (Gitterman & van Eck, 1993). A
strong coupling, caused in particular by the dense salt water and relatively large detonation depth,
implies execution of effective calibration explosions, which are lacking in the region (e.g.
Baumgardt, 1996) and would be necessary for improvement of Comprehensive Test Ban Treaty
monitoring.
The bubbling actually affects all types of recorded seismic waves throughout the entire
seismogram, but should be better manifested in the initial body-wave portion (P and P-coda) of the
seismogram which is virtually free from reverberations and low-velocity channeling effects:
Figure 4a shows a complete harmonic series (3, 6, and 9 Hz), due to the bubbling effect. The
spectra of the whole signal are much more complicated and mixed with reverberations maxima
(Fig. 4b), however, low-frequency spectral modulation minima (4.5 and 7.5 Hz), corresponding to
the bubbling interference, are sharpened and appear more distinctly. Similar plots for offshore
earthquakes demonstrate an irregular, azimuth-dependent character of the spectral shapes and
minima (Fig. 3).
Using ground truth information (Table 1) and taking into account the high density of salt water in
the Dead Sea (1.236 gr/cm3 compared to the usual 1.03 gr/cm3), we estimated from Eq. 4 the
fundamental bubbling frequency for explosion EX1 as fb2.9 Hz. This corresponds well to the
observed value (Fig. 4a).
In contrast to the complete harmonic series caused by the bubbling effect, water reverberations
produce two odd harmonic series with fundamentals fH and fd (Weinstein, 1968):
fH = V0/4H, fnH = (2n-1) fH, n = 1, 2,..., (5)
fd = V0/4d, fnd = (2n-1) fd, n = 1, 2,..., (6)
where H is the water depth in meters and d is the detonation depth in meters.
The sound velocity V0 in the Dead Sea is approximately 1770 m s-1, the average value of H in the
initial (southern) part of the refraction profile is approximately 310 m (Neev & Hall 1979), and
d = 70 m; we estimated fH 1.4 Hz and fd 6.3 Hz from Eqs. (5) and (6). The spectral modulation
caused by reverberations is better manifested in the spectra of the coda window shown on Fig. 2a.
The spectra on Figs. 4c and 4d demonstrate distinct maxima, fitting (on average) the estimated
values of fH, the odd harmonics 5fH and 7fH and the less visible harmonic 3fH. A noticeable
maximum at about 6 Hz probably contains joint contributions of the close 5fH, fd and 2fb
harmonics.
@seismicisolation
@seismicisolation
Figure 2. Seismograms of underwater explosions: (a) the largest explosion, EX1, with charge weight W =304 kg;
arrows show a coda window used in the analysis (see Fig. 4c); (b) a small explosion, EX13, showing some
earthquake-like wave features (e.g. the largest amplitudes are those of the S/Rayleigh arrivals at each station).
Amplitudes in this and the following figures are not scaled.
@seismicisolation
@seismicisolation
Figure 3. Examples of the Dead Sea earthquakes: (a) ‘typical’ seismograms of event QS3; (b) spectral shapes at
different stations, showing non-coherency; (c) explosion-like wave features for earthquake QS12 (the largest
amplitudes are in the P group, for the five closest stations, at distances less than 90 km).
@seismicisolation
@seismicisolation
Figure 3. (Continued.)
Correct interpretation of spectral features is useful in evaluating the dependence of the energy
distribution in the frequency domain on different effects and on parameters of the source and
medium, and thus in estimating the performance of spectral discriminants.
DISCRIMINATION RESULTS
Both proposed discriminants were applied through an interactive procedure, requiring the
preliminary choice of good-quality station recordings and computation of instrument-corrected,
smoothed amplitude spectra.
We used the relatively low-frequency (1-6 Hz) and high-frequency (6-11 Hz) bands to calculate
the seismic energy ratio (this choice divides the registration frequency range roughly in half),
which had provided good discrimination results for the Mediterranean events (Gitterman &
Shapira, 1993). The ratios, averaged over the ISN subnetwork of 10 stations (Fig. 1), showed a
significant overlap between the two populations of UWEs and earthquakes: the ratios were in the
range 0.7-7.5 for explosions and 0.8-3.2 for earthquakes, with the exception of one event QS9 (see
Fig. 5a).
Testing of other frequency bands, i.e. (1-3)/(6-8 Hz), which were optimal for the classification of
local quarry blasts (Shapira et al., 1997), did not improve the performance of the discriminant (Fig.
5b). This result can be explained by the high-frequency spectral content for most of the UWEs
used in the study, due to small charge weights (W=16 and 24 kg) and the depth (H=70 m).
Comparison of earthquake and explosion spectra (Figs. 3b and 6) shows a similar energy
distribution in the frequency domain. In this case the multivariate discriminant function analysis
for the evaluation of optimal frequency bands is not effective, whereas application of such an
analysis to the discrimination of quarry blasts and earthquakes with significantly different spectral
content was successful (Pinsky et al., 1997). According to Eq. 4, the predicted dominant frequency
@seismicisolation
@seismicisolation
f1b ≈ 6.7 Hz is close to the observed value of about 6.0-6.5 Hz of the maximum (see Fig. 6), which
can also be influenced by the fd harmonic (6.3 Hz) related to the detonation depth. It should be
noted that the evident spectra similarity for these small events confirms the high stability of the
blasting parameters for different explosions (Fig. 6b).
Figure 4. Smoothed spectra for EX1 at several stations for various time windows: (a) the first 8 s of P and P-coda
waves; (b) the whole signal (about 60 s); (c) a coda window (about 30 s), marked on Fig. 2(a); (d) average spectrum
for the coda window. A large spectral modulation is observed, due to the bubbling effect (in parts a and b) and
reverberations (c and d).
@seismicisolation
@seismicisolation
Figure 5. Discrimination results for the Dead Sea underwater explosions and earthquakes: (a), (b) averaged energy
ratio versus semblance for different frequency bands; (c) semblance versus cross-correlation (, earthquakes; ,
UWEs). Events with apparently anomalous waveforms, labelled on the plot, were correctly discriminated, the only
exception being the outlier QS9.
@seismicisolation
@seismicisolation
Figure 6. Spectra of small explosions (W=24 kg): (a) for three events (EX2, EX3, EX4) recorded at station PRNI; (b)
at different stations for a fixed event (EX2). The observed similarity of the spectra confirms the high stability of the
blasting parameters for different explosions.
It turned out, that for small explosions, the reverberation effect caused by the water depth H was
not revealed and, therefore, did not contribute to the low-frequency energy in the band 1-3 Hz
(unlike the largest EX1, see Fig. 4), resulting in the discriminant’s low performance. We guess that
compressional waves from small explosions are scattered more strongly (and lose energy) during
reverberatory multiple reflections from the uneven bottom, because of smaller wave lengths ( =
V0/fb = 1770/6.7 = 260 m), as compared with the explosion EX1 (=610 m).
The application of the additional spectral discriminant, i.e. semblance and cross-correlation
statistics, based on the coherency of spectra modulated by bubble oscillations, significantly
enhanced the resolving power. The results are shown on Fig. 5, demonstrating the almost complete
separation of the two classes of seismic events. (As expected, a clear equivalence between the two
statistics is observed on Fig.5c). Application of a different frequency range (1-8 Hz), instead of
(1-12 Hz), did not change the discriminant performance a great deal (Fig. 5b). In general, we
suggest that seismograms yielding a semblance value (in the frequency band 1-12 Hz) greater than
0.65 are associated with UWE.
Only one event would appear to be an outlier - the earthquake QS9, which has anomalously high
values of the energy ratio, semblance and cross-correlation. (It should be noted that all other events
- one explosion and five earthquakes - which showed the previously mentioned anomalous P/S
ratio at close distances, were successfully classified). This event was located in the southern Dead
Sea basin at the Jordanian seashore (see map on Fig. 1) at a very shallow depth evaluated as
H=0.0±1.4 km (mislocation estimates are very small, dX, dY1 km). Relocation computations did
not change the hypocenter location. Seismograms and spectra of the outlier, presented on Fig. 7 for
the same stations as the closest underwater explosion EX1 (RMN station was excluded because of
strong spikes), do not show similar obvious spectral modulation (see Fig. 4) and, thus, it could not
be identified as a UWE. This nocturnal event could hardly be ascribed to any land single-fired
blast, industrial or military. Application of the velogram analysis, based on kinematic features of
seismic waves (Pinsky et al., 1996), also suggests that QS9 is a natural event. It is possible that the
very shallow source of this earthquake led to the low-frequency spectral content of the
seismograms, because of partial energy propagation in thick subsurface sediments, or in the Dead
Sea water layer and hence caused anomalous spectral features.
@seismicisolation
@seismicisolation
Figure 7. (a) Seismograms and (b) spectra of the whole signal (about 50 s) of the outlier earthquake QS9 recorded at
different stations, showing the absence of a noticeable spectral modulation of the kind observed for the closest
underwater explosion, EX1 (Figs 4a and c).
CONCLUSIONS
The ISN seismograms of the single-shot underwater explosions in the Dead Sea and their
corresponding smoothed spectra, demonstrate a major feature which was utilized for
discrimination analysis: distinct spectral minima and maxima, coherent over the ISN stations. This
feature is mainly related to the bubbling effect, combined with the symmetric radiation
characteristic of an explosion (an actual point source).
@seismicisolation
@seismicisolation
These observations provide the basis for application of the newly developed multichannel
discriminant: spectral “semblance” and “cross-correlation” statistics which measure the coherency
of smoothed spectra at different ISN stations.
The seismic energy spectral ratio between the low (1-6 Hz) and high (6-11 Hz) frequency bands
was efficient in the case of Mediterranean UWE. However, in the case of the Dead Sea it showed
overlapping of the populations of the earthquakes and the Dead Sea underwater explosions. After
the application of spectral coherency statistics, utilizing azimuthal invariance of explosion spectra,
resolving power was enhanced and the two populations were separated. These results are
consistent with the successful application of the two spectral discriminants to more than 200
quarry blasts and earthquakes in different areas of Israel and Jordan (Shapira et al., 1997).
The ground-truth information for the selected controlled explosions provided the basis for the
interpretation of spectral features of the seismograms and of the discrimination results, and for the
future execution of effective calibration explosions. The results show that the spectral semblance
can be considered a successful discriminant, providing reliable identification of source induced
low-frequency spectral modulation of suspected underwater explosions, thus contributing to a
detailed study of seismicity parameters in the active Dead Sea Transform zone.
ACKNOWLEDGMENTS
The authors wish to thank Dr. A. Shapira for his support and for fruitful discussions and remarks
which significantly improved the manuscript. We thank the anonymous referees for their valuable
comments and suggestions. This study was partially supported by the Minerva Dead Sea Research
Center, Tel Aviv University and by U.S. DOE Contract F19628-95-K-0006, monitored by the U.S
Air Force Phillips Laboratory.
REFERENCES
Baumgardt, D.R., 1996. Characterization of regional-phase propagation seismic discriminants for
the Middle East. Proceedings of 18th Annual Seismic Research Symposium on Monitoring a
Comprehensive Test Ban Treaty, 474 483.
Baumgardt, D.R. & K.A. Zigler, 1988. Spectral evidence of source multiplicity in explosions:
application to regional discrimination of earthquakes and explosions, Bull. Seism. Soc. Am.,
78:1773-1795.
Blandford, R.R., 1995. Regional seismic event discrimination, in: E.S. Husebye and A.M. Dainty
(eds.) Monitoring a Comprehensive Test Ban Treaty, NATO ASI Series, Series E: Applied
Sciences, 303:689-719.
Gitterman, Y. & T. van Eck, 1993. Spectra of quarry blasts and microearthquakes recorded at local
distances in Israel, Bull. Seis. Soc. Am., 83:1799-1812.
Gitterman, Y. & A. Shapira, 1993. Spectral discrimination of underwater explosions, Isr. J. Earth
Sci.; 42:37-44.
Gitterman, Y. & A. Shapira, 1994. Spectral characteristics of seismic events off the coast of the
Levant, Geophys. J. Int., 116:485-497.
Gitterman, Y., Pinsky, V. & Shapira, A., 1996a. Discrimination of seismic sources using the Israel
Seismic Network, Scientific Report No.1, PL-TR-96-2207, Phillips Lab., MA, 98 pp.
Gitterman, Y., A. Shapira & V. Pinsky, 1996b. Spectral semblance statistics as effective regional
discriminants of seismic events in Israel, Proceedings of the 27th Nordic Seminar on Detection
Seismology, XXV ESC General Assembly, September 9-14, 1996, Reykjavik, Iceland, 587-592.
Hedlin, M., J.G. Minster & J.A. Orcutt, 1989. The time-frequency characteristics of quarry blasts
and calibration explosions recorded in Kazakhstan, USSR, Geophys. J. Int., 99:109-121.
Neev, D. & J.K. Hall, 1979. Geophysical investigation in the Dead Sea, Sediment. Geol., 23, 209-
238.
@seismicisolation
@seismicisolation
Neidell, N.S. & M.T. Taner, 1971. Semblance and other coherency measures for multichannel
data, Geophysics, 36:482-497.
Pinsky, V., A. Shapira & Y. Gitterman, 1996. Multi-channel velogram analysis for discriminating
between earthquakes and quarry blasts, Proceedings of the 27th Nordic Seminar on Detection
Seismology, XXV ESC General Assembly, Reykjavik, Iceland, 581 586.
Pinsky, V., Y. Gitterman & A. Shapira, 1997. Multivariate discrimination of regional and
teleseismic events, using the Israel Seismic Network recordings, Proceedings, 19th Annual
Seismic Research Symposium on Monitoring a Comprehensive Test Ban Treaty, 432 441.
Pomeroy, P.W., W.J. Best & T.V. McEvilly, 1982. Test ban treaty verification with regional data -
a review, Bull. Seis. Soc. Am., 72, S89-S129.
Shapira, A., 1988. Magnitude scales for regional earthquakes monitored in Israel, Isr. J. Earth Sci.,
37:17-22.
Shapira, A. & V. Avirav, 1990. ISDA - Israeli Seismic Data Acquisiton System, User’s Guide,
Version I, IPRG Rep. Z1/567/79(76).
Shapira, A., V. Pinsky, V. & Y. Gitterman., 1997. Discrimination of seismic sources using the
Israel Seismic Network, Final Report, PL-TR-97-2207, Phillips Lab., MA, 78 pp.
Suteau-Henson, A. & T.C. Bache, 1988. Spectral characteristics of regional phases recorded at
NORESS, Bull. Seis. Soc. Am., 78, 708-725.
Weinstein, M.S., 1968. Spectra of acoustic and seismic signals generated by underwater
explosions during Chase experiment, J. Geoph. Res., 73, No. 16, 5473 5476.
Willis, D.E., 1963. Seismic measurements of large underwater shots, Bull. Seism. Soc. Am., 53,
789.
@seismicisolation
@seismicisolation
CHAPTER FIFTEEN
NEAR-SOURCE AUDIOVISUAL, HYDROACOUSTIC, AND SEISMIC

OBSERVATIONS OF DEAD SEA UNDERWATER EXPLOSIONS 1
YEFIM GITTERMAN
Submerging of the 5-ton charge using the 12-ton crane placed on the iron raft; the steel buoy
(8,000 liters) was XVHGWR¿[WKHFKDUJHWRDGHSWKRIP (photo by Y. Gitterman).
Abstract: Results of a study of hydroacoustic, acoustic, and seismic H൵HFWV from a series of three
large-scale chemical H[SORVLRQV of 0.5, 2, and 5 tons, conducted in November, 1999 in the Dead
Sea, are presented. The shots were detonated at the water depth of m (485 m below the ocean
level). The main objective of the H[SHULPHQW was calibration of seismic stations of the International
Monitoring System in the Middle East, using accurate travel times and source phenomenology
features of underwater H[SORVLRQV The largest shot provided magnitude about 4 and was recorded
at distances up to 3500 km. Near-source seismic and hydroacoustic observations obtained were
utilized to estimate source parameters of the conducted H[SORVLRQV Based on the curve-¿W equation
of the time-pressure measurements, the direct shock wave energy was estimated as of the
total H[SORVLYH energy. The TNT equivalent to the 5000 kg charge of the H[SORVLYH used
(Chenamon) was determined as 4010 kg, corresponding to the manufacturer’s estimate of the
Chenamon energy as ~ of TNT.
Key words: underwater H[SORVLRQ hydroacoustics, seismic waves, shock wave
1
Originally published in Combustion, Explosion, and Shock Waves, Vol. 45, No. 2, 2009. Translated from Fizika
Goreniya i Vzryva, Vol. 45, No. 2, March–April, 2009. Original article submitted March 20, 2008.
@seismicisolation
@seismicisolation
Near-6RXUFH$XGLRYLVXDO+\GURDFRXVWLFDQG6HLVPLF2EVHUYDWLRQVRI'HDG6HD8QGHUZDWHU([SORVLRQV
INTRODUCTION
A unique seismic and hydro-acoustic H[SHULPHQW involving large-scale underwater chemical
H[SORVLRQV was conducted in the Dead Sea during November 8–11, 1999 by the Geophysical
Institute of Israel (GII). This H[SHULPHQW was part of the research project to provide calibration
sources lacking in the region, improved monitoring, and YHUL¿FDWLRQ of the Comprehensive Test
Ban Treaty (CTBT) in the Middle East [1].
The goals of the H[SHULPHQW were to calibrate regional travel times and propagation paths of
seismic waves across the Middle East and the Eastern Mediterranean region and, thus, improve
location accuracy of seismic events; to calibrate local, regional, and International Monitoring
System (IMS) stations, and to provide data for source characterization to improve IMS detection,
location, and discrimination capabilities.
Based on a series of small-scale underwater shots in 1993 [2], it was estimated that a 5000 kg
H[SORVLRQ would correspond to a seismic event of magnitude 4. It was H[SHFWHG that events of this
order of magnitude would be by seismic stations at far-regional distances, i.e., within 20° from the
Dear Sea.
Apart from the “deadness” of the Dead Sea, the choice of this site was also caused by unique
properties of super saline Dead Sea water: density 1.236 g/cm3 and acoustic velocity of sound
m/sec [3], much higher than in the ocean. It is well-known that such properties of water
enhance the seismic H൵HFW of an underwater H[SORVLRQ
Although there is no aquatic life in the Dead Sea, it is unique with its numerous nearby natural
phenomena such as mineral springs, desert oasis, wild life reservations, etc. The H[SORVLRQ point
was located a few kilometers from local spas, settlements, and an important engineering
installation, the Dead Sea Works dam. Therefore, all conditions of safety for the people in the sea,
property, and local environment were estimated and analyzed before the H[SHULPHQW However, in
order to verify the assumptions upon which the safety analysis is based, GII began the H[SHULPHQW
by detonating 0.5 and 2 ton H[SORVLRQV
The H[SHULPHQW was video-taped. A home camera was placed on a raft, at a distance of several
hundred meters from the shot site (Fig. 1). The video and audio tracks provided observations and
DSSUR[LPDWH time measurements of the physical phenomena of interest. Special hydroacoustic
measurements were provided to record water shock waves at close distances (§ m) from two
H[SORVLRQV 2 and 5 tons, by two piezoelectric sensors. Such records have never been made before
in hyper-saline water with unique properties as in the Dead Sea.
Table 1: Parameters of Calibration Underwater Explosions in the Dead Sea
([SHULPHQW Date Total charge, kg Origin time, GMT Location

1 08.11.99 500 13:00:00.33 31.5330° N 35.4406° E
2 10.11.99 2060 13:59:58.21 31.5338° N 35.4400° E
3 11.11.99 5000 15:00:00.80 31.5336° N 35.4413° E
@seismicisolation
@seismicisolation
328 Chapter Fifteen
Fig. 1. Location of the H[SHULPHQW site and close recording stations.
The near-source seismic and hydroacoustic observations obtained were utilized to estimate source
parameters of the conducted H[SORVLRQV All measurements and results were presented in detail in
[1].
EXPLOSION DESIGN AND EXPERIMENT LOCATION

Details of the charge design and sea operations are presented in [4]. The waterproof high H[SORVLYH
(HE) Chenamon (density of 1.3–1.4 g/cm3) based on ammonium nitrate was used as a primer
blasting agent. ([plosion parameters are listed in Table 1. The actual charge depth was h = –
m, variable due to underwater currents. The origin (detonation) time was measured with accuracy
better than 5 msec. The coordinates were measured by the precise GL൵HUHQWLDO GPS system;
however, due to strong variable underwater currents and winds, we estimated the actual location
accuracy as 20–30 m.
A detailed bathymetric map of the Dead Sea [5] shows that the sea depth in the H[SHULPHQW area
was H § 260–265 m (Fig. 2). The site was located about 9 km from the jetty, and the raft with the
H[SORVLYH was towed by a wooden boat.
During the ¿rst H[SORVLRQ (0.5 ton), the measurement systems for determining the GPS
coordinates, origin time, and hydroacoustic data were placed on the boat at r § 900 m from the
shot point. During the two larger shots, the systems were deployed on the raft, which is more
resistant to the shock wave impact; consequently, the distance to the shot point could be shortened
to §600 m.
@seismicisolation
@seismicisolation
Near-6RXUFH$XGLRYLVXDO+\GURDFRXVWLFDQG6HLVPLF2EVHUYDWLRQVRI'HDG6HD8QGHUZDWHU([SORVLRQV 329
Fig. 2. Bathymetric chart of the Dead Sea near the shot point: the water surface level is - 115 m (November, 1999); the
D[HV are local coordinates in km.
AUDIO-VISUAL OBSERVATIONS
The video and audio tracks provided observations and rough time measurements of interesting
physical phenomena, such as a “cavitation hat,” arrivals of shock waves, and bubble pulsations
near the raft. The video clips also provided a rough estimation of the shock wave propagation time
WRWKHVXUIDFHDQGWRWKHUDIWDQGFRQWULEXWHGWRYHUL¿FDWLRQRIWKHFKDUJHGHSWKDQGH[SORVLRQ-to-
UDIWGLVWDQFH$VSHFL¿FVXUIDFHSKHQRPHQRQDZKLWH³SOXPH´ZDVREVHUYHGDWWKH¿UVWH[SORVLon
epicenter (Fig. 3).
Fig. 3. Aerial view from the Ofeq airplane of a “plume” in the epicenter several seconds after the ¿rst H[SORVLRQ
(contributed by the Survey of Israel).
For the video-recording rate of 25 frames per second used, the time measurement accuracy can be
considered as the time between two successive frames which is 40 msec. Figure 4 shows snapshots
of the 2-ton shot, where the following phenomena can be observed:
@seismicisolation
@seismicisolation
330 Chapter Fifteen
(a) “cavitation hat,” created upon the direct shock wave arrival to the water surface, can be
revealed at t § 40 msec, according to the propagation time tc = m/ m/sec) = 39.5
msec. The frames at t = 80, 120, and 200 msec show the evolution of the cavitation process;
(b) the camera focus is distributed by the shock wave on the raft t § 440 msec, yielding an
H[SORVLRQ-to-raft distance r § m, which corresponds well to the estimation from
hydroacoustic recordings;
(c) the focus distortion is repeated after 500 msec (at the t = 920 msec snapshot) due to the
pulsation of the gas bubble containing detonation products. This time is close to the
predicted value and to the bubble pulse period tb measured from seismic spectra;
(d) a water uplift by gases rising at 9–14 sec;
(e) appearance of yellow gases H[SORVLYH detonation products), during 35–40 sec, in the area to
the right from the epicenter, evidences strong underwater currents in the Dead Sea, which
prevented more accurate than 30 m location of the charge placed on the depth h = m.
The video-record of the largest H[SORVLRQ (Fig. 5) included an impressive soundtrack (the audio
sample rate was 8192 kHz) with three VSHFL¿F successive audible signals produced by the gas
bubble pulsations. The spectral analysis of the soundtrack showed repeating interval of the signals
įW 0.8 sec (Fig. 6), corresponding well to the estimation of the bubble pulse period of sec from
seismic records (Table 2).
HYDROACOUSTIC MEASUREMENTS
Unique water shock wave observations were provided for the two largest H[SORVLRQV of 2 and 5
tons. Such records have never been made before in super-saline water of the Dead Sea.
Recording System, Estimation of Travel Times and Shock Wave Speed

The water pressures were measured by 138A01 piezoelectric gages (produced by PCB
Piezotronics). Two gages were located near the raft, at a distance of 600–800 m from the
H[SORVLRQV initiated at a depth of 25–30 m. The pressure–time history (with a duration of sec)
was recorded by a data acquisition system with a 250-kHz response for each channel used; for the
largest H[SORVLRQ only one gage record was obtained (Figs. and 8).
The accurate zero time for triggering the recording was provided by an electric fuse parallel to the
charge detonators, thus, facilitating accurate determination of the shock wave travel times between
the charge and the sensors necessary to estimate the high speed of the waves. Figure shows the
pressure–time history for the 2-ton H[SORVLRQ recorded by two gages spaced at dr = 25 m, which
provides an estimate of the shock wave speed u = m/sec.
The 23 m/sec H[FHVV of the acoustic velocity (c = .6 m/sec) is related to an error in
measuring the gage spacing and to H[WUHPHO\ high pressures in the shock wave at short distances.
The increment of speed on the leading edge of the shock wave is determined as
¨C = ¨P/ȡF (1)
For the measured peak pressure pPD[ § kPa (Fig. at a distance r = m, density ȡ = 1.236
kg/m3, and acoustic velocity c = .6 m/sec, Eq. (1) gives only ǻc = 3.2 m/sec. However, at a
distance r = 100 m, where the pressure is estimated as 5.2 MPa [see Eq. (4) below], the increase in
speed can reach 24 m/sec.
An additional hydroacoustic measurement system based on MP-24L3 Geospace hydrophones
(natural frequency 10 Hz) was used for determining (via GPS) the absolute arrival times of the
shock waves, thus, providing YHUL¿FDWLRQ of the origin time. For the 2-ton H[SORVLRQ, the shock
wave arrival time is ta = 0.445 sec (see Fig. and the detonation duration of the detonating cord
is about 10 msec; then, the origin time was estimated as 13:59:58.645 (0.445 0.010) = 13:59:58.21
@seismicisolation
@seismicisolation
GMT (see Fig. 9a and Table 1).
Fig. 4. Sample video snapshots of the 2-ton H[SORVLRQ taken on the raft from a distance of m H[SORVLRQ No. 2):
note the horizontal strip on the water surface at t = 80–200 msec (“cavitation hat”) and the camera focus disturbance at
t = 440 msec, caused by the shock wave. @seismicisolation
@seismicisolation
332 Chapter Fifteen
Fig. 5. Sample video snapshots of the 5-ton H[SORVLRQ taken on the raft from a distance of 620 m H[SORVLRQ No. 3):
note the horizontal strip on the water surface at t = 0.2 sec (“cavitation hat”); the spot on top of the water uplift is a
steel buoy (shown at t = 0) 2.2 m high and 2.4 m long, used to ¿[ the charge to a depth of m.
@seismicisolation
@seismicisolation
Fig. 6. Spectrum of the soundtrack recorded during the H[SORVLRQ
Fig. Pressure–time history of the shock wave in water for the H[SORVLRQ of the HE charge of 2060 kg (10.11.1999) at
a distance of m from the observation point (the distance between gages 1 and 2 is dr § 25 m).
Fig. 8. Pressure–time history of the shock wave in water for the H[SORVLRQ of the 5000 kg charge (11.11.1999) at a
distance of 620 m from the observation point. @seismicisolation
@seismicisolation
334 Chapter Fifteen
Table 2: Shot &RQ¿JXUDWLRQ Parameters and %XEEOH Periods tb Estimated from Shock Wave
Travel Times
([SHULPHQW m, kg h, m H, m r, m tb, sec Comments
1 500 - - 900 0.390 no precise piezoelectric measurements

2060 - r is for two gages separated by 25 m
2
5000 265 618 H estimate depends on phase
3 (212) interpretation
Fig. 9. Measurements of the origin time and shock wave arrival time for the H[SORVLRQ of the HE charge of 2060 kg:
arrival of the signal due to the detonating cord H[SORVLRQ and 2) arrival of the signal due to the main charge H[SORVLRQ
(a) measurements near the raft; (b) measurements at the HOF station (see Fig. 1) at a distance of 4.3 km from the
H[SORVLRQ
Joint recordings of a hydrophone and a seismometer located near the jetty, a few meters from the
shore, provided clear LGHQWL¿FDWLRQ of high-frequency and high-amplitude seismic waves (Fig. 9b),
generated by the impact of water shock waves arriving to the sea shelf.
@seismicisolation
@seismicisolation
Fig. 10. Spectral analysis of seismic pulsations for the H[SORVLRQ of the 5000 kg charge (11.11.1999) recorded by local
short-period ISN stations. A bubble pulse period is s (2.56/2 = 1.28 Hz) or s (1.31 Hz).
Table 3: 0HDVXUHG Arrival (Travel) Times of 'L൵HUHQW Shock Waves in Water
Sensor Arrival time from the initiation moment, sec

Sensor type depth,
No. P1, P2, surface P3 P4 (t4) 1st 2nd bubble
hs, m
direct reflection (t3) bubble pulse
wave (td) (ts) pulse
1 hydrophone 2 0.510 - - - 0.931 1.289
hydrophone 2 - - - - 1.095 1.602
2
30 0.44546 0.44854 0.5035 - - -
piezoelectric
25 0.45933 0.46180 - - -
hydrophone 2 0.349* - - - 1.922
3
piezoelectric 30 0.35915 0.36304 0.40248 0.432 - -
Note. The asterisk means the time from the detonation moment.
Types of Water Shock Waves

Three phases, P1, P2, and P3, are observed in the piezoelectric gage records of the 2-ton H[SORVLRQ
(see Fig. one more phase P4 is found for the 5-ton H[plosion (see Fig. 8). The direct wave P1
and surface-UHÀHFWHG P2 waves can be easily LGHQWL¿HG whereas interpretation of P3 and P4 is
uncertain in a trade-R൵ between bottom-UHÀHFWHG bottom-refracted, and surface-bottom-UHÀHFWHG
phases.
The cepstral simulation and inversion algorithm for analysis of regional seismic recordings of
underwater H[SORVLRQV developed [6], allow the band modulation of the surface UHÀHFWLRQ to be
@seismicisolation
@seismicisolation
336 Chapter Fifteen
LGHQWL¿HG For the narrow band (20 Hz), EIL and MRNI data of the Dead Sea shots give a negative
cepstral peak at 0.15 sec. The optimal search inversions program returned very low values of the
best matching UHÀHFWLRQ FRH൶FLHQW sometimes of the order of (0.3–0.4) [6].
Considering the high acoustic velocity in the Dead Sea (c = .6 m/sec), we can easily interpret
the 0.15 sec cepstral peak as really caused by the surface UHÀHFWLRQ The fundamental frequency in
this case is estimated as fr = c/4h = 6.32 Hz, then the appropriate period is tr = 0.158 sec, ¿tting
very well to the cepstral peak (much better than for the ocean acoustic velocity c0 §1500 m/sec).
The prominent azimuth and distance independent spectral modulation caused by the bubble pulse
H൵HFW observed in all seismograms recorded by Israel Seismic Network (ISN), especially for the
5-ton H[SORVLRQ (Fig. 10), provided an estimation the bubble frequency as fb 1.28 Hz (period tb =
0. sec; also Table 2). In the ISN-spectra, supposedly, the surface UHÀHFWLRQ PD[LPXP at fr is
merged with the 5th harmonic of the bubble fundamental frequency fb5 = fb u 5 = 1.28 u 5 = 6.4
Hz, causing an enhanced PD[LPXP at about 6.46 Hz (Fig. 10).
Water pressure measurements (see Figs. and 8) show that the amplitude of the surface UHÀHFWHG
wave is about 0.5 of the primary shock wave amplitude. This observation corresponds in some
way to the low reÀection coefficient estimated from the inversion algorithm [6].
9HUL¿FDWLRQ of Distances and Depths

Based on the H[SHULPHQW FRQ¿JXUDWLRQ and arrival times of GL൵HUHQW phases measured from
recordings of shock waves, we tried to verify the estimations of the source-sensor distance r,
charge depth h, and sea depth H in the shot point area. The travel times of different phases, as
measured from recordings of piezoelectric sensors and hydrophones, are presented in Table 3. The
required parameters were estimated from combinations of simple kinematic equations, under the
assumption that the shock wave velocity is constant along the entire propagation path:
ri/c = ti í ǻt, (2)
where the subscript i corresponds to GL൵HUHQW waves, ri is the propagation path depending on h and
H, ti is the arrival time (relative to the initiation moment), and ǻt is the shot delay caused by the
detonation cord of unknown length, which connects the initiation point on the water surface and
the charge on the depth; the detonation velocity of the cord is more than 6500 m/sec. The detailed
bathymetric map of the Dead Sea [5] shows that the sea depth in the H[SHULPHQW area is H = 260–
265 m (see Fig. 3). The phase analysis and interpretation results can be summarized as follows:
1) For both shots, the delay ǻt was estimated at about 0.01 sec, a reasonable value considering
the detonation cord length of – m and detonation velocity of – m/sec. This is
FRQ¿UPHG by a similar GL൵HUHQFH in times of the direct wave arrival to hydrophone and
piezoelectric sensor for shot No. 3 (Table 3);
2) The interpretation of the P3 phase (Fig. 11) as a bottom-transmitted wave and P4 as a
bottom-UHÀHFWHG wave provides a sea depth estimate H = 265 m, which corresponds well to
the bathymetric map estimation. Very high amplitudes and short duration of the P3 phase,
however, contradict the theoretical description of the bottom-transmitted wave as a slow-
rising and low-amplitude wave, and correspond much better to the bottom-UHÀHFWHG wave
(see, e.g., >@ Thus, this interpretation is based only on the kinematic features of the
observed phases.
3) The interpretation of the P3 phase as a bottom-UHÀHFWHG ZDYH DQG 3 DV D VXUIDFH-bottom
UHÀHFWLRQSURYLGHVDUDWKHUGL൵HUHQWHVWLPDWHRIWKHVHDGHSWKH PDQGWKHGL൵Hrence
cannot be related to measurement errors. This interpretation is based on both kinematic and
G\QDPLF IHDWXUHV RI WKH REVHUYHG ZDYHV EXW WKHUH LV VLJQL¿FDQW RSSRVLWLRQ WR WKH
bathymetric map. Another discrepancy is that the surface-ERWWRP UHÀHFWLRQ VKRXOG EH
QHJDWLYHGXHWRUHÀHFWLRQIURPWKHZDWHUVXUIDFHKRZHYHUDSRVLWLYH3Shase is observed
(see Fig. 8). @seismicisolation
@seismicisolation
Near-6RXUFH$XGLRYLVXDO+\GURDFRXVWLFDQG6HLVPLF2EVHUYDWLRQVRI'HDG6HD8QGHUZDWHU([SORVLRQV
4) A slightly increased depth estimation h .5m for the 5-ton shot seems to be reasonable,
possibly due to long-drown rope and a charge-center shift.
Fig. 11. &RPSOH[ form of the P3 phase for the H[plosion of the 5000 kg charge.
Fig. 12. Estimated peak water pressure versus the distance for underwater H[SORVLRQV of the 5000 (1) and 2060 kg (2)
TNT charges in ocean water and measurements of the Dead Sea shots (points).
The estimates of the shot FRQ¿JXUDWLRQ parameters and the bubble period (as the average for the
arrival times of the 1st and 2nd bubble pulses) are presented in Table 2.
Eneva et al. [8] analyzed the shock wave phases observed for the largest H[SORVLRQ using the
REFMS code for modeling surface UHÀHFWLRQV bottom UHÀHFWLRQV and refractions due to sound-
velocity gradients. An equivalent TNT charge weight of 4000 kg was considered. According to
this analysis, P4 is obviously the bottom UHÀHFWLRQ whereas P3 is the “precursory arrival”
(according to the REFMS terminology). The latter notion is used for any shock ray that appears
before a wave obeying Snell’s UHÀHFWLRQ law [called in seismology as “head” or refracted wave
propagating along the water–bottom interface). The bottom sound velocity vb that ¿ts best the
width of P3 is 3500 m/sec (if vb < 3200 m/sec, P3 becomes too wide; if vb>3600 m/sec, then P3
disappears).
@seismicisolation
@seismicisolation
338 Chapter Fifteen
Enhanced Peak Pressures

For the largest 5000 kg shot, the main blasting agent Chenamon was about (of the total
charge weight). According to the manufacturer’s VSHFL¿FDtions, its H[SORVLYH energy is about
of TNT. Nevertheless, the measured peak pressures VLJQL¿FDQWO\ H[ceeded the H[SHFWHG values
from an equal TNT charge in ocean water (see Table 4 and Fig. 12).
In a form similar to that provided by Cole [10], empirical relationships for estimation of shock
wave parameters were derived in > 9]:
p(t) = pPD[ H[S>í(t í ta)/Q], (3)
pPD[ = 53.1Rsí1.13, (4)
ta = r/cw, (5)
Qs = 9.2 · 10í5W 1/3Rs0.18, (6)
Rs = r/W 1/3.
Here pPD[ [MPa] is the peak pressure, ta is the arrival time, Q is the time constant, W [kg] is the
TNT equivalent weight, detonated in ocean/lake water, r [m] is the distance from the shot point, Rs
[m/kg1/3] is the scaled distance, and cw [m/sec] is the sound (acoustic) velocity in water.
The peak pressure curves in Fig. 12 are calculated from Eq. (4). This observation of the enhanced
pressures for the Dead Sea shots (far H[FHHGLQJ the estimated values) can be attributed to a
GL൵HUHQFH in the acoustic impedance Z = rc of the supersaline Dead Sea water being higher
than in the ocean (see below) and, thus, yielding a much stronger shock waves impact.
Estimation of the Shock Wave Energy for the 5-Ton Shot

The shock wave energy ÀX[ density is calculated as [10]
ିଵ
‫ܧ‬஽ௌ =ܼ஽ௌ ‫݌ ׬‬ଶ (‫ݐ݀)ݐ‬ (8)
where ZDS is the Dead Sea water acoustic impedance, ZDS = ȡDSucDS = 1236 kg/m3 m/sec =
2.u106 kg/(m2sec) and p(t) is the curve-¿W equation DSSUR[LPDWLQJ the H[SHULPHQWDOO\
measured pressure– time SUR¿OH (Fig. 13) with an estimated time
· constant QDS = 0.0018458 sec:
PDS (t) = 1.23 H[S [-(t-0.3592)/QDS] (MPa) (9)

The integral in Eq. (8) is calculated from the interval time ‫ݐ‬௔ = 0.3592 sec to t = ‫ݐ‬௔ + 4 [10]
ta 6. Q
1
³ >1.23 u 10 @
6
E DS u H[S (t 0.3592) / Q ) 2 dt 1219 J / m 2
Zw
ta
The full shock wave energy (at r = 618 m) is calculated as Es = EDSu4ʌU2 = 0.585•10 erg.
@seismicisolation
@seismicisolation
7DEOH(VWLPDWHGDQGPHDVXUHGSHDNSUHVVXUHVLQWKHVKRFNZDYHV
No. TNT r, m Sensor depth, Peak pressure, kPa
equivalent m
estimated measured Ratio
weight, kg
from Eq.(4) measured/estimated
2 1660 30 1.41
25 458 654 1.43
3 4010 618 30 1230 1.45
Fig. 13. Time evolution of pressure in primary and surface-UHÀHFWHG shock waves (1): curve 2 is the apSUR[LPDWLRQ of
the pressure SUR¿OH in the primary wave by Eq. (9); the charge weight is 5000 kg.
The VSHFL¿F energy of the Chenamon H[SORVLYH used in the H[SHULPHQW according to the
manufacturer’s VSHFL¿FDWLRQV is 900 cal/g = 3.8u1013 erg/kg, and the full H[SORVLYH energy is Ee =
·
1.9u10 erg. Then, the share· energy estimate for the primary shock wave is
Es/Ee = 30..
This value seems to be reasonable, although it slightly H[FHHGV the free-water estimate (23–
presented in shallow reservoir research [8] and about one-fourth estimate at distances more
than 15 charge radii [11].
Estimation of the TNT Equivalent

It is assumed that the primary shock wave energy ÀX[ for the largest Dead Sea shot (W = 5 ton of
Chenamon) EDS equals the shock wave energy ÀX[ EOC for the “equivalent” TNT charge WT
H[SORGHG in the ocean water:
EDS = EOC. (10)
The value EOC is calculated from Eqs. (4), (6), and (8) as
EOC = Z-1OC ³ [53.1 Rsí1.13 H[S -t/QOC)]2 dt (11)

where the impedance for the ocean water is ZOC = ȡOCucOC = 1030 kg/m3u1536 m/sec = 1.582u106 ·
kg/(m2sec) and QOC depends on WT as in Eq. (6). The value EDS is calculated from Eqs. and
(8):
ିଵ
‫ܧ‬஽ௌ =ܼ஽ௌ (െt/ܳ஽ௌ )]ଶ ݀‫ݐ‬
‫[׬‬1.23exp@seismicisolation (12)
@seismicisolation
340 Chapter Fifteen
Finally, the TNT equivalent to the 5000 kg charge of Chenamon H[SORGHG in the Dead Sea is
determined from Eqs. (10)–(12) as WT = 4010 kg; this value corresponds to the manufacturer’s
estimation of the H[SORVLYH Chenamon energy as ~ TNT.
CONCLUSIONS
The Dead Sea calibration H[SHULPHQW provided a variety of unique data. The video and audio
tracks presented interesting phenomena such as the “cavitation hat”, arrivals of the shock waves,
bubble pulsations at the raft, and water uplift at the detonation point. The video clips also provided
a rough estimation of the shock wave propagation time to the surface and to the raft and
contributed to YHUL¿FDWLRQ of the charge depth and H[SORVLRQ-to-raft distance.
The measured peak pressures VLJQL¿FDQWO\ H[FHHGHG the H[SHFWHG values from an equal TNT
charge in ocean water due to enhanced acoustic impedance of the supersaline Dead Sea water. An
analysis of GL൵HUHQW phases and arrival times on the records facilitated the LGHQWL¿FDWLRQ of surface
(“ghost”) and bottom reÀections and contributed to the YHUL¿FDWLRQ of the H[SHULPHQW
FRQ¿JXUDWLRQ The shock wave energy was calculated utilizing the wave energy ÀX[ density
equation, recorded pressure–time functions, and VSHFL¿F properties for the Dead Sea water. These
observations could also provide preliminary estimations of the shock energy share relative to the
total H[SORVLYH energy and the TNT equivalent of the shots.
The high magnitudes under relatively low charge weights FRQ¿UPHG the high seismic H൶FLHQF\ of
underwater H[SORVLRQV in the Dead Sea, VLJQL¿FDQWO\ H[FHHGLQJ the upper limit curve for all
known chemical and nuclear H[SORVLRQV in hard rock. The observed magnitudes and amplitudes of
seismic waves verify that the whole charge was properly detonated in all cases.
The calibration H[SHULPHQW was supported by the US Department of Defense under contract
DSWA01- -C-0151. Mr. S. Kobi, of ELITA Security Ltd. and Eng. E. Hausirer of ([SORVLYHV
Manufacturing Industries Ltd. provided the design, assembling of HE charges, and
conducting of the H[SORVLRQV Mr. M. Gonen, of Gonen Marine Works Ltd. provided the boat and
the raft for all sea operations. Dr. L. Sadwin of Sadwin Engineering Consultancy, Israel, conducted
measurements of shock waves by piezoelectric sensors donated by the Israel Ministry of National
Infrastructures. Special thanks to Dr. M. Eneva of 0D[ZHOO Technologies and Dr. D. Baumgardt of
ENSCO, Inc., who supplied valuable results of the complicated analysis and interpretation of the
data obtained. The management and VWD൵ of the Geophysical Institute of Israel provided technical
support throughout the H[SHULPHQW the whole H[SHULPHQW could not be realized without support
and H൵RUWV of Dr. A. Shapira.
REFERENCES
1. Y. Gitterman, V. Pinsky, and A. Shapira, “Improvements in Monitoring the CTBT in the
Middle East by the Israel Seismic Network,” Final Report, DTRA- TR-01-35, September
to March 2001, sponsored/monitored by the U. S. Defense Threat Reduction Agency
(2003).
2. Y. Gitterman, Z. Ben-Avraham, and A. Ginzburg, “Spectral analysis of underwater
H[SORVLRQV in the Dead Sea,” Geophys. J. Intern., 134, 460– (1998).
3. D. A. Anati, “The hydrography of a hypersaline lake,” in: T. M. Niemi, Z. Ben-Avraham,
and J. R. Gat (eds.), The Dead Sea: the Lake and its Setting, 2[IRUG Univ. Press, New York–
2[IRUG pp. 89–103. 2[IRUG monographs on geology and geophysics No. 36.)
4. Y. Gitterman and A. Shapira, “The Dead Sea gives life to a unique seismic calibration
H[SHULPHQW´ EOS, Trans. Amer. Geophys. Union, 82, No. 6 (2001).
5. J. K. Hall, “The GSI digital terrain model (DTM) completed,” in: R. Bogoch and Y. Eshet
(eds.), GSI Current Research, V., Jerusalem (1993), pp. –50.
6. D. Baumgardt and A. Freeman, “Characterization of underwater H[SORVLRQV by cepstral
analysis, modeling and inversion,” J.@seismicisolation
Acoust. Soc. Amer., 109, No. 5, 2496 (2001).
@seismicisolation
Underwater Blast Monitoring, Eng. Tech. Lett. (FR), No. 1110-8-11, U. S. Army, Corps of
Engineers (US-ACE), Washington (1991).
8. M. Eneva, J. L. Stevens, J. Murphy, and B. D. Khristoforov, “Effect of charge depth in
Russian hydroacoustic data from nuclear and HE H[SORVLRQV´ in: Planning for Verification
of and Compliance with the Comprehensive Nuclear-Test-Ban Treaty, Proc. of the 22nd
Annu. DoD/DoE Seismic Research Symp. (September 12–15, 2000), New Orleans (2000).
9. C. E. Joachim and C. R. Welch, “Underwater shocks from blasting,” in: Proc. of the 23rd
Annu. Conf. on Explosive and Blasting Technique, Int. Soc. of ([SORVLYH Engineers, Las
Vegas–Cleveland pp. 526–536.
10. R. H. Cole, Underwater Explosions, Princeton Univ. Press (1948).
11. J. W. Pritchett, “An evaluation of various theoretical models for underwater H[SORVLRQ
bubble pulsation,” Report No. IRA-TR-2- Information Research Associates Inc.,
Berkeley
@seismicisolation
@seismicisolation
CHAPTER SIXTEEN
UNDERWATER EXPLOSION (UWE) ANALYSIS OF THE ROKS CHEONAN

INCIDENT 1
SO GU KIM AND YEFIM GITTERMAN
The Yeonhwari littoral coast of Baengnyeong-do showing some submerged rock which may be
related to a running aground prior to the underwater explosion of the ROKS Cheonan Sinking on
March 26, 2010.
ABSTRACT
The underwater explosion (UWE) resulting in the sinking of the South Korean warship, ROKS
Cheonan occurred on March 26 2010. Raw data was analyzed from several 3-component
stations—Baengyeong-do Korea Meteorological Administration (KMA) station (BAR), Ganghwa
KMA station (GAHB), Incheon Incorporated Research Institutions for Seismology (IRIS) station
(INCN), the short-period station—Deokjeok-do KMA station (DEI), as well as from the seismo-
acoustic array Baengyeong-do Korea Institute of Geoscience and Mineral Resources (KIGAM)
station (BRDAR). The ROKS Cheonan incident has been investigated by both the Multinational
Civilian-Military Joint Investigation Group (Ministry of National Defense, 2010) and Hong (Bull
Seism Soc Am 101:1554–1562, 2011). Their respective methods and conclusions are also
presented in this study. One of the main differences between their findings and ours is that we
deducted that the fundamental bubble frequency was 1.01 Hz* with a subsequent oscillation of
1.72 Hz. Also, in contrast to findings by the MCMJIG and Hong, our analysis shows the first
reverberation frequency to be 8.5 Hz and the subsequent one to be &25 Hz. The TNT-equivalent
1
Originally published in Pure Appl. Geophys. (2013). 170 (4). * accurate bubble pulse 1.012 Hz
(0.988 s) is calculated in the updated study.
@seismicisolation
@seismicisolation
Underwater Explosion (UWE) Analysis of the ROKS Cheonan Incident 343
charge weight (seismic yield) and seismic magnitude were estimated from an observed bubble
frequency of Hz and the analytical model of a bubble pulse. From the data analyzed, we deducted
that the seismic yield would be about 136 kg of TNT, which is equivalent to the individual yield of
a large number of land control mines (LCM) which were abandoned in the vicinity of the ROKS
Cheonan incident by the Republic of Korea (ROK) Navy in the 1970s (Ministry of National
Defense 2010). Also, whereas both the MCMJIG and HONG estimated the local magnitude at 1.5,
our findings came to the conclusion of a local magnitude of approximately 2.04 based on the
bubble frequency of 1.01 Hz measured on the vertical component of BAR station data considering
the empirical relationship between charge weight (TNT yield) and underwater explosion
magnitude. Strong high-frequency signals collected at the 3-component BAR station
approximately 30 s after P-wave arrivals and infrasound records at BRDAR clearly indicate
powerful acoustic phases and N-waves caused by a relatively shallow UWE. T-phases are also
observed on seismograms and spectra at 15–17 Hz on the DEI, GAHB, and INCN stations.
Key Words: Bubble pulse, bubble collapse, toroidal bubble, whipping, T-phase, N-wave
INTRODUCTION
The underwater explosion (UWE) incident vis-a`- vis the ROKS Cheonan occurred on March 26
2010 with an origin time of 12:21:57 (UTC) and location at 37.93°N and 124.60°E. The
Multinational Civilian- Military Joint Investigation Group (MCMJIG) concluded that the UWE
had a TNT-equivalent yield of 250 kg of explosives and occurred at a depth of 6–9 m with a
seismic magnitude of 1.5 (MCMJIG 2010). This TNT-equivalent yield was in line with the yield
of the North Korean CHT-02D torpedo. Analysis was taken from the 3-component records of the
BAR, DEI, and GAHB stations of KMA, INCN station of IRIS and the BRDAR station of
KIGAM. The seismometers utilized for data collection are the STS- 2 (BB) for the BAR and
INCN stations, the CMG-3T (borehole BB) for the GAHB station, the CMG40T-1 (SP) for the
DEI station and short period (SP) and infrasound arrays for the BRDAR station (see Table 1; Fig.
1). The sampling rate was 100 sps, except for the INCN station with 40 sps.
A shallow UWE was empirically confirmed by strong high-frequency features of acoustic phase
signals at seismograms
from the BAR station, observed approximately 30 s after P-wave arrivals and infrasound records
also showed N-waves (SOR- RELS et al. 2002) clearly indicating a sonic boom. The main objective
of this paper is to identify the source of the UWE vis-a`-vis the ROKS Cheonan incident via
spectral analysis of the seismic waves related to the bubbling and reverberation effects of the
UWE, including T-phase.
Table 1: Coordinates and distances
Station, site Agency Symbol Latitude N° Longitude E° Distance
(km)
UWE site – * 37.92917 124.60056 0
BRDAR KIGAM D1 37.9677 124.6303 5.39
BAR KMA D2 37.9771 124.7142 11.43
DEI D3 37.256 126.105 152.4
GAHB D4 37.708 126.447 164.4
(borehole)
INCN IRIS D5 37.4833 126.6333 186.0
@seismicisolation
@seismicisolation
344 Chapter Sixteen
The ROKS Cheonan incident site (star) and seismic stations (triangles) are shown with distances
from the epicenter (origin time 12:21:57 UTC time)
Figure 1: Map of the incident site (indicated by star) and station locations (indicated by triangles)
This study analyzed identical data sets as well as additional infrasound records and empirical data
and also utilized different methodologies of the bubble dynamics of physics. Our conclusion is
also vastly different as for the cause of the UWE compared to the one set forth in the MCMJIG
report.
METHODS OF ANALYSIS
b
The magnitude-charge size relations play important roles in forensic seismology as well as in
Comprehensive-Test-Ban Treaty (CTBT) monitoring (KOPER et al. 2001; BOWERS and SELBY 2009).
JACOB and NIELSON (1977) analyzed intensively all body wave magnitudes (M*) collected from
underwater experiments conducted off the UK coast and North Sea and converted them into local
magnitudes (ML). The empirical relationship between local magnitude (ML) and TNT equivalent
charge weight W (kg) using a series of 38 underwater explosions was derived as follows: ML =
0.753 log (W) + 0.436 with a coefficient of determination R2 = 0.856 which should be close to 1 in
order to be a good degree of fit (see Eq. 1 in Fig. 2).
As shown in Fig. 2, the empirical relationships between charge size and magnitude are given in
Eqs. 1 and 2 for underwater explosions, and 3 and 4 for land explosions. A simple relationship
between local magnitude ML and TNT equivalent charge weight W (Eq. 2) was developed for deep
(70 m) underwater explosions as follows: ML = 0.285 + log10(W). The equation was verified during
the series of experimental underwater explosions in the Dead Sea for calibration of seismic
stations of the International Monitoring System (GITTERMAN and SHAPIRA 2001). Nevertheless, this
equation may be in an overestimation due to the high salinity (33.7 %) and high density of the
Dead Sea as well as the charge size and detonation depth. Therefore, another empirical formula
(Eq. 1) for the average ocean salinity (3.5 %) at shallow water is used for this study. Equations 3
and 4 represent land explosions for an empirical upper limit magnitude in hard rock (KHALTURIN et
al. 1998) and a coda magnitude (BROCHER 2003) in California and Nevada, respectively.
@seismicisolation
@seismicisolation
Figure 2: Reported magnitude for the underwater and land detonations versus the charge size.
Empirical relations between local magnitude and underwater explosions from this study (Eq. 1)
(solid line), by GITTERMAN and SHAPIRA (2001) for Dead Sea underwater explosions (Eq. 2) (dashed
line), by KHALTURIN et al. (1998) the empirical upper limit magnitude of chemical land explosions
in the hard rock, fitting nuclear explosions (Eq. 3) (dotted and dashed lines) and coda magnitude
of land explosions in California and Nevada, including most of coda duration magnitudes
(BROCHER 2003) (Eq. 4) (dotted line)
In general the magnitudes of a UWE near free surface or the bottom will be reduced due to the
deficit of energy efficiency by ill-coupling. The large variations of magnitude are mainly due to
inaccuracies in evaluating charge weight and variation in the detonation depth. Scattered
distributions at around 200 kg TNT yield are due to different depths of the UWE’s (Fig. 2). An
optimum depth designed for seismic efficiency shows the largest magnitudes for a given charge
(JACOB 1975). Underwater explosions are more seismically efficient than explosions on land for a
given charge weight.
The oscillation time of the gas bubble is calculated from the modified Rayleigh-Willis formula: Tb
= kW1/3/(P0 + qgd)5/6, where k is the constant, W is the product of volume and pressure of air gun in
work energy for marine seismology (BARGER and HAMBLEN 1980; ZIOLKOWSKI et al. 2003), P0 is the
atmospheric pressure at the sea surface (1.01325 9 105 Pa), q is the density of sea water (1,025.52
kg/m3) at 3 °C and 3.2 % salinity), g is the gravity constant (9.81 m/s2), and d is the detonation
depth (m). A more commonly used simplified approximation model using the SI units is derived as
follows: Tb = KT W1/3/(10.1 + d)5/6 where KT slightly varies (2.10–2.11), depending on the
experimental and approximation methods utilized. The analytical dependence of bubble pulse
frequency on explosive yield and detonation depth was investigated by many researchers
(RAYLEIGH 1917; COLE 1948; WILLIS 1963; CHAPMAN 1985; GITTERMAN and SHAPIRA 2001;
GIT- TERMAN 2002; REYMOND et al. 2003; GONG et al. 2010; ZHANG and ZONG 2011), Utilizing
this simplified approximation model, the fundamental frequency fb and the spectral harmonic
series are represented as follows: fb = 1/Tb; fnb = nfb, n = 1, 2…
In contrast to the completely harmonic series caused by the bubbling effect, reverberation effects
of acoustic waves in the water layer (between the bottom and the surface) produce an odd
harmonic series in seismic records with the fundamental frequency fH (cutoff frequency),
@seismicisolation
@seismicisolation
346 Chapter Sixteen
according to the equation: fH = V0/4H; fnH = (2n-1)fH for n = 1, 2, … where V0 is the water sound
velocity (m/s), H is the water depth (m) (OFFICER 1958; WEINSTEIN 1968). This reverberation effects
equation is valid in the case of shallow water media when the maximal radius Rb of the explosive
gas bubble is smaller than the distance to any boundary in the vicinity of a UWE. The radius Rb is
estimated from the equation: Rb = KR[W/ (d + 10)]1/3 (WILLIS 1963; KLASEBOER et al. 2005), where
KR is a constant, from 3.36 (GONG et al. 2010) to 3.38 (KLASEBOER et al. 2005; ZHANG and ZONG,
2011) depending on the experimental methods. For this paper KR = 3.38 was used. The UWE near
the vessel causes local damage via shock wave, while global damage can be due to the water
bubble jet from bubble collapse and the whipping motion from the bubble pulse. For plane wave
and TNT explosive equivalents, peak pressure, decay time and pressure function are derived as
follows (SWISDAK 1978; SAVAGE and HELMBERGER 2001; MCMJIG 2010): Pmax = 52.4
(W1/3/D)1.13MPa; θ =0.084W1/3(W1/3/D)-0.23 and P=52.4(W1/3/D)1.13x exp[(t-t0) / θ ] where W is the
charge weight (kg), D is the stand-off distance (m), h is the decay time in seconds (the time it takes
for the pressure to decay to 1/e, t0 is the initial time at which the shock wave arrives at the distance
D, t is the time elapsed since the shock wave arrived at the distance D.
SPECTRAL ANALYSIS OF SEISMIC WAVEFORMS AND YIELD ESTIMATION
Figure 3: a Smoothed spectra for equal time window 22 s of seismic signals. Bubble pulse
frequencies appear at 1.01 and 1.72 Hz, including pre-signal noise for the equal time window
without these spectral features. b Smoothed normalized spectra of seismic signal of time window
15 s, with clear reverberation spectral maxima at 8.5 and &25 Hz (green arrows). c High
resolution spectra using 25 s time window in the frequency range (0.6–2.1 Hz). The maximum
spectrums can be determined at 1.01 (2,750 U) and 1.72 Hz (1,950 U) on the vertical component
(black line). P, S, LR, LQ and B indicate P, S, Rayleigh, Love waves and bubble pulses,
respectively. d Waveforms of the vertical component at the BAR station. Initial P-wave observed
@seismicisolation
@seismicisolation
at 21:21:59.24 with compression (?), a bubble pulse arrival B at 21:22:0.24 with rarefaction (-),
and S-wave and Rayleigh wave arrivals at 21:22:1.4 (S) and at 21:22:3.2 (LR), respectively. More
detailed information about bubble pulse, bubble jet and toroidal bubble is given in Appendix.
We used a software package developed by the Geophysical Institute of Israel for spectral analysis.
The source spectrum for bubble pulse, water reverberation column, transfer function of the
propagation medium and the recording system was theoretically calculated by WIELANDT (1975).
KOPER et al. (2001) and BAUMGARDT (2003) estimated yield and detonation depth of the Kursk
explosion using bubble pulse and water column reverberation from the time-independent
scalloping of the spectra of seismic waves. In the case of Cheonan incident the spectra are
calculated for broad windows of 22–90 s including all seismic signal waveforms: P, S and a
delayed wave train (for the remote DEI and GAHB stations), composed of T-phase arrivals. Using
this large window provides accumulated and enhanced spectral features (especially the bubbling
source effect), imprinted in all observed seismic phases. The spectra are smoothed using
convolution with a moving triangle window. In order to separate signal and noise spectral
maximums, the spectra of the same length pre-signal noise window are also calculated. In some
cases the spectra are presented in two versions: frequency axis in both log and linear scale,
providing better observation of spectral features at low and high frequencies, respectively. The
bubble pulse frequencies are estimated from the spectral peaks at 1.01 Hz (2,750 U) and at 1.72 Hz
(1,950 U) on the vertical component of the BAR station since the initial motion of an explosion is
the only compressional source (as shown Fig. 3a, c). The bubble pulse frequencies which consist
of P and Rayleigh waves (P, LR) are well indicated at 1.01 and 1.72 Hz using the short frequency
domain (0.6–2.1 Hz) of the vertical component, while the higher amplitude of the NS component
at 1.78 Hz (Fig. 3c) indicates salient Love wave arrivals (LQ) due to a toroidal bubble and the jet
impact at the source which are formed via rebound and bubble collapse (PRITCHETT 1966;
KLASEBOER et al. 2005; ZHANG et al. 2008).
The rarefaction (dilatation) impulse arrival of a bubble pulse (B) with about 1.0 s time lag after P-
wave onset in Fig. 3c, d is associated with an impact penetrating the bubble when water fills the
gas bubble at high speed. The compressional polarity of the P-wave first motion evidently
indicates an underwater explosion (Fig. 3d). The peak amplitude spectra of the vertical and the NS
components from the high resolution spectra are most likely related to the jet impact and a toroidal
bubble which caused large Rayleigh waves as well as unexpected S waves and strong Love waves.
Despite an underwater explosion, the phenomenon that we can observe involving large Love and
Rayleigh waves (Fig. 3c, d) is attributed to the effects of a toroidal bubble and the upward
migration of jet impact (PRITCHETT 1966), including the interaction of gas bubble and the ship’s
hull. Spectra of pre-signal noise do not show these modulation features (Fig. 3a).
The most possible source of the ROKS Cheonan incident was a non-contact underwater explosion
and not any collision in the light of the spectral analysis and apparent damage feature. We infer
that the most probable cause of this UWE would be either a land control mine (LCM), or the
DPRK’s torpedo of 250 kg TNT asserted by the MCMJIG (2010) because other possibilities are
very unlikely in the range of 140–250 kg TNT charge size. Also too shallow of detonation depth
cannot be accepted due to the absence of bubble pulses. However, the latter assertion clearly does
not fall within the parameters set forth by the observed data. For example, at about 167 kg TNT
yield at a detonation depth 9 m or 190 kg TNT yield at a detonation depth at 10 m, the maximum
radiuses of the gas bubble would be about 7 m. Those UWE’s would not breach the hull structure
with the jet impact and shock waves which are consistent with the observed data. The MCMJIG
asserts that the UWE had a TNT yield of 250 kg at a depth of 6–9 m. This is simply not feasible
according to the data collected (Table 2; Fig. 4). The closest and most likely UWE source is a land
control mine (LCM) previously mentioned, of which many were abandoned in the late 1970s in
that vicinity with an individual TNT yield of 136 kg which, according to the analyzed data, had a
1.01 Hz (0.990 s) of a bubble pulse frequency and a detonation depth of approximately 8 m (Table
@seismicisolation
@seismicisolation
348 Chapter Sixteen
2; Figs. 3a, c and 4). Table 2 shows bubble pulse periods for Cole’s method (COLE 1948) and a
boundary element method (BEM) (ZHANG, et al. 2008; ZHANG 2012, personal comm) with other
analytical parameters in case of 136 and 250 kg. The Cole’s method is based on s spherical bubble,
ignoring the jet impact on the fluid field, while BEM is determined at the moment of the jet
penetrating the bubble surface based on the nonlinearity of various shape boundaries. As indicated
in Table 2, the bubble pulse periods near free surface by BEM (asterisked) are closely consistent
with those of Cole’s method, larger than those of the Cole’s method. The smallest relative error
margin of bubble pulse periods of BEM with respect to Cole’s method is found at 136 kg TNT
yield with a detonation depth of 8 m (Table 2). Figure 4 shows how a UWE with a 136 kg TNT
yield at a depth of 8 m falls within the acceptable parameter values, but a UWE with a 250 kg TNT
yield at a depth of 6–9 m (MCMJIG, 2010) falls outside these parameters.
Considering the characteristics of the acoustic phases, infrasound records, seismic data, and the
impact location of the UWE on the ROKS Cheonan at approximately 3 m portside from the center
of the hull bottom (MCMJJIG 2010), the ocean surface had to be barely breached, which narrows
down the source possibilities in terms of TNT yield and detonation depth. Analysis of observed
data as well as the damage feature to the ROKS Cheonan clearly indicate damage inflicted via the
effects of bubble pulse and bubble collapse from the UWE. The 136 kg TNT at a depth of 8 m
generates the maximum bubble radius of 6.63 m and contain both a bubble pulse and a bubble
collapse which would breach the ship’s draft of 2.9 m.
The charge size of 136 kg TNT at the detonation depth of 8 m with a bubble pulse period of 0.967
s (0.9764 s for BEM) has the boundary characteristics of bubble pulse and bubble collapse to just
breach the ocean surface and is consistent with the parameters set forth by the observed bubble
pulse period of 0.990 s (Table 2). The bubble pulse period of 250 kg TNT at the depth of 9 m was
estimated at 1.132 s by Cole’s method and 1.150 s by BEM (Fig. 4; Table 2). According to the
simulation of bubble shapes for the fluid- structure interaction with 136 kg TNT at detonation of 8
and 3 m portside from the center of the ship (ZHANG 2012, personal comm.), the bubble pulse
period is found to be much longer (1.058 s) than that of the Cole’s method by 9.41 % because of
the influence of the hull. The case for the 250 kg TNT yield detonated at around depth 6–9 m
would have an oscillation time much greater than 1.1 s which is contrary to the bubble dynamics
of physics.
Table 2: Bubble pulse periods (T and T*) and maximum radiuses of gas bubbles (Rb) of shallow
underwater explosions vis-a`-vis charge size (TNT) and detonation depth (d)
TNT Depth (m) Rb T (s) T* (s) Depth/Rb Error (%)
(kg) (m)
136 7.0 6.76 1.014 1.0279 1.04 1.37
136 8.0 6.63 0.967 0.9764 1.06 0.97
250 8.0 8.12 1.184 1.2082 0.99 2.04
250 9.0 7.98 1.132 1.15 1.13 1.59
T* is calculated by a boundary element method (ZHANG et al. 2008; ZHANG 2012, personal comm.). Error = 100 (T* - T)/T. The
minimal error of 0.97 % is observed in the case of a UWE with a 136 kg TNT yield detonated at a
depth of 8 m indicating by close correlation between calculations utilizing Cole’s method and cal-
culations utilizing BEM
Bold lettered values indicate inconsistencies between observation data analysis and calculations
utilizing Cole’s method and BEM
@seismicisolation
@seismicisolation
Figure 4: The relationship between charge weight versus pulse bubble oscillation time (bubble
pulse period) with different UWE depth. Distribution of estimated bubble pulse periods utilizing
the spectral analysis from observation data (closed red star), utilizing Cole’s method (open red
star) and utilizing a boundary element method (BEM) (open red circle) for 136 kg TNT at a depth
of 8 m. The distribution of bubble pulse periods by the Cole’s method (closed black star) and by
BEM (open black circle) for 250 kg TNT at a detonation depth of 9 m
In Fig. 4, a closed red star and an open red star indicate an observed bubble pulse period and the
analytical bubble period by the Cole’s method from 136 kg TNT yield at a depth of 8 m which fall
within the acceptable value parameters but a UWE with a 250 kg TNT yield (closed black star) at
a depth of 9 m (MCMJIG 2010) falls outside these parameters.
The water column reverberation frequencies are observed at 8.5 Hz and at around 25 Hz on the
spectra of the BAR station data (Fig. 3b). The water column reverberation is a multiple reflection
of sound waves from surface to bottom, blending into a more or less steady oscillation in shallow
water. Clear water column reverberation spectral maximums at 7.5, 8.0, and 8.0 Hz from remote
stations INCN, DEI and GAHB, respectively, (Figs. 5, 6 and 7). At the remote station INCN, the
fundamental frequency maximum is dominated by low-frequency noise, but the second harmonic
bubble frequency is observed at ≈ 1.8 Hz. The reverberation frequency at ≈7.5 Hz (Fig. 5a, b)
indicates a relatively deep oceanic path of ≈ 54 m. The spectral peak at 8.5 Hz from the BAR
station corresponds to the fundamental frequency of the water column reverberation in accordance
with the sea depth at the UWE site estimated at 44 m based on sound velocity at about 1,500 m/s
(MEDWIN 1975). This falls in the range of the reported sea depth H = 40–50 m at the UWE site
(MCMJIG 2010).
@seismicisolation
@seismicisolation
350 Chapter Sixteen
Figure 5: a Seismograms at remote station INCN and smoothed spectra for equal time windows
(40 s) of seismic signal where reverberation spectral maxima at &7.5 Hz and pre-signal noise
without this spectral feature (left). b Seismograms (lower) after the narrow band-pass filter of 6–8
Hz
@seismicisolation
@seismicisolation
Figure 6: a Original and b filtered (2–20 Hz) seismograms at DEI station (time windows for
spectra calculations indicated by colors), c spectra of noise and d all signal waveforms in log and e
linear frequency scale. The water column reverberation is observed at 8.0 Hz
The spectral peak at fH ≈ 25 Hz from the BAR station matched the second harmonic odd
reverberation frequency series fnH = 3fH, for n = 2. However, the maximums of even reverberation
frequency series at ≈17 Hz (2fH) and ≈ 34 Hz (4fH) (Fig. 3b red arrows) as presented by HONG
(2011) are not congruent with our findings for the reverberation effect in the sea water. The
spectral peaks with small spectral nulls at 17-18 Hz and its second harmonic frequency at 34-
35 Hz on E-W and Z components can be attributed to the nonlinear reflections from the hull
structure mixing with direct P-wave arrivals along with overlapping readings from later
arriving T-phases. One can estimate source depth at around 8 using spectral null at 18.5 Hz and
less than 300 m/s of P-wave velocity within the bubble.
At the DEI station, reverberation water columns are clearly observed at 8.0 and ≈ 24 Hz although
it is adulterated by noise Fig. 6d, e). There are known spectral anomalies at ≈ 28 and ≈ 34 Hz. This
station is situated on an island with its propagation path mainly in the sea. At the GAHB station,
the reverberation effect is indicated by spectral maximums at 8.0 and ≈ 24 Hz (3fH) and there
appears to be unexplained frequency series at ≈ 22 and ≈ 34 Hz (Fig. 7d, e). It is a more complex
spectral picture than from the DEI station attributable to a more complicated propagation path with
some islands in the way and a variable sea depth (see the map on Fig. 1).
T-PHASE ARRIVALS
T-phase or T-waves are hydroacoustic waves from large undersea earthquakes or underwater
explosions that are characterized by travel within the ocean as sound waves, which are then
converted into P, S, and surface waves as they travel on the continent. Some of the spectral
maximums of the recorded signals are related to arrivals of T-phases that traveled from the UWE
site through the Yellow Sea to a seismic station having a very short adjacent continental segment
resulting in cylindrical propagating energy loss (1/r). Variability in T-phase velocity is reported in
the findings due to different terrestrial dispositions in the propagation path resulting in dispersion,
later modes, and reverberation with the earliest arrivals at 1.65 km/s and a large central spike at the
velocity of sound wave in the SOFAR channel at≈1.49 km/s (NORTHROP 1962). These features
correspond to delayed (relative to the P, S phases) wave trains clearly observed at the remote
@seismicisolation
@seismicisolation
352 Chapter Sixteen
stations DEI and GAHB after the optimal filtration of 2–20 Hz (Figs. 6a, 7a) have been applied
(Figs. 6a, b, and 7a, b). The group velocity for these arrivals varies from ≈ 1.47 to 1.77 km/s (1.64
km/s for the central maximum) for station DEI to a little higher values of ≈ 1.55–1.81 km/s (1.68
km/s) for GAHB most likely due to a longer continental segment in its propagation path (see the
map in Fig. 1).
Figure 7: a Original and b filtered (2–20 Hz) seismograms at GAHB station, c spectra of noise and
d all signal waveforms in log and e linear frequency scale. The water column reverberation is
observed at 8.0 Hz
At the DEI station, the bubbling maximum at ≈ 1 Hz overlaps with pre-signal noise but at a low
amplitude (Fig. 6c–e); for the GAHB station this is not the case and it is dominated by large low-
frequency micro-seismic noises with its maximum at ≈ 0.3 Hz (Fig. 7c, d). Spectra for DEI,
GAHB, and INCN stations were analyzed with narrow windows containing only T-phase arrivals
with observed dominant (compared to spectra for all waveforms) spectral maximums at 15–17 Hz
(Fig. 8a–c), implying that the seismic source is a UWE.
ACOUSTIC PHASE AT BAR STATION AND INFRASOUND SIGNALS AT BRDAR
Band-pass filtering at 30–40 Hz revealed rather strong high-frequency signals at 3C seismograms
at the BAR station, about 32 s after P-wave arrivals, with the largest amplitude at the NS
component (Fig. 9a, b). The spectral maximum for this signal is found at ≈ 33 Hz (Fig. 9b, c). It is
a powerful acoustic phase with group velocity 341 m/s (corresponding to the sound velocity in the
air). This phase was induced in the seismometer by the shock wave produced by detonation
explosives venting to the atmosphere due to a relative shallow depth detonation. The loss of the
explosive energy to the air-shock wave resulted in the reduced seismic magnitude of the explosion
(GITTERMAN and KIM 2010).
The time-separation between two peaks of the UWE and N-waves (SORRELS et al. 2002) was about
1.1 s on infrasound records of the BRDAR array (Fig. 10) with acoustic phase velocity of 337 m/s.
The strong sonic booms of this UWE were clearly observed as N-waves 15.4 s after P-wave
arrivals on the seismograms from Fig. 10, where two big pressure pulses were detected with a 1.1 s
interval from stations. The time-separation from the infrasound signals is slightly delayed as
@seismicisolation
@seismicisolation
compared with the bubble pulse period (0.990 s) due to the sonic boom arrival after the bubble
pulse. The two impulsive signals indicate the source explosion and N-waves venting to the
atmosphere as a sonic boom which occurs when the speed of the mine shell fragment is faster than
that of sound in the underwater and at the atmosphere (supersonic velocity of TNT is 6,900 m/s).
The peak amplitude in the high resolution spectrum and the strong Love waves at the NS
component of the BAR station may be a result of the jet impact and a toroidal bubble motion
blending with the sonic boom by the advent of the shock wave (Figs. 3c, 10). The bottom line of
the acoustic phase and infrasound signal appearance clearly indicates that this UWE is very
shallow and/or near the surface, but not in contact with or in very close proximity to the ship’s hull
avoiding the cavitation (WIELANDT 1975). The peak pressure and decay time to reach 37 % of the
initial pressure were estimated at 37 MPa and 0.46 s for this UWE. At 37 MPa, the shock waves
are equivalent to a force of 31 m/s2 accelerating the ship which would produce the nonlinear
deformation onto the hull’s structures as observed (SWISDAK 1978; ZHANG and ZONG 2011).
Figure 8: Spectra of delayed wave trains, composed of T-phase arrivals, clearly showing spectral
maximum at 15–17 Hz for stations a DEI, b GAHB and c INCN
DISCUSSION
For a UWE occurring almost directly beneath and/or very close to a vessel (Table 2) two main
contributors to subsequent vessel damage are the bubble collapse in conjunction with the jet
impact to the ship’s hull and the bubble pulse causing whipping damage to the ship’s girders. At
shallow depth, the gas bubble is likely to vent into the atmosphere, throwing water into the air and
considerably reducing the seismic energy transmitted downwards by expansion and contraction of
the bubble. The jet impact and the toroidal bubble are formed by a non-spherical shape caused by
@seismicisolation
@seismicisolation
354 Chapter Sixteen
the rebound of a bubble and a bubble collapse. The bubble jet causes the rapid acceleration
upwards of the spherical bubble with a speed of 130–170 m/s (PLESSET and CHAPMAN 1971)
along with the influence of inertia and buoyancy. Moreover, a high-speed bubble jet penetrating a
bubble can easily cut through the ship even with a relatively small TNT yield (KHOO 2010,
personal comm.) because of the very rapid rise in its plume.
Given the circumstances at the time surrounding the incident: relatively shallow sea depth of 40–
50 m, no detection of any targets (torpedo or mine), joint Republic of Korea (ROK)-US naval
exercises ongoing about 120 km south of the UWE site with naval combatant vessels at its
periphery, etc., the chance of this being a cut and dry torpedo attack is minimal. Also, the ROKS
Cheonan clearly showed signs of navigational ‘‘difficulties’’—e.g., fishing nets caught around the
right screw, the right screw being inwardly bent and breached at its edges, fresh scratches on its
rudder and hull inconsistent with the assertion that this incident was solely due to a UWE, etc. It is
impossible to pinpoint with absolute certainty the UWE source given the available data and
evidence. However, given the possible ‘‘culprits’’, the abandoned land control mines (LCM)
which were cut off nearly 30 years ago and literally lost track of, are at the top of the candidate
list. The detonation may have been triggered by the mine cables being dragged out with the mines
during the ROKS Cheonan navigation problems along with magnetic signature, acoustic
properties, water pressure changes, or even, Impressed Current Cathodic Protection (ICCP) acting
sensory triggers. The approximately 136 kg TNT yield attributable to this specific type of mine
(ROK LCM) (MCMJIG 2010) along with a detonation depth of 8 m, maximum bubble radius of
6.6 m, oscillation time of 0.967 s, and a local magnitude 2.04 with a bubble frequency of 1.01 Hz
fits the seismic parameters set forth in the analyzed data. Others have scientifically refuted the
MCMJIG’s assertion that the UWE source was a DPRK 250 kg TNT yield CHT-02D torpedo
using different methods (LEE and SUH 2010) and our seismo-acoustic analysis likewise clearly
disproves this assertion.
Figure 9: . a Acoustic phase at station BAR after the band-pass filter of 30–40 Hz. The high
acoustic phase is observed on the NS component 31 s after P-wave arrivals. b Raw data of 3-
component of the BAR (left) and the maximum spectra observed at around 33 Hz of the record
window (7 s)
@seismicisolation
@seismicisolation
Figure 10: Seismograms and infrasound signals record at BRDAR by 3C sismometer and
infrasound sensor array (H. LEE, courtesy of KIGAM). The second peak between the time-
separation of about 1.1 s is called the N-wave (SORRELS et al. 2002) which is due to the sonic boom
of the underwater explosion
CONCLUSIONS
The observation of oscillation time of 0.990 s corresponds to 0.967 s of the bubble pulse period via
the Cole’s method with an estimated yield of 136 kg TNT explosives detonated at around 8 m of
depth with a seismic magnitude 2.04. This is in accord with the TNT yield of previously existing
land control mines (LCM) in the area (MCMJIG, 2010) but a 250 kg TNT yield of CHT-02D
DPRK’s torpedo proposed by the MCMJIG (2010) fails to fit the seismic as well as the analytical
determination by Cole’s method (1948) because its oscillation time is far beyond 1.1 s.
A local magnitude of 1.5 submitted in the MCMJIG report (2010) and a body wave magnitude of
1.46 proposed by HONG (2011) would mean a UWE with a TNT yield of about 20–30 kg which
would be too small to cause the kind of destruction sustained by the ROKS Cheonan. Also, the
application of a terrestrial-based magnitude determination formula to a UWE is inappropriate
because for a given charge a UWE produces a larger magnitude than a terrestrial explosion. The
magnitude of a UWE is always greater than that of a land chemical explosion (BROCHER 2003;
BOOTH 2009). When an explosion is detonated at a depth which is sufficient to prevent immediate
venting of the gas bubble at the surface as in the case of the ROKS Cheonan UWE, the energy of
the explosion is well-coupled to the surrounding transmitting medium and through surface
reflection most of it directed downward so that there is efficient generation of seismic waves. On
the other hand, a greater portion of the energy from a land explosion is converted to vertical and
lateral displacement of the surrounding rock. Therefore, for a given charge size, a UWE produces
a lager magnitude than land explosions. In general, the seismic efficiency of a UWE is about ten
times that of explosions in hard rock (BOOTH 2009). KHALTURIN et al. (1998) stated that in their
chemical explosion study, a UWE produces a magnitude excess rather than a deficit because of
super-efficient coupling in water. So 1.5 magnitude determined by land explosion formulas is not
appropriate for this UWE. The most feasible scenario would be that the seismic yield of the ROKS
Cheonan UWE is a 136 kg TNT yield, corresponding to a seismic magnitude of 2.0 at the
detonation depth of about 8 m, taking into account the observed bubble pulse period of 0.990 s,
@seismicisolation
@seismicisolation
356 Chapter Sixteen
which is equivalent to a LCM (local land control mine)*, but not a 250 kg TNT yield
corresponding to DPRK’s torpedo CHT-02D.
ACKNOWLEDGMENTS
The authors thank Boo Cheong Khoo (National University of Singapore), and Aman Zhang
(Harbin Engineering University) for the insightful and informative discussions regarding this
work. The bubble pulse periods by a boundary element method were calculated by Aman Zhang,
Harbin Engineering University. Special appreciation is expressed to Aman Zhang for his
contribution bubble shape simulation concerning a boundary element method and the bubble shape
simulation.
REFERENCES
BARGER, J. E. and HAMBLEN, W. R. (1980), The air gun impulsive underwater transducer, J. Acoust.
Soc. Am, 68, 1038-1045.
BAUMGARDT, D. R, (2003), Seismic characterization of underwater explosions and undersea
earthquakes using spectral/cepstral modeling and inversion, in Proc. 21st Seismic Res. Symp.
Technologies for Monitoring the Comprehensive Nuclear-Test- Ban Treaty, Sep. 2003,
Available: http://www.rdss.info.
BOOTH, D. C. (2009), The relation between seismic local magnitude ML and charge weight for UK
explosions, British Geological Sur- vey, Earth Hazards Programme, Open Report OR/09/062,
21 pp.
BOWERS, D. and SELBY, N. D. (2009), Forensic seismology and the Comprehensive-nuclear-Test-
Ban Treaty, Annual Review of Earth and Planetary Sciences, 37, 209-236.
BROCHER, T. M. (2003), Detonation charge size versus coda magnitude relations in California and
Nevada, Bull. Seism. Soc. Am., 93, 2089-2105.
CHAPMAN, N. R. (1985), Waveform parameters of explosive charges, J. Acoust. Soc. Am. 78, 672-
681.
COLE, R. H. (1948), Underwater Explosions, Princeton Univ. Press, Princeton, New Jersey.
GITTERMAN, Y. (2002), Implications of the Dead Sea experiment results for analysis of seismic
recordings of the submarine ‘‘Kursk’’ explosions, Seism. Res. Lett., 73, 14-24.
GITTERMAN, Y. and KIM, S. G. (2010), Interpretation of seismic and acoustic observations from
underwater explosion accidents, The 8th General Assembly of the Asian Seismological
Commission, Hanoi, Vietnam, 8-10 November, 2010.
GITTERMAN, Y. and SHAPIRA, A. (2001), Dead Sea seismic calibration experiment contributes to
CTBT monitoring, Seism. Res. Lett., 72, 159-170.
GONG, S. W., OHL, S. W., and KLASEBOER, E. (2010), Scaling law for bubbles induced by different
external sources: theoretical and experimental study, Physical Review E 81, 056317, 1-11.
HONG, T. (2011), Seismic Investigation of the 26 March 2010 Sinking of the South Korean Naval
Vessel Cheonanham, Bull. Seism. Soc. Am., 101, 1554-1562.
JACOB, A. W. (1975), Dispersed shots at optimum depth- an efficient seismic source for
lithospheric studies, J. Geophysics 41, 63-71.
JACOB, A. W. B. and NEILSON, G. (1977), Magnitude Determination on LOWNET, GSU, Report 86,
Institute of Geological Sciences, Edinburgh, Scotland, U.K., 40 pp.
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.
KHOO, B. C. (2010), Personal communication with B. C. Khoo, Fluid Mechanics Group,
Department of Mechanical Engineering, National University of Singapore, Singapore.
KLASEBOER, E., KHOO, B. C., and HUNG, K. C. (2005), Dynamics of an oscillating bubble near a
floating structure, J. of Fluids and Structure, 21, 395-412.
KOPER, K. D., WALLACE, T. C., TAYLOR, S. R., and HARTSE, H. E. (2001), Forensic seismology and
the sinking of the Kursk, EOS 92, no. 4.
LEE, S. and SUH, J. J. (2010), Rush Judgment: Inconsistencies in South Korea’s Cheonan Report,
@seismicisolation
@seismicisolation
The Asia-Pacific Journal, 28-1- 10, July 12, 2010.

MCMJIG (Multinational Civilian-Military Joint Investigation Group) (2010), Joint Investigation
Report: On the Attack against ROKS Cheonan, Ministry of National Defense, Seoul, 313 pp.
MEDWIN, H. (1975), Speed of sound in water for realistic parameters, J. of Acoust. Soc. Am., 58,
1318-1319.
NORTHROP, J. (1962), Evidence of dispersion in earthquake T- phases, J. Geophys. Res., 67, 2823-
2830.
OFFICER, C. B. (1958), Introduction to the theory of sound transmission, with application to the
ocean, McGraw-Hill Book Company, New York, 284 pp.
PLESSET, M. S. and CHAPMAN, R. B. (1971), Collapse of an initially spherical cavity in the
neighbourhood of solid boundary, J. of Fluid Mechanics, 47, 283-290.
PRITCHETT, J. W. (1966), Explosion product redistribution mechanisms for scaled migrating
underwater explosion bubbles, USNRDL-TR-1044, 23 May 1966.
RAYLEIGH, J. W. S. (1917), On the pressure developed in a liquid during the collapse of a spherical
cavity, Philos. Mag., 34, 94-98.
REYMOND, D., Hyvernaud, O., TALANDIER, J., and OKAL, E. A.(2003), T-wave detection of
two underwater explosions off Hawaii on 13 April 2003, Bull. Seism. Soc. Am., 93, 804-816.
SAVAGE, B and HELMBERGER, V. (2001), Kursk Explosion, Bull. Seism. Soc. Am., 91, 753-759.
SORRELS, G., BONNER, J., and HERRIN, E. T. (2002), Seismic pre-cursors to space shuttle shock
fronts, Pure Appl. Geophys., 159, 1153-1181.
SWISDAK, M. M. (1978), Explosion effects and properties: part II— explosion effects in water,
NSWC/WOL TR 76-116, 22 Feb 1978.
WEINSTEIN, M. S. (1968), Spectra of acoustic and seismic signals generated by underwater
explosions during Chase experiment, J. Geophys. Res., 73, 5473–5476.
WIELANDT, E. (1975), Generation of seismic waves by underwater explosions, Geophys. J. R. astr.
Soc., 40, 421-439.
WILLIS, D. E. (1963), Seismic measurements of large underwater shots, Bull. Seism. Soc. Am., 53,
789-809.
ZHANG, A. (2012), Personal communication with A. Zhang, College of Shipbuilding Engineering,
Harbin Engineering University, Harbin, China.
ZHANG, A. M., YAO, X. I., and LI, J. (2008), The interaction of an underwater explosion bubble and
an elastic-plastic structure, Applied Ocean Research, 30, 159-171.
ZHANG, N. and ZONG, Z. (2011), The effect of rigid-body motions on the whipping response of a
ship hull subjected to an underwater bubble, J. of Fluids and Structures, 27, 1326-1336.
ZIOLKOWSKI, A., HANSSEN, P., GATLIFF, R., AKUBOWICZ, H., DOBSON, A., HAMPSON,
G., LI, X., and LIU, E., (2003), Use of low frequencies for sub-basalt imaging Geophys.
Prospect. 51, 169-182.explosions, Geophys. J. R. astr. Soc., 40, 421-439.
WILLIS, D. E. (1963), Seismic measurements of large underwater shots, Bull. Seism. Soc. Am., 53,
789-809.
ZHANG, A. (2012), Personal communication with A. Zhang, College of Shipbuilding Engineering,
Harbin Engineering University, Harbin, China.
ZHANG, N. and ZONG, Z. (2011), The effect of rigid-body motions on the whipping response of a
ship hull subjected to an underwater bubble, J. of Fluids and Structures, 27, 1326-1336.
ZIOLKOWSKI, A., HANSSEN, P., GATLIFF, R., AKUBOWICZ, H., DOBSON, A., HAMPSON,
G., LI, X., and LIU, E., (2003), Use of low frequencies for sub-basalt imaging Geophys.
Prospect. 51, 169-182.
____________________
• The accurate bubble pulse was found to be 1.012 Hz (0,988 s) from the updated study.
@seismicisolation
@seismicisolation
CHAPTER SEVENTEEN
ESTIMATING DEPTH AND EXPLOSIVE CHARGE WEIGHT FOR AN

EXTREMELY SHALLOW UNDERWATER EXPLOSION OF THE ROKS
CHEONAN SINKING IN THE YELLOW SEA 1
SO GU KIM, YEFIM GITTERMAN AND ORLANDO CAMARGO RODRIGUEZ
Air wave (>30 Hz) record at Baengnyeongdo station (BAR) after 31 sec from the ROKS Cheonan Sinking underwater
explosion at the Yellow Sea in the Korean Peninsula
ABSTRACT
We estimated the detonation depth and net explosive weight for a very shallow underwater
explosion using cutoff frequencies and spectral analysis. With detonation depth and a bubble pulse
the net explosive weight for a shallow underwater explosion could simply be determined. The ray
trace modeling confirms the detonation depth as a source of the hydroacoustic wave propagation in
a shallow channel. We found cutoff frequencies of the reflection off the ocean bottom to be 8.5 Hz,
25 Hz, and 43 Hz while the cutoff frequency of the reflection off the free surface to be 45 Hz
including 1.01 Hz for the bubble pulse, and also found the cutoff frequency of surface reflection to
well fit the ray-trace modeling. We also attempted to corroborate our findings using a 3D bubble
shape modeling and boundary element method. Our findings led us to the net explosive weight of
the underwater explosion for the ROKS Cheonan sinking to be approximately 136 kg TNT at a
depth of about 8 m within an ocean depth of around 44 m.
Keywords: Cutoff frequency, total reflection, cycle distance, bubble pulse, boundary element
method
1
Originally published in Methods in Oceanography (2014), 11, 29-38.
@seismicisolation
@seismicisolation
Estimating Depth and Explosive Charge Weight for an Extremely Shallow Underwater Explosion of the 359
ROKS Cheonan Sinking in the Yellow Sea
INTRODUCTION
The underwater explosion (UWE) vis-à-vis the ROKS Cheonan took place off the coast of
Baengnyeong Island in the Yellow Sea of the Korean peninsula on March 26, 2010 (see Fig. 1).
Considerable efforts have been devoted to estimate the net explosive weight of this UWE using
spectral analysis and analytical approach including simulation of boundary element method (BEM)
(Kim and Gitterman, 2013; Kim, 2013). These attempts have typically used ad hoc models of the
relationship between bubble pulse period and net explosive weight or have been based on 3D
shape simulation by boundary element method (BEM). We attempted to estimate and interpret the
source depth and a net explosive weight using underwater acoustics (hydroacoustics) as well as
hydrodynamics. This paper, especially, presents the relationship between the cutoff frequencies
and the detonation depth resulting in obtaining the net explosive weight, including application of
ray-trace modeling for confirmation of estimation. We utilized cutoff frequencies to estimate the
detonation and ocean depths including the bubble pulse period for the extrapolated net explosive
weight. We also verified whether our estimated source depth fitted the observed one using ray-
trace model in the shallow channel. The compelling reason of this study is to estimate the net
explosive weight (NEW) for a very shallow underwater explosion (<50m) using only cutoff
frequencies and Rayleigh-Willis equation (Kim and Gitterman, 2013). The NEW estimation is
possible using a cutoff frequency and the bubble pulse period from spectral analysis relating to the
Rayleigh-Willis equation since it is a function of detonation depth and NEW.
Fig. 1. Map of the UWE site and Baengyeong-do. The star, triangle and circles indicate the UWE
site, seismic station and ports, respectively.
CUTOFF FREQUENCIES AND NORMAL MODES FOR A SHALLOW UNDERWATER
EXPLOSION
Shallow water (<200m) sound fields are defined in terms of a normal mode propagation which
oscillates with resonant frequencies in series of harmonics. The normal mode propagation without
attenuation are those for which water depth is greater than one-quarter wavelength (H>λ/4,
H=water depth, λ=wavelength). The frequency corresponding to H = λ/4 is termed the cutoff
frequency of the waveguide (as the critical frequency to build a waveguide). Waves with
frequencies lower than the cutoff frequency are propagated in the channel only with attenuation
and are not effectively trapped in the duct of the layer. There is no mode propagation below the
@seismicisolation
@seismicisolation
360 Chapter Seventeen
cutoff frequency.
Note Snell’s law from fundamental physics: C2sin θ1 = C1sin θ2 (C1 and C2 are velocity of upper
layer and lower layer respectively and θ1 and θ2 are angle of incidence and transmission
respectively). Taking one particular incident angle θC called a critical angle which is the
transmission limit angle (90º), when the incident angle is greater than θC, all the incident waves are
reflected in the water layer and no energy is transmitted in the sediment layer (Nicolas et al., 2013).
The velocity with which the wave front progresses is dependent on the incident angle θ1 and will
always be greater than the medium velocity C1 of each downward or upward ray. To focus on the
critical angle θC= θ1, the cutoff frequency (critical frequency) for each mode is simplified. If the
incident angle is greater than the critical angle, the normal mode propagation starts in a waveguide.
The wave propagates by multiple reflections at an incident angle between the grazing angle and
critical angle for total reflection under the condition of constructive interference in the reflection
off the ocean bottom (Nicolas et al., 2003) in case of C1<C2.
According to Urick (1983), the cutoff frequency (critical frequency) for ocean depth (H) is
presented as follow:
fH = C1(2n-1)/4H [1-(C1/C2)2]-1/2 n=1,2,… (1)

where C1 is the downward velocity of the water layer of thickness H with mode number n and C2 is
the velocity at the rigid bottom (sediment or basement). This equation is reduced to fH = C1/4H
when C1<<C2 which is the cutoff frequency of the first mode for the water column reverberation;
i.e. the ocean depth for the cutoff frequency is H = C1/4fH.
For the upward velocity (C1) with a water layer of thickness d and C0 the acoustic velocity in the
atmosphere, in the case of incident waves in water reflecting off an air-water interface, C1 > C0
yields θt < θi where θt is a maximum transmitted angle and the transmitted wave is refracted closer
to the normal (13.1°) in case of grazing incidence angle of 90˚. According to Snell’s law, i. e. sin
θi/sin θt =C1/C0 and the grazing incidence (θi=90˚) and there are small transmitted waves with θtmax
= arcsin (C0/C1), resulting in reflected acoustic waves. Therefore we take into account the cutoff
frequency of the surface reflection at the boundary without transmission, especially as sound
velocity increases with depth in the case of isothermal velocity profile during the cold season
(Lurton, 2002). The cutoff frequency for the detonation depth (d) is presented as follows:
fd = C1(2n-1)/4d [1-(C0/C1)2]-1/2 n=1,2,… (2)

The equation is also reduced to fd = C1/4d for the first mode when C0 << C1. In case of a reflection
coefficient = -1, there is a total reflection. The transmission is negligible and the incident waves
are always reflected at the free surface and become channel waves neglecting evanescent modes.
Therefore, the detonation depth for the first mode is d=C1/ 4fd. The more detailed explanation for
a very shallow UWE is given in Figs. 2 and 3.
RAY-TRACE MODELING FOR HYDROACOUSTIC WAVES
When temperature does not vary with depth (isothermal profile), the sound velocity linearly
increases with depth due to the hydrostatic pressure. This special case is going to be applied to the
ray-tracing modeling and this underwater explosion. All the ray paths are therefore refracted
upward and propagate by successive bounces off the surface (see Figures 2 and 3). This
configuration is called a surface channel (Lurton, 2002) in the hydroacoustic waves. Such
phenomena are often found in the polar sea or enclosed seas during the cold season like the
maritime environment as this study. Ray-trace modeling (in the context of underwater acoustics) is
a computational technique based on the calculation of propagating rays which correspond to the
normal direction to the wavefront (Rodriguez et al., 2012). The calculation of the rays can be used
to estimate the acoustic pressure induced in the underwater waveguide by the pressure of an
@seismicisolation
@seismicisolation
acoustic source (Porter and Bucker, 1987). We can observe the predominant surface channel of
hydroacoustic waves at around 8 m by ray-tracing. From spectral analyses, we also observe
corresponding small peaks at around 45 Hz since nondispersive hydroacoustic waves with a large
quality factor are converted into seismic waves with high attenuation due to a small quality factor
of the island. The amplitudes steeply decay passing through the FIR filtration near Nyquist
frequency (50Hz) during the formation of seismic waveform data (see Figs. 4 and 5). The use of
seismic data with high sampling rate (e.g. 200 Hz) or the hydrophone data with high frequency
signals would produce larger amplitude of spectral anomalies at 45 Hz in this case. Shapira (1981)
found the quality factor, Q of the sea water is found to be more than 10 times larger than that of
the land from underwater explosions off the coast of Israel. These results can also be explained by
the fact that the island formation of a small quality factor Q contributed to small amplitudes of the
spectra of this UWE due to high attenuation of seismic waves (Fig. 5).
For the isothermal configuration (Lurton, 2002) with the linear law C(z) = 1458.53 + 0.9425z as a
function of depth z in this study, the rays of hydroacoustic waves would be arcs of circles reflected
on the surface showing a periodic structure with a horizontal cycle distance DC (Lurton, 2002) for
a ray bouncing off the surface with a grazing angle u (Fig. 2). We can derive a horizontal cycle
distance (DC), propagation time over one cycle (TC), and bottoming depth (ZC) using Snell’s law in
differential form and a radius of curvature to model the ray path contingent on the isothermal
configuration of the cold sea pertinent to this study.
𝒃𝒃
DC = 2∫𝒂𝒂 𝑹𝑹(𝒖𝒖)𝒄𝒄𝒄𝒄𝒄𝒄𝒄𝒄𝒄𝒄𝒄𝒄 (3a)
𝒃𝒃
ZC = ∫𝒂𝒂 𝑹𝑹(𝒖𝒖)𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔 (3b)
𝒃𝒃
TC = 2∫𝒂𝒂 𝑹𝑹(𝒖𝒖)𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔 (3c)
where R(u)=curvature of radius and R(u) = -Cv/g when Cv is a vertex sound velocity at a grazing
angle u = 0 in this case, g= velocity gradient, and a and b are grazing angles at depth Z=0 and
Z=vertex.
Fig. 2. Ray geometry for an isothermal sound velocity profile that increases with depth due to the
hydrostatic pressure.
@seismicisolation
@seismicisolation
Provided the attenuation is not too large and dispersion is little, we estimated a horizontal cycle
distance (DC) at about 325 m with group velocity of 1460 m/s at a bottoming depth (ZC) of about 8
m which was estimated by Kim and Gitterman (2013). The horizontal cycle distance is clearly
visible (see Fig. 3) and is strongly associated with the group velocity carrying energy along a ray
or beam shift (Tindle et al., 1980; Weston, 1994). There are about 10 wavelengths in each
horizontal cycle distance implying that there must be the total reflection (reflection coefficient of
−1) on the surface (Urick, 1983). Fig. 3 disambiguates cutoff frequency of this hydroacoustic wave
propagation at around 45 Hz taking into account about 33m wavelength in this ray-trace modeling.
It is notable that surface reflected arrivals usually suffer very little loss in the horizontal cycle
distance with group velocity. As a result, we could prove that the source (detonation) depth at 8 m
which was 2.5 km away from the coast contributed to the shallow reflection from the free surface
using the ray-trace modeling.
Fig. 3. Ray-tracing based on hydroacoustic wave speed in the early spring season with a UWE
source detonated at a depth of 8 m near the Baengnyeong-do in the Yellow Sea. Rays refract away
from the sea surface and are going up to the surface due to the increase of sound velocity with
depth in the maritime environment of the cold season.
RESULTS OF SPECTRAL ANALYSIS AND DISCUSSION
We vindicated the non-contact UWE by not only the observations of bubble and reverberation
effects, but also boundary element method (BEM) in the previous studies (Kim, 2013). We
estimated a net explosive weight using the bubble pulse period of 0.990 s (1.01 Hz) and Rayleigh-
Willis equation without prior information of a source depth. Here we estimated the source depth
first and then we simply determined a NEW with known bubble pulse period. Spectral analysis
was applied for P-wave and the summation of P-wave and surface wave trains to find the
corresponding frequencies to the physical characteristics of the UWE. We estimated the cutoff
frequencies for the reflection off the ocean bottom and surface as well as the bubble pulse period.
Cutoff frequencies from the ocean bottom and surface as well as the bubble pulse frequency are
observed both on the P-wave spectra and the summation of P and Rayleigh wave spectra (in Figs.
@seismicisolation
@seismicisolation
4 and 5). Cutoff frequencies are 8.5, 25, 43 Hz from the bottom reflection and 45 Hz from the
surface reflection. The bubble pulse (1.01 Hz) which was determined from the previous study
(Kim and Gitterman, 2013) is not displayed on the left side in Fig. 4 but it appears on the right side
in Fig. 4 because the bubble pulse (1.01 Hz) comes 0.990 seconds after P-wave arrival at
12:21:59.24; i.e. 12:21:59.4 + 0.990 = 12:22:0.390, which is beyond the time window (see left in
Fig. 4) and we cannot find the appearance of the bubble pulse arrival. On the other hand in Fig. 5,
we can observe non-linear reflections from the hull of the ship at about17.5 Hz and 35 Hz in the
broad band as well as 8.5 Hz, 25 Hz, and 43 Hz (cutoff frequencies of reflection off the bottom)
and 45 Hz (cutoff frequency of reflection off the free surface) as well as 1.01 Hz (bubble pulse).
There are more salient spectral peaks in Fig. 5 due to amalgamation of P waves and surface waves
(Rayleigh waves). The advent of the bubble pulse and bubble collapse starts 0.990 s after P-wave
arrival disambiguating this non-contact UWE as well as corroborating cause of the ROKS
Cheonan sinking (see Fig. 6). Furthermore the strong energy of the bubble collapse on the N-S
component indicates that the bubble pulse collapsed in the N-S direction on the portside of the ship
near the surface.
Fig. 4. P-wave spectra for 1 s time window (left) and P-wave spectra for 2.1 s time window (right).
The bubble pulse (1.01 Hz) is not observed for the short time window on the left side, but is
observed for the longer time window on the right side because the bubble pulse arrives 0.990
seconds later. 12:21:59.4 + 0.990 s = 12:22:0.39 (bubble pulse arrival time). The appropriate time
window selection is very important to pick signals.
@seismicisolation
@seismicisolation
Fig. 5. Spectra of the summation of P-wave arrival and surface wave for a shorter time window (6
s) and a longer time window (12 s). Blue, green, red and purple arrows indicate a bubble pulse,
cutoff frequencies reflected from bottom, non-linear reflected hydroacoustic waves (Z & E-W
comp.) from the hull of the ship and mixed with T phases, and a cutoff frequency from the free
surface at around 45 Hz, respectively. The longer the time window is the more noise signals the
spectra obtain.
Fig. 6. The bubble pulse and bubble collapse were revealed 0.990 s later after P-wave arrival at
12:22:0.39 on the vertical component of the Baengyeong-do seismic station (BAR) which is 11.3
km away from the site and were also mixed with Rayleigh waves (Kim and Gitterman, 2013). 1.01
Hz and 1.72 Hz represent the first bubble pulse and the second bubble pulse frequencies.
@seismicisolation
@seismicisolation
We can see many different acoustic phase arrivals 31 seconds after P-wave arrivals from the
spectrogram of the Z component implying that the UWE took place at a very shallow underwater
(see Fig. 7). In particular, The broad amplitude spectra from Z and E-W components at around
17.5 and 35 Hz indicate that the non-linear hydroacoustic waves of the reflected P waves from the
hull of the ship were mixed with T phases (Figs. 4 and 5), whereas the bubble pulses (1.01, 1.72
Hz) (Kim and Gitterman, 2013) are well revealed on the vertical component after 0.990 s since the
bubble collapse occurred off the portside near the surface by means of the buoyant force and the
Bjerknes effect (Li et al., 2012). There are many different acoustic phases 31 seconds after P-wave
arrival from the spectrogram of the Z component implying that the UWE took place at a very
shallow underwater (see Fig. 7). We found the acoustic phase reflected from free surface around
45 Hz, direct acoustic phase at around 33 Hz, the second reflected acoustic phase from bottom at
25 Hz, the first reflected from bottom at 8.5 Hz, and the acoustic phases reflected from the hull and
mixed with T phases at 17.5 and 35 Hz including the acoustic phase due to a bubble pulse at 1.01
Hz (see Fig. 7).
Fig.7. Acoustic phases are also observed on the high resolution spectrogram of Z component 31 s
after P-wave arrivals in series of different types of paths at the Baengnyeong Island station (BAR).
We can observe an acoustic phase reflected from free surface at around 45 Hz, reflections from
hull with direct waves at around 17 Hz and 34 Hz, the first and the second reflected acoustic
phases from bottom at 8.5 and 25 Hz and the acoustic phase due to a bubble pulse at 1.012 Hz.
The bubble pulse (1.01 Hz), bottom reflection cutoff frequencies (8.5 Hz, 25 Hz, and 43 Hz), and
the free surface reflection cutoff frequency (45 Hz) are clearly observed (see Figs. 5 and 6). The
broad and strong peaks at around 17.5 and 35 Hz on the vertical and radial components (Z and E-
W) are the nonlinear reflected waves from the hull of the ship. Ocean depth (H), detonation depth
(d), and net explosive charge weight (W) ─ H ≈ 44 m [H= (1500m/s)/4*8.5Hz], d ≈ 8 m [d=
(≈1500m/s)/4*45Hz], and W ≈ 136 kg TNT [T=2.1W1/3/(10.1 + d)] where T= bubble pulse period
0.990 s ─ were estimated from the cutoff frequencies of the first harmonic mode from both the
bottom and surface reflections as well as @seismicisolation
the bubble pulse period from previous studies (Kim and
@seismicisolation
Gitterman, 2013; Kim, 2013) neglecting free surface effect due to the UWE off beneath the hull of
the ship. Provided the detonation depth via the cutoff frequency and the bubble pulse period are
determined from spectral analysis, it is possible directly to estimate the amount of the NEW
without testing the detonation depth piecemeal for each different case which was shown in the
early study.
BUBBLE PULSE PERIOD ANALYSIS USING BOUNDARY ELEMENT METHOD (BEM)
The time it takes for bubble expansion and contraction is called the bubble pulse period which was
found to be 0.990 s from the seismic observation, whereas it was 0.967 s from the analytical
solution using the Rayleigh-Willis equation (Kim and Gitterman, 2013). We also calculated the
bubble pulse period using the boundary element method (BEM) and 3D bubble shape simulation
(Figs. 7 and 8). We have attempted to find the best-fitting results using various charge weight,
depth as well as the portside from the hull centerline. Using the best estimated bubble pulse period
of 1.030 s by BEM and the Rayleigh-Willis equation aka the modified Rayleigh-Willis equation
(Kim and Gitterman, 2013; Kim, 2013), the UWE occurred at a depth of 8m approximately 5m
portside of the hull centerline with a 136 kg NEW detonation (Kim, 2013). It is worth noting that
the UWE did not take place on the free surface and there must have been the interaction between
the hull of the ship and the gas bubble. Although we used the Rayleigh-Willis equation which was
originated from the free surface, the surface effect for this shallow underwater explosion does not
influence the calculation of NEW because relative errors of bubble pulse periods between both
observation versus free surface (2.32%) and observation versus 3D simulation with BEM (4.04%)
are insignificant. Even if the non-contact UWE occurred off the portside of the ship, the nearer the
bubble pulse from the hull of the ship, the longer the life time (bubble pulse period) sustains due to
the Bjerknes effect assuming that the non-contact UWE occurred off the portside of the ship
(Zhang, 2008; Li et al., 2012).
Fig. 8 represents the approximate frame of the ROKS Cheonan for 3D bubble shape modeling and
boundary element method (BEM). The post-detonation bubble expands and achieves its maximum
volume at t=0.500 s and the bubble impinges on the hull before starting its collapsing phase in Fig.
9. At the same time a jet impact penetrates the bubble at t=0.932 s and again at t=0.947 s,
generating a toroidal bubble as shown in Fig. 9 (5) and 9 (6). At t=1.030 s the bubble achieves its
minimum volume and ends its first pulsing period in Fig. 9 (8).
(c) W = 10 m
(a) L = 88 m
(b) D = 2.9 m
Fig. 8. Sketch of the ship’s dimension: (a) Bottom view, (b) Side view, (c) Front view.
@seismicisolation
@seismicisolation
(1) t=0.000s (2) t=0.089s (3) t=0.500s
(4) t=0.791s (5) t=0.932s (6) t=0.947s
(7) t=1.007s (8) t=1.030s (9) t=1.122s
Fig. 9. Bubble shape evolution near the ship’s hull with a charge of 136 kg TNT at a depth of 8 m
and 5m portside. The behavior of the 3D bubble shape simulation taking into account the
interaction between the bubble and the ship’s hull based on BEM.
CONCLUSIONS
Cutoff frequencies of the reflections off the ocean bottom and the surface as well as the bubble
pulse were analyzed for this shallow underwater explosion via spectral analysis based on normal
modes and ray-tracing method. The cutoff frequency of 45 Hz on the surface reflection indicates a
NEW of approximately 136 kg TNT at a detonation depth of 8 m vis-à-vis the ROKS Cheonan.
Our findings are also espoused by the ray-trace modeling as well as the boundary element method
including that the pressure impulse on the hull of the ship during jet impact and the vortex of the
torodial bubble indicate that the high-speed liquid jet has a significant potential for causing severe
damage to the ship, notwithstanding a small size of TNT detonation (Zhang et al., 2008).
REFERENCES
Kim, S. G., Gitterman, Y., 2013. Underwater explosion (UWE) analysis of the ROKS Cheonan
Incident, Pure Appl. Geophysics., 170 (4), 479-484.
Kim, S. G., 2013. Forensic seismology and boundary element method application vis-à-vis ROKS
Cheonan underwater explosion,” J. Marine Sci. Appl., 12 (4), 422-433.
Li, Z., Sun, L., et al. 2012. Some dynamical characteristics of a non-spherical bubble in proximity
to a free surface, Acta Mech. 223, 2331-2355.
Lurton, X., 2002. An Introduction to Underwater Acoustics: Principles and Applications Praxis
Publishing, Chichester, UK. , pp. 13-58.
Nicolas, B., Mars, J., et al., 2003. Geoacoustical parameters estimation with impulsive and boat
noise sources, IEEEJ. of Oceanic Engineering, 28 (3), 494-503.
@seismicisolation
@seismicisolation
Porter, M. B., Bucker, H. P., 1987. Gaussian beam tracing for computing ocean acoustic Fields, J.
Acoust. Soc. Am., 82 (4), 1349-1359.
Rodriguez, O. C., Collis, J., et al., 2012. Seismo-acoustic ray model benchmarking against
experimental tank data, J. of Acoust. Soc.Am. 132 (2), 709-717.
Shapira, A., 1981. T phases from underwater explosions off the coast of Israel, Bull. Seism. Soc.
Am. 71 (4), 1049-1059.
Tindle, T. C., Weston, D. E., et al., 1980. Cycle distances and attenuation in shallow water, J.
Acoust. Soc. Am. 68 (5), 1489-1492.
Urick, R. J., 1983. Principles of Underwater Sound, 3rd edn, McGraw-Hill Book Company, New
York, pp. 172-201.
Weston, D. E., 1994. Waves shifts, beam shifts, and their role in modal and adiabatic Propagation,
J. Acoust. Soc. Am. 96 (1), 406-417.
Zhang, A., Yao, X., et al., 2008. Some dynamical characteristics of a bubble jet near a vertical wall,
J. Marine Sci. Appl., 7 (1), 1-10.
Zhang, A., Yao, X, et al., 2008. The interaction of an underwater explosion bubble and an elastic
plastic structure, Appl. Ocean Research, 30, 159.-171.
@seismicisolation
@seismicisolation
APPENDIX A: PROBABILITY CALCULATION FOR LCM
The ROKS Cheonan Sinking was also proved by probability that it was due to the Land Control
Mine (LCM).
The cause of Sinking of the ROKS Cheonan can be proved that it is attributed to the sea mine
(LCM) not the North Korean torpedo, The JIG (Civilian-Military Joint Investigation Group
Report, 2010) concluded that the ROKS Cheonan was split and sunk due to shockwave and bubble
effects generated by the underwater explosion of a torpedo. The detonation location was 3 m to the
port from the center of the gas turbine room and at a depth of 6~9 m. The weapon system used was
a CHT-02D torpedo with approximately 250 kg of explosives manufactured by North Korea. First
of all, however, using 1.1 s of a bubble pulse period is not correct because the determination of a
bubble pulse period came from two peaks which were measured from the interval of a burst of the
underwater explosion and a sonic boom.
P(A∣B) = P(A)P(B∣A)/P(B) = P(A)P(B∣A)/[P(A)P(B∣A) + P(not A)P(not B∣ A)]
where P(A∣B) = posterior probability(final probability after information) P(A) = prior
probability(initial probability)
P(B∣A) = conditional probability, B occurs if A is true(likelihood) P(B) = normalization
constant
P(not A) = probability in case of not P(A), [1 - P(A)]
P(not A)P(not B∣ A) = probability in case of not P(A)P(B∣A), [1 - P(A)P(B∣A)] Numerical
Computation:
LCM (land control mine) has a seismic yield of 136 kg equivalent TNT which was modified from
MK-6 (JIG Report, MD, 2010) , North Korean torpedo is assumed to have a seismic yield of 250
kg equivalent TNT from the wrong bubble pulse period (1.1 s). Choosing either a LCM or a North
Korean torpedo, so each probability is 1/2 (0.5) using seismic, hydroacoustic, infrasound waves
and BEM analyses from the underwater explosion. P(A) = 0.5
P(B∣A) = 0.99999(if A is a LCM, it(B) is 136 kg TNT from Joint Investigation Report of the
JIG (Ministry of Defense, Republic of Korea, 2010)
P(A∣B) = 0.5 * 0.99999/[0.5*0.99999 +0.5*0.00001] = 0.99999
Therefore, the odds that the cause of sinking of the ROKS Cheonan was due to a LCM whose
probability is 99.999 % (for sure LCM).
@seismicisolation
@seismicisolation
APPENDIX B: SUPPLEMENTARY MATRIALS FOR FORENSIC EVIDENCE FOR

CAUSES OF THE ROKS CHEONAN SINKING
Figure S1. The coast line of the Baengnyeong Island consists of hard diatomaceous Earth beach
and rocks (reefs) covered with many oyster shells. The ship tried avoid some reefs and ran aground
prior to the underwater explosion on March 26, 2010 at 21:21:57.
Figure S2. The beach consists of mostly hard sand powder in diatomaceous earth (diatomite) and
some oyster-shells adhered rock. Especially the large diatomaceous earth beach was once used as a
natural runway because it was too hard to leave any footprints. There are also some reefs
(unknown submerged rock) in this region.
@seismicisolation
@seismicisolation
Figure S3. Seismograms and accelograms recorded at Baengnyeong-do (Island) from the ROKS
Cheonan incident clearly approve that this must have been a non-contact underwater explosion on
March 26, 2010. P, S, BP, BJ and PP indicate P-wave first arrivals (up). S-wave arrival, gas bubble
pulse, bubble jet and peak pressure. LR (weak arrow) and LQ represent Rayleigh wave and Love
waves. The records consist of 20 and 100 samples per second (SPS) for acceleration and velocity,
respectively.
We can ensure that the ROKS Cheonan sinking must have been an underwater explosion
observing the positive higher amplitude of the first P-wave arrival (+) on the vertical component
(HHZ) including well generation of Love waves on N-S component. We can also clarify where the
ROKS Cheonan underwater took place from the station measuring vector analysis of 3
components (NS, EW and Z) resulting in in the direction of SW (BAZ=225.86⁰) in an aspect of
seismology. Furthermore we can observe the down-going (rarefaction) of bubble pulse phases
(BP) and bubble jets (BJ) in low samples (20 Hz) as well as in high samples (100 Hz) of the
seismograms in Figure S3.
@seismicisolation
@seismicisolation
Figure S4. The underwater nuclear explosion on October 23, 1961 at 08:31:05 (GMT) was
conducted at the Bay Chernaya of Novaya Zemlya Island at a depth of 101 m with 1.56 kt TNT
(recalculated from M 5.10, not 10 or 4.8 kt). P, S, LR, BP, BJ, RF and SR indicate P- and S- wave
arrivals, Rayleigh wave, bubble pulses, bubble jets and reverberation frequencies reflected from
the sea bottom and surface at 3.61 s after P-wave onset which were recorded on the broadband
(0.2-30 sec) at town Belushya, Novaya Zemlya 120 km away from the epicenter.
A principal feature of underwater explosions is a distinct rarefaction impulse (negative) arrival
(BP) with amplitude 2 times larger than that of the first P-wave arrival. The first bubble pulse
period is 4.82 sec (0.207 Hz) and the second one is about 2.53 sec (0.395 Hz). Using bubble period
4.82 s and the modified Rayleigh-Willis Equation [T=2.1 W1/3/(d + 10)5/6], we calculated the
detonation (actual) depth at about 101 m with a bubble radius of about 92 m even if the reported
depth was 20-50 m (Khristoforov, 1996). Using the seafloor reverberation frequency (1.93 Hz) from
the bottom in the record, the sea depth is estimated at 194 m in the Barents Sea. It should be also
noted that one can see a reflected phase from the surface at 3.61 s (SR) from the P-wave arrival onset.
Figure S5. Navy soldiers heard two sounds: the first weak sound ‘Kung’ and around 1 second
later, the second thundering and wrecking sound ‘Kwang’. The first sound is the burst of an
underwater explosion and the second one is a sonic boom.
@seismicisolation
@seismicisolation
Figure S6a. Schematic of torpedo’s motor and propeller and recovered rear section of torpedo
(JIG, 2010).
Figure S6c. The recovered torpedo propeller which was not so rusty because Ministry of Defense
provided us with a new one that was safely kept right after having collected from the sea.
@seismicisolation
@seismicisolation
Figure S7. The tangled fishing nets on the propeller axis and some charred hull surface indicate the
running aground and strong electric current flow on the hull by ICCP before the main underwater
explosion took place. Also it cannot be ruled out that the soot might have been related to the
burning during the burst of the underwater explosion.
Russian investigators said that the rear part of the torpedo was older than 6 months according to
the rust and the minor accident (running aground) also took place before the major underwater
explosion according to the stopping of a clock at 21:17:03 in the ship. There are also some
communications between sailors and their acquaintances during 21:15-21:17 implying that the Key
Resolve(anti-submarine) war exercise was being conducted searching for submarines between ROK
and US Navy at that time. The ship may have navigated too close to the littoral land resulting in
running aground touching the coast bottom.@seismicisolation
After the underwater explosion the exercise stopped.
@seismicisolation
Figure S8. Structure and operating principle of the Land Control Mine (LCM). The land control
mine (LCM) was a modification of the hydroacoustic pressure MK-6 depth charge (JIG, 2010).
The design and specification are shown in the above figure (JIG, 2010). Mr. Kim, Dong-Hyeong,
developed LCMs from MK-6 and deployed LCMs near the Baengnyeong Island against the North
Korean invasion between July and October 1979. Unfortunately the technician, Mr. Kim, Dong-
Hyeong has not kept any records and/ or reports regarding LCM development and its installation
near the Baennyeong-Island at that time.
Using voltage (5 V) and current (150 mA), the principle in which current flow (0.75 W) from
ship’s ICCP spike can electrically power an exposed naval mine (LCM) cable in sea water
resulting in an underwater explosion triggering the detonation cable connected to a LCM at
21:21:57 since the ship tried to have run away from running aground at 21:15-21:17.
Figure S9. Principle of ICCP (Impressed Current Cathodic Protection)
@seismicisolation
@seismicisolation
Figure S10. The International Hydro-acoustic Workshop (IHW2015) was held on June 29-30,
2015 at Vienna. Austria, by CTBTO. The author (the third from the front row) presented a paper
on the ROKS Cheonan Sinking by an Underwater Explosion.
@seismicisolation
@seismicisolation
CHAPTER EIGHTEEN
FORENSIC SEISMOLOGY AND BOUNDARY ELEMENT METHOD

APPLICATION VIS-À-VIS ROKS CHEONAN UNDERWATER EXPLOSION 1
SO GU KIM
ROKS Cheonan before and after split of PCC-722 (displacemenet=1200 t, length = 88 m and draft
= 2.9m) from Joint Investigation Report of Civilian Military Joint Investigation Group (2010).
ABSTRACT
On March 26, 2010 an underwater explosion (UWE) led to the sinking of the ROKS Cheonan. The
official Multinational Civilian-Military Joint Investigation Group (MCMJIG) report concluded that
the cause of the underwater explosion was a 250 kg net explosive weight (NEW) detonation at a
depth of 6-9m from a DPRK “CHT-02D” torpedo. Kim and Gitterman (2012a) determined the NEW
and seismic magnitude as 136 kg at a depth of approximately 8m and 2.04, respectively using basic
hydrodynamics based on theoretical and experimental methods as well as spectral analysis and
seismic methods. The purpose of this study is to clarify the cause of the UWE via more detailed
methods using bubble dynamics and simulation of propellers as well as forensic seismology.
Regarding the observed bubble pulse period of 0.988 s, 0.976 s and 1.030 s are found in case of a
136 NEW at 8m detonation depth using boundary element method (BEM) and 3D bubble shape
simulations derived for a 136 kg NEW detonation at a depth of 8m approximately 5m portside of
the hull centerline. Here we show through analytical equations, models and 3D bubble shape
simulations that the most probable cause of this underwater explosion is a 136 kg NEW detonation
at a depth of 8 m attributable to a ROK littoral “land control” mine (LCM).
Keywords: cepstrum, spectrogram, bubble pulse, toroidal bubble, boundary element method, ICCP
1
Originally published in J. Marine Sci. Appl. (2013). 12, 422-433.
https://www.youtube.com/watch?v=EHFC30bbhhc @seismicisolation
@seismicisolation
378 Chapter Eighteen
INTRODUCTION
The underwater explosion (UWE) vis-à-vis the ROKS Cheonan sinking took place on March 26,
2010 with an origin time of 12: 21: 57 (UTC) and location at 37.93°N and 124.60 ° E (see Fig. 1).
Fig. 1 Map of the UWE site. The star, triangle and circles indicate the UWE site, seismic station and ports, respectively.
The Multinational Civilian-Military Joint Investigation Group (MCMJIG) (2010) concluded that the
UWE had a net explosive weight (NEW) of 250 kg and occurred at a depth of 6-9m with a seismic
magnitude of 1.5 (MCMJIG, 2010) ; the MCMJIG also asserted that this yield was equivalent to the
North Korean “CHT-02D” torpedo. Kim and Gitterman (2012a and 2012b) studied this UWE
through spectral analysis vis-à-vis bubble pulse periods as well as acoustic-phase, T-phase and infra-
sound records and concluded that the detonation source had a NEW of 136 kg and detonated at a
depth of 8m with a seismic magnitude of 2.04. The 136 kg NEW is equivalent to the yield of ROK
“land control” naval mines (LCM) installed in the late 1970’s in the vicinity of the ROKS Cheonan
sinking. In this paper we will show how we came to our conclusion regarding the cause of the UWE
vis-à-vis the ROKS Cheonan through forensic seismology and bubble dynamics analysis.
@seismicisolation
@seismicisolation
Forensic Seismology and Boundary Element Method Application vis-à-vis ROKS Cheonan Underwater 379
Explosion
Fig. 2 Reported magnitude for the underwater and land explosions versus the net explosive weight (From Kim and
Gitterman, 2012a).
The empirical relationship between local magnitude (ML) and net explosive weight W (kg) for
shallow underwater explosions were analyzed in previous studies (Jacob and Nelson, 1977;
Khalturin et al.; 1998; Brocher, 2003). Equations 1 and 2 of Fig. 2 (Kim and Gitterman, 2012a) are
more seismically efficient for undersea explosions than explosions on land for a given net explosive
weight.
SPECTRAL AND CEPSTRAL ANALYSES OF SEISMIC WAVES
Bubble pulse (BP) frequencies are observed at 1.01 Hz (0.990 s) and 1.72 Hz (0.581 s) while
reverberation (RH) frequencies are observed at 8.5Hz for the first harmonic and 25 Hz for the second
harmonic via spectral analysis of the Baengnyeong-do (BAR) station records (Kim and Gitterman,
2012a). The higher amplitude of the NS component at 1.78 Hz indicates that the unexpected S
wave and large Love waves (LQ) are related to the interaction between the bubble at its maximum
volume and the ship’s hull with its accompanying jet impact and toroidal bubble (Klaseboer et al.,
2005; Zhang et al., 2001; Kim and Gitterman, 2012a). The two small bubble peaks near the bubble
pulse period on the cepstrum ( in Fig. 3) indicates the jet impact penetrating the bubble along with
gas recompression. So we can hardly see the cepstral peak which normally shows up on the cepstrum
in this very shallow underwater explosion (Cohen,1970), but there is a high amplitude of BP which
arrived later on the spectrogram.
@seismicisolation
@seismicisolation
Fig. 3 The small bubble pulses on the cepstrum may be due to the jet impact penetrating the gas bubble and a gas
recompression by a very shallow underwater explosion. BP, RH and Rd represent the bubble pulse period (0.990 s=1.01
Hz), the water column reverberation on the bottom (0.118 s=8.5 Hz) and reflection on the surface water (0.022 s=45
Hz) respectively. BP, RH and Rd are parameters are used to estimate the bubble pulse period, the sea depth, detonation
depth for this UWE.
However, distinct water column reverberations from the bottom (RH), surface reflections (Rd), and
bubble pulses (BP) can be observed from the spectrogram and cepstral analysis (Figs 3 and 4). The
bubble pulse period, detonation depth, and depth of the sea are represented by BP, Rd and RH,
respectively (Weinstein, 1968; Baumgardt and Der, 1988; Kim and Gitterman, 2012a). If the UWE
is located almost directly underneath and/or close to a vessel, the bubble collapse could significantly
contribute to the subsequent damage to the vessel by producing a high speed water jet and also via
whipping damage caused by the bubble pulse. The spectral anomalies (NDS) at 17.5 Hz and 35 Hz
saliently observed on the radial direction (vertical and EW components) may be attributed to the
nonlinear shock waves associated with the UWE (Fig. 4; Kim and Gitterman, 2012a).
@seismicisolation
@seismicisolation
Explosion
Fig. 4 Spectrograms for the 3-component from the BAR station. Acoustic phases (33Hz) and T-phases (15-17Hz)
appeared as well as bubble (1.01 Hz) and reverberation (8.5 Hz) effects (Kim and Gitterman, 2012a). NDS (17.5 and
35 Hz) indicates “non-defined spectrogram” that may be attributed to the nonlinear shock waves from the UWE.
Our modified Rayleigh-Willis equation based on the pressure developed in a liquid during the
collapse of a spheriFDOFDYLW\ G5max) where d is the detonation depth and Rmax is the maximum
bubble radius is more applicable as an initial approximation model than straight up BEM and 3D
bubble shape simulations (Krieger and Chahine, 2005; Gong et al., 2010; Kim and Gitterman,
2012a).
Swisdak (1978) stated his correction equation (3) should be taken into account “for the
configurations which amount to 15% at 2 bubble radii” to the surface, the bottom or both, but in this
case we do not have a bubble due to bubble collapse. The updated study of Krieger and Chahine
(1978) recently noted that underwater explosion data are unavailable below approximately the
normalized depth C (standoff distance/maximum bubble radius) = 0.5 - where standoff distance is
the detonation depth - as the spark-generated bubbles vent their contents to the atmosphere in the
region, lose their intensity, and fail to emit bubble pulses. More detailed studies for the ratio of
detonation depth to maximum bubble radius have been carried out to investigate the distorted bubble
by proximity to the free surface, the bottom, or to a solid object using recent experimental and
empirical data (Klaseboer et al., 2005; Zhang et al., 2008). Klaseboer et al. (2005) also studied the
boundary condition at free surface using scaled depth C in the range of 0.5<C< 3.0 and found that
generally in 0.5<C<1.5 the bubble “works” and for C<1.5 it is nonlinear but that as the maximum
bubble radius becomes greater than detonation depth in the range of 0.5 <C<1.0, the top bubble rises
above the equilibrium free-surface level but the surface is not breached – it maintains its integrity
and exhibits behavior resembling an elastic membrane and a water layer remains between the bubble
and the free surface. Near C=1, we can see the jet impact and toroidal bubble formation during the
@seismicisolation
@seismicisolation
initial bubble collapse (Krieger and Chahine, 2005 2XU&YDOXH FOHDUO\VDWLVILHVWKHHDUOLHU

studies of Swisdak (1978), Krieger and Chahine (2005), and Klaseboer et al., (2005).
Fig. 5 Nomograph of the first bubble pulse period and maximum bubble radius of an underwater gas bubble (thick black
line) from Swisdak (1978). The thin blue line represents the detonation of 136 kg TNT at a depth of 8m with the bubble
pulse period of 0.990 s and the maximum bubble radius of 6.63 m in our paper (KIM and GITTERMAN 2012a). The
bubble pulse period of 0.990 s and 250kg TNT produces the maximum bubble radius of 8m and the detonation depth of
15 m (dark- red dashed line). The thick line indicates the detonation of 100 kg TNT at a depth of 500m with the bubble
pulse period of 0.054 s and the maximum bubble radius of 2.033 m (SWISDAK, 1978).
The initial motion of the P-wave arrival with a compressional motion on the vertical component and
corresponding bubble pulses and reverberation effects (Kim and Gitterman, 2012a) from spectral
analysis as well as acoustic-phase, T-phase, and infrasound signals verify that the UWE was a non-
contact, shallow depth UWE. Just like we can estimate the bubble pulse period and the detonation
depth via forensic seismology, we can also estimate the net explosive weight and the maximum
bubble pulse radius using the Swisdak nomograph (1978) as shown in Fig. 5. The thin blue line
represents the detonation of 136 kg NEW at a depth of 8 m with a bubble pulse period of 0.990 s
and a maximum bubble radius of 6.63 m (Kim and Gitterman, 2012a). A bubble pulse period of
0.990 s and 250 kg NEW produces a maximum bubble radius of 8 m and a detonation depth of 15
m (dark- red dashed line) in Fig. 5. The case of a 250 kg NEW DPRK torpedo obviously fails to
correspond to the observed data.
@seismicisolation
@seismicisolation
Explosion
BASIC THEORY OF BEM ON BUBBLE DYNAMICS
BEM has been widely used in bubble dynamics for its advantages in accuracy (Klaseboer et al.,
2005), efficiency, and interface capture. For an underwater explosion bubble, it is reasonable for the
fluid field to be assumed incompressible (Magnaudet and Eames, 2000; Klaseboer at al., 2005).
Based on the potential flow theory, the Laplace equation can be rewritten through the Green’s
Function Method as utilized by Plesset and Prosperetti (1977), Klaseboer et al. (2005) and Gonzalez-
Avila et al. (2011).
§ w< w ·
c p < p ³S © wn
¨ G p , q < G p , q ¸ ds
wn ¹ (1)
Here, ୠ is the velocity potential, S is the boundary surface of the fluid field including the bubble
surface, p and q are the fixed point and integral point on the boundary, respectively, c(p) is the solid
angle at the location p, G(p,q)=1/rpq is the Green function, and n is the unit normal vector pointing
out of the fluid field.
For a violent bubble generated by an underwater explosion, the surface tension is neglected. Thus,
the hydrodynamic pressure at the bubble surface would be equal to the inner bubble pressure
(Klaseboer et al., 2005, Wang and Blake, 2011).
J
§V ·
Pin Pc P0 ¨ 0 ¸
©V ¹ (2)
Here Pc is the saturated vapor pressure of the condensable gas, P0 and V0 are the initial pressure and
volume of the bubble, respectively, and Ȗ is the ratio of the specific heat of the gas inside the bubble
(1.25 for an underwater explosion).
The boundary condition of the interface of fluid and submerged structure can be expressed as:
w<
Vs n
wn (3)
Here, Vs is the velocity of the structure.

The dynamic boundary conditions of the interfaces including bubble surface and the other surfaces
are given by the Bernoulli Equation:
w< Pf Pin 1 2
< gz
wt U 2 (4)
Here, z is the vertical component of spatial coordinates, P is the undisturbed far field pressure at
z=0, g is the gravity acceleration, and ʌ is the fluid density.
@seismicisolation
@seismicisolation
Toroidal bubble
When the jet impact penetrates the bubble surface, forming a toroidal bubble, the potential on the
bubble surface is not unique anymore because of the first bubble collapse. So a ring inside the bubble
is introduced in the simulation to solve the problem. Then the velocity potential can be decomposed
as shown by Wang et al. (1996) and Zhang et al. (2001).
< total < res < ring

(5)
Here ୠring is the velocity potential induced by the vortex ring and associated with the circulation
generated by the impact, ୠres is the residual potential on the bubble surface which is smooth in the
entire fluid domain.
The strength of the vortex ring equals the difference of jump in the velocity potential across the
contacting points during the impact process on the jet point and impact point which can be obtained
as follows:
3 <i < j
(6)
Following the Biot-Savart law, the velocity and velocity potential induced by the vortex ring can be
expressed as follows:
3 r ( p, q ) u d l q
v ring ( p, t ) ³
4S r3 , where the integral represents ‫݌(ݎ ׯ‬, ‫ܫ݀ݔ)ݍ‬௤ /‫ ݎ‬ଷ .
3:( p )
< ring ( p )
4S
where
w §1·
: p ³ ds
Sc wn ¨ r ¸
© ¹
is the solid angle of the vortex surface Sc to the point on the bubble surface.
Following Eq. (4), the residual velocity potential can then be updated with the following Bernoulli
Equation (Wang et al., 1996):
w< res Pf P 1 2
vring (vring < res ) < res vring gz
wt U 2 (7)
@seismicisolation
@seismicisolation
Explosion
Numerical procedure
(a) (b) (c) (d) (e)
Fig. 6 Meshes on bubble surface for 3D bubble shape simulation. (a) mesh 3(92 nodes, 180 elements), (b) mesh 4(162
nodes, 320 elements), (c) mesh 5(252 nodes, 500 elements), (d) mesh 6(362 nodes,720 elements), (e) mesh 7(492 nodes,
980 elements).
We use triangular planes to mesh the bubble surface for our simulation. The bubble surface is
meshed in different grid as shown in Fig. 6, which also includes different mesh information. Using
the mesh described in Fig. 6, we discuss the convergence of mesh and the time step. According to
Wang et al. (1996), the time step is controlled with Eq. (1). The time step changes according to
bubble surface velocity and its inner pressure. dߚ is used to control the bubble time-step. Here we
choose dߚ =0.03, 0.06, 0.10, and 0.20 to validate the program with different time step
Gt min^'t1 , 't 2 ` (8-a)
dE
't1
1 U 2 U gz Pin
max 1 v
2 Pf Pf Pf
(8-b)
dE
't1
1 U 2 U gz
max v
2 Pf Pf
(8-c)
J
§V ·
Pin P0 ¨ 0 ¸
where ©V ¹ .
Lee et al. (2007) putted forward a method which considered the energy loss during the bubble
pulsation. In their study, they assume that the total mechanical energy is conserved in each bubble
cycle. The total energy is constant in each cycle contains the potential energy and the kinetic energy,
which is written below.
V V 1 2
E ³ Pin dV ³ Pf dV U ³ v dS
V0 V0 2 S (9)
The initial inner pressure at each bubble pulsation stage can be estimated from the following
equation.
3 3
§ T2 · § Rm 2 · P02
¨ ¸ ¨ ¸ |
© T1 ¹ © Rm1 ¹ P01 (10)
@seismicisolation
@seismicisolation
Where the subscript 1 and 2 means the bubble’s first and second pulsation. T means the bubble
pulsation period, Rm the bubble maximum radius, P0 the initial inner pressure at each pulsation stage.
From Eq.(9), the initial inner pressure in each bubble pulsation period can then be got. Based on this
equation, the following results are presented. According to the experimental data of bubble pulsation
period and maximum radius in its first and second cycle, the ratio P021/P01=0.21.
Fig. 7 Bubble radius with different time steps.
Considering the cases of 136 kg NEW exploding at the depth of 8 m, its maximum radius is 6.63 m.
Here we validate it by comparing it with the analytic method (Rayleigh, 1917) in which the buoyancy
effect is ignored. The mesh chosen is Mesh 6, with 362 nodes and 720 elements in Fig. 6. Fig. 7
indicates that the Rayleigh equation does not take the energy loss into consideration, resulting in the
maximum radius not decreasing in the second bubble pulse period. It is also notable that the
numerical method converges with decrease dߚ, and in case of dߚ =0.30, the error is larger. This is
because the larger dߚ means a larger time step and can cause a relatively large error whereas dߚ
=0.03, 0.06, and 0.10 gives a relatively smaller error. So in the following simulation, we choose
dߚ=0.03.
Fig. 8 shows bubble radius with different triangular plane mesh count sizes. From the figure, we
can see that the mesh size corresponds well with the analytic method (Rayleigh, 1917). Fig. 8 gives
the comparison of bubble radius with time and with different triangular plane mesh size (from Mesh
3 to Mesh 7). From the figure, we can see that in the first bubble pulsation period, a higher mesh
level gives a better agreement with the analytic method (Rayleigh 1917), while in the second period
the bubble radius differs because the Rayleigh Equation does not consider the energy loss.
Considering the computational efficiency, we use mesh 6 for our simulation in the following
simulations. Figs 7 and 8 also tell that the numerical method adopted in this paper offers good
accuracy.
@seismicisolation
@seismicisolation
Explosion
Fig. 8 Bubble radius with time and with different triangular plane mesh size
Bubble pulse period analysis

The time it takes for bubble expansion and contraction is called the bubble pulse period which in
this case was 0.990s and through this we can determine the net explosive weight and explosion
depth. It is in line with the calculations of 0.967 s via Rayleigh-Willis equation (Cole, 1948;
Willis, 1963; Barger and Hamblen, 1980) also known as the modified Rayleigh-Willis equation
(Kim and Gitterman, 2012a), 0.976 s via boundary element method (BEM) (Zhang et al., 2008)
and 1.030 s via 3D bubble shape simulation (Zhang et al., 2008) derived for the case of a 136 kg
NEW detonation at a depth of 8m approximately 5m portside of the hull centerline (Fig. 9 and
Table 1).
a
@seismicisolation
@seismicisolation
(1) t=0.000s (2) t=0.100s (3) t=0.507s
(4) t=0.773s (5) t=0.908s (6) t=0.920s
(7) t=0.972s (8) t=1.058s (9) t=1.140s

c
(1) t=0.000s (2) t=0.110s (3) t=0.509s
(4) t=0.904s (5) t=1.031s (6) t=1.074s
@seismicisolation
@seismicisolation
Explosion
(7) t=1.130s (8) t=1.189s (9) t=1.220s

d
(1) t=0.000s (2) t=0.089s (3) t=0.500s
(4) t=0.791s (5) t=0.932s (6) t=0.947s
(7) t=1.007s (8) t=1.030s (9) t=1.122s

e
(1) t=0.000s (2) t=0.118s (3) t=0.509s

@seismicisolation
@seismicisolation
(4) t=0.813s (5) t=0.938s (6) t=0.955s
(7) t=1.034s (8) t=1.172s (9) t=1.286s
Fig. 9 The behavior of 3D bubble shape simulation taking into account the interaction between the bubble and the
ship’s hull based on the BEM method. a. (1) bottom view (L=88m) (2) side view (D=2.9 m) (3) front view
(W=10m). b. Bubble shape evolution near a ship hull with 136 kg TNT depth at a depth of 8 m and 3 m portside. c.
Bubble shape evolution near ship with 250 kg TNT at a depth of 9 m and 3 m portside. d. Bubble shape evolution
near ship hull with charge of 136kg TNT at a depth of 8 m and 5 m portside. The bubble pulsing period is 1.030 s
which is close to the observed bubble pulse period (0.990s) e. Bubble shape evolution near ship hull with charge of
250 kg TNT at a depth of 9 m and 5m portside. The bubble pulsing period is 1.172 s in this case.
The post-detonation bubble expands and achieves its maximum volume at t=0.500 s and the
bubble impinges the hull generating a toroidal bubble at 0.932 s before starting its collapsing
phase. At the same time a bubble jet impact penetrates the bubble at t=1.007 s with vortex ring
(130-170 m/s), finally reaching a bubble pulse at t=1.030 s, At t=1.030 s the bubble achieves its
minimum volume and ends its first pulsing period.
THE PROPELLER SIMULATION
The propeller blades plunge into the sand/seabed when running aground (Lake and Gibson, 1987)
as evidenced in Fig. 10
@seismicisolation
@seismicisolation
Explosion
a b
Fig. 10 Damage to propeller and shaft. a. Fore side view of the starboard propeller showing deformation due to collision
against the seabed. Shaft axis is also off-set. b. Aft view of the starboard propeller blades bent opposite of its rotation
(clockwise). The blade edges were severely bent backward during collision with the seabed. The clean-cut of the bottom
blade was due to collision against the deck when the stern part was taken from the sea.
As the ship ran aground, high pressure is applied to the parts of blade plunged into the sand, about
0.2 m2 for each blade. These plunged parts of blades took all the power coming from the engine.
Taking 0.3 as the propulsive efficiency (0.5-0.6 for normal propulsion), the pressure applied on the
plunged blade part would be:
F 1 PSK 1 8e5 u 725 u 0.3

P 2.3e7 Pa
5S 5S U 5 u 0.2 7.7 (11)
where F is the propeller thrust, is the area of plunged part of each blade, PS is the power supply, Ș
is the propulsive efficiency, and U is the advance speed.
The dynamic response of the propeller is solved with Explicit Dynamic FEM of ABAQUS. The
discrete equations of Explicit Dynamic FEM can be expressed as:
ui M 1 Fi Ii
(12)
where ‫ݑ‬ሷ is the acceleration vector, the subscript i indicates its variable at the time of increment i,
and M, F and I are the diagonal lumped mass matrix, the applied load vector, and the internal force
vector, respectively. Then the nodal displacement and velocity are updated with the following
expressions in time domain:
'ti 1 'ti
ui 1 u 1 u
i
i
2 2 (13)
ui 1 ui 'ti 1u 1
i
2 (14)
where ¨t is the time increment. The finite element (FE) model of the propeller is shown in Fig. 11.
@seismicisolation
@seismicisolation
Fig. 11 a. Original undamaged starboard screws FE model: view of the screw from directly behind, view of the screw
from in front at an oblique angle, view of the screw from behind at an oblique angle (first row). b. Screw damaged due
to running aground: numerical results from view of the screw from in front at an oblique angle, by Mises contour (von
Mises stress distribution) and by PEEQ contour (accumulated fatigue damage by inelastic strain) (second row). Screw
damaged due to running aground: numerical results from view of the screw from behind at an oblique angle, by Mises
contour and by PEEQ contour (third row).
CALCULATIONS OF BUBBLE PULSE PERIODS AND ERRORS

Calculations made based on various explosive yields and depths show that a 136kg NEW detonation
at 8m in depth is the best match to produce the bubble pulse period in the observed seismic data.
This is equivalent to a local magnitude of 2.04 (Kim and Gitterman, 2012a) and corresponds to the
net explosive weight of one of many LCM naval mines (MCMJIG, 2010) known to have been
abandoned in that area (Fig. 1). The LCM was a MK-6 moored contact naval mine which was
heavily modified by the ROK Navy to not be detonated by contact but rather by remote detonation
electrically from shore. It was also redesigned to be a bottom mine - retaining its original 300lb
(136 kg) net explosive payload - to be deployed in shallow waters in the vicinity of the ROKS
Cheonan sinking. They were installed in that area in 1979 but in 1985 they were deactivated by
simply cutting all the electric mine cables. In 2008 due to mounting public sentiment the ROK
@seismicisolation
@seismicisolation
Explosion
government undertook an operation to collect and dispose of these particular naval mines, however,
only a few were successfully recovered. The vast majority of these naval mines were not recovered
and though they were tactically useless and deemed “functionally useless”, they still retained the
capability to detonate especially because they were sealed in Fiber Reinforced Polymer (FRP) boxes.
Bubble pulse period calculations for the case of a 250 kg NEW detonated at a depth of 9m were
estimated at 1.123 s by Rayleigh-Willis equation, 1.150 s by BEM and 1.172 s by bubble shape
simulation indicating non-correlation with the observed bubble pulse period (see Table 1).
Table 1: Relative errors are estimated from the observed bubble period (T0) vis-à-vis
calculations derived via Rayleigh-Willis equation (TR), BEM (TB) and calculations utilizing
simulations with detonation occurring at 3m (T3) and 5m (T5) portside of the hull centerline
where the relative eUURU Ň>7 7R, TB, T3, T5) – T0@Ň70 * 100, d and Rm represent
detonation depth and the maximum radius of gas bubble respectively.
TNT d Rm TR Error TB Error T3 Error T5 Error
d/Rm
kg m m sec % sec % sec % sec %
LCM 136kg TNT
136 7.0 6.76 1.014 2.42 1.0279 3.83 n/a n/a n/a n/a 1.04
136 8.0 6.63 0.967 2.32 0.9764 1.37 1.058 6.87 1.030 4.04 1.21
CHT- 02D 250kg TNT
250 8.0 8.12 1.184 19.60 1.2082 22.04 n/a n/a n/a n/a 0.99
250 9.0 7.98 1.132 14.34 1.15 16.16 1.189 20.10 1.172 18.38 1.13
The analytical empirical model of the bubble pulse period vis-à-vis NEW and detonation depth (Fig.
12) has been investigated by many researchers (Rayleigh, 1917; Cole, 1948; Willis, 1963; Kim and
Gitterman, 2012a, etc.). Considering the impact location of the UWE on the ROKS Cheonan at
approximately 5m portside from the hull centerline, the ocean surface was barely breached which
narrows down the source possibilities in terms of NEW and detonation depth. Analysis of observed
data as well as the damage features of the ROKS Cheonan clearly indicate damage inflicted via the
effects of bubble pulse and bubble collapse from the UWE. A 136 kg NEW detonation at a depth
of 8m generates a maximum bubble radius of 6.63 m and contains both a bubble pulse and a bubble
collapse which would breach the ship’s draft of 2.9 m. A 250 kg NEW detonation fails to fit not
only the observed bubble pulse period but also analytical and simulated bubble pulse periods (Kim
and Gitterman, 2012a). At a depth of 15m, a 250 kg NEW detonation would have yielded a bubble
pulse period close to the observed period of 0.990 s, however, at this depth the maximum radius of
the gas bubble would have been 8 m which could not have breached the ship’s hull resulting in a jet
impact with a toroidal bubble following bubble collapse (Figs 5 and 12). At this distance a whipping
motion would have been generated by the bubble pulse but would not have been able to inflict the
kind of damage that the ROKS Cheonan sustained.
@seismicisolation
@seismicisolation
Fig. 12. Distribution of possible bubble pulse periods by spectral analysis from observed data (star), Rayleigh-Willis
equation (TR), BEM (TB), 3D simulation (T3, T5) for 136 kg net explosive weight at a depth of 8 m at 3 m and 5 m
portside of the hull centerline and for 250 kg net explosive weight at adepth of 9 m at 3 m and 5 m portside of the hull
centerline.
RESULTS AND DISCUSSION

The toroidal bubble and jet impact are formed in a non-spherical way by the bubble collapse with
the jet penetrating the bubble. The bubble jet causes the rapid acceleration upwards of the non-
spherical bubble with a speed of 130-170 m/sec (Plesset and Chapman, 1971) along with the
influence of inertia and buoyancy.
The detonation was most likely triggered when the mine cables were dragged out as the ROKS
Cheonan ran aground. Evidence of serious navigational problems are manifested by the backward
bent propeller with serrated blades as well as the fishing net and rope (later removed) caught around
the starboard propeller shaft (Figs 10 and 11). The ship’s ICCP (Impressed Current Cathodic
Protection) (Ditchfield et al., 1995; Wu et al., 2009) was most likely the catalyst for triggering the
naval mine (LCM). The vessel operates four anodes with two reference cells which send currents
to prevent corrosion, usually maintaining 10-20A and 0.45-0.85V but depending on damage to the
hull this could reach a maximum of 300A and 24V. The minimum current required to ignite the
LCM naval mine is 0.45A (MCMJIG, 2010). The largest contributors to a ship’s ICCP demand are
the area of exposed hull and propellers as well as the ship’s speed (Huber and Wang, 2012). The
ROKS Cheonan definitely impinged the seabed or likewise underwater obstacle as evidenced by the
non-UWE related damage to its starboard screw, propeller shaft, and hull which could have led to
an ICCP spike or overflow (DeGiorgi et al., 1998). One possible scenario is that as the ROKS
Cheonan was maneuvering to free itself after having run aground off the coast of the nearby island,
an exposed LCM mine cable made contact or came in very close proximity to the ship’s hull enabling
its ICCP power to trigger detonation. Due to the high conductivity of seawater, even without full
contact between the exposed mine cable and the ship’s hull, it is possible for a sufficient current to
be conducted to trigger detonation of the electrically triggered naval mine ( see Fig. 13).
@seismicisolation
@seismicisolation
Explosion
a b
c d
Fig. 13. Experiment with electric current in salt water (3.5%). a, Preparation for experiment with battery (12V/7AH),
salt water, copper wire, midget lamps (4). b, Two midget lamps are not lighted when wire ends from battery and those
from lamp are in the plain water, no lamps are turned on (no electric current through the electric circuit. c, Three midget
lamps are lighted on when the ends of wires from battery and those from lamps are even more distant in the salt water
(3.5%). d, Measurement of voltage (5V) and current (150mA) in this electric circuit using four lighted lamps. This is
the same principle in which current flow from a ship’s ICCP spike can electrically power an exposed electrical naval
mine (LCM) cable in sea water resulting in detonation.
The approximately 136kg NEW attributable to this specific type of naval mine (ROK LCM)
(MCMJIG, 2010) along with a detonation depth of 8m, maximum bubble radius of 6.63 m and a
seismic yield of 2.04 fits the seismic parameters set forth in the analyzed data. On the contrary, the
MCMJIG’s assertion that the UWE’s cause was a DPRK torpedo with a yield of 250 kg NEW,
detonated at a depth of 6-9 m, 3 m “left of the center of the gas turbine room” (equivalent to 3 m
portside of hull centerline) and producing a seismic magnitude of 1.5 – this terrestrial based
magnitude estimate is inappropriate because a UWE will always have a greater magnitude than a
land explosion (Booth, 2009; Kim and Gitterman 2012a and 2012b) – simply does not correlate with
the seismic parameters of the observed and analyzed data.
CONCLUDING REMARKS
Our findings differ greatly from those of the MCMJIG (2010) which found the cause of the sinking
to be a DPRK “CHT-02D” torpedo with 250 kg NEW exploding at a depth of 6-9 m producing a
magnitude of 1.5. Our calculations based on analyzing various yields and depths using forensic
@seismicisolation
@seismicisolation
seismology and the Rayleigh-Willis equation, BEM, and 3D bubble shape simulations found that an
explosion of 136 kg NEW (magnitude 2.04) at a depth of 8 m and 5 m portside would produce an
equivalent bubble pulse period to the one in the observed seismic data. 136 kg NEW is equivalent
to the net explosive weight of LCM naval mines which were deployed in the late 1970’s and
discarded. From our analyses and observations the detonation was most likely caused by the mine
cables being dragged out along with the mine during the ROKS Cheonan navigational problems
which allowed the current flow of the ship’s Impressed Current Cathodic Protection (ICCP) to
trigger mine detonation.
ACKNOWLEDGEMENTS
I gratefully acknowledge Yefim Gitterman of Geophysical Institute of Israel and Mark Williams
and Anne Trehu of Oregon State University for helping to analyze the underwater waveforms. I
am much obliged to Dr. Olivier Hyvernaud (Laboratoire de Géophysique, Tahiti, French
Polynesia) for the help of Cepstral Analysis in spite of difficulties finding a bubble pulse for a very
shallow underwater explosion. I also owe many thanks to Aman Zhang, Harbin Engineering
University, China who calculated the boundary element method in this study and Dong Hyoung
Kim of East Seoul College for conducting seawater electrical current flow experiment (ICCP)
along with providing first hand in-depth information about the heavily modified MK-6 mines
(antisubmarine bombs) aka ROK naval land control mines (LCM) with which he directly
participated in installing on the seabed in July-October 1977 (from communication with D. H.
Kim).
REFERENCES
Barger JE, Hamblen WR (1980). The air gun impulsive underwater transducer. J. Acoustic. Soc.
Am. 68 (4), 1038-1046.
Baumgardt DR, Der Z (1988). Identification of presumed shallow chemical explosions using land-
based regional arrays. Bull. Seism. Soc. Am. 88 (2), 581-595.
Booth DC (2009). The relation between seismic local magnitude ML and charge weight for UK
explosions, Keyworth, Nottingham, BGS, UK, Brithsh Geological Survey, Earth Hazards
Programme, OR/09/062, Open Report pp. 25.
Brocher TM (2003). Detonation charge size versus coda magnitude relations in California and
Nevada. Bull. Seism. Soc. Am. 93 (5), 2089-2105.
Cohen TJ (1970). Source-depth determination using spectral, pseudo-autocorrelation and cepstral
analysis. Geophys. J. R. astr. Soc. 20, 223-231.
Cole R H (1948). Underwater Explosions. Princeton, NJ, USA, Princeton Univ. Press, pp.437.
DeGiorgi VG, Thomas ED, Lucas KE (1998). Scale effects and verification of modeling of ship
cathodic protection systems. Eng. Anal. Bound. Elements. 22, 41-49.
Ditchfield RW McGrath JN Tighe-Ford DJ (1995). Theoretical validation of the physical scale
modeling of the electrical potential characteristics of marine impressed current cathodic
protection. J. Applied Electrochemistry. 25, 54-64.
Gong SW, Ohl, SW, Klaseboer, E (2010). Scaling law for bubbles induced by different external
sources: theoretical and experimental study. Physical Review. E 81 056317, 1 -11.
Gonzalez-Avila SR, Klaseboer E, Khoo BC, Ohl, C (2011). Cavitation bubble dynamics in a liquid
gap of Variable height. Journal of Fluid Mechanics. 682, 241-260.
Huber T, Wang Y (2012). Effect of propeller coating on cathodic protection current demand: sea
trial and modeling studies. Corrosion. 68, 441-448.
Jacob AWB, Neilson G (1977). Magnitude determination on LOWNET. Edinburgh, UK, Institute
of Geological Sciences, GSU, Report pp.86.
Jacob AWB (1975). Dispersed shots at optimum depth-an efficient seismic source for lithospheric
studies. J. Geophysics. 41, 63-71.
Khalturin VI, Rautian TG, Richards PG (199). The seismic signal strength of chemical explosions.
Bull. Seism. Soc. Am. 88 (6), 1511-1524.@seismicisolation
@seismicisolation
Explosion
Kim SG, Gitterman Y (2012a). Underwater explosion (UWE) analysis of the ROKS Cheonan
Incident. Pure and Appl. Geophys. 169, doi:10.1007/s00024-012-0554-9.
Kim SG, Gitterman Y (2012b). A source verification of the underwater explosion (UWE) for the
ROKS Cheonan accident, the 9th General Assembly of the ASC, September 17-20, 2012,
Ulaanbaatar, Mongolia, pp. 296.
Klaseboer E, Khoo BC, Hung KC (2005). Dynamics of an oscillating bubble near a floating
structure. J. of Fluids and Structure. 21, 395-412.
Klaseboer E, Hung KC, Wang C, Wang, CW, Khoo B C, Boyce P, Debono S, Charlier H
(2005).Experimental and numerical investigation of the dynamics of an underwater explosion
bubble near a resilient/rigid structure. Journal of Fluid Mechanics. 537, 387-413.
Krieger JR, Chahine GL (2005). Acoustic signals of underwater explosion near surfaces. J.
Acoustic. Soc. Am. 118 (5), 2961-2974.
Lake JR, Gibson DC (1987). Cavitation bubbles near boundaries. Annual Review of Fluid
Mechanics.19, 99-123.
Lee M., Klaseboer E, Khoo, BC (2007). On the boundary integral method for the rebounding bubble.
J. Fluid Mechanics. 570, 407-429.
Magnaudet J, Eames I (2000). The motion of high-Reynolds-number bubbles in inhomogeneous
Flows. Annual Review of Fluid Mechanics, 32, 659-708.
MCMJIG (Multinational Civilian-Military Joint Investigation Group) (2010). On the Attack Against
ROK Ship Cheonan. Seoul, ROK, Joint Investigation Report, Ministry of National Defense, pp.
313.
Plesset MS, Chapman RB (1971). Collapse of an initially spherical cavity in the neighbourhood of
solid boundary. J. of Fluid Mechanics. 47, 283-290.
Plesset MS, Prosperetti A. (2000). Bubble dynamics and cavitation. Annual Review of Fluid
Mechanics, 19, 145-185.
Rayleigh L (1917). On the pressure developed in a liquid during the collapse of a spherical cavity.
Philosophical Magazine Series 6, 34, 94-98.
Swisdak MM (1978). Explosion effects and properties: part III-explosion effects in water. Dahlgren,
VA, USA, NSWC/WOL, TR 76-116, pp.109.
Wang QX, Yeo KS, Khoo BC, Lam KY (1996). Nonlinear interaction between gas bubble and free
surface. Computer & Fluids, 25(7), 607-628.
Wang QX, Blake JR (2011). Non-spherical bubble dynamics in a compressible liquid. Part 2.
Acoustic Standing wave. J. of Fluid Mechanics. 679, 559-3581.
Weinstein MS (1968). Spectra of acoustic and seismic signals generated by underwater explosions
During Chase Experiment. J. of Geophysical Research, 73 (16), 5473-5476.
Willis DE (1963). Seismic measurements of large underwater shots. Bull. Seism. Soc. Am. 53 (4),
789- 809.
Wu J, Xing S, Yun F (2009). The influence of coating damage on ICCP cathodic protection Effect.
Simulation of Electrochemical Processes III, WIT Trans. on Eng. Sc. 65, 89-96.
Zhang YL, Yeo KS, Khoo BC, Wang C (2001). 3D jet impact and toroidal bubbles. J. Comput. Phys.
16(6), 336-360.
Zhang AM, Yao XI, Li J (2008). The interaction of an underwater explosion bubble and an elastic-
plastic structure. Applied Ocean Research. 30, 159-171.
@seismicisolation
@seismicisolation
So Gu Kim Currently Director of Korea Seismological Institute and his research has covered a very
wide orange of topics in seismology, marine geophysics and oceanography (underwater acoustics
and signal processing). He has more than 100 publications in the international journals as well as
more publications in the national journals including an author of four professional books. He was
an invited professor at Hamburg University (Germany), University of New England (Australia),
GFZ Potsdam (Germany), Hokkaido University (Japan) and Peking University (China) as well as
Professor of Physics and Geophysics at Hanyang University, Korea for 30 years. His research
highlights are in using his expertise to analyze the complicated problems which are involved in
Seismology (passive and active), Oceanography (physical. geophysical and hydroacoustics),
Applied Physics including Petroleum Exploration and Geotechnology.
@seismicisolation
@seismicisolation
CHAPTER NINETEEN
REPLY TO COMMENT ON “UNDERWATER EXPLOSION (UWE) ANALYSIS

OF THE ROKS CHEONAN INCIDENT” BY K. S. KIM1
SO GU KIM AND YEFIM GITTERMAN
A mini-submarine at Hatfield Marine Science Center, OSU, Newport, Oregon USA. Overall
length = 4.1 m, Test Depth (safe depth)=106.7 m, weight = 1814.4 Kg, normal dive duration = 1-4
hours. By courtesy of Maureen Collson, Hatfield Marine Center, OSU. It is exhibited on account
of a concept of the Test Depth for a submarine. Any submarines cannot submerge in the Yellow
Sea because the average depth of the Yellow Sea is about 50 m in which normal submarines
cannot dive freely. The ROKS Cheonan sinking was due to an underwater explosion by one of the
abandoned land control mine (LCM).
On choosing two net explosive weights (NEW) (250 kg and 136 kg) to primarily focus on: 250- and
136-kg net explosive weights—the case of a DPRK torpedo and the case of an ROK LCM—were
researched by the MCMJIG (2010) with the conclusion that the UWE source had to be a 250-kg
NEW according to the evidence. During the course of our research we have been led to have
serious doubts on the authenticity and veracity of this evidence provided by the MCMJIG
(including chemical residue and the torpedo submitted as proof). Coupled with the often
overlooked or ignored fact that at least two antisubmarine warfare (ASW) warships—the ROKS
Cheonan (PCC-772) itself and the ROKS Sokcho (PCC-778); the PCC class series 761–785 were
specifically designed for ASW)—were in the area at the time of the UWE incident and both failed
1
Originally published in Pure Appl. Geophys (2013), 170 (3), 479-484, Use of UWE formulas are discussed and
updated. @seismicisolation
@seismicisolation
400 Chapter Nineteen
to detect either a DPRK sub or a torpedo travelling through the water, this makes it extremely
unlikely that a torpedo fired from a DPRK sub is the culprit. Another important fact is that there is
ample evidence of the Cheonan having serious navigational problems (i.e., partially running
aground) leading up to the incident. Now there is sufficient evidence that there was some sort of
underwater explosion (UWE) from seismo-acoustic analyses. The only two plausible UWE source
possibilities are a mine or a torpedo. Of course it could also theoretically be a hybrid, i.e., a
torpedo launched from a mine. However, if it is not a DPRK launched torpedo, then it has to be a
mine. A friendly- fire torpedo has also been raised as a possibility, but the ROK-US Joint Exercise
(Key Resolve/Foal Eagle) was already finished (between March 8 and March 18, 2010) and it was
impossible for a torpedo to reach that area because the joint military exercise was 170 km away
from the site even if it was there, also that involves many more players and would be much more
difficult to cover up, especially in a very wired country such as the ROK. Our premise is that the
most likely source of this UWE is a ROK littoral ‘‘land control mine’’ (LCM) naval mine. This
premise is the basis of our hypothesis, but of course it cannot be definitively concluded due to the
lack of access and transparency regarding this incident. Various net explosive weight (NEW)
variables were hypothetically tested in this paper, with 250 kg being pointed out for being the
NEW asserted by MCMJIG and 136 kg being pointed out for being the NEW for the ROK LCM
naval mine. In the case of the 136-kg NEW of the ROK LCM, the variables fall within the
allowable parameters and we are asserting this as the most likely UWE source. In the case of the
250-kg NEW DPRK torpedo, the variables fail to fall within the parameter limits.
On the question of the pertinent equation to use: there seems to be some confusion regarding the
equations. The commentator stated that we had to use Eq. (3) of the SWISDAK (1978) report.
However, the third equation is a ‘‘correction equation’’ that is not valid in our case because the
bubble size is small relative to the total water and is not closer than about 10 bubble radii to either
the surface or the bottom. The underwater occurred near the surface. The commentator misused
formulas provided by Swisdak (1978) which are only valid for relatively deep underwater
explosions compared to a bubble size, i.e. to depths and charge weight such that the bubble is
closer than about 10 bubble radii to either the surface or the bottom. For cases where either the
surface, the bottom, or both begin to influence the bubble, the correction equations should be used.
K = T Z5/6/W1/3 (1)
J = Rmax(Z/W)1/3 (2)
0:651φ(y)W1/3=DZ1/3K2 - K + TZ5/6/W1/3 = 0 (3)
K bubble period coefficient (Swisdak used 2.11; we used KT = 2.1), T first bubble period (s), Z
hydrostatic pressure (d + 10.1), W charge weight (kg), J bubble radius coefficient (SWISDAK
used 3.50; we used KR = 3.38), Rmax maximum bubble radius (m), d charge depth (m), in correction
equation (3), D total water depth, y = d/D, φ(y) function related to the bottom characteristics and
φ(y) vs. y is presented in Fig. 10 from SWISDAK’s report (Swisdak, 1978). In particular, it should
be noted that the correction equation (3) is void to apply it to most of the underwater explosions as
shown on Fig. 17 in Chapter Fourteen of this book.
@seismicisolation
@seismicisolation
Reply to Comment on “Underwater Explosion (UWE) Analysis of the ROKS Cheonan Incident” 401
by K. S. Kim
Figure 1: Nomograph of the first bubble pulse period and maximum bubble radius of an underwater gas bubble (TNT)
from SWISDAK (1978) The blue dotted line, red dotted line and thick black lines indicate the bubble pulse period,
detonation depth and charge weigh of (8.0 m, 0.99 s, 136 kg), (15 m, 0.99 s, 250 kg) and ( 500 m, 0.054 s, 100 kg).
To check the influence of the surface the correction factor, Eq. (3) should be used to estimate K
prior to determining the bubble pulse period. The commentator’s Figs. 1 and 2 are derived from a
misapplication of the correction Eqs. (3) and (4) to estimate the bubble period and must have
misused a bubble pulse period of 0.81 s which is shown in Cole’s Fig. 8.21 without considering
the depth presentation of (d + 33) plot. The commentator seems to have based his figures on
Cole’s Fig. 8.21 which is very close to our result of a bubble period (0.988 s) taking into account
the plot of bubble period with [(d + 33 (10.1 m)] adding 33 feet (10. 1 m) to the charge depth. Our
modified Rayleigh–Willis equation based on the pressure developed in a liquid during the collapse
of a spherical cavity (d ≥ Rmax) (KIM and GITTERMAN 2012b; GONG et al. 2010) has more reasonable
results, even if we ignore surface effects compared to BEM and 3D bubble shape simulations.
Even theoretical periods including surface effects lead to non-definitive conclusions due to
extensive use of approximations in the procedure and the assumption of unknown amounts of
energy dissipated via shock waves.
@seismicisolation
@seismicisolation
Figure 2: Distribution of possible bubble pulse periods by spectral analysis from observed data (star), Cole’s method
(TR), BEM (TB), 3D simulation (T3, T5) for 136-kg net explosive weight at a depth of 8 m, at 3 m and 5 m portside of
the hull centerline, and for 250-kg net explosive weight at a depth of 9 m, at 3 m and 5 m portside of the hull
centerline
SWISDAK (1978) stated that Eqs. (1) and (2) in his paper (Eq. 1 in this study) should be used only to
free water explosions, to depths and weights such that the bubble is not closer than about 10
bubble radii to either the surface or the bottom, whereas the correction equation Eqs. (3) and (4)
are used for case where the detonation depth is closer than about 10 bubble radii to either surface,
the bottom or both begin to influence the bubble. Eq. (3) is not valid in our study because our
study belongs to a free water explosion. The updated study of KRIEGER and CHAHINE (2005) recently
noted that UWE data are unavailable below approximately the normalized depth C (standoff
distance/maximum bubble radius) = 0.5—where standoff distance is the detonation depth—as the
spark-generated bubbles vent their contents to the atmosphere in the region, lose their intensity and
fail to emit bubble pulses. More detailed studies for the ratio of detonation depth to maximum
bubble radius have been carried out to investigate the distorted bubble by proximity to the free
surface, the bottom, or to a solid object using recent experimental and empirical data (KLASEBOER et
al. 2005; ZHANG et al. 2008). KLASEBOER et al. (2005) also studied the boundary condition at free
surface using scaled depth C in the range of 0.5 ˂ C ˂ 3.0 and found that generally in 0.5 ˂ C ˂ 1.5
the bubble ‘‘works’’ and for C ˂ 1.5 it is nonlinear but that as the maximum bubble radius
becomes greater than detonation depth in the range of 0.5 ˂ C ˂ 1.0, the top bubble rises above the
equilibrium free surface level but the surface is not breached—it maintains its integrity and
exhibits behavior resembling an elastic membrane and a water layer remains between the bubble
and the free surface. Near C = 1, we can see the jet impact and toroidal bubble formation during
the first bubble collapse (KRIEGER and CHAHINE 2005). Our C value clearly satisfies the earlier
studies of KRIEGER and CHAHINE (2005) and KLASEBOER et al. (2005) as well as SWISDAK (1978).
Therefore, our results are correct.
On Commentator’s Figs. 1 and 2 vis-a`-vis bubble pulse periods: misapplication of the correction
formula (3) resulted in the commentator’s Fig. 1 having low values which do not fit the Rayleigh-
Willis equation (BARGER and HAMBLEN 1980), SWISDAK (1978) equations or his nomograph.
Likewise for the same reason the commentator’s Fig. 2 is off. From the commentator’s Fig. 2, 320
kg TNT at a depth of 8 m produces the bubble pulse period of 0.99 s which is inconsistent with the
SWISDAK (1978) study (nomograph of Fig. 1 in this reply). The thick line indicates the detonation of
@seismicisolation
@seismicisolation
by K. S. Kim
100 kg NEW at a depth of 500 m with a bubble pulse period of 0.054 s and a maximum bubble
radius of 2.033 m (SWISDAK 1978). The thin blue line represents the detonation of 136 kg TNT at a
depth of 8 m with a bubble pulse period of 0.99 s and a maximum bubble radius of 6.63 m in our
paper (KIM and GITTERMAN 2012a). The bubble pulse period of 0.99 s and 250 kg TNT produces the
maximum bubble radius of 8 m and a detonation depth of 15 m (dark-red dashed line) in our paper
(KIM and GITTERMAN 2012a, b).
Through the bubble pulse period of 0.990 s obtained via spectral analysis we can determine the
approximate net explosive weight and explosion depth by narrowing down the possible parameters
along with supplementary estimations of the bubble pulse periods of 0.967 s via Rayleigh–Willis
equation, 0.976 s via BEM and 1.030 s via 3D bubble shape simulation derived for the case of a
136-kg net explosive weight detonation (Fig. 2 in this reply).
@seismicisolation
@seismicisolation
Figure 3: The behavior of 3D bubble shape simulation taking into account the interaction between the bubble and the
ship’s hull based on the BEM method. a. (1) bottom view (L = 88 m) (2) side view (D = 2.9 m) (3) front view (W = 10
m). b. Bubble shape formation near ship’s hull: t = 0.000, 0.089, 0.500, 0.791, 0.932, 0.947, 1.007, 1.030, 1.122 s
with 136-kg net explosive weight detonation at a depth of 8 and at 5 m portside of the hull centerline
On free surface and hull effects: the Rayleigh– Willis equation for a bubble sphere for free surface
ignores its deformation, buoyancy and the nonlinearity of the boundaries while the BEM takes the
aforementioned factors into account, but may not factor in the interaction between the bubble and
complex boundaries such as a ship’s hull. On the other hand, 3D bubble shape simulations (not
included in the referenced paper) based on the BEM can take into account the interaction effect of
the ship’s hull on the motion of the bubble to estimate the UWE’s bubble pulse period. We can see
the jet impact and toroidal bubble formation as the bubble shape changes in 3D simulations (Fig. 3
below).
The post-detonation bubble expands and achieves its maximum volume at t = 0.500 s, and the
bubble impinges the hull before starting its collapsing phase. At the same time, a jet impact
penetrates the bubble at t = 0.932 s and again at t = 0.947 s, generating a toroidal bubble as shown
in Fig. 3b. At t = 1.030 s the bubble achieves its minimum volume and ends its first pulsing period
which is larger than observations from spectral analysis of seismic waves (see Figs 9 and 10 in
Chapter 14 for a bubble pulse phase and an accurate bubble pulse frequency of the underwater
explosion for the ROKS Cheonan Sinking).
On accuracy of BEM: the accuracy of the forward BEM solution depends on the degree of
boundary discretization, i.e., number of panels representing the bubble shape. GUMEROV and
CHAHINE (2000) found the accuracy of the bubble shapes increasing with the number of panels
using the boundary element code 2DynaFSTM, and KRIEGER and CHAHINE (2005) also used the
bubble pulse period Eq. (1) to estimate charge yield and depth of an explosion in conjunction with
3DynaFATM BEM code, as well as including 3D simulations. The bubble pulse periods using BEM
have been intensively studied by other researchers (KLASEBOER et al. 2005; ZHANG et al. 2008).
The bubble pulse periods from the Rayleigh– Willis equation, BEM and 3D simulations are given
in Fig. 2. 3D simulations of the bubble from a 136-kg net explosive weight detonation at a depth of
8 m and 5 m portside of the hull centerline (KIM and GITTERMAN 2012b)—rather than the ‘‘official’’
3 m portside of the hull centerline—is most consistent with the observed bubble pulse period.
On TNT being asserted as the explosive: we have never asserted TNT as being the explosive.
‘‘TNT’’ referred to in this paper has always been in the context of ‘‘TNT equivalent yield’’,
otherwise known as net explosive weight (NEW). However, the ROK LCM naval mine is
basically a MK 6 moored contact mine heavily modified to be an electrically triggered bottom
mine, which originally contained 300 lb (136 kg) net explosive weight in TNT.
On questions regarding the exact detonation location on the hull: the exact detonation site on the
ship’s hull will be very difficult to ascertain, especially with the strict restrictions on further
investigation. Even with access it may never be confirmed. The official MCMJIG report states that
the detonation site was 3 m portside of the centerline of the gas turbine room, which is equivalent
to 3 m portside of the hull centerline. If a ROK LCM was the UWE source the detonation site
might have been closer to 5 m portside of the hull centerline. When a full reinvestigation is
launched we may get some definitive answers, but until then it seems we will only be ascertaining
the feasibility and likelihood of hypotheses. However, we are sure that the cause of the ROKS
Cheonan Sinking was due to neither a running aground nor a collision with a submarine and a
torpedo attack by a submarine based on the scientific investigation. It must have been associated
with a ROK LCM.
REFERENCES
BARGER, J. E. and HAMBLEN, W. R. (1980). The air gun impulsive underwater transducer, J.
@seismicisolation
@seismicisolation
by K. S. Kim
Acoustic. Soc. Am. 88 (4), 1038-1046.
COLE, R. H. (1948), Underwater Explosions, Princeton Univ. Princeton, New Jersey.
GONG, S. W.,OHL, S. W. and KLASEBOER, E. (2010), Scaling law for bubbles induced by different
external sources: theoretical and experimental study, Physical Review E81, 056317-1-056317-
11.
GUMEROV, N. A. and CHAHINE, G. L. (2000), An inverse method for the acoustic detection, location,
localization and determination of the shape evolution of a bubble, Inverse Problem, 16, 1741-
1760.
KIM, S. G. and GITTERMAN, Y. (2012a), Underwater explosion (UWE) analysis of the ROKS
Cheonan incident, Pure Appl. Geophys., doi:10.1007/s00024-012-0554-9.
KIM, S. G. and GITTERMAN, Y. (2012b), A source verification of the underwater explosion (UWE)
for the ROKS Cheonan accident, the 9th General Assembly of the ASC, September 17-20, 2012,
Ulaanbaatar, Mongolia, 296.
KLASEBOER, E., KHOO, B. C. and HUNG, K. C. (2005), Dynamics of an oscillating bubble near a
floating structure, J. of Fluids and Structure, 21, 395-412.
KRIEGER, J, R, and CHAHINE, G. L. (2005), Acoustic signals of underwater explosion near surfaces,
J. Acoustic. Soc. Am. 118 (5), 2961-2974.
MCMJIG (Multinational Civilian-Military Joint Investigation Group) (2010), Joint Investigation
Report: On the Attack against ROKS Cheonan, Ministry of National Defense, Seoul, 313 pp.
SWISDAK, M. M. (1978), Explosion effects and properties: part II— explosion effects in water,
NSWC/WOL TR 76-116, 22 Feb 1978.
@seismicisolation
@seismicisolation
CHAPTER TWENTY
DEPTH COMPUTATION AND SOURCE CHARACTERISTICS OF DPRK’S

2016-01-06, 2016-09-09 AND 2017-09-03 NUCLEAR TESTS 1
SO GU KIM , YEFIM GITTERMAN, SEOUNG-KYU LEE AND SANG-MO KOH
DPRK’s six nuclear explosions recorded (normalized) on GERES Array, BGR, Germany by
curtesy of L. Ceranna and on FINES, University of Helsinki, Finland by courtesy of T. Tiira
ABSTRACT
The Democratic People’s Republic of Korea (DPRK aka North Korea) conducted underground
nuclear explosions on January 6, 2016 (2016J), September 9, 2016 (2016S) and September 3, 2017
(2017) since the 2006, 2009 and 2013 nuclear tests. The principal objective of this paper is to
present depth determination and source characteristics for the North Korean nuclear tests using
body wave and surface wave spectra. The source depths for the 2016J, 2016S and 2017 nuclear
tests were estimated at 2.11 km, 1.99 km and 1.96 km, respectively using spectral nulls of P- and
S- wave and Rg-wave spectra. It should be noted that our over-burial determination is significantly
greater than expected from the standard experiment practice of nuclear nations. In particular we
found the Rg-wave spectral nulls in North Korean nuclear tests to appear due to a most general
source type of pure dip-slip motion with about 45°dip. We assume that the North Korean nuclear
test source may be an inverted conical volume dip-dip slip motion which could be a vertically
distributed source in vertically contained shafts through mutually connected-deep tunnels. The
over-burial determination would affect the interpretation of MS-mb discriminants and the
estimation of the seismic yield as well.
Keywords: CTBT, depth phase, spectral null, Rayleigh wave, Mt. Mantap
1
Some parts presented in Horizons in Earth Science Research, Vol. 18 (2018), 120-160. eds. by B. Veress and J.
Seigethy, NOVA, Science Publishers, Inc. New York
@seismicisolation
@seismicisolation
Depth Computation and Source Characteristics of DPRK’s 2016-01-06, 2016-09-09 and 2017-09-03 407
Nuclear Tests
INTRODUCTION
Source depth is pertinent for ensuring enforcement of the Comprehensive Nuclear-Test-Ban-
Treaty (CTBT) as nuclear explosions are unlikely to be fired at depths of more than a few
kilometers (Bakun and Johnson, 1973; 2001; Bonner et al., 1986; Bowers et al., 2001). Two basic
techniques of spectra using body waves and surface waves are utilized for depth calculation. Depth
phases such as pP (sP) and pPn (sPn) are used to compute a source depth. Surface wave spectra
can be used to determine depths, as mode excitation reflects the behavior of the eigenfunctions
with depth. The primary limitation on this study, however, arises from lateral heterogeneity of
velocity structure and mechanism. The spectral nulls by pP + P and sP + P are applied to estimate
depths by the time separation between the direct P-wave arrivals and near-source surface
reflections. These studies can be limited by a partial trade-off between depth and source time
function duration, but the trade-off with source time function duration is found not to generally
produce major difficulties for small events like nuclear explosions (Stein and Wiens, 1986)
especially North Korean nuclear explosions within the rigid granite source. The DPRK conducted
the first nuclear test (mb 4.3) on October 9, 2006, the second one (mb 4.7) on May 25, 2009 and the
third one (mb 5.1) on February 12, 2013, the fourth one (mb 5.1) on January 6, 2016, the fifth one
(mb 5.3) on September 9, 2016 and sixth one (mb 6.3) on September 3, 2017 beneath the vicinity of
Mt. Mantap in the Punggye-ri region of northeastern DPRK (Table 1). The burial depth is also
determined through the measured time differences between the refracted P-wave (Pn) from Moho
discontinuity and the reflected pPn and/or sPn depth phases from the free surface. To provide
robustness of the result is to consider teleseismic and regional arrays at a wide range of better
azimuthal distribution. For a simple structural model, simple formulas are developed to estimate
the source depth from identified spectral nulls.
The strength of a seismic signal and the body-wave magnitude (mb) depends strongly on the
medium in which the explosion is detonated. Explosions in hard rock, like granite or tuff give
considerably stronger amplitudes than explosions in unconsolidated rock such as alluvium.
Seismic signal strength depends not only on seismic yield and the medium but also on the wave
transmission properties of the Earth. These properties vary considerably for different source-
receiver paths. The range of the yield would be a factor of about 10. Taking this uncertainty into
account, we correlate magnitude 5.0 roughly with 100 kt in dry unconsolidated rock and with 10 kt
for wet hard roc (Harjes et al., 1985). These crude relations can be considered as indicating an
upper and lower yield limit for a given magnitude. Furthermore, various researchers derived
different magnitude-yield relations depending on the detonation site (Table 1). For example, using
mb = 3.92 + 0.81 logY for wet hard rock (Murphy, 1981), it resulted in a 29 kt yield for the 2013
nuclear test while it alternatively estimated at a 14 kt yield when using mb = 4.25 + b logY for a
fully coupled explosion where Y is the explosive yield in kilotons (kt), with b = 0.75 for Y 1,
and b = 1.0 for Y < 1 which is suitable for conservatively estimating the yield threshold of a
potential violation of the CTBT in the Novaya Zemlya region (Bowers et al., 2001). However
Ringdal et al. (1992) used a larger constant of 4.45 to determine the seismic yield of the Soviet
underground nuclear explosions at the Shagan River test site, resulting in 292.86 kt which is much
less than 867.50 kt (Murchy, 1981) and 505.48 kt for the 2017 test (Bowers et al., 2001).
Seismic signals from an explosion can also be reduced by detonating an explosion in a large
underground cavity due to decoupling effects (Harjes et al., 1985; Murphy et al., 2017). Since
emplacement conditions are unknown for North Korean nuclear explosions, it is most difficult to
estimate seismic yields of underground nuclear explosions conducted surreptitiously. The seismic
yield, depth of burial and medium around the emplaced nuclear device including configuration
play significant roles in verifying the observed seismic signals of underground nuclear explosions.
@seismicisolation
@seismicisolation
408 Chapter Twenty
Table 1. Source parameters of North Korean nuclear tests. Origin times, locations and
magnitudes were extracted from the United States Geological Survey (USGS) and IDC
(International Data Center, CTBTO), Seismic yield from Murphy (1981) and Bowers et al.
(2001) in the brackets, respectively. Depth in the brackets refer to determination from Kim
et al. (2015), Kim et al. (2016) and Kim et al. (2017) studies, respectively.
Date Origin time1 Location2 mb Yield (kt) Depth
Event Lat.º N, Lon. º E USGS/IDC (km)
10/09/2006 01:35:28 41.294, 129.094 4.3/4.1 1.67 (1.17) 0 (2.12)
2006 41.3119,129.0189
05/25/2009 00:54:43 41.303, 129.037 4.7/4.5 9.18 (3.98) 0 (2.06)
2009 41.3110,129.0464
02/12/2013 02:57:51 41.299, 129.004 5.1/4.9 28.63 (13.59) 0 (2.05)
2013 41.3005,129.0652
01/06/2016 01:31:01 41.326, 129.010 5.1/4.8 28.63 (13.59) 0 (2.11)
2016J 41.3039,129.0481
09/09/2016 00:30:01 41.323, 128.987 5.3/5.1 50.55 (25.12) 0 (1.99)
2016S 41.2992,129.0491
09/03/2017 03:30:01 41.333, 129.056 6.3 /6.1 867.50 (505.48) 1.04 (1.96)
2017 41.3296,129.0267 6.43
1
UTC (Universal Time Coordinated). 2The upper and lower locations determined by USGS and IDC/CTBTO,
respectively. 3 BGR, Germany estimate. 4 GEOFON, GFZ Germany estimate.
DEPTH ESTIMATION USING DELAY TIMES OF DEPTH PHASE

The usual way to estimate the depth of an event is to measure the time difference between the
arrival time of the direct P wave and that of the surface reflections pP or sP using teleseismic
waves. pP and sP are signals that travel upward from the focus as P and S waves (Douglas et al.,
1972; Massé, 1981; Okal, 1992) and then travel steeply downwards following the same path as to
the receiving stations of the P waves. The relative amplitudes and polarities of these phases depend
on the source mechanism while the time delay depends on the velocity structure and the ray paths.
For shallow sources, identification of individual arrival times is difficult in the time domain
because the arrival pulses interfere with each other. However, the destructive interference has a
strong effect on the amplitude spectra of seismic arrays that may be exploited to estimate the
depth.
To provide waveforms for spectral analysis, we used stacked P-wave signals of the vertical
components from global teleseismic seismic arrays (Figure 1a). Such stacking increases the S/N of
the P-wave signals by suppressing the effects of local noise. We performed spectral analyses
taking 10 or 30 second time-windows of short-period vertical components for the ASAR (Alice
Springs, Australia), FINES (FINESS, Finland), KURK (Kurchatov, Kazakhstan), NOA and
NORESS (Norway), NVAR (Mina, USA), PDAR (Pinedale, USA) including South Korean local
stations and KSRS (Korea Seismological Research Station, Wonju, South Korea), USRK
(Ussuriysk, Russia) and broadband records for WRA (Warramunga, Australia) and YKA
(Yellowknife, Canada), ISN (Israel Seismic Network) and DSN (Dongbei Seismic Network,
China) along the Border between northeastern China and North Korea (Figure 1a). The effects of
absorption and the recording system can be removed by utilizing a source in the same high
velocity layer and the same array stations. The spectral nulls (minima) from the destructive
interference of pP + P and sP + P spectra are equivalent to a reciprocal of pP-P and sP-P lag time
so that the pP delay times estimated from the spectral analysis are compatible with the pP phase
arrivals in the time domain. The high quality factor Q in the granite source rock (1000) from
amplitude A(t) = A (0)exp(-ʌIW4 and low attenuation of source in the high velocity layer clearly
reveal high frequency arrivals of pP and sP in the teleseismic seismograms of the seismic arrays.
However, as the depth phase sP or sPn is very sensitive to the above-source structure and source
@seismicisolation
@seismicisolation
Nuclear Tests
configuration, use of sP or sPn often results in inaccurate depths for underground nuclear tests.
However we found the use of sP or sPn to be good enough in North Korean nuclear tests because
the North Korean underground nuclear explosions generate considerable shear motions compared
with other nuclear tests.
a b
Figure 1. a) Closed squares represent global arrays of the ASAR, FINES, ISN (PRNI), KURK, NOA, NVAR, PDAR,
WRA and YKA in better azimuthal coverage and regional arrays KSRS and USRK including MDJ. Closed triangles
represent U of China and South Korean stations used in this study. The red star indicates the underground nuclear test
site. b) Ray trajectories of P, pP and sP at a source of the nuclear test. i and j are take-off angles for pP and sP ray-
paths. G, G', R and R' indicate the contact points on the surface and on the ray-path by pP and sP. S and O represent
seismic stations and the center of the Earth, respectively using the flat Earth model.
The interference produces spectral modulation (scalloping) – minima (spectral nulls) at

fundamental and multiple frequencies (Bakun and Johnson, 1973; Cohen, 1970; Kulhanek, 1971;
Husebye et al., 2013; Heyburn et al., 2013). The spectral null f01 of the fundamental mode is
represented as
f01 = Vp/2cos (i)*h, f0n = n f01,
n = 1, 2, … (1)
where Vp is an average compressional wave velocity of the medium between the source and the
surface (overburden velocity), h and i are source depth and take-off angle. We identified spectral
nulls of the stacked spectra from the respective array traces to enhance seismic signals in noise
(Figure 2).
In case of a small take-off angle of pP for a long range epicentral distance (ο ൒ 75.0°), we may
approximate equation (2) as h ൎ pP-P*Vp/2 which was utilized by many researchers (Douglas et
al., 1972; Kulhanek, 1971; Gitterman et al., 2015). Ideally, the pP-P time would be obtained from
the time difference between the onset of P and that of the downswing depth phase of pP in the time
domain. However, those onsets are difficult to pick so an alternative is to use spectral nulls of the
average spectra because the depth phases are always 180° out of phase with respect to the
incoming P-wave arrivals. i.e., we measured the time difference between the maximum positive
amplitude P-wave arrival and the maximum negative amplitude of pP assuming P and pP have an
identical shape with negative polarity and interference between the two pulses (Douglas et al.,
1987). The absence of depth phase is the result of scattering and defocusing of P waves by lateral
variation of medium above the source (Douglas et al., 1990). However we can minimize the
nonlinear topographic effects of the mountainous source and the azimuthal variations of pP
waveforms (Burdick et al., 1984) using the average spectra obtained from a broad and uniformly
@seismicisolation
@seismicisolation
410 Chapter Twenty
azimuthal distribution of teleseismic arrays (Figure 1a). We used spectral nulls of the average
spectra of multi-pathing arrays and utilized the full depth formula with take-off angles assuming
the average overburden P-wave velocity to be Vp = 5.1 km/sec (Burdick et al., 1984’ Bonner et al.,
2008) at the nuclear test site. All spectra of single array show similar ‘spectral nulls’ indicating that
the psedudo-1D radiation in one direction is constant and nothing about the contribution of
different source components. So data from a test in the middle of a mountain will be even
lookalike in all directions since an onset time of P-wave arrivals at each element of an array
follows Fermat Principle and Snell’s Law that a raypath is along a minimum of travel time and that
an incident angle (take-off angle) is always equal to a reflected angle of a single raypath within a
limited time window. It should be noted that there may be all but no surface elevation difference
between a source and a reflector with a shallow dip as the closest approximation possible on an
average in the present state of evidence. Consequently we do not count on the unknown and
complicated terrain effects of the Mt Mantap and neglected the trade-off between our applied
implications and the claimed accuracy of the results from the nonlinear topographic effects
considering the four sides’ azimuthal coverage by means of the global seismic arrays from the
source.
From the relationship between ray paths and travel times, W[ = sin (i)/v0 = 1/vz for flat earth,
where i = a take-off angle for the reflected pP wave, v0 = P-wave velocity at the focal depth, vz =
the bottoming average Vp velocity (vav) where the ray returns at i = 90° where 1/Vz = 1/vapp =
߲ܶ/߲ο (travel-time curve slope). Using the take-off angle i, where i = arcsin (v0/vav), the time
delay for pP-P from the spectral null (a reciprocal of spectral null), and P-wave velocity at the
focal depth, v0 as 5.1 km/sec, we can determine depth from equation (2). Lay (1991) stated that the
source depth estimated by pP-P delay times may be overestimated by biasing effects on depth
phase propagation for very short lag times with limited bandwidth due to the slapdown phase at
the surface in case of the shallow nuclear tests. However, we offset these effects by using the
average spectra of stacked signals of the deep North Korean nuclear tests from different azimuthal
propagation of seismic arrays and take-off angles. Using pP-P in Figure 1a, the source depth h:
h = (pP-P)Vp/2cos (i) (2)
The estimated depth using sP-P delay times is derived using Snell’s law as follows (Figure 1b):
h = (sP-P)VpVs/[Vp cos (j) + Vs cos (i)] (3)
where j is the take-off angle of sP calculated from arcsin [Vs/Vp sin (i)].
We performed spectral analyses using short-period vertical components from the ASAR, FINES,
NOA, NVAR, PDAR and broadband vertical components from the WRA and YKA. We used not
only the pP + P depth phases, but also sP + P phases in this study.
Source depth can be estimated for a simple single-layer over a halfspace model from the time
delay between Pn and pPn using the following equation. The source depth by means of utilizing
pP-Pn can be derived considering the take-off angle for the Moho discontinuity with its velocity
Vn and the overburden average velocity Vp as follows (see Fig. S1 in APPENDIX A of Chapter
Twenty-one) :
h = (pPn-Pn)Vp/2cos(Į) (4)
Using sPn-Pn delay times, the depth can be derived as follows:
h = (sPn-Pn)/F (Vp, Vs, Vn, Į ȕ) (5)
where F(Vp, Vs, Vn, Į ȕ = 9S9QFRVĮ + 9V9QFRVȕ – 9S9VFRVĮVLQ ȕ –
9S9VFRVȕVLQ Į 9S9V9QFRVĮFRV
@seismicisolation
@seismicisolation
Nuclear Tests
where Į and ȕ are take-off angles for pPn and sPn phases. Į = arcsin (Vp/Vn) and ȕ = arcsin
[Vs/Vp sin (Į)] according to Snell’s law. We used the average spectra from spectral nulls for
destructive interference pPn + Pn since the ray paths should pass through the heterogeneous and
anisotropic crustal structures to reach each station of the array. Using equation (4), Vp = 5.1
km/sec, Vs = 2.9 km/sec (Bonner et al., 2008; Chung, 1995; Kim and Bae, 2006; Phillips et al.,
2007) and Vn = 7.8 km/sec from the refracted P-wave velocity of ray paths through the central part
of the Chugaryeong Rift Zone and the subcontinental crust of the East Sea (Chung. 1995; Kim and
Bae, 2006) whereas Vn = 8.0 km for USRK array in the Sikhote Alin region (Phillips et al., 2007;
Rodnikov et al., 2008).
DEPTH INTERPRETATION BY DELAY TIMES OF DEPTH PHASE
The spectral nulls are used to determine depths taking the average spectra with the band-pass
filtered between 0.5 - 3.0 Hz or 0.5 - 5.0 Hz. Here the only spectra of WRA are presented as a
representative teleseismic array for the spectral analysis.
The depth phase method is most successful when P-wave arrival has a signal-to-noise ratio (SNR)
greater than 4-7 and the depth phase pP exhibits a SNR (signal-to-noise ratio) greater than ~2
(Bonner at al., 2002). The most of nuclear tests were valid for these conditions, but the 2006 test
was hampered by weak signals and noise (Figure 2). It is often not certain if the difference in
spectral nulls is due to intrinsic source differences or because of its path within the limited time
window for the near-field. Nonetheless, the delay times of the teleseismic arrays are very sensitive
to the propagation path effects of the depth phase pP. The nonlinear topographic effects are
assumed not to be included in this study since a surface elevation of the source is almost equal to
that of each reflector from the source taking into account a raypath in 1D direction from the source
in the middle of the mountain which is obtained from the teleseismic arrays distributed in
uniformly azimuthal coverage.
The apparent delay times of pP-P are often significantly greater than the predicted from the
explosion depth and the P-wave velocity in the material above the shot point (Douglas and
Hudson, 1990) because the corresponding pP arrivals would be expected to be reduced in
amplitude and possibly delayed relatively to what would be expected for true pP in case of the
damage effect zone above an explosion. The reflected signals containing the fracture and crushing
associated with the explosion would be also reduced and delayed resulting in the slapdown phase
as a secondary source (Lay, 1991). However these phenomena of depth delay times would be
excluded in most of North Korean nuclear tests at the deep source rock of granite except for large
ones like 2016S and 2017 nuclear tests, but not associated with the surface spallation, possibly
subsurface slapdown phase in the different location. We estimated the source depth at 2.11 km for
the 2016J nuclear test using spectral nulls of pP+P and sP+P in Figure 2 and Table 2. The low
frequency of spectral nulls at the ASAR (Table 2) may be more closely related to the low velocity
layer of the large aquifer beneath Alice Springs area in the upper crust and possibly including the
low P-wave velocity layer in the upper mantle of the Late Palaeozoic Alice Springs Orogeny
beneath the array according to the tomography study (Van Der Hilst et al., 1988). We also
estimated the source depth at 1.99 km for the 2016S test using depth phases showing Figure 3 and
Table 3. The source depth of 1.92 km for the 2017 test is shallower than that of the 2016S test
resulting in many sP phases implying the more shear motions which may be related to the source
damage near the source by a slapdown. We also slapdown phases for the 2017 test after 1.75 s
(0.57 Hz) from the onset at ARCES and EKA in Figure 4 and Table 4. We estimated a depth at
1.92 km for the 2017 test despite lack of depth phases of pP because of considerable damage and
fractures near the source the large yield of the thermos-nuclear explosion (M 6.3 and 0.9 Mt)
@seismicisolation
@seismicisolation
412 Chapter Twenty
Figure 2. ARCES, ASAR, EKA, KURK, NVAR, PDAR, WRA and YKA teleseiamic arrays were used to determine
the depth of of the fourth North Korean nuclear test on January 6, 2016. The source depth is estimated by pP-P delay
times from the destructive interference (pP + P) of P and the surfece reflectd P wave (pP).
@seismicisolation
@seismicisolation
Nuclear Tests
Table 2. The source depth of the 2016J nuclear test was determined via pP-P delay times
with take-off angles using 10 seismic arrays. Distance and AZ (direction to the station) are
measured in degrees. The asterisk (*) indicates the representative station PRNI of the Israel
Seismic Network.
Array Names (distance, AZ) Spectral Null Take-off Angle Source Depth
(°) (Hz) (°) (km)
ASAR (64.8, 175.1) 1.1 17.57 2.43
ARCES(56.4, 335.8) 1.25 18.91 2.16
EKA (75.5, 334.1) 1.35 15.56 1.96
ISN (73.5, 295.6)* 1.28 16.13 2.07
KURK (35.6, 302.8) 1.22 22.86 2.27
NORESS(66.2,331.9) 1.25 17.15 2.14
NVAR (79.7, 47.5) 1.25 (0.98) 14.62 (8.25) 2.11 (1.92)
PDAR (81.0, 39.5) 1.25 (1.02) 14.33 (8.09) 2.11 (1.85)
WRA (61.1, 174.3) 1.25 18.40 2.15
YKA (64.7, 27.4) 1.25 17.57 2.14
Average depth, km 2.11±0.15
Table 3. The source depth of the 2016S nuclear test was determined via pP-P delay times
with take-off angles using 8 seismic arrays. Distance and AZ (direction to the station) are
measured in degrees.
Array Names (distance, AZ) Spectral Null Take-off Angle Source Depth
(°) (Hz) (°) (km)
ASAR (64.8, 175.1) 1.1 17.57 2.43
ARCES(56.4, 335.8) 1.48 18.91 1.82
EKA (75.5, 334.1) 1.37 (0.98) 15.56 (8.77) 1.93 (1.93)
FINES(60.3, 327.4) 1.56 (0.94) 16.13 (9.09) 1.70 (2.01)
NVAR (79.8, 47.4) (0.98) 14.62 (8.25) (1.92)
PDAR (81.0, 39.5) 1.26 (0.92) 14.33 (8.09) 2.09 (2.03)
WRA (61.1, 174.3) 1.26 18.40 2.13
YKA (64.7, 27.4) 1.40 17.57 1.91
The numbers in the brackets refer to spectral nulls and depths via sP-P delay times including a take-off angle of sP
phase.
@seismicisolation
@seismicisolation
414 Chapter Twenty
Figure 3. The depth-phase travel times for the 2016S nuclear test are obtained via spectral nulls of the average spectra
of pP showing various spectral nulls. The numbers in the brackets refer to sP-P delay times.
@seismicisolation
@seismicisolation
Nuclear Tests
Figure 4. Red and blue arrows indicate spectral nulls of pP + P and sP + P using teleseismic arrays for the 2017 test.
The spectral nulls for the 2017 test present low frequencies and do not appear at some arrays like ASAR and NVAR.
The slapdown phases appear at 0.57 Hz (after 1.75 seconds from the onset) from the spectra of ARCES and EKA
records. The vertical black, red, blue and brown bars in ARCES indicate P-wave, pP, sP and slapdown phase arrivals.
Table 4. The source depth of the 2017 nuclear test was determined via pP-P delay times with
take-off angles using 8 seismic arrays. Distance and AZ (direction to the station) are
Array names (distance, AZ) spectral null take-off angle source depth
(¨°, AZ°) (Hz) (°) (km)
ASAR (64.8, 175.2) none 17.57 none
ARCES (56.4, 335.8) 1.47 (0.95) 18.91 (10.62) 1.83 (2.01)
EKA (75.5, 334.1) 1.40 15.56 1.89
FINES (60.3, 327.4) 1.57 (0.97) 18.62 (10.46) 1.71 (1.95)
NORESS (66.2, 332.0) 1.35 17.14 1.98
PDAR (81.0, 39.6) 1.33 14.33 1.98
WRA (61.2, 174.4) 1.43 18.40 1.88
YKA (64.7, 27.4) 1.24 17.57 2.05
Average depth, km 1.92± 0.10
Note: The numbers in the brackets refer to spectral nulls and depths via sP-P delay times including a take-off angle of
sP phase. ¨° = 1° = 111.19 km
Note: The numbers in the brackets refer to sP-P delay times. NORESS is a sub-array of NOA.
@seismicisolation
@seismicisolation
416 Chapter Twenty
From a known firing depth of 0.463 km in granite and a P-wave velocity of 4.8 km/sec, they found
it impossible to see the predicted delay times of 0.2 sec from pP-P, because the pP-P delay times
were found to be 0.8 sec (corresponding to a depth at about 2 km) from the Pile Driver explosion
(Kulhanek, 1971; Douglas and Rivers, 1988) since it was conducted at a shallow depth (around
500 m) of granite resulting in damage of fracture and crushing from a spalling. However in this
study, it should be noted that North Korean nuclear explosions were conducted in the deep
homogeneous granite without the surface spallation (Douglas and Hudson, 1990). There is a good
example of a deep-depth estimate using the depth phase pP-P for a deep atomic explosion (Bolt,
1976). On May 21, 1968 a 47-kiloton (mb = 5.6) atomic explosion (detonated in salt) was fired at a
depth of 2.45 km near Bukhara, USSR to seal an oil well. It was observed by pP-P delay times of
1.7 seconds at YKA (79°) indicating that source depth was estimated at about 2.6 km which was in
very good agreement with the actual depth. For another example Douglas et al. (1987) estimated at
1791 m of a depth below the free surface (HOB5= –1700 m) (Johnston, 2005) corresponding to the
observed spectral nulls of 0.94 Hz (1.06 s) by pP+P spectra using single channels at GBA, EKA
and YKA stations from the Cannikin explosion of the Amchitka Island explosions (mb=6.8,
yield=5Mt) on November 6, 1071. The low spectral null of 0.94 Hz or a longer pP-P delay time
(1.06 s) also attests to overestimation by an overshoot of pP pulses (Douglas et al., 1987). It is due
to the slapdown phase associated with near surface spallation reducing P-wave velocity in the
crushed rock from the explosion in the island site.
The data are best fit by a depth since the depth determinations by body waves are robust to
uncertainties to other parameters (Stein and Wiens, 1986; Douglas and Hudson, 1990). Figures 5a
and 5b represent synthetic seismograms and spectra of the vertical components for the near-field
with little inelastic attenuation showing pPn-Pn and for the far-field including effects of anelastic
attenuation with showing pP-P (Burdick et al., 1984; Van Der Hillst, 1998). The earth structure
models for synthetics are used by Korea Model (Kim and Bae, 2006; Hermann, 2016) in the near-
field and AK135 (Kennett, 1995; Montagner and Kennett, 1996) in the far-field for the teleseismic
arrays taking a short span of the first P-wave arrivals for 5-6 seconds. The preferred pP-P and pPn-
Pn delay times are in very close agreement with observed results from Figure 5. The synthetic
seismograms and spectra are produced using Hudson96 and Herrmann programs (Herrmann,
2016). The synthetic spectral nulls of short-period P waves for the near-field and those of long-
period P waves for the far-field are considerably consistent with the observed spectral nulls
(Burdick et al., 1984; Douglas and Hudson, 1990).
We also found that spectral nulls of the spectra for synthetic seismograms are well consistent with
those of observations in Figure 5. The data are best fit by a depth near 2.0 km since the depth
determinations by spectral nulls of body waves are robust to uncertainties to other parameters
(Stein and Wiens, 1986; Douglas and Rivers, 1988). The earth structure models for synthetics used
Korea Model (Kennett et al., 1995) for the local stations and AK135 for the teleseismic arrays. The
effect of anelastic attenuation using t* which is the ratio of travel time and a specific quality factor
Q is assumed to be independent of frequency. Several researchers (Burdick et al., 1984; Douglas
and Rivers, 1988) found that the short-period P wave synthetic seismograms showing two troughs
include the effect of anelastic attenuation computed using t* whereas no effects of anelastic
attenuation are included in the long-period P wave synthetic seismograms showing a trough in in
Figure 5. The synthetic seismograms and spectra are produced using Hudson96 and Herrmann
programs (Tsai and Aki, 1971). The spectral nulls of P-wave spectra due to the destructive
interference by the direct and reflected P waves from the surface are reasonably consistent with
those of our observations for the near-field and the far-field.
________
5
HOB = Height of Burst
@seismicisolation
@seismicisolation
Nuclear Tests
a
Figure 5. Synthetic seismograms and spectra of the vertical component for regional and teleseismic data to account for
a) a spectral null (minimum) due to the destructive interference at 1.75 Hz showing pPn + Pn for the near-field (441
km) and b) at a spectral null at 1.25 Hz showing pP + P for the far-field (distance 81°) in case of a depth at 2.0 km
assuming the flat Earth model.
DEPTH ESTIMATION USING SPECTRAL NULLS OF RAYLEIGH WAVES

The notches (spectral nulls) of the amplitude spectra of fundamental-mode Rayleigh waves
corroborate the estimated depths extrapolated via body waves. Rayleigh wave excitation is
sensitive to source depth, especially at intermediate and short periods due to the approximate
exponential decay of Rayleigh wave displacements with depth. The frequency-dependent “spectral
null” in the Rayleigh wave amplitude spectra are most pronounced for the pure (vertical) strike-
slip types and the reverse faulting mechanisms with about 45° dip. The spectral null of the
fundamental-mode Rayleigh wave spectrum is dependent on source mechanism, depth and source-
receiver azimuth but does not vary much@seismicisolation
with the shot medium and the shot yield (Tsai and Aki,
@seismicisolation
418 Chapter Twenty
1971; Douglas et al., 1972; Langston, 1980; Okal, 1992; Fox et al., 2012; Heyburn et al., 2013).
The conical dip-slip reverse faulting may accompany a deep-seated tensile failure occurring at a
depth above shot point (Massé, 1981; Patton and Taylor, 2008) generating the strong Rg wave
radiation from a vertically oriented CLVD 2 source (Rodgers et al., 2010). We observed spectral
nulls at 0.138 Hz for DSN (Dongbei Seismic Network) and 0.13 Hz at BJT, INCN, MDJ and SEO
for the 2006 test (Figures 6a and 6b). No spectral nulls for the 2009 test at DSN may be related to
the different azimuth from the source to stations and lack of Rg-wave generation in the very
shorter distance less than 200 km in the range 150-350 km. We also observed spectral nulls at
0.14 Hz at BJT and INCN, but 0.17 Hz at SEO and no spectral nulls at MDJ (Figure 6b) for the
2017 test. The spectral nulls may be related to a reverse faulting dipping at about 45° dip which is
consistent with findings from other researchers (Tsai and Aki, 1971; Douglas et al., 1972;
Langston, 1980; Okal, 1992; Fox et al., 2012; Heyburn et al., 2013) whereas no clear spectral nulls
were found for non-pure strike-slip faulting motions and the oblique reverse fault mechanisms
(Heyburn et al., 2013). The source mechanism for the North Korean nuclear explosions may be
assumed to be the conical dip-slip volume accompanying a reverse faulting motion as a vertically-
distributed source in a shaft. We found the Rg-wave spectral nulls varying depending upon
frequency and azimuth of the nodal line (strike) of mechanism (Fox et al., 2012; Heyburn et al.,
2013). A spectral null (minima) is caused by an excitation null for short-period fundamental-mode
Rayleigh waves, termed Rg waves as a CLVD (compensated linear vector dipole) source which
correlates with normal mode theory in the form of resonant frequency (Massé, 1981; Heyburn et
al., 2013; Fox et al., 2012; Langston, 1980). Bonner et al. (2003) have examined the performance
of MS scales on 7-sec Rayleigh waves recorded distances less than 500 km from Nevada Test Site.
The spectral null (minimum) from the CLVD depth (9DYU\þXN and Kim, 2014) correlates with
source depths derived from pP-P and pPn-Pn delay times. The conical dip-slip reverse faulting
may accompany a deep-seated tensile failure occurring at a depth above shot point (Massé, 1981;
Patton and Taylor, 2008) generating the strong Rg wave radiation from a vertically oriented CLVD
source (Cho et al., 2016; Udias, 1999) which is a kind of a spallation-like source function, but it is
not associated with the surface spall because it was conducted at granite in the deep tunnels. The
estimated depth via the CLVD model also disambiguates the estimated source depth by body
waves.
The conical dip-slip faulting may accompany a deep-seated tensile failure occurring at a depth
above the shot point as a CLVD source generating Rg waves. Rg-waves resulting in the distinctive
interference by P an SV waves reflected from the surface is induced from the tensile failure from a
vertically oriented CLVD source showing spectral nulls of Rg wave spectra. There are more
elliptical motions of Rg and the ‘retrograde‘ reverses ‘prograde’ at a depth of d = 0.192
wavelength which represents a spectral null (hole) at the spectra. Therefore, a source depth for the
2006 test is estimated at 2.09 km (h = 0.192 * 7.4627 s * 1.46 km/sec = 2.09 km) using spectral
nulls at 0.134 Hz from Figures 6a and 6b. We also estimated source depth for the 2017 test at 2.00
km using 0.14 Hz using BJT and INCN (h = 0.192 *1/0.14* 1.46 = 2.00 km). The little higher
spectral null of 0.17 Hz at SEO may be due to an azimuth change from a source to the station and
Rg wave velocity reduction through the sub-continental path of the Chgaryeong Rift Zone of the
central part of the Korean Peninsula. We need further study in order to verify these phenomena of
Rg-wave spectra for the North Korean nuclear explosions utilizing more high-quality data. Rg
wave spectral nulls tightly constrain the source depth when spectral nulls often occur for a limited
range of azimuth because this method of depth estimation is dependent on the orientation of the
mechanism and the station locations. Rg-wave velocity of 1.46 km/sec is obtained from Rayleigh
wave equation assuming a source rock to be a Poisson solid (Udias, 1999). This source mechanism
of CLVD source could be assumed to be a pure dip-slip (reverse-faulting) and the absence of
2
There are source changes in volume (isotropic), shear fracture (double-couple) and sudden changes in rigidity (shear
modulus) in moment tensor inversion. The sudden@seismicisolation
change in shear modulus is called a CLVD source.
@seismicisolation
Nuclear Tests
spectral nulls or slight change is due to the change of azimuth from the source to the stations
through the complicated crustal structures in the shorter distance in this region.
a
Figure 6. a). The fundamental-mode Rayleigh waves and displacement spectra for the BHZ band-pass
filtered (0.02-0.1 Hz) with amplitude in counts. a) A spectral null at 0.138 Hz was found at Dongbei
Seismic Network (DSN) data only for the 2006 but no spectral null for the 2009 test. b). A spectral null
(minima) at 0.13 Hz was observed for the 2006 test at BJT, INCN MDJ and SEO whereas spectral nulls at
0.14 Hz at BJT and INCN and 0.17 Hz at SEO, but no nulls at MDJ except for the 2006 test indicating that
the 2006 test is located in the quite different place from other tests.
The synthetic seismograms were produced based on Green function and the modified AK135
Model (Kennett et al., 1995; Montagner and Kennett, 1996) taking into account the average depth
of 2 km calculated from Rg-wave spectra and a general nuclear explosion mechanism of a reverse
faulting with 45°dip, i.e. depth = 2 km, rake = 90° and, dip = 45° but strike varies with every 15°
using a narrow band-pass filter between 0.02 and 0.1 Hz for in Figure 7. Various researchers
(9DYU\þXN and Kim, 2014; Cho et al., 2016; Ford et al., 2009; Shin et al., 2010; Barth, 2014) have
found different source mechanisms from moment tensor inversion with some uncertainties for the
North Korean nuclear explosions. In particular there are some trade-offs between spectral nulls
and moment tensor inversion for underground nuclear explosions owing to DC and CLVD
components in North Korean nuclear tests (Kim et al., 2015; 9DYU\þXN and Kim, 2014; Cho et al.,
2016; Shin et al., 2010). We found that a theoretical spectral null of 0.14 Hz at a depth of around 2
km is in good agreement with observations in Figure 7.
@seismicisolation
@seismicisolation
420 Chapter Twenty
Figure 7. The synthetic seismograms and spectra of Rg-waves are band-pass filtered between 0.02-0.1 Hz to obtain the
fundamental-mode based on the Korea model considering a general mechanism for an explosion with dip 45° and rake
90° varying with every 15° of strike to 90°. The higher amplitude at an azimuth of 90° attributes the large radiation
pattern of Rayleigh wave and it decreases again up to 180°. We found the most fitting depth (2 km) at 0.14-0.15 Hz
from an azimuth of 30°which corresponds to the tectonic stress (ENE) in this region (9DYU\þXN and Kim, 2014).
APPLICATION OF RASPBERRY SHAKE SEISMOGRAPHS

It is very interesting and worthwhile that we have found the source depth and characterization for
the 6th nuclear test of DPRK on September 3, 2017. We used 14 global RS (Raspberry Shake)
stations and two RS local stations which consist of the vertical components (Figure 8).
@seismicisolation
@seismicisolation
Nuclear Tests
a
Figure 8. a) Raspberry Shake stations for 10 teleseismic stations including 2 local stations in Korea. INCN, MDJ,
USRK and KSRS are also used in this study. b) Response and phase curves for a Raspberry Shake seismograph.
The Raspberry Shake Personal Seismograph integrates a vertical velocity sensor, the digitizer, the
hyper damper and the computer into a single box. The seismograph has a single-component 4.5 Hz
geophone with sampling rate of 50 Hz and a digitizer with 24-bit ADC and 144 dB. The amplitude
of frequencies between 0.8 and 23 Hz are flat and fit to use spectral analysis in Figures 8a and 8b.
In the R8086, R9F1B, R8633 and R869F @seismicisolation
records, there is a shoulder in the second downswing, but
@seismicisolation
422 Chapter Twenty
in the R4F51 and R9475 records there are double peaks in the second upswing which can be all
attributed to a slapdown phase of this particular nuclear explosion. The source site and size of the
sixth nuclear explosion of DPRK so called a thermonuclear are very similar to the Milrow nuclear
explosion of the United States which was conducted in the Amchitka Island on October 2, 1969 at
a depth of 1220 m, with 1000 kt (M 6.4). We also observed some shoulders in the second upswing
or downswing phases which are producing pP in the many teleseismic seismograms in the Milrow
explosion (Bakun and Johnson, 1973; Burdick et al., 1984) which attest to the slapdown phases
associated with near surface spallation from the strong thermonuclear explosion which was also
shown in the sixth North Korean nuclear explosion. A slapdown is a complex process of inelastic
and nonlinear interactions, involving swarm of rock falling events in an extended time interval. It
is also noted that the spectral nulls in the Australian continent are lower than other observations
due to the lower velocity layer of the upper mantle at the sites (Van Der Hilst et al., 1998).
The local stations in Figure 9, there is strong variation in velocity-sensitive parameters such as the
ray paths and crustal structures with different phases resulting in incongruity with the estimates of
pPn-Pn and sPn-Pn delay times. However, the source depths are more reliably estimated using a
local seismic array which is located in the stable region with high Q in the Korean Peninsula and
can enhance a good SNR by super positioning of signals, resulting in strong average spectral nulls
of regional P-wave spectra in KSRS array (Wonju, Republic of Korea) with a low Moho
discontinuity velocity of Pn (7.8 km/sec) (Kim and Bae, 2006; Phillips et al., 2007; Kim et al.,
2016) whereas a high Moho discontinuity velocity of Pn (8.0 km/sec) in USRK (Ussuriysk,
Russia) (Phillips et al., 2007). Consequently the irregularly dipping mountainous topographic
effects do not significantly influence the depth estimation.
It is also worthwhile to use the Raspberry Shake stations to estimate source depth for the 2017 test.
We estimated the depth at 2.0 km using regional (pPn and sPn) and teleseismic depth phases (pP
and sP) for the 2017 test. In general, the spectral nulls for the 2017 test are higher (retarding depth
phase delay times) compared to the previous nuclear tests indicating that there may have been a
spallation at thr subsurface. If this is true, the source depth for the 2017 test must be less than 2 km
depending upon the reduction rate of the source rock velocity. It is important to note that we can
find some slap down phases at 1.75 s after detonation in the time domain of Gobongsan (R720B)
and INCN stations (Figures 9 and 10). The slapdown phases also appear at the second upswing for
R4F51, but at the second downswing for R8086, R9F1B, R8633 and R869F records including
double peaks at R869F and R9475 teleseismic records of the Raspberry Shake seismographs which
are attributed to the inelastic and nonlinear process. In general, the stations in UK and Europe
show less noise signals than those in USA and Australia which may be located in the urban
environments. It is important that we have to consider a velocity reduction for the surface-reflected
P waves because of the slapdown. King et al. (1974) found that the apparent average overburden
velocities are approximately 15% lower than the velocities from on-site measurements from the
Longshot, Milrow and Cannikin Nuclear Explosions. The propagation of reflected waves from the
free surface could be slowing down even at the greater distance because of increase of the inelastic
volume radius (Springer, 1974). However we can see the slapdown phase at the near-field stations
such as Gobongsan and INCN because the the radius of inelastic region is increased at the hard
rock source like granite (Springer, 1974). Therefoere, the observed burial depth for the 2017 test of
North Korea could be less than the expecting depth (greater than 2 km) taking into account a
velocity reduction in the inelastic zone owing to the slapdown. The behavior of the spectral null of
depth implication does not appear to vary with azimuth in a smooth way, which is most easily
explained in terms of lateral variations near the receivers. The spectral null which varies rapidly
with azimuth from the source cannot, however, be ruled out owing to a largely variation of the
receiver sites such as ASAR and FINES. Therefore it is important to use the average spectral null
of the teleseismic arrays which are located in a uniformly azimuthal distribution on the globe.
@seismicisolation
@seismicisolation
Nuclear Tests
It should be also noted that sP phases can be explained in terms of the non-isotropic radiation of
the North Korean nuclear tests from the tectonic stress by the nearby source zone (9DYU\þXN and
Kim, 2014) rather than shear dislocation faulting. The tensile faulting associated within the
opening or closing of a crack is described by nonzero ISO and considerable components of DC
and CLVD resulting in SH and Love waves by asymmetry of the explosive source for the North
Korean nuclear tests (9DYU\þXN and Kim, 2014; Cho et al., 2016). The deviatoric components of
DC and CLVD evidence that the non-isotropic radiation is related to tectonic stress in the
surrounding rock rather than shear faulting triggered along preexisting nearby fault structures. It is
worthwhile to note that Zhao et al. (2017) recently determined surface wave magnitudes (Ms)
using Rayleigh waves well generated from five nuclear tests of North Korea including the 2016S.
Dreger (2017) found DC and CLVD components for the 2017 nuclear test to be 34% and 24%
from waveform modeling mechanism using some regional long-period records including 34
teleseismic P-wave polarity observations. The deviatoric stress caused probably asymmetric shape
of the cavity during the detonation resulting in an asymmetric radiation for the shear wave
generation.
Table 5. Spectral nulls and depth using pPn-Pn and sPn-Pn delay times. There are
unidentified events (brown arrows) after 1.75 seconds from the onset of P-wave arrivals
which may be related to the slapdown phase of a thermonuclear explosion on September 3,
2017 (Figures 9 and 10).
Stations ǻ$= pPn + Pn / sPn + Pn/ Comments
code (°) depth (Hz/km) depth (Hz/ km)
R720B 4.31/208.3 1.71/1.97 1.18/2.02 Slapdown and/ or
collapse event
R3930 4.13/203.6 1.74/1.94 1.07/2.20
MDJ 3.30/6.9 1.67/1.97 1.07/2.20
INCN 4.29/206.6 1.76/1.91 1.17/2.04 Slapdown and.or
Collapse event
KSRS 3.94/193.3 1.72/1.96 1.19/2.00
USRK 3.61/36.1 1.58/2.09
Average 1.97±0.06 2.09±0.09 Mean DOB = 2.0
depth, km
@seismicisolation
@seismicisolation
424 Chapter Twenty
Figure 9. Spectral nulls using pPn + Pn and sPn + Pn using local stations for the sixth North Korean nuclear test in the
upper two rows. The spectral holes between the pPn + Pn and sPn + Pn may be due to pP*+ P* (refracted from the
Conrad Discontinuity). Spectral analyses of the North Korean nuclear explosion on September 3, 2017 using
Raspberry Shake stations in UK and Europe and USA and Australia in the lower last row. The slapdown phases
appear at 0.57 Hz (after 1.75 seconds from the onset in the local stations (Gobongsan and INCN) as well as teleseismic
Raspberry Shake stations’ records in the second downswing or upswings.
@seismicisolation
@seismicisolation
Nuclear Tests
Table 6. Spectral nulls and depths using the global Raspberry Shake seismographs for the
sixth nuclear explosion of DPRK. There are some shoulders at the second downswing or
upswing phases implying that a source has a slapdown (spall-closure) phase at 0.57 Hz (after
1.75 s from the onset of the explosion).
Raspberry ǻAZ pP + P/ depth sP + P/ depth Take-off Comments
Shake (°) (Hz)/(km) (Hz)/(km) angle (i, j) Slapdown and/or
codes (°) Collapse event
R4F51 77.11/ 32.4 1.19/2.22 15.19 2nd upswing
slapdown
R8086 76.50/331.5 1.25/2.11 15.29 2nd downswing
slapdown
R9F1B 76.74/331.2 1.38/1.92 1.08/1.75 15.25, 2nd downswing
8.60 slapdown
R8633 76.04, 326.6 1.19/2.22 15.37 2nd downswing
slapdown
R898B 72.94/319.8 1.25/2.12 16.13 3rd upswing
slapdown
RB7A9 79.85/48.4 1.31/2.01 14.65
R6EE9 84.45/50.7 1.36/1.94 0.99/1.90 13.35,
7.54
RCCE6 74.63/162.2 1.17/2.30 15.87
RB69F 77.49/161.3 1.09/2.42 15.13 2nd double peak
slapdown
R9475 74.62/162.2 1.17/2.27 0.99/1.91 15.87, 2nd double peak
8.95 slapdown
Average 2.15± 0.15 1.85± 0.07 mean DOB = 2.0
depth, km
@seismicisolation
@seismicisolation
426 Chapter Twenty
Figure 10. The analogue seismograms of the North Korean nuclear test on September 3, 2017 recorded on the
Raspberry Shake seismograph at Gobongsan (R720B: 37.6936°N, 126.7852°E) which is 448.6 km away from the site
We see strong P-wave arrivals (esp. Pg) and de minimus S-wave arrivals (esp. Sg) of short duration, while surface
wave (Rg-wave) duration is also short compared to the equivalent size of an earthquake. Maximum amplitudes are
clipped because of the over - scale of analogue records. The body wave magnitudes are determined as 6.3 for the
nuclear test and 4.1 for the non-tectonic event after about 8.5 minutes. The seismic yield estimated at 867.5 kt
(Murphy, 1981) and 541.2 kt (Bowers et al., 2001).
DISCUSSION AND RESULTS

The firing depth of nuclear explosions may be associated with the optimum depth to ensure the
containment of fallout exposure on the surface. In the case of underground nuclear explosions, the
scaled depth of burial or burst (DOB) should not be less than 122Y1/3 (Glasstone and Dolan (1977)
where depth is measured in meters and yield in kilotons (kt). Therefore a depth of about 2000 m
would allow containment of an underground nuclear explosive event having a yield of up 4.4
megatons which may be equivalent to a yield of a hydrogen bomb explosion. However, this scaled
depth concept does not fit the North Korean nuclear tests in the light of the observed magnitude.
Therefore it seems sure that the over-burial depths of the North Korean nuclear tests were made
beyond the economic cost and the ordinary tunneling technology.
@seismicisolation
@seismicisolation
Nuclear Tests
a b
Figure 11-a-b. a) Geological map of the nuclear test site and surrounding area which is reproduced from the
Geological Society of Korea (1987). 1. The Mesozoic intrusive granite rock in Kwanmo Massif, 2. Quaternary
volcanic rocks in Kwanmo Massif, 3. Kilju-Myongchon Graben (Cenozoic sedimentary rocks), 4. Machollyong super
group (Proterozoic rock), 5. Rangnim Massif (Proterozoic-Archaeozoic), 6. Triassic intrusives, 7. Hayang group
(Cretaceous), 8. Daedong super group (Jurassic-Triassic), 9. Sangwon system (Proterozoic-Archeozoic). The star
indicates the site of the nuclear detonation. b) Satellite image of Mt. Mantap (2205 m) and surrounding area near the
test sites was produced by French Satellite ‘Spot-5’ on May 14, 2009 before the second test implying that tunnels may
have been built before tests. We presume that three portals may be related to the entrances of old metallic mines (Mo)
as shown in Figure 10c. The scale is 1:25,000.
Three white spots at the bottom are the tunnel entrances AKA West, South and East portals
(Pabian and Hecker, 2012; Israelsson, 2016), but they were built before the 2009 and 2013 nuclear
tests implying pre-existed portals (Figures 11a and 11b). These portals may have been used to
explore the mineral deposits of molybdenum in Figure 11c. Consequently the possibility using pre-
existed mining tunnels could be possible to easily excavate deep tunnels for underground nuclear
explosions in this area. In the light of the excavation of deep tunnels, a few piled spoils appeared
in front of portals because most of the spoil must have been removed into the old mine tunnels or
somewhere else (see Figure 11c). Very recently Pabian and Coblentz (2017) described rather a
large piled spoil in front of west and north portals implying that a large nuclear explosion is
impending. However, the piled spoil from the satellite image is actually nothing to do with the true
source locations as well as the seismic yield because the exact source location is undisclosed.
@seismicisolation
@seismicisolation
428 Chapter Twenty
Figure 11-c. c) Geological structure and mineralized deposits near nuclear sites (Science Academy Press, 1964). 1
Archean rocks 2. Proterozoic rocks 3. Upper Proterozoic rocks 4. Upper Paleozoic rocks 5. Late Mesozoic Formation
6. Lower Proterozoic intrusive rocks 7. Upper Paleozoic-Lower Mesozoic intrusive rocks 8. Mesozoic intrusive rocks
(granite) 9. Late Mesozoic to Early Tertiary intrusive rocks 10. Mesozoic to Tertiary extrusive rocks 11. Metallogenic
boundary 12. Structural metallogenic boundary 13. Deposit boundary 14. Structural line 15. Fe mineralized belt 16.
Cu-Ni mineralized belt 17. W-Mo mineralized belt 18. Molybdenum 19. Au-Cu-Co mineralized belt 20. Au
mineralized belt 21. Cu mineralized belt 22. Pb-Zn mineralized belt 23. Fe-sulfide mineralized belt 24. Dolomite-skarn
mineralized belt 25. Graphite mineralized belt. NTS is the nuclear test site of DPRK.
Despite need of a large cost and an advance tunneling technology, North Korea conducted nuclear
tests by means of the over-burial detonation which is very unlikely for certain keen seismologists
to believe. There are some controversies regarding the increase of delay times of depth phase due
to the slapdwon phase associated with near surface spallation in the Longshot, Milrow, and
Cannikin nuclear explosions in the sedimentary rock (tuff) of the Amchitka Island (Bakun and
Johnson, 1974; King et al., 1974; Lay, 1991). However we cannot expect the slapdown phase
resulting from spallation of the Earth’s surface near ground zero from the North Korean nuclear
tests because the source rock is a homogeneous rigid granite and the cavity and source radii are
located at the over-burial depth whose stress wave energy is too deep to reach the surface
spallation. Nonetheless it should be noted that the large one like the 2017 test may be associated
with the subsurface spallation generating collapse events in the different location. We presumed
that it did not significantly influence the delay times of depth phase from the major crushed and
disturbed volume of material to some elastic radii at the surface spallation since the North Korean
underground nuclear explosions were conducted at the rigid source rock of granite.
@seismicisolation
@seismicisolation
Nuclear Tests
It seems that North Korea did not take into account cost and manpower for deep tunnels which
were not problematic in North Korea. Furthermore it cannot be ruled out that there may have been
a possibility using pre-existing mining tunnels related to the mineral deposits (Mo) at the test site
in Figure 11c. It seems that the distributions of nuclear test sites are close together as a cluster
except for the isolated tests of 2006 and 2017 tests according to IDC/CTBTO determination which
is reasonably appropriate in this study. As a result, we infer that the source radii including the
inelastic volume for the North Korean tests are much less than 500 m (King, et al., 1974). In
particular a cavity radius including the inelastic zone for the 2017 test should be larger than any
other tests because of its larger yield and magnitude. The cavity radius of the 2017 test was found
to be 56.8 m using the calculated parameters from this study (Closmann, 1969; Springer, 1974).
Figure 12. Source locations of North Korean nuclear explosions determined by various researchers showing elevations
and depths (negative) in km. The prefixes of I, M, PH, ZW and TYW represent the works from several researchers
(Israelsson 2016; Murphy et al., 2013; Pabian and Hecker, 2012; Zhang and Wen, 2013; Tian, Yao and Wen, 2018).
Dark red and yellow symbols before year indicate source locations estimated by International Data Center
(IDC)/CTBTO and National Earthquake Information Center (NEIC), United States Geological Survey (USGS) for the
event year.
CONCLUSION
The source depths for the 2006, 2009, 2013, 2016J, 2016S and 2017 nuclear tests were estimated
at 2.12 km, 2.06 km, 2.05 km, 2.11 km, 1.99 km and 1.96 km, respectively (Figure 12) using
spectral nulls of depth phases by teleseismic and regional seismic arrays which are located in the
uniformly azimuthal coverage on the globe including of Rg-wave spectral nulls.
Our over-burial depth determination of the North Korean nuclear tests is reasonably consistent
with the previous studies (Gitterman et al., 2015; Kim et al., 2015; Kim et al., 2016; Kim et al.,
@seismicisolation
@seismicisolation
430 Chapter Twenty
2017) even if it disagrees with other studies based on satellite image and the relative method using
the 2006 test as the master event including the scaled depth (Zhao et al., 2017, Murphy et al.,
2013; Zhang and Wen, 2013; Rougier et al., 2011). The relative method may result in some
problems unless the source location of the master event is evident compared to the absolute
method. We did not use ostensible satellite data and unclear scaled depth relation which cannot be
applied for North Korean nuclear tests because North Korea is assumed not to have considered
cost and manpower to excavate deep tunnels. Notwithstanding those complicated stories, it should
be clearly noted in this study that DPRK conducted nuclear tests at the over-burials at around 2 km
beneath the vicinity of Mt. Mantap with an elevation of 2205 m implying possibilities using the
pre-existing mining (Mo) tunnels at the test site resulting in some difficulties to determine the
accurate seismic yield as well as the executability of a preemptive strike. The radionuclide
observations were detected for the only first 2006 test, possibly the 2013 test even though 59 days
later which was skeptical. It was also inferred from Rg-wave wave generation and spectral nulls
that possibly the nuclear tests may have been conducted as vertically-distributed sources (Viecelli,
1973) in the shaft through mutually connected-deep tunnels well generating Rayleigh waves (Rg
waves) from a spectral null of a CLVD source (Langston, 1980; Patton and Taylor, 2008; Rodgers
et al., 2010) including SH and Love waves induced from tectonic stress (Nuttli, 1969; Toksoz and
Kehrer, 1971) and/or new cracks in addition to Rayleigh waves
ACKNOWLEDGMENTS
We wish to thank Robert Herrmann for reviewing this manuscript and offering very constructive
criticism. We would like to thank Paul Richards for his providing Dongbei Seismic Network
(DSN) data and Rick Benson (IRIS) and David Jepson (Geoscience Australia) for the teleseismic
data. We also appreciate Albert Brouwer as well as the Executive Secretary of CTBTO, Lassina
Zerbo and Director of IDC, CTBTO for their collaboration to collect IDC data in this study. One
of author (SGK) would greatly appreciate CTBTO’s funding to participate in SnT2015, SnT2017
and SNT2019, CTBTO, Vienna, Austria. We acknowledge Director of ISC (International
Seismological Centre), Dmitry Storchak for his providing phase reading data and KMA (Korea
Meteorological Administration) for waveform data release, including Alex Leon (Gempa
Geoservices S. A.) for Raspberry Shake Seismograph Data information in this study.
REFERENCES
Bakun, B. H. & Johnson, L. R. (1973). The convolution of teleseismic P waves from explosions,
Milrow and Cannikin, Geophys. J. R. astr. Soc., 34, 321-342.
Barth, A. (2014). Significant release of shear energy of the North Korean nuclear test on February
12, 2013, J. of Seismology18 (3), 605-615.
Bolt, B. A. (1976). Nuclear Explosions and Earthquakes The Parted Veil. W. H. Freeman and
Company. San Francisco CA.
Bonner, J. L. Harkrider, D. G. Herrin, E. T. Shumway, R. H. Russell, S. A. & Tibuleac, I. M.
(2003). Evaluation of short-period, near-regional MS scales for the Nevada Test Sit, Bull. Seism.
Soc. Am., 93 (4), 1773-1791
Bonner, J. L. Herrmann, R. B. Harkrider, D. & Pasyanos, M. (2008). The surface wave magnitude
for the October 2006 North Korea Nuclear explosion, Bull. Seism. Soc. Am., 98 (5), 2498- 2506.
Bonner, J. L. Reiter, D. L. & Shumway, R. H. (2002). Application of cepstral F statistic for
improved depth estimation, Bull. Seism. Soc. Am., 92 (2). 1675-1693.
Bowers, D., Marshall, P. D. & Douglas, A. (2001). The level of deterrence provided by data from
the SPITTS seismometer array to possible violations of the Comprehensive Test Ban in Novaya
Zemlya Region, Geophys. J. Int., 46, 425-438.
Burdick, L. J. Wallace, T. & Lay, T. (1984). Modeling near-field and teleseismic observations
from the Amchitka test site, J. Geophys. Res., 89 (B6), 4373-4388.
@seismicisolation
@seismicisolation
Nuclear Tests
Cho, C. Shin, J. S. & Kim, G. (2016). Comparison of results of relative location methods and
2016AGU Fall Meeting, San Francisco, 10-12, 2016.
Chung, T. W. (1995). A quantitative study on the crustal structure of the Korean Peninsula based
on the Earthquakes from 1991 to 1994, J. Kor. Earth Science Soc., 16 (2), 152-157.
Closmann, P. J. (1969). On the prediction of cavity radius produced by the underground nuclear
explosion, J . Geophys. Res., 74 (15), 3935-2939.
Cohen, T. J. (1970). Source-depth determination using spectral, pseudo-autocorrelation and
cepstral analysis, Geophys. J. R. astr. Soc., 20, 223-231.
Douglas, A, Corbishley, D. J. Blamey, C. & Marshall, P. D. (1972). Estimating the firing depth of
underground explosions, Nature 237, 26-28.
Douglas, A. & Hudson, J. A. (1990). The effect on the teleseismic P of the zone of damage created
by an Explosion, Geophys. J. Int., 103, 111-133.
Douglas, A. & Rivers, D. W. (1988). An explosion that looks like an earthquake, Bull. Seism. Soc.
Am., 78 (2), 1011-1019.
Douglas, A. Marshall, P. D. & Young, J. B. (1987). The P waves from the Amchitka Island
explosions. Geophys. J. R. Astr. Soc., 90, 101-117.
Douglas, A. Richardson, L. & Hutchins, M. (1990). Surface reflections and S-to-P conversions on
P Seismograms, Geophys. J. Int., 100, 303-314.
Dreger, D. (2017). Updated UC Berkeley solution for the mb 6.3 (Mw 5.2) suspected DPRK
nuclear test September03, 2017 03: 30:00 UTC, [Douglas Dreger, UC Berkeley] Source-type
inversion result for the DPRK nuclear test in
https://ds.iris.edu/ds/nodes/dmc/specialevents/2017/09/03/2017-north-korean- nuclear-test/.
Ford, S. R. Douglas, D. S. & Walter, W. R. (2009). Source analysis of the Memorial Day
Explosion Kimchaek North Korea, Geophys. Res. Lett., 36 (21), L21304.
Fox, D. B. Selby, D. N. Heyburn, R. & Woodhouse, J. H. (2012). Shallow seismic source
parameter determination using intermediate-period surface wave amplitude spectra, Geophys. J.
Int., 191, 601-615.
Geological Society of Korea (1987). Geological Map of Korea, Geological Society of Korea,
Seoul, Korea.
Gitterman, Y. Kim, S. G. & Hofstetter, R. (2015). Spectral modulation effect in teleseismic P-
waves from North Korean nuclear tests recorded in broad azimuthal range and possible source
depth estimation. Pure Appl. Geophys., DOI: 10.1007/s00024-015-1169-8.
Glasstone, S. & Dolan, P. J. (1977). (compiled & edited) The Effects of Nuclear Weapons, United
States Department of Defense and the Energy Research and Development Administration, 657
pp.
Harjes, H. Henger, M. Mulle, G. W. & Wilhelm, H. (1985). A system design for the gradual
improvement of seismic monitoring and verification capabilities for a Comprehensive Nuclear
Test Ban Conference on Disarmament, CD/624, GE. 85-63293, 26 July 1985 Federal Republic
of Germany 68 pp.
Herrmann, R. B. (2016). Computer Programs in Seismology. Current Version 3.30 is dated May 8,
2016 (NP330. May-08-2016.tgz), Saint Louis University, Saint Louis, MO. 186 pp.
Heyburn, R. Neil, D. & Fox, B. (2013). Estimating earthquake source depths by combining surface
wave amplitude spectra and teleseismic depth phase observations, Geophys. J. Int., 94 (2),
1000-1010.
Husebye, E. S. Matveeva, T. & Fedorenko, Y. (2013). Focal-depth estimation using Pn-coda
phases including pP, sP, and PmP, Bull. Seism. Soc.Am., 103 (3), 1771-1783.
Israelsson, H., (2016). A note on the location of the North Korean nuclear test on Jan 6, 2016,
Technical Note 2016-01, Seismic Infra Research, Washington DC, 17 pp.
Johnston, W. R (2005). Database of nuclear tests, United States: Part 2, 1964-1972, complied and
last modified 19, June 2005.
@seismicisolation
@seismicisolation
432 Chapter Twenty
Kennett, B. L. N., Engdahl, E. R. & Buland, R. (1995) Constraint on seismic velocities in the
Earth from traveltimes, Geophys. J. Int., 221, 108-124.
Kim, S. G. & Bae, H. S. (2006). Investigation of post-sites using local seismic tomography in the
Korean Peninsula, Kr. Soc. Econ. Environ. Geol., 39 (2), 111-128.
Kim, S. G. Gitterman, G. Lee, S., 9DYU\þXN V. & Kim, M. (2015). Estimating depth and source
characteristics of nuclear tests by the Democratic People’s Republic of Korea in 2006, 2009 and
2013 using regional and teleseismic network, Science and Technology, CTBTO, SnT2015,
Vienna, Austria, 22-26, 2015.
Kim, S. G. Gitterman, Y. Lee, S. & Vavrycuk, V. (2016). Depth determination and source
characteristics of the North Korean nuclear tests (2006, 2009, 2013 and 2016) using local and
teleseismic arrays, Final Paper Number: S34A-02, AGU Fall Meeting, San Francisco,
December 12-16, 2016.
Kim, S. G. Gitterman, Y. Lee, S. & 9DYU\þXN, V. (2017). Accurate depth determination and source
characteristics of the DPRK nuclear tests (2006, 2009, 2013, 2016J (01/06/2016) and 2016S
(09/09/2016) using regional and teleseismic arrays, Science and Technology, T2.1-01, SnT2017,
CTBTO, Hofburg Palace, Vienna, Austria, 20-24 June 2017.
King, C. Y. Abo-Zena, A. M. & Murdock, J. N. (1974). Teleseismic source parameters of the
Longshot, Milrow, and Cannikin nuclear explosions, J. Geophys. Res., 79 (5), 712-718.
Kulhanek, O. (1971). P-wave amplitude spectra of Nevada underground nuclear explosions,
pageoph 88, 121-136.
Langston, C. A. (1980). A note on spectral nulls in Rayleigh waves, Bull. Seism. Soc. Am., 70 (4),
1409-1414.
Lay, T. (1991). The teleseismic manifestation of pP: problems and paradoxes. In Explosion Source
Phenomenology, Geophysical Monograph 65, 109-125.
Massé, R. P. (1981). Review of seismic source models for underground nuclear explosions, Bull
Seism Soc Am 71 (4), 1249-1268.
Montagner J. P. & Kennett, B. L. N. (1996) How to reconcile body-wave and normal mode
reference Earth models? Geophys. J. Int., 125: 229-248.
Murphy, J. R. (1981). P wave coupling of underground explosions in various geologic media. In
Identification Seismic Sources, ed. Huseby ES, Mykkeltveit SD. Reidel Publishing Company,
pp. 201- 205.
Murphy, J. R. Kitov, I. Rime, N. & Barker, B. (2017). Further studies of the seismic
characteristics of Russian explosions in cavities implication for cavity decoupling of
underground nuclear explosions PL- TR-96- 2017 Phillips Laboratory Hanscom AFB MA 66
pp.
Murphy, J. R. Stevens, J. L. Kohl, B. C. & Bennett, T. J. (2013). Advanced seismic analyses of the
source characteristics of the 2006 and 2009 North Korean Nuclear Tests, Bull. Seism. Soc. Am.
103 (3), 1640-1661.
Nuttli, O. W. (1969). Travel times and amplitudes of S waves from nuclear explosions from
Nevada, Bull. Seism. Soc. Am.,, 59, 385-398.
Okal, E. A. (1992). A student’s guide to teleseismic body wave amplitudes. Seism. Res. Lett., 63
(2), 169-180.
Pabian, F. & Coblentz, D. (2017). North Korea’s Punggye-ri Nuclear Test Site: Analysis Reveals
Its Potential for Additional Testing with Significantly Higher Yields. 38 NORTH (03/16/2017).
http://38north.org/2017/03/punggye031017/.
Pabian, F. &. Hecker, S. (2012). Contemplating a third nuclear test in North Korea, Bull Atomic
Sci August. http://www.thebulletin.org/web-edition/features/contemplating-third-nuclear-test-
north-korea.
Patton, H J. & Taylor, S. R. (2008) Effects of shock-induced tensible failure on mb - MS
discrimination: contrasts between historic nuclear explosions and the North Korean test of 9
October 2006, Geophys. Res. Lett., 35 (14). DOI: 10.1029/2008GL034211.
@seismicisolation
@seismicisolation
Nuclear Tests
Phillips, W. S. Begnaud, M. L. Rowe, C. A. Steck, S. C. Myers, M. E. Pasyanos, M. & Ballard, S.
(2007). Accounting for lateral variations of the upper mantle gradient in Pn tomography studies.
Geophys. Res. Lett., 35 (14), DOI: 1029/2008GL034211.
Ringdal, F., Marshall, P. D., & Alewine, R. W. (1992). Seismic yield determination of Soviet
underground nuclear explosions at the Shagan River test site, Geophys. J. Int. 109, 65-77.
Rodgers, A. J. Petersson, N. A., & Sjogreen, B. (2010). Simulation of topographic effects on
seismic waves from shallow explosions near the North Korean nuclear test site with emphasis
on shear wave generation, J. Geophys. Res., 115 (B11309):1-27 (2010). doi:10.1029/
2010JB007707.
Rodnikov, A. G., Sergeyeva, N. A., Zabarinskaya, L. P. Filatova, N. I. Piip, V. B. & Rashidov, V.
A. (2008). The deep structure of active continental margins of the Far East (Russia), Russian J.
Earth Sciences, 10, ES4002, doi:10.2205/2007ES000224.
Rougier, E. Patton, H. W. Knight, E. E. & Bradley, C. R. (2011). Constraints on burial depth and
yield of the 25 May 2009 North Korea test from hydrodynamic simulation in a granite medium,
Geophys. Res. Lett., 38 (16). DOI:10.1029/2011GL048269.
Science Academy Press The Geological structure and underground resources of NE Korea and
Southern Primorsky Krai of the Maritime Province, USSR (in Korean), (1964) The Geological
Institute, Pyongyang, DPRK and The Far East Geological Institute, Vladivostok, Siberia
Branch, AS, USSR.
Shin, J. S. Sheen, D. & Kim, G. (2010). Regional observation of the second North Korean nuclear
test on 2009 May 25, Geophys. J. Int., 180, 243-250.
Springer, D. L. (1974). Secondary sources of seismic waves from underground nuclear explosions,
Bull. Seism. Soc. Am., 64 (3), 581-594.
Stein, S. & Wiens, D. Depth determination for shallow teleseismic earthquakes: methods and
results, Reviews Geophys., 24 (4). 806-832.
Tian, D., Yao, J. & Wen, L. (2018). Collapse earthquake swarm after North Korea’s 3
September2017 nuclear test. Geophys. Res. Lett., 45. 1-8. DOI:org/10.1029/2018GL077649
7RNVऺ] M. N. & Kehrer, H. H. (1971). Underground nuclear explosion: tectonic utility and
dangers. Science 173, 230-233.
Tsai, Y. B. & Aki, K. (1971). Amplitude spectra of surface waves from small earthquakes and
underground nuclear explosions, J. Geophys. Res., 76 (17), 3940-3952.
Udias, A. (1999). Principles of Seismology, Cambridge University Press, Cambridge, UK.
Van Der Hilst, R. D. Kennett, B. L. N. & Shibutani, T (1998). Upper mantle structure beneath
Australia from portable array deployments. Geodynamics, 26, 39-57.
9DYU\þXN V. & Kim, S. G. (2014). Non-isotropic radiation of the 2013 North Korean nuclear
Explosion, Geophys. Res. Lett., 41 (20), 7048-7056. DOI: 10.1002/2014GL06126.
Viecelli, J. A. (1973). Topography and Rayleigh wave generating efficiency of buried explosive
sources, J. Geophys. Res., 78 (17), 3334-3339.
Zhang, M. & Wen, L. (2013). High-precision location and yield of North Korea’s 2013 nuclear
test, Geophys. Res. Lett., 40, 2941-2946.
Zhao, L. F. Xie, X. B. Wang, W. M. Fan, N. Zhao, X. and & Yao, Z. X. (2017). The 9 September
2016 North Korean underground nuclear test, Bull. Seism. Soc. Am., 107 (6), 3044-3051.
@seismicisolation
@seismicisolation
CHAPTER TWENTY-ONE
SOURCE DEPTH AND CHARACTERISTICS FOR DPRK’S 2016- 01-06, 2016-

09-09 AND 2017-09-03 NUCLEAR TESTS USING BODY AND RAYLEIGH
WAVE SPECTRA AND POLARIZATION OF SURFACE WAVES1
S. G. KIM, Y. GITTERMAN, S. LEE AND H. BAE
The North Korean thermo-nuclear explosion on September 3, 2017 recorded on R720B,

Gobobgsan (distance=448.6 km, M=6.3, yield =541.2/867.5 kt) with a non-tectonic event (m=4.1).
ABSTRACT
North Korea has conducted six underground nuclear explosions so far. In this study, we
determined source depth and characterization for the 2016J, 2016S and 2017 tests which were
conducted on January 6, 2016 (2016J), September 9, 2016 (2016S) and September 3, 2017 (2017)
respectively. It has been difficult to ascertain the accurate depth of North Korean nuclear
explosions due to paucity of data and information. We explore the depth calculation for the North
Korean nuclear tests based on detailed depth phases using teleseismic and regional arrays. We
present the coherent spectral nulls from the average spectra of pP+P, sP+P, pPn+Pn and sPn+Pn
which correlate with the depth phases showing 180° phase reversals with the P-wave arrivals. We
also verify source depth and characteristics using spectral minima of the fundamental-mode
Rayleigh wave amplitude spectra and polarization of surface waves. We estimated the burial
depths at 2.06, 2.05, 1.97 km for the 2016J, 2016S and 2017 nuclear tests, respectively, We
1
Some parts published in Pure and Applied Geophysics (2018). Depth calculation for the January 06, 2016, the
September 09, 2016 and the September 03, 2017 nuclear tests of North Korea from detailed depth phases using
regional and teleseismic arrays @seismicisolation
@seismicisolation
Source Depth and Characteristics for DPRK’s 2016- 01-06, 2016-09-09 and 2017-09-03 Nuclear Tests 435
Using Body and Rayleigh Wave Spectra and Polarization of Surface Waves
anticipate our absolute findings to be significant since in the past depth estimates for North Korean
nuclear tests have been inconclusive and unclear owing to not only paucity of data but also trade-
offs of the relative assessment based on the satellite images between the true source location and
the tunnel entrance for the 2006 nuclear test used as a reference event.
Keywords: CTBTO; spectral null; take-off angle; Fermat Principle; Mantapsan (Mt. Mantap)
INTRODUCTION
Fig. 1. Closed squares represent teleseismic arrays with uniformly azimuthal coverage for the average spectra. Open
triangles represent the regional seismic network including KSRS (Korea Seismological Research Station, Wonju,
South Korea) and USRK (Ussuriysk, Russia) arrays. The red star indicates the nuclear test site. The map plots were
made using the Generic Mapping Tools version 4.2.1 (Wessel and Smith, 1998) at www.soest.hawaii.edu/gmt.
North Korea conducted its fourth nuclear test (M 5.1) on January 6, 2016, fifth (M 5.3) on
September 9, 2016 and sixth (M 6.3) on September 3, 2017 near Mt. Mantap in the Punggye-ri
region of northeastern North Korea. Many researchers have studied source mechanisms (e.g.,
Bonner et al., 2008; Patton and Taylor, 2008; Ford et al., 2009; Vavrycuk and Kim, 2014; Dreger,
2017) and source depth estimations (Zhang and Wen, 2013; Murphy et al., 2013; Gitterman et al.,
2015; Kim et al., 2015; Zhao et al., 2017) with uncertainty due to the paucity of data and relative
assessment of trade-offs between the surface-looking interpretation from satellite images and true
depth based on the reference event of the 2006 test. In this paper we stress the point that the
accurate depths for last three North Korean nuclear tests (2016J, 2016S and 2017S) were
determined comparing them with previous North Korean nuclear tests utilizing the teleseismic
arrays and regional arrays. We estimated the source depth using depth phases pP and sP as well as
pPn and sPn taking into account take-off angles for further corroboration and improvement. The
detonation source depths are pertinent vis-à-vis enforcement of CTBTO as nuclear explosions are
unlikely to be fired at depths of more than a few kilometers (Douglas et al., 1972; Bakun and
Johnson, 1973; Bowers et al., 2001; Heyburn et al., 2013). However, absolute depth
determinations are presently based on the delay times of pP-P which often present the problem of
free surface reflection, pP from the direct signal P waves in shallow nuclear explosions. The usual
way to estimate the depth of an event using teleseismic waves is to measure the time difference
@seismicisolation
@seismicisolation
436 Chapter Twenty-One
between the arrival time of the direct P wave and the onset of the surface reflection phase of pP
(180º phase reversal). The pP phases are signals that travel upward from the focus as P waves from
the explosions (Cohen, 1970; Kulhanek, 1971) and then travel steeply downwards following a path
to the same receiving stations as the P waves. Regarding depth phase identification, the single
station is complicated by ambient noise and elastic effects (multi-pathing, scattering, etc.). The
teleseismic array data which are located in a better azimuthal coverage can improve these effects
suppressing noise signals and reducing the non-linear topographic effects enhancing a signal-to-
noise ratio (SNR) of the depth phases of pP, sP, pPn and sPn. In particular we applied the global
arrays which are roughly located in unformly azimuthal coverage to minimize noise signals and
the nonlinear topographic effects assuming that a surface elevation of the source is almost equal to
that of each reflector from the source taking into account a raypath in 1D direction from the source
in the middle of the mountain. We performed the teleseismic spectra analyses of short-period
vertical components for ASAR (Alice Springs, Australia), ARCES (Karasjok, Norway), EKA
(Eskdalemuir, UK), KURK (Kurchatov, Kazakhstan), NOA and NORESS (NORSAR, Norway),
NVAR (Mina, USA), PDAR (Pinedale, USA), and Korea Seismological Research Station (KSRS,
Wonju, Korea), USRK (Ussuriysk, Russia) and broadband records of ISN (Israel Seismic
Network), WRA (Warramunga, Australia) and YKA (Yellowknife, Canada) arrays (Fig. 1). We
also removed the effects of absorption and recording system taking into account a source in the
same high velocity layer and the same array stations eliminating path and instrument effects. We
determined spectral nulls from destructive interference of pP+P (sP+P) and pPn+Pn (sPn+Pn)
spectra corresponding to reciprocals of pP-P (sP-P) and pPn-Pn (sPn-Pn) lag times.. Futhermore
the high quality factor Q in the granite source rock (~1000) of this test clearly reveals high
frequency arrivals of pP (sP) and pPn (sPn) in teleseismic and regional seismograms due to low
attenuation of amplitude in the high velocity layer from the amplitude relation A(t) = A0exp(-πft/Q)
in this study.
BASIC CONCEPT OF DEPTH PHASE AND ITS DELAY TIME
The depth phases are useful for accurate determination of a focal depth finding its delay times or
spectral nulls. Stein and Wiens (1986) found that the depth determination of body waves is robust
to typical uncertainties in focal mechanism and near-source structure for moderate-size events
(M<6.5). The depth phases of pP, sP, pPn and sPn are clearly presented in the waveforms
produced by superposition with respective channels of the array which are 180 º out of phase
relatively to the first P-wave arrivals (Massé, 1981; Udias, 1999) (Refer to APPENDIX A).
The delay times for depth phases present the difference between the onsets of P-wave arrivals and
those of the downswing depth phases of pP (sP) in the time domain. In the frequency domain, the
delay time for the depth phase produces a sharp scalloping (hole) of the amplitude spectrum,
yielding minima at periods Tn=T/n for n=1, 2, … where T is the delay time for pP-P or sP-P (pPn-
Pn, sPn-Pn) (Frasier, 1972; Burdick et al., 1984; Stein and Wiens, 1988; Stead and Helmberger,
1988). The onsets of pPn-wave arrivals appear on the zero lines in KSRS arrays whereas those of
pP-wave arrivals appear at the peak in the teleseismic arrays because of the overshoot of the
teleseismic P-wave propagation path whereas there are no overshoots on the regional seismograms
of KSRS and USRK.
The noticeable depth phase of sP at PDAR and NVAR probably indicates a relatively high Q path
with low attenuation of sP wave through the mantle beneath the high alitude arrays compared to
pP with low Q of high attenuation. The broad and slower depth phase of pP and sP at ASAR
inducing the low frequency spectral null at the ASAR may be more closely related to the low
velocity layer of the large aquifers beneath Alice Springs area in the upper crust and possibly
including the low P-wave velocity layer in the upper mantle of the Late Palaeozoic Alice Springs
Orogeny beneath the array according to the tomography study (Van Der Hilst et al., 1998). It
should also be noted that the spectral nulls calculated from spectra of teleseismic and regional
arrays well fit the delay times of depth phases in the time domain.
@seismicisolation
@seismicisolation
CALCULATION OF BURIAL DEPTHS
The effect on depth determination is less pronounced for the long-period data since some of the
additional small arrivals are filtered out by the instrument. Therefore, short-period records used in
this study are more conducive to determine accurate source depth. Furthermore, use of body wave
data is often more sensitive for depth analysis than source parameters such as strike, dip and rake
(Stein and Wiens, 1986) as well as the surface topography. The depth via pP-P and sP-P can be
estimated using equations (1) and (2) in APPENDIX A. In particular for the mountainous
topography like the North Korean nuclear test site, the travel time of different reflection point is
not always representative of the depth of the corresponding reflectors. However, each incident
angle always equals each reflected angle (take-off angle) with respect to the normal at the reflector.
As a result the arrival times of successive reflections are representative of the depths of the
corresponding reflectors taking into account the average spectra of global seismic arrays. In this
study we used the average spectra from the seismic arrays with uniformly azimuthal coverage
from all directions at a source in the middle of the mountain in order to enhance the signal-to-
noise-ratio minimizing the nouse signals and the the topographic effects (see Figs. 2, 3 and 4).
Table 1. The source depth of the 2016J nuclear test was determined via pP-P and sP-P delay
times with take-off angles using 10 seismic arrays. Distance and AZ (direction to the station)
are measured in degrees. The asterisk (*) indicates the representative station PRNI of the
Israel Seismic Network.
(°) (Hz) (°) (km)
ASAR (64.8, 175.1) 1.1 17.57 2.43
ARCES(56.4, 335.8) 1.25 18.91 2.16
EKA (75.5, 334.1) 1.35 15.56 1.96
ISN (73.5, 295.6)* 1.28 16.13 2.07
KURK (35.6, 302.8) 1.22 22.86 2.27
NORESS(66.2,331.9) 1.25 17.15 2.14
NVAR (79.7, 47.5) 1.25 (0.98) 14.62 (8.25) 2.11 (1.92)
PDAR (81.0, 39.5 ) 1.25 (1.02) 14.33 (8.09) 2.11 (1.85)
WRA (61.1, 174.3) 1.25 18.40 2.15
YKA (64.7, 27.4) 1.25 17.57 2.14
Average depth (km) 2.11±0.15
Values in the brackets refer to sP+P spectral null and depth including a take-off angle of sP phase.
@seismicisolation
@seismicisolation
Fig. 2. The delay times for depth phases for the 2016J nuclear test of North Korea are obtained via spectral nulls of the
average spectra of pP+P (red arrow) and sP+P (blue arrow) from the teleseismic arrays. The seismograms and spectra
were band-pass filtered between 0.5-3.0 Hz.
@seismicisolation
@seismicisolation
Fig. 3. The depth-phase travel times for the 2016S nuclear test are obtained via spectral nulls of the average spectra of
pP+P (red arrow) and sP+P (blue arrow). The seismograms and spectra were band-pass filtered between 0.5-3.0 Hz.
@seismicisolation
@seismicisolation
Table 2. The source depth of the 2016S nuclear test was determined via pP-P and sP-P dealy
times with take-off angles using 8 seismic arrays. Distance and AZ (direction to the station)
are measured in degrees. The numbers in the brackets refer to spectral nulls and depths via
sP-P delay times including a take-off angle of sP phase.
(°) (Hz) (°) (km)
ASAR (64.8, 175.1) 1.1 17.57 2.43
ARCES (56.4, 335.8) 1.48 18.91 1.82
EKA (75.5, 334.1) 1.37 (0.98) 15.56 (8.77) 1.93 (1.93)
FINES (60.3, 327.4) 1.56 (0.94) 16.13 (9.09) 1.70 (2.01)
NVAR (79.8, 47.4) (0.98) 14.62 (8.25) (1.92)
PDAR (81.0, 39.5 ) 1.26 (0.92) 14.33 (8.09) 2.09 (2.03)
WRA (61.1, 174.3) 1.26 18.40 2.13
YKA (64.7, 27.4) 1.40 17.57 1.91
Average depth (km) 1.99±0.15
Ideally, the pP-P time would be obtained from the time difference between the onset of P and that
of the downswing depth phase of pP in the time domain. However, those onsets are difficult to
pick so an alternative is to use spectral nulls of the average spectra because the depth phases are
always 180°out of phase with respect to the incoming P-wave arrivals. i.e. we measured the time
difference between the maximum positive amplitude P-wave arrival and the maximum negative
amplitude of pP assuming P and pP have an identical shape with negative polarity and interference
between the two pulses (Douglas et al., 1987). The absence of depth phase is the result of
scattering and defocusing of P- and S- waves by lateral variations in the crust above the source
(Douglas et al., 1990). However we can minimize the nonlinear topographic effects of the
mountainous source and the azimuthal variations of pP waveforms (Burdick et al., 1984) using the
average spectra obtained from a broad and good azimuthal distribution of teleseismic arrays (Fig.
1). We used spectral nulls of the average spectra of multi-pathing arrays and utilized the full depth
formula with take-off angles assuming the average overburden P-wave velocity to be Vp=5.1
km/sec (Bonner et al., 2008) and Vs=2.9 km/sec (Burdick et al., 1984) at the DPRK nuclear test
site. All spectra of single array show similar ‘spectral nulls’ indicating that the psedudo-1D
radiation in one direction is constant and nothing about the contribution of different source
components. So data from a test in the middle of a mountain will be even lookalike in all
directions since an onset time of P-wave arrivals at each element of an array follows Fermat
Principle and Snell’s Law which indicate that a raypath is along a minimum of travel time and that
an incident angle (take-off angle) is always equal to a reflected angle of a single raypath within a
limited time window. It should be noted that there is no elevation difference between a source
surface and a reflector with a shallow dip as the closest approximation possible on an average in
the present state of evidence. Consequently we do not count on the unknown and complicated
terrain of the Mt Mantap and neglected the trade-off between our applied simplifications and the
claimed accuracy of the results from the nonlinear topographic effects considering the uniformly
azimuthal coverage by means of the globally distributed seismic arrays from the source. It should
be also noted that sP phases can be explained in terms of the non-isotropic radiation of the North
Korean nuclear tests from the tectonic stress by the nearby source zone (Vavrycuk and Kim, 2014)
rather than shear dislocation faulting. .The tensile faulting associated within opening or closing of
a crack is described by nonzero ISO and considerable components of DC and CLVD resulting in
not only Rg wave but also SH and Love waves by asymmetry of the explosive source for the North
Korean nuclear tests (Vavrycuk and Kim, 2014, Cho et al. 2016). The deviatoric components of
DC and CLVD evidence that the non-isotropic radiation is related with tectonic stress in the
surrounding rock rather than shear faulting triggered along preexisting nearby fault structures. It is
worthwhile to note that Zhao et al. (2017) recently determined surface wave magnitudes (MS)
@seismicisolation
@seismicisolation
using Rayleigh waves well generated from five nuclear tests of North Korea including the 2016S.
Dreger (2017) also found DC and CLVD components for the 2017 nuclear test to be 34% and 24%
from waveform modelling mechanism using some regional long-period records including 34
teleseismic P-wave polarity observations. The deviatoric stress caused probably asymmetric shape
of the cavity during the detonation resulting in an asymmetric radiation for the shear wave
generation.
The burial depths for the 2016J, 2016S and 2017 tests were estimated at 2.11 km, 1.99 km and
1.92 km from the average spectral nulls in Figs 2, 3 and 4 including Tables 1, 2, and 3. We have
also found that the spectral null for the 2017 are very variable and low compared to those of the
2016J and 2016S tests. The slapdown phase also appears at 0.57 Hz (after 1.75 seconds from the
onset) from the ARCES and EKA records in the 2017 test. It cannot be ruled out that the delay
times for the 2017 may be a velocity reduction for the surface-reflected P waves in the inelastic
source region because of the slapdown (spall closure) as a secondary source (Springer, 1974).
King et al.(1974) found that the apparent average overburden velocities are approximately 15%
lower than the velocities from on-site measurements from the Longshot, Milrow and Cannikin
Nuclear Explosions. Therefoere, the actual burial depth for the 2017 test of North Korea should be
greater than 2 km in the light of a slapdown phase. Since the slapdown phase is due to the inelastic
and nonlinear process near the source, we should take into account a velocity reduction for the
surface-reflected P waves resulting in the shallow depth for the 2017 nuclear test.
@seismicisolation
@seismicisolation
Fig. 4. Red and blue arrows indicate spectral nulls of pP+P and sP+P for the 2017S nuclear test using teleseismic
arrays. Black, red, blue and brown bars at ARCES indicate P, pP, sP and slapdown phases in the time domain. The
slapdown phases also appear at spectra of ARCES and EKA records.
measured in degrees. The numbers in the brackets refer to spectral nulls and depths via sP-P
delay times including a take-off angle of sP phase.
(°) (Hz) (°) (km)
ASAR (64.8, 175.2) none 17.57 none
ARCES (56.4, 335.8) 1.47 (0.95) 18.91 (10.62) 1.83 (2.01)
EKA (75.5, 334.1) 1.40 15.56 1.89
FINES (60.3, 327.4) 1.57 (0.97) 18.62 (10.46) 1.71 (1.95)
NORESS (66.2, 332.0) 1.35 17.14 1.98
PDAR (81.0, 39.6 ) 1.33 14.33 1.98
WRA (61.2, 174.4) 1.43 18.40 1.88
YKA (64.7, 27.4) 1.24 17.57 2.05
Average depth (km) 1.92± 0.10
Note: NORESS is a sub-array of NOA.
Table 4. Depth determination via depth phases of pPn+Pn and sPn+Pn using KSRS, USRK
and MDJ records. The numbers in the brackets refer to spectral nulls and depths via sPn-Pn
delay times including a take-off angle of sPn phase.
Event KSRS USRK MDJ
spectral null (Hz) spectral null (Hz) spectral null (Hz) Averge depth
depth (km) depth (km) depth (km) (km)
2016J 1.72 (1.14) 1.56 1.71 (1.10)
1.96 (2.13) 2.12 1.94 (1.92) 2.01±0.09
2016S 1.72 (1.17) 1.54 (0.92) 1.71 (0.95)
1.96 (2.08) 2.15 (2.30) 1.94 (2.23) 2.11±0.13
2017 1.72 (1.19) 1.58 1.67 (1.07)
1.96 (2.04) 2.09 1.97 (1.98) 2.01±0.05
@seismicisolation
@seismicisolation
Fig. 5. Spectral nulls via pPn+Pn (red arrow) and sPn+Pn (blue arrow) using KSRS and USRK arrays for the 2016J, 2016S and
2017 nuclear tests of North Korea. It should be noted that there is a slapdown phase at 0.74 Hz (1.35 s) and 0.57 Hz (1.75 s) after
onsets of the P-wave arrivals in the USRK2016S and USRK2017S records (brown arrow).
@seismicisolation
@seismicisolation
The source depths for the 2016J, 2016S and 2017 nuclear tests were estimated at 2.01, 2,11 and
2.01 Km, respectively using the average spectral nulls of pPn+Pn and sPn+Pn from KSRS, USRK
and MDJ (Fig. 5 and Table 4) based on equations (3) and (7) in Appendix A. The Pn-wave
velocity for USRK is used as 8.0 km/sec in the Sikhote Alin region which is obtained from Phillips
et al. (2007) and Rodnikov et al. (2008) whereas that for KSRS is 7.8 km/sec which is obtained
from various researchers (Chung, 1995: Kim and Bae, 2006; Phillips et al., 2007). However, the
spectral nulls for the USRK are lower than those of KSRS due to the low velocity layer overlying
the subducting slab of the Pacific Plate.
It is also significant that we could find slapdown phases at 0.74 Hz (1.35 s) and 0.57 Hz (1.75 s)
after onsets of the P-wave arrivals in the USRK records for the 2016S and 2017 tests (brown
arrow). The slapdown – P time interval of 1.75 s (0.57 Hz) for the 2017 test is larger than that of
1.35 s (0.74 Hz) for the 2016S test since the radius of inelastic region for the larger size of the
2017 test is increased in hard or fully-saturated materials such as granite source (Springer. 1974).
It is also interesting to compare the slapdown –P time intervals for the 2016S and 2017 occurred
within 1.35 s (0.74 Hz) and 1.75 s (0.57 Hz) after detonation with that of the Milrow
thermonuclear explosion (1000 kt) whose slapdown time interval was 1.35 s which can be
explained in terms of the granite source rock of 5.1 km/sec for the North Korean nuclear
explosions. The average overburden source rock of the Milrow thermonuclear explosion (Bakun
and Johnson, 1973; King et al., 1974) was found to be 3.1 km/sec (reduced from 3.7 km/sec
according to drill core data and direct P-waves due to the inelastic source region by shot) since the
radius of the inelastic volume (idealized) for hard or fully-saturated materials (granite, wet tuff,
etc.) is much greater than that of the lower overburden velocity (Closmann, 1969; Springer, 1974).
Nevertheless, it is worth noting that radii of cavity (15-57 m) and the inelastic volume (damage)
zones which are about (3-5) times of cavity radius are distributed not to interrupt each test and do
not influence depth estimates taking into account source locations except for slapdown phases for
the 2017 test (Closmann, 1969; Springer, 1974) (Refer to APPENDIX C).
a b
Fig.6. Synthetic seismograms and spectra of the vertical components with little anelastic attenuation showing pPn-Pn
for the near-field and b) including effects of anelastic attenuation showing pP-P for the far-field (Burdick et al., 1984;
Douglas and Hudson, 1990) in the flat Earth model. The arrows of pPn+Pn (1.72 Hz) and pP+P (1.25 Hz) in spectra
correspond to the 2 km depth in the seismograms.
uncertainties to other parameters (Stein and Wiens, 1986; Douglas and Hudson, 1990). The earth
structure models for synthetics are used by Korea Model (Kim and Bae, 2006; Herrmann 2016) in
the near-field and AK135 (Kennett et al., 1995; Montagner and Kennett, 1996) in the far-field for
the teleseismic arrays taking a short span of the first P-wave arrivals for 5-6 seconds. The preferred
@seismicisolation
@seismicisolation
pP-P and pPn-Pn delay times in synthetics are in very close agreement with observed results from
spectra (Fig. 6). The synthetic seismograms and spectra are produced using Hudson96 and
Herrmann programs (Herrmann, 2016). The synthetic spectral nulls of short-period P waves for the
near-field and those of long-period P waves for the far-field are considerably consistent with the
observed spectral nulls (Burdick et al.,1984; Douglas and Hudson, 1990).
We also found that spectral nulls for pPn+Pn and pP+P for synthetic seismograms are well
consistent with those of observations in 2016J, 2016S and 2017 tests. The effect of anelastic
attenuation using t* which is the ratio of travel time and a specific quality factor Q is assumed to
be independent of frequency. Burdick et al.(1984) and Douglas and Hudson (1990) found that the
short-period P wave synthetic seismograms showing two troughs includes the effect of anelastic
attenuation computed using t* in Fig. 6a whereas no effects of anelastc attenuation are included in
the long-period (broad-band) P wave synthetic seismograms showing one trough without an
overshoot in Fig. 6b. The spectral nulls of 1.72 Hz and 1.25 Hz at a depth of about 2.0 km in Figs
6a and 6b reasonably fit the observed spectral nulls in Figs 2, 3, 4 and 5.
explosion depth and the P-wave velocity in the material above the shot point (Douglas and Hudson,
1990; Lay 1991) because the corresponding pP arrival would be expected to be reduced in
amplitude and possible delayed relatively to what would be expected for true pP in case of the
effect of a zone of damage above an explosion. The reflected signals containing the fracture and
crushing associated with the explosion would be reduced and delayed resulting in the slapdown
phase as a secondary source (Springer, 1974). However, in this study we should note that North
Korean nuclear explosions were conducted in the deep homogeneous granite without the surface
spalling as shown on the satellite images except for some collapse events for the 6th nuclear test.
As a consequence we neglected the diffraction effects of depth phases from the edge of a spalling
for North Korean nuclear explosions except for the 6th nuclear test (Douglas and Hudson, 1990).
SOURCE DEPTH ESTIMATE AND CHARACTERISTICS BY THE FUNDAMENTAL-
MODE RAYLEIGH WAVE AMPLITUDES AND POLARIZATION OF SURFACE
WAVES
exponential decay of Rayleigh wave displacements with depth (Refer to Appendix B). The
frequency-dependent “spectral null” in the Rayleigh wave amplitude spectra are most pronounced
for the pure (vertical) strike-slip faultings and the reverse faulting mechanisms with 45° dip. The
spectral null of the fundamental-mode Rayleigh wave spectrum is dependent on source
mechanism, depth and source-receiver azimuth but does not vary much with the shot medium and
the shot yield (Tsai and Aki, 1971; Douglas et al., 1972; Langston, 1980; Okal, 1992; Fox et al.,
2012; Heyburn et al., 2013). The conical dip-slip reverse faulting may accompany a deep-seated
tensile failure occurring at a depth above shot point (Massé, 1981; Patton and Taylor, 2008)
generating the strong Rg wave radiation from a vertically oriented CLVD source (Rodgers et al.,
2010). We also estimated source depths for North Korean nuclear tests using spectral nulls of the
fundamental-mode amplitude spectra of Rayleigh waves. We found spectral nulls at 0.13 Hz and
0.146 Hz using INCN and BJT through the pure continental-path data for DPRK’s nuclear tests in
Fig. 7. However, we found no spectral nulls at ERM which follows dispersion with higher mode
Rayleigh waves along the subduction zone. The spectral nulls may be related to a reverse faulting
dipping at about 45° dip which is consistent with findings from other researchers (Tsai and Aki,
1971; Douglas et al., 1972; Langston, 1980; Okal, 1992; Fox et al., 2012; Heyburn et al., 2013)
whereas no clear spectral nulls were found for non-pure (vertical) strike-slip faulting motions and
the oblique reverse fault mechanisms (Heyburn et al., 2013). The source mechanism for the North
Korean nuclear explosions may be assumed to be the conical dip-slip volume accompanying a
@seismicisolation
@seismicisolation
reverse faulting motion as a vertically-distributed source in a shaft. A spectral null (minima) is

caused by an excitation null for short-period fundamental-mode Rayleigh waves, termed Rg waves
as a CLVD (compensated linear vector dipole) source which correlates with normal mode theory
in the form of resonant frequency (Massé, 1981; Heyburn et al., 2013; Fox et al., 2012; Langston,
1980). Bonner et al. (2003) have examined the performance of MS scales on 7-sec Rayleigh waves
recorded distances less than 500 km from Nevada Test Site. Using equation (17) for Appendix B,
we estimated at 2.15 km for the 2006 test and 2.01 km for the rest of nuclear tests using the
spectral nulls of the fundamental-mode Rayleigh wave amplitude spectra which are in good
agreement with body wave studies. We estimated the spectral nulls for the North Korean nuclear
tests at 0.14 Hz at BJT, CHC and INCN whereas at 0.17 Hz at SEO and HIA (Fig. 7). The reason
why we found the higher spectral nulls at SEO and HIA is due to the low velocity path of Rg
waves. We observe the Rg-waves at SEO which pass through the Chgaryeong Rift Zone of a low
velocity layer which is running through the central part of the Korean Peninsula (Kim and Li, 1998)
and HIA also observes Rg-waves through the low velocity layer due to the tectonic depression in
the Northeastern Manchuria, China (see Fig. 7). The spectral null (minimum) from the CLVD
depth by waveform modeling mechanism (Vavrycuk and Kim, 2014) correlates with source depths
derived from pP-P and pPn-Pn delay times. The conical dip-slip reverse faulting may accompany a
deep-seated tensile failure occurring at a depth above shot point (Massé, 1981; Patton and Taylor,
2008) generating the strong Rg wave radiation from a vertically oriented CLVD source (Cho et al.,
2016; Udias, 1999) which is a kind of a spallation-like source function, but it is not associated with
the surface spall because it was conducted at granite in the deep tunnels. The estimated depth via
the CLVD model also disambiguates the estimated source depth by body waves.
@seismicisolation
@seismicisolation
@seismicisolation
@seismicisolation
Fig. 7. The fundamental-mode Rayleigh waves and displacement spectra for the BHZ band-pass filtered (0.02-0.1 Hz)
with amplitude in counts. The spectral nulls for North Korean nuclear tests are estimated at 0.14 Hz at BJT, INCN and
CHC whereas 0.17 Hz at HIA and SEO except for the 2006 test which shows around 0.13 Hz.
We found the particle motions of surface waves to be the striking similarity indicating that source
depth and mechanism including path are identical with all nuclear tests except for the 2006 test
which is the poor S/N of a small size (Fig. 8). We can see almost the same type of particle motions
of Rayleigh waves on the radial component (R) and Love waves on the transverse component (T)
in six North Korean nuclear explosions (Fig. 8). Since Rayleigh waves have no transverse
component, they are polarized in the vertical plane. The vertical and horizontal components are
shifted in phase by 90⁰ and the particle motion is elliptical with the vertical major axis and
retrograde to the wave propagation. For the pure-continental path stations such as MDJ, HIA, BJT
including those in the Korean Peninsula (INCN, CHC, SES and SEO) show weak Love waves as
well as strong Rayleigh waves. In general Love waves always appear only when there is a low
velocity layer overlying a high velocity medium since SH waves are totally reflected at a free
surface and trapped when an incident angle is greater than the critical angle (sin jc =Vs1/Vs2) and
the particle motion is transverse to the direction of wave propagation. In case of nuclear explosions,
Love waves are always weak due to paucity of interaction of SH waves. We should could observe
poor generation of Love waves from the stations (YSS, MAJO and ERM) since they are located in
the subduction zone of the Pacific slab with the low Q path with high attenuation from a low
velocity layer overlying a high velocity layer in the subduction zone (no high frequency) including
ocean noise effects and some unknown events near the Pacific slab. The anomalous polarization
patterns for 2006 HIA may be related to the instrumental malfunction whereas 2006 INCN may be
incorrect due to the noise effects in the ocean including mixing an unknown event.
It is also noticeable that the 2013 nuclear test reveals large Love waves as well as strong Rayleigh
waves compared to other events. Vavryčuk and Kim (2014) found that the appearance of distinct
Love waves for the 2013 test may be due to the cracks and/ or block movements in the pre-existing
tectonic stress in this region resulting in an oblique-reverse faulting mechanism. Our findings of
the non-isotropic radiation for the 2013 test shows horizontal and transverse motion of evident
Love waves on a tangential component as well as strong Rayleigh waves on a radial component
taking into account the same constraints of depth and mechanism as the other tests (Vavryčuk and
Kim, 2014).
@seismicisolation
@seismicisolation
@seismicisolation
@seismicisolation
Fig. 8. Particle motions for surface waves using the velocity records which were band-pass filtered between 0.02-0.1
Hz for the 2006, 2009, 2013, 2016J, 2016S and 2017 nuclear tests using YSS, MDJ, HIA, BJT, CHC, SEO, SES and
INCN and 0.04-0.1 Hz for ERM and MAJO stations.
There are some representative examples of a deep-depth estimate using the depth phase pP-P for a
deep atomic explosion (Douglas et al., 1972). On May 21, 1968 a 47-kiloton (mb =5.6) atomic
explosion (detonated in salt) was fired at a depth of 2.45 km near Bukhara, USSR to seal an oil
well. The geology in the vicinity of the explosion is a sedimentary basin in which an overburden
velocity is 3.0 km/sec (Douglas et al., 1972, Bolt, 1976 ). It was observed by pP-P delay times of
1791 m of a depth below the free surface (HOB= –1700 m) (Johnston, 2005) corresponding to the
observed spectral nulls of 0.94 Hz (1.06 s) by pP+P spectra using single channels at GBA, EKA
and YKA stations from the Cannikin explosion of the Amchitka Island explosions (mb=6.8,
@seismicisolation
@seismicisolation
yield=5Mt). The low spectral null of 0.94 Hz or a longer pP-P delay time (1.06 s) also attests to
overestimation by an overshoot of pP pulses (Douglas et al., 1987). It is due to the slapdown phase
from spalling at some depth of subsurface reducing P-wave velocity in the crushed rock around the
explosion in the island site.
Fig. 9. The pin symbols represent North Korean nuclear test locations with elevations and depths (negative) in km.
The prefixes of I, M, PH, ZW and TYW indicate the works from several researchers (Israelsson 2016; Murphy et al.
2013; Pabian and Hecker, 2012; Zhang and Wen, 2013; Tian, Yao and Wen, 2018). IDC and NEIC prefixes before
year indicate source locations estimated by International Data Center (IDC)/CTBTO and National Earthquake
Information Center (NEIC), United States Geological Survey (USGS) for the event year.

The behavior of the spectral null of depth implication does not appear to vary with azimuth in a
smooth way, which is most easily explained in terms of lateral variations near the receivers. The
spectral null which varies rapidly with azimuth from the source cannot, however, be ruled out
owing to a largely variation of the receiver sites such as ASAR and FINES. Therefore it is
important to use the average spectral null of the teleseismic arrays which are located in a
uniformly azimuthal distribution on the globe. The burial depths of the 2016J, 2016S and 2017
nuclear tests were 2.06, 2.05 and 1.97 km respectively using pP+P (sP+P) and pPn+Pn (sPn+Pn)
which correlate with findings for the 2006, 2009 and 2013 tests from previous studies (Vavrycuk
and Kim, 2014; Gitterman et al., 2015; Kim et al., 2015; Kim et al., 2016) indicating that North
Korean nuclear tests were conducted at approximately the same depth in tunnels (shaft and drift)
underneath the vicinity of Mt. Mantap (2.205 km elevation). However, our depth estimations were
quite different from previous studies by other researchers which concluded depth estimations of
200 m (2006 test)/550 m (2009 test) by Murphy et al.(2013) and 610 m (2009 test)/430 m (2013
test) by Zhang and Wen (2013). In those studies, source depths were estimated by subtracting the
west portal elevation of the 2006 test as a reference event from the elevation of the respective
ground-zero (GT0) location of the other @seismicisolation
tests using Google Earth digital terrain data (Patton and
@seismicisolation
Pabian, 2014). Several other researchers (Wen and Long, 2010; Pabian and Hecker, 2012; Paton
and Pabian, 2014; Israelsson, 2016; Zhao et al., 2017) have also proposed quite different locations
for the 2006 test from IDC and NEIC (see Fig. 9). The source locations for the 2006, 2009, 2013,
2016J, 2016S and 2017 tests were close each other maintaining a certain distance without giving
damage beneath the high elevation of around the Mantapsan by IDC which looks like forming a
cluster of mutually connected-deep tunnels except for 2006 and 2017 tests. Nonetheless the largest
2017 test of 6.3 (mb) with the yield of the detonation of 541 kt (Bowers et al., 2001) to 868 kt
(Murphy, 1981) did not cause any damage and surface spallation owing to the deep source rock of
granite. According to Fig. 10, the seismic yield for the 2017 nuclear test of North Korea would be
around 1 Mt in the upper bounds taking into account the wet hard rock as the source rock. The
2017 test is close to the Milrow thermonuclear explosion detonated at the shot medium of pillow
lava, Amchitka Island on October 2, 1969, which has known parameters such as magnitude (6.5),
yield (1 Mt) and depth (1218 m). The overburden P-wave velocity (3.7 km/sec from drill core data
and from on-site direct P-wave measurement) in the pillow lava (King et al., 1974) may be a likely
wet hard rock (sedimentary rock) like the 2017 nuclear test of North Korea in Fig. 10.
We infer that from the absence of radioisotopes in the atmosphere except for the first 2006 test that
North Korean nuclear tests may have been conducted as vertically distributed sources (Viecelli,
1973) in the shafts through mutually connected-deep tunnels beneath the mountainous region. The
more accurate source locations should be disclosed for further study since the source locations by
the absolute methods of IDC and NEIC are quite different from those by the above-mentioned
relative method (Fig. 9). The relative method begins with the assumption that the reference event
is on the same entrance elevation as the reference event by the Google Earth image including the
absence of the spoil excavated in front of the tunnel entrance. However we assume that the
individual test location is independent of the entrance elevation considering a plausible reason why
the mass spoil excavated disappeared compared with the deep burial of depths. We could presume
from the absence of the large piled spoils in front of entrances that those spoils could be backfilled
into the levels of the tunnels and/or old mines or possibly removed during the night. Therefore the
absolute method could be more objective and reliable based on the depth phases from regional and
well-coverage teleseismic arrays and the enhancement of S/N by the average spectra from the
records. As a result, we concluded that the burial depths of the 2016J, 2016S and 2017 nuclear
tests were estimated at 2.06, 2.05 and 1.97 km by depth phases which are almost identical with the
detonation depths of the 2006, 2009 and 2013 tests estimated from previous studies (Vavrycuk and
Kim, 2014; Gitterman et al., 2015; Kim et al., 2015; Kim et al., 2016) and fundamental-mode
Rayleigh wave amplitude spectra, indicating that all North Korean nuclear tests were conducted at
depth of about 2 km near Mt. Mantap.
@seismicisolation
@seismicisolation
Fig. 10. Magnitude versus yield relation for different geological settings. Semipalatinsk, Kazakhastan (1), Nevada,
USA (2), dry unconsolidated rock (3), cavities, salt (4), and cavities, granite (5) (Gaebler and Ceranna, 2017). The
DPRK’s nuclear explosions belong to Wet hard rocks (2) implying almost 1 Mt for the North Korean nuclear test on
September 3, 2017 (mb=6.3 by USGS and 6.4 by BGR, Germany ).
ACKNOWLEDGMENTS
criticism. We also appreciate the Executive Secretary of CTBTO, Lassina Zerbo and Director of
International Data Centre (IDC), for their collaboration in IDC data collection for this study. One
of authors (SGK) would greatly appreciate CTBTO’s funding to participate in SnT2015, SnT2017
and SnT2019, CTBTO, Vienna, Austria. We are also grateful to Dmitry Storchak (Director of
International Seismological Centre, UK) for providing phase reading data for this study. We would
acknowledge the collaboration for waveform data with IRIS (USA) and KMA (Korea
Meteorological Administration) in this research.
REFERENCES
Bakun, W. H, and Johnson, L. R. (1973). The convolution of teleseismic P waves from explosions
Milrow and Cannikin, Geophys. J. R. astr. Soc.. 34, 321-342.
Bolt, B. A. (1976). Nuclear Explosions and Earthquakes: the Parted Veil, W. H. Freeman and
Company, San Francisco, 309 pp.
Bonner, J. L., Harkrider, D. G., Herrin. E. T., Shumway, R. H., Russell, S. A. and Tibuleac, I, M.
(2003). Evaluation of short-period, near-regional MS scales for the Nevada Test Site. Bull.
Seism. Soc. Am. 93(4), 1773-1791.
Bonner, J., Herrmann, R. B., Harkrider, D. and Pasyanos, M. (2008). The Surface wave magnitude
for the October 2006 North Korean nuclear explosion, Bull. Seism. Soc. Am. 98 (5), 2498-2506.
@seismicisolation
@seismicisolation
Bowers, D., Marshall, P. D. and Douglas, A. (2001). The level of the deterrence provided by data
from The SPITTS seismometer array tp possible violation of the Comprehensive Test Ban in
Novaya Zemlya Region, Geophys. J. Int., 46, 425-438.
Burdick, L. J., Wallace, T. and Lay, T. (1984). Modeling near-field and teleseismic observations
from the Amchitka test site, J. Geophys. Res., 89 (B6), 4373-4388.
Cho, C. Shin, J. S. & Kim, G. (2016). Comparison of results of relative location methods and
2016AGU Fall Meeting, San Francisco, 10-12, 2016.
Chung, T. W. (1995). A quantitative study on the crustal structure of the Korean Peninsula based
on the Earthquakes from 1991 to 1994, J. Kor. Earth Science Soc., 16 (2), 152-157.
Cohen, T. J. (1970). Source-depth determination using spectral, pseudo-autocorrelation and
cepstral Analysis, Geophys. J. R. astr. Soc., 20, 223-231.
Closmann, P. J. (1969). On the prediction of cavity radius produced by an underground nuclear
explosion, J. Geophys. Res. 74 (15), 3935-3939.
Douglas, A., Corbishley, D. J., Blamey, C. and Marshall, P.D. (1972). Estimating the firing depth
of underground explosions, Nature, 237, 26-28.
Douglas, A., Marshall, P. D. and Young, J. B. (1987). The P waves from the Amchitka Island
explosions, Geophys. J. R. astr. Soc., 90, 101-117.
Douglas, A. and Hudson J. A. (1990). The effect on the teleseismic P of the zone of damage
created by an explosion. Geophys. J. Int., 103, 111-133.
Douglas, A. Richardson, L. and Hutchins, M. (1990). Surface reflections and S-to-P conversions
on P seismograms, Geophys. J. Int., 100, 303-314.
Dreger, D. (2017). Updated UC Berkeley solution for the mb 6.3 (Mw 5.2) suspected DPRK
nuclear test September 03, 2017 03: 30:00 UTC, [Douglas Dreger, UC Berkeley] Source-type
inversion result for the DPRK nuclear test in
https://ds.iris.edu/ds/nodes/dmc/specialevents/2017/09/03/2017-north-korean-nuclear-test/.
Ford, S. R., Douglas, D. S. and Walter, W. R (2009) Source analysis of the Memorial Day
Explosion Kimchaek North Korea. Geophys. Res. Lett., 36 (21), L21304.
Fox, D. B., Selby, D. N., Heyburn, R. and Woodhouse, J. H. (2012). Shallow seismic source
parameter determination using intermediate-period surface wave amplitude spectra, Geophys. J.
Int., 191, 601-615.
Frasier, C. W. (1972). Observations of pP in the short-period phases of NTS explosions recorded
at Norway, Geophys. J. R. astr. Soc. 31, 99-109.
Gaebler, P. J. and Ceranna, L. (2017). The seismic network of the International monitoring system
(IMS), 69-90, Pilger, Ceranna and Bonnoennemann (eds): Monitoring Compliance with the
Comprehensive Nuclear-Test-Ban Treaty (CTBT), Federal Institute for Geosciences and
Natural Resources (BGR), Hannover, Germany, 328pp.
Gitterman, Y., Kim, S. G. and Hofstetter, R. (2015). Spectral modulation effect in teleseismic P-
waves from North Korean nuclear tests recorded in broad azimuthal range and possible source
depth estimation, Pure and Applied Geophysics, 173 (4), 1157-1174 (2016),
DOI:10.1007/s00024-015-1169-8.
Herrmann, R. B. (2016). Computer Programs in Seismology. Current Version 3.30 is dated May 8,
2016 NP330.May-08-2016.tgz), Saint Louis University, Saint Louis, MO. 186 pp.
Heyburn, R., Neil, D. and Fox, B. (2013). Estimating earthquake source depths by combining
surface wave amplitude spectra and teleseismic depth phase observations, Geophys. J. Int., 94
(2), 1000-1010.
Israelsson, H. (2016). A note on the location of the North Kotran nuclear test on Jan 6, 2016,
Technical Note 2016-01, SeismicInfra Research, Washington DC.
Johnston, W. R (2005). Database of nuclear tests, United States: Part 2, 1964-1972, complied and
last modified 19, June 2005.
Kennett, B. L. N. Engdahl, E. R. and Buland, R. (1995). Constraint on seismic velocities in the
Earth from traveltimes, Geophys. J. Int., 221, 108-124.
@seismicisolation
@seismicisolation
Kim, S. G. and Li, Q. (1998). 3-D crustal velocity tomography in the Central Korean Peninsula,
Econ. Environ. Geol., 31 (3), 235-247.
Kim, S. G. and Bae, H. S. (2006). Investigation of post-sites using local seismic tomography in the
Korean Peninsula, Kr. Soc. Econ. Environ. Geol, 39 (2), 111-128.
Kim, S. G. Gitterman, G. Lee, S. Vavrycuk, V. and Kim, M. (2015). Estimating depth and source
characteristics of nuclear tests by the Democratic People’s Republic of Korea in 2006, 2009 and
2013 using regional and teleseismic network Science and Technology, CTBTO, SnT2015,
Vienna, Austria, 22-26, 2015.
Kim, S. G., Gitterman, Y. Lee, S. and Vavrycu, V. (2016). Depth determination and source
characteristics of the North Korean nuclear tests (2006, 2009, 2013 and 2016) using local and
teleseismic arrays, Final Paper Number: S34A-02, AGU Fall Meeting, San Francisco,
December 12-16, 2016.
King, C. Y., Abo-Zena, A. M., & Murdock, J. N. (1974). Teleseismic source parameters of the
Longshot, Milrow, and Cannikin nuclear explosions, Journal of Geophysical Research, 79(5),
712–718.
Kulhanek, O. (1971). P-wave amplitude spectra of Nevada underground nuclear explosions,
Pageoph, 88, 121-136.
Lay, T. (1991). The teleseismic manifestation of pP : problems and paradoxes, in Explosion
Source Phenomenology, Geophysical Monograph 65, 109-125.
Langston, C. A. (1980). A note on spectral nulls in Rayleigh waves. Bull. Seism. Soc. Am., 70 (4),
1409-1414.
Massé, R. P. (1981). Review of seismic source models for underground nuclear explosions, Bull.
Seism. Soc. Am., 71 (4), 1249-1268.
Montagner J. P, Kennett, B. L. N. (1996) How to reconcile body-wave and normalmode reference
Earth models? Geophys. J. Int., 125: 229-248.
Murphy, J. R. (1981). P wave coupling of underground explosions in various geologic media. In
pp. 201-205.
Murphy, J. R., Stevens, J. L., Kohl, B. C. and Bennett, T. J. (2013). Advanced seismic analyses of
the source characteristics of the 2006 and 2009 North Korean Nuclear Tests, Bull. Seism. Soc.
Am. 103 (3), 1640-1661.
Okal, E. A. (1992). A student’s guide to teleseismic body wave amplitudes. Seism. Res. Lett., 63
(2), 169-180.
Pabian, F. and Hecker, S. (2012). Contemplating a third nuclear test in North Korea, Bull. Atomic.
Scientists, Opinion, 6 August, 2012.
http://www.thebulletin.org/web-edition/features/contemplating- third-nuclear-test-north-korea.
Patton, H. J. and Pabian, F. K. (2014). Comment on “Advanced seismic analyses of the source
Characteristics of the 2006 and 2009 North Korean tests,” by J. R. Murphy, J/ L. Stevens, B. E.
Kohl. T. J. Bennett. Bull. Seism. Soc. Am., 104 (4). 2104-2110.
Patton, H. J. and Taylor, S. R. (2008). Effects of shock-induced tensile failure on mb - MS
Discrimination contrast between historic nuclear explosions and the North Korean test of 9
October 2006. Geophys. Res. Lett., 35 (14), DOI: 10.1029/2008GL034211.
Phillips, W. S., Begnaud, M. L., Rowe, C. A., Steck, L.K., Myers, S. C., Pasyanos, M. E. and
Ballard, S. (2007). Accounting for lateral variations off the upper mantle gradient in Pn
tomography studies, Geophys. Res. Lett., 34, L14312. doi:10.1029/2007GL029338, 1 of 5-5 of
5.
Richter, C. F. (1958). Elementary Seismology, W. H. Freeman and Company, San Francisco and
London, 708 pp.
Rodgers, A. J., Petersson, N. A. and Sjogreen, B. (2010). Simulation of topographic effects on
seismic waves from shallow explosions near the North Korean nuclear test site with emphasis
on shear wave generation, J. Geophys. Res., 115 (B11309):1-27 (2010).
doi:10.1029/2010JB007707.
@seismicisolation
@seismicisolation
Rodnikov, A. G., Sergeyeva, N. A., Zabarinskaya, L. P. Filatova, N. I. Piip, V. B. and Rashidov, V.

A. (2008). The deep structure of active continental margins of the Far East (Russia), Russian J.
Earth Sciences 10, ES4002, doi:10.2205/2007ES000224,
Sead, R. and Helmberger, D. V. 1988). Numerical-analytical interfacing in two dimensions with
applications to modeling NTS seismograms, PAGEOPH, 128 (1/2), 157-193.
Springer, D. L. (1974). Secondary sources of seismic waves from underground nuclear explosions,
Bull. Seism. Soc. Am., 64 (3), 581-594.
Stein, S. and Wiens, D. (1986). Depth determination for shallow teleseismic earthquakes: methods
and results, Reviews Geophys, 24 (4), 806-832.
Stump, B. W. (1985). Constraints on explosive sources with spall from near-source waveforms,
Bull. Seis, Soc. Am. 75 (2). 361-377.
Tian, D., Yao, J. and Wen, L. (2018). Collapse earthquake swarm after North Korea’s 3
September2017 nuclear test. Geophys. Res. Lett., 45. 1-8. DOI:org/10.1029/2018GL077649
Tsai, Y. -B. and Aki, K., (1971). Amplitude spectra of surface waves from small earthquakes and
underground nuclear explosions, J. Geophys. Res., 76 (17), 3940-3952.
Udias, A. (1999). Principle of Seismology, Cambridge University Press, United Kingdom, 475 pp.
Van Der Hilst, R. D., Kennett, B. L. N. and Shibutani, T. (1998), Upper mantle structure beneath
Australia from portable array deployments, Geodynamics, 26, 39-57.
Vavryčuk, V., Kim, S. G. (2014). Non-isotropic radiation of the 2013 North Korean nuclear
explosion, Geophys. Res. Lett., 41 (20), 7048-7056. DOI: 10.1002/2014GL06126.
Viecelli, J. A. (1973). Topography and the Rayleigh wave generating efficiency of buried
explosive Sources, J. Geophys. Res., 78 (17), 3334-3339.
Wen, L.and Long, H. (2010). High-precision location of North Korea’s 2009 nuclear test, Seis.
Res. Lett., 81, 26-29.
Wessel, P. and Smith, W. H. F. (1998). New, improved version of the Generic Mapping Tools
released, EOS Trans. AGU 79, 579. doi:10.1029/98EO00426.
Zhang, M. and Wen, L. (2013). High-precision location and yield of North Korea’s 2013 nuclear
test, Geophys. Res. Lett., 40, 2941-2946.
Zhao L-F. Xie, X-B. Wang, W-M. Fan, N, Zhao, X. and Yao, Z-X. (2017). The 9 September 2016
North Korean underground nuclear tes, Bull. Seism. Soc. Am., 107 (6), 3044-3051.
@seismicisolation
@seismicisolation
APPENDIX A
Basic Equations for Depth Calculation
a b
Fig. S1. a) Ray trajectories of P, pP and sP at a source of the nuclear test. i and j are take-off angles for pP and sP ray-
paths. G, G', R and R' indicate the contact points on the surface and on the ray-path by pP and sP. S and O represent
seismic stations and the center of the Earth, respectively using the flat Earth model. b) Ray trajectories of P, pPn and
sPn at a source of F. α is a crtical angle Pn (take-off angle of pPn) and β is a take-off angle of sPn. Depth can be
calculated using travel time delays of pPn-Pn and sPn-Pn
The interference produces spectral modulation (scalloping) – minima (spectral nulls) at the
fundamental and multiple frequencies of the spectra for depth phase waveforms. The spectral null
of the fundamental mode is represented as the delay time difference between the onset of the P-
wave arrival and that of the downswing depth phase (pP-P) in the fundamental ray-path theory. So
the source depth is derived as follows:
d = (pP-P)Vp/2cos (i) (1)
where Vp is the average compressional wave velocity of the medium between the source and the
surface (the overburden velocity), d and i are source depth and take-off angle, respectively. In case
of a small take-off angle of pP for the long range epicentral distance (∆≥ 75.0°), we approximate
equation (1) as d ≈ pP-P *Vp/2 where pP-P is a dealy time difference between P and pP. The
estimated depth using sP-P delay times is derived as follows:
d = (sP-P)VpVs/[Vp cos (j) + Vs cos (i)] (2)
where sP-P is the delay time between the onset of P-wave arrival and that of the downswing depth
phase of sP and j is the take-off angle of sP (j <i).and obtained from j=arcsin [Vs/Vp sin (i)]
We estimated spectral nulls of the average spectra via stacking traces to determine the source
depth. From Snell’s law, i=arcsin[V0/Vapp], where i=take-off angle for the focal sphere, V0= P-
wave velocity at the focal depth assuming that it is the average P-wave velocity between the
source and the surface (Vp=5.1 km/sec). Vapp = the bottoming Vp velocity where the ray returns
at i=90° where Vapp is an apparent velocity at distance, a reciprocal of the travel-time curve slope
(𝜕𝑇/𝜕∆ ). Using the take-off angle i, the depth of burial for the 2016J, 2016S and 2017 tests were
estimated in Figs 2, 3, and 4 and Tables 1, 2 and 3. The source depth d is estimated by delay times
pPn-Pn using take-off angles of pPn heading for the Moho discontinuity and Snell’s law as follows:
@seismicisolation
@seismicisolation
From Fig. S1b,

pPn-Pn = 2d/Vpcosα -2dtanα/Vn = 2d/cosα (1/Vp – tan α/Vn) (3)
= 2d/cos α (Vn-Vpsinα)/VpVn = 2d/Vpcos α (1 – Vp/Vnsinα)
= 2d/Vpcosα (4)
Where α = arcsin (Vp/Vn)
Therefore, depth d = (pPn-Pn)Vp/2cosα (5)
sPn-Pn = d/(cos β Vs)+ d/(cosαVp) – (dtanβ/Vn + dtanα/Vn)
= d(1/cosβVs + 1/cos αVp – tan β/Vn –tan α/Vn) (6)
Where β = arcsin (Vs/Vp sin α)
Therefore , depth d =(sPn-Pn)/F(Vp, Vs, Vn, α, β) (7)
where
F(Vp, Vs, Vn, α, β) = (VpVncosα + VsVncosβ – VpVscosαsin β – VpVscosβsin α)/
VpVsVncosαcos β = [cosα (VpVn-VpVs sinβ)+ cos β (VsVn-VpVs sin α)]/
cos αcosβVpVsVn (8)
For example: Vp=5.1 km/sec, Vs=2.9 km/sec, Vn=7.8 km/sec, F=0.4110747
α =arcsin (5.1/7.8)=40. 83⁰, β =arcsin (Vs/Vpsin α)=21.83⁰ according to Snell’s law.
α =arcsin (5.1/7.8)=40. 83⁰, β =arcsin (Vs/Vpsinα)=21.83⁰
Vn is the Moho discontinuity velocity and Vs is the overburden average S-wave velocity between
the source and the surface. Vn depends upon the Moho discontinuity of the different crustal
structures such as the Central part of the Korean Peninsula (Vn=7.8 km/sec) and the the northern
continental path-stations such as MDJ and USRK (Vn=8.0 km/sec) in Fig. 5 and Table 4.
@seismicisolation
@seismicisolation
APPENDIX B
Relationship between a spectral null of fundamental-mode Rayleigh wave and depth in a half-
space.
It should be noted that the particle motion of the fundamental-mode Rayleigh waves becomes
prograde below a depth of about at z = - 0.192 λ, (λ =wavelength). There is a value for the
horizontal displacement u1 is null, z = -0192 λ, whereas the vertical displacement u3 is never null.
The spectral null in Rayleigh wave excitation is due to a zero-crossing of the horizontal
displacement eigenfunction. This special feature has been used for estimating source depths of
underground explosions as well as earthquakes (Tsai and Aki, 1971; Fox et al., 2012). At a depth
of the nodal plane u1 is null and vanishes, Its amplitude changes sign and below the depth of the
nodal plane and the particle motion is prograde. With increasing depth, the amplitudes of u1 and u3
decrease exponentially, with u3 always larger than u1. The purpose of this section is to find the
spectral null that is a combination of displacement eigenfunction and its derivatives of the
fundamental-mode Rayleigh wave controlled by CLVD source.
It is worth noting that the velocity of Rayleigh wave does not depend upon the frequency of the
disturbance since they are nondispersive in this study.
For Rayleigh waves of frequency ω that propagate in the x direction with velocity c, potential φ
(pure dilatational) and ψ (pure rotational) in homogeneous and isotropic half–space,
φ =f(z) exp[ik(x-ct)]
ψ=g(z) exp[ik(x-ct)] (9)
where f(z) and g(z) present an amplitude, i, k and c indicate √(−1) , wave number (2πf/c) and
Rayleigh wave velocity, respectively. by inserting φ and ψ into wave equations as below:
∂2f/∂z2 + kr2f(z) = 0
∂2g/∂z2 + ks2g(z) = 0 (10)
where r=(c2/Vp2-1)1/2, s=(c2/Vs2-1)1/2
The solutions of equation (10)
f(z)=A exp (ikrz) + D exp (-ikrz)
g(z)=B exp(iksz) + E exp (-iksz) (11)
D=0 and E=0 since the values of f(z) and g(z) decrease with depth.
The equation (9) can be rewritten as follows:
φ =A exp[ikrz +ik (x-ct)]
ψ=B exp[iksz + ik(x- ct)] (12)
Using boundary conditions at free surface (P31=P32=P33=0), numerals for r=0.8475i, s=0.3933i and
B=-1.4675iA the horizontal displacement eigenfunction u1 and vertical displacement eigenfunction
u3 are as follows:
u1 = ∂φ/∂x – ∂ψ/∂z
u3 = ∂φ/∂z + ∂ψ/∂x (13)
from equation (12)
@seismicisolation
@seismicisolation
u1 = Aik[exp(-0.8475kz – 0.5773exp (-0.3933kz)]exp[ik(x-ct)]

u3 = Aik[0.8475exp(-0.8475kz – 1.4679exp(-0.3933kz)]exp[ik(x-ct)] (14)
Considering taking real parts, the horizontal and vertical displacement eigenfunctions for the
fundamental-mode Rayleigh waves are as follows:
u1 = -Aksin (kx- ωt) [exp (-0.8475kz) – 0.5773exp (-0.3933kz)]
u3= -Akcos (kx- ωt) [0.8475exp (-0.8475kz) – 1.4679exp (-0.3933kz)] (15)
where A= amplitude, ω=2πf and k=2πfh/c when z=d
From equation 14, at the depth where u1 is null, its amplitude changes sign.
1-0.5733exp(0.4542kz) = 0 (16)
Therefore d = 0.1923c/fnull (17)
where c=Rg-wave velocity, 1.46 km/sec from Rayleigh Equation (Udias, 1999) and fnull
spectral minimum of the Fundamental-mode Rayleigh wave amplitude spectra.
APPENDIX C
Calculation of a Cavity Radius for Underground Nuclear Explosions (Closmann, 1969)
Cavity radius Rc=21.0W0.306E0.514ρ-0.244μ-0.576h-0.161, where Rc = cavity radius in meters, W = yield
in kt, E and μ=Young’s and shear moduli in megabars, ρ = overburden density in grams per cubic
centimeters, and h = depth of burial in meters.
APPENDIX D
Presentation at SnT2019, CTBTO, Vienna, Austria on 24-28 2019.
https://www.youtube.com/watch?v=VWdBS6jS4jM&t=491s
@seismicisolation
@seismicisolation
CHAPTER TWENTY-TWO
DEPTH ESTIMATE OF THE DPRK’S 2006-10-09, 2009- 05-25 AND 2013-

02-12 NUCLEAR TESTS USING SPECTRAL NULLS OF BODY WAVES
AND FUNDAMENTAL- MODE RAYLEIGH WAVE AMPLITUDE SPECTRA 1
S. G. KIM, Y. GITTERMAN, S-K. LEE AND V. V$95<ý8.
DPRK's underground nuclear explosion sketch at a depth of around 2 km with gas cavity’s radius
of around 15-57 m
ABSTRACT
nuclear explosions on October 9, 2006 (mb 4.3), May 25, 2009 (mb 4.7) and February 12, 2013 (mb
5.1). We expanded standard depth calculation methods to encompass full depth utilizing spectral
nulls of spectra using body waves and Rayleigh waves from local and teleseismic arrays. The depths
of burial were calculated from the fundamental-mode Rayleigh wave spectra as well as the average
spectra of depth phase (pP+P/sP+P, pPn+Pn/sPn+Pn) considering take-off angles based on the flat
Earth model. The average depths of the 2006, 2009 and 2013 nuclear tests were thus estimated at
2.12 km, 2.06 km and 2.05 km, respectively. It should be noted that those depths are significantly
greater than expected from the standard experiment practice. The possibility of the over-buried
nuclear tests would affect the interpretation of magnitudes as well as the estimation of the yield.
1
Some parts presented in J. Asian Earth Sciences, 163 (2018), 249-269. Depth estimate of the DPRK’s 2006-10-09, 2009-05-
25 and 2013-2-12 underground nuclear tests using local and teleseismic arrays by S. G. Kim, Y. Gitterman and S-K. Lee
@seismicisolation
@seismicisolation
462 Chapter Twenty-Two
Keywords: CTBT, depth phase, spectral null, Rg-wave, CLVD, Mt. Mantap
INTRODUCTION
Source depth is pertinent for ensuring enforcement of the Comprehensive Nuclear-Test-Ban-Treaty
(CTBT) as nuclear explosions are unlikely to be fired at depths of more than a few kilometers (Bakun
and Johnson, 1973; Bowers et al., 2001; Bonner et al., 2002). Two basic techniques of body waves
and surface waves are utilized for depth analysis. Surface wave spectra can be used to determine
depths, as mode excitation reflects the behavior of the eigen functions with depth. The primary
limitation on this study arises from lateral heterogeneity of velocity structure and mechanism. The
waveforms of body waves are also used to estimate depths by the time separation between the direct
P-wave arrivals and near-source surface reflections. These studies can be limited by a partial trade-
off between depth and source time function duration, but the trade-off is found generally not
producing major difficulties for small nuclear explosions (Stein and Wiens, 1986). The DPRK
conducted the first nuclear test (mb 4.3) on October 9, 2006, the second one (mb 4.7) on May 25,
2009 and the third one (mb 5.1) on February 12, 2013 beneath the vicinity of Mt. Mantap in the
Punggye-ri region of northeastern DPRK (Fig. 1a and Table 1). Three white spots at the bottom (Fig.
1b) are considered as the tunnel entrances AKA West, South and East portals (Pabian and Hecker,
2012; Israelsson, 2016), but they were built before the 2009 and 2013 nuclear tests implying pre-
existed tunnels.
Despite the excavation of deep tunnels, the small and little spoil pile appeared in front of portals
because the spoil may have been removed during night or may have been backfilled in the old mine
tunnels. Most recently Pabian and Coblentz (2017) described a rather large piled spoil in front of
west and north portals with a detailed topography map from the satellite image implying that a large
nuclear explosion is impending.
A b
Fig. 1. a) Geological map of the nuclear test site and surrounding area (reproduced from the Geological Society of Korea,
Kim et al., 2017). 1. The Mesozoic intrusive granite rock in Kwanmo Massif contains the nuclear test sites, 2. Quaternary
volcanic rocks in Kwanmo Massif, 3. Kilju-Myongchon Graben (Cenozoic sedimentary rocks), 4. Machollyong supergroup
(Proterozoic rock), 5. Rangnim Massif (Proterozoic-Archaeozoic), 6. Triassic intrusives, 7. Hayang group (Cretaceous), 8.
Daedong supergroup (Jurassic-Triassic), 9. Sangwon system . (Proterozoic-Archeozoic). b) Satellite image of Mt. Mantap
(2205 m) and surrounding area near the test site was produced by French Satellite ‘Spot-5’ on May 14, 2009 before the
second test. The scale is 1:25,000. Stars and arrow marks represent the nuclear test sites in granite rock determined by
IDC/CTBTO and three tunnel entrances in red arrow marks in Fig. 1b.
@seismicisolation
@seismicisolation
Depth Estimate of the DPRK’s 2006-10-09, 2009- 05-25 and 2013-02-12 Nuclear Tests Using Spectral Nulls 463
of Body Waves and Fundamental- mode Rayleigh Wave Amplitude Spectra
Although depth determination through the measured time differences between the direct P-wave (P
and Pn) and the depth phases (pP, sP, pPn and sPn) is difficult for shallow sources, spectral analysis
techniques can be effectively used by the identification of spectral nulls at seismic arrays owing to
the destructive interference between the direct and surface reflected arrivals. To provide robustness
of the result, we consider mainly teleseismic arrays and local array of KSRS including some regional
stations which include broad observations at a wide range of azimuths from the source to the stations
(Fig. 2a). For a basic raypath theory, simple formulas are developed to estimate the source depth
from identified spectral nulls in this study.
Table 1. Parameters of North Korean nuclear tests. Origin times, magnitudes and depths were
extracted from United States Geological Survey (USGS) and International Data Centre (IDC,
CTBTO). Seismic yields in M/B were calculated from methods utilized by Murphy (1981) and
Bowers et al. (2001), respectively.
Date OriginLocation** mb Yield (kt) Depth***(km)
M/D/Y time* Lat, ºN, Lon. ºE USGS/IDC M/B
10/09/2006 01:35:28
41.294, 129.094 4.3/4.1 1.67/1.17 0 (2.12)
41.3119, 129.0189
05/25/2009 00:54:43 41.306, 129.029 4.7/4.5 9.18/3.98 0 (2.06)
41.3110, 129.0464
02/12/2013 02:57:51 41.308, 129.076) 5.1/4.9 28.63/13.59 0 (2.05)
41.3005, 129.0652
a
UTC (Universal Time Coordinated) by USGS. **The lower and upper locations determined by
USGS and IDC. ***Depths in the brackets refer to this study.
Explosions in hard rock, like granite or tuff give considerably stronger amplitudes than explosions
in unconsolidated rock such as alluvium. The strength of a seismic signal (magnitude) strongly
depends on not only the medium in which an explosion is detonated but also on the wave
transmission properties of the Earth for different source-receiver paths.
a b
Fig. 2. a) Open squares represent teleseismic arrays in almost uniform azimuthal coverage including KSRS. Open
triangles represent DSN (Dongbei Seismic Network), MDJ and BJT (closed triangle) of China and South Korean
stations. The red star indicates the underground nuclear test site of North Korea. b) Ray trajectories of P, pP and sP at
source of the nuclear test. i and j are take-off angles for pP and sP ray-paths based on the flat Earth model. G, G', R and
R' indicate the contact points on the surface and on the ray-path by pP and sP. Symbols S and O represent seismic
stations and the center of the Earth, respectively.
@seismicisolation
@seismicisolation
Various researchers derived different magnitude-yield relations depending on the detonation site
(Table 1). For example, using mb=3.92 + 0.81 logY for wet hard rock (Murphy, 1981) gives a 29 kt
yield for the 2013 nuclear test, while we alternatively obtained a 14 kt yield, when using mb = 4.25
+ b logY for a fully coupled explosion, where Y is the explosive yield in kilotons (kt), with b=0.75
IRU<DQGE IRU<ZKLFKLVVXLWDEOHIRUFRQVHUYDWLYHO\HVWLPDWLQJWKH\LHOGWKUHVKROGRID
potential violation of the CTBT in the Novaya Zemlya region (Bowers et al., 2001). Seismic signals
from a nuclear explosion can also be reduced by detonating an explosion in a large underground
cavity due to decoupling effects (Harjes et al., 1985; Murphy et al., 1996). Because emplacement
conditions are unknown for North Korean nuclear explosions, it is most difficult to estimate seismic
yields of underground clandestine nuclear explosions. It is very worth noting that seismic yield,
burial depth, medium and configuration play significant roles in the seismic signals of underground
nuclear explosions.
DEPTH ESTIMATION USING pP-P AND sP-P DELAY TIMES
The usual way to estimate the depth of an event using teleseismic waves is to measure the time
difference between the arrival time of the direct P wave and that of the surface reflections pP and
sP. The phases pP and sP are signals that travel upward from a focus as P and S waves (Douglas et
al., 1972; Massé, 1981; Okal, 1992) and then P and the converted P waves from SV travel steeply
downwards following the same path as to the receiving stations of the P waves (Fig. 2b). The relative
amplitudes and polarities of these phases depend on the source mechanism while the time delay
depends on the velocity structure and the ray paths. For shallow sources, identification of individual
phase arrivals is difficult in the time domain because the arrival pulses interfere with each other.
However, a spectral hole (null), due to the destructive interference, has a strong effect on the
amplitude spectra that may be exploited to estimate the depth.
To provide waveforms for spectral analysis, we used P-wave signals of the vertical components
from nine teleseismic arrays (Fig. 2a). Such stacking increases the S/N of the P-wave signals by
suppressing the effects of random seismic noise. We performed spectral analyses taking 10 or 30
second time-windows of short-period vertical components for the ASAR (Alice Springs, Australia),
FINES (FINESS, Finland), KURK (Kurchatov, Kazakhstan), NOA (NORSAR, Norway), NVAR
(Mina, USA), PDAR (Pinedale, USA) including South Korean local stations and KSRS (Korea
Seismological Research Station, Wonju, South Korea) and broadband records for WRA
(Warramunga, Australia) and YKA (Yellowknife, Canada), ISN (Israel Seismic Network) and
Dongbei Seismic Network (DSN) along the border between northeastern China and North Korea
(Fig. 2a). We can remove the effects of absorption and the recording system by utilizing a source in
the same high velocity layer and the same array stations. We determined spectral nulls from
destructive interference of pP+P and sP+P phases equivalent to a reciprocal of pP-P and sP-P lag
time so that the pP delay times estimated from the spectral analysis are compatible with the pP phase
arrivals in the time domain. The high quality factor Q in the granite source rock (1000) from
amplitude A(t) = A(0)exp(-ʌIW4 DQGORZDWWHQXDWLRQRIVRXUFHLQWKHKLJKYHORFLW\OD\HUFOHDUO\
reveal high frequency arrivals of pP and sP in the teleseismic array's seismograms.. However, as the
depth phase sP is a conversion of P to SV wave and very sensitive to the above-source structure and
source configuration, it appears infrequently in seismic waves radiated by an explosion. However,
in the North Korean nuclear tests we often see sP phases, thus, utilizing the sP phase compensates
for lack of pP phase data for estimating the depth.
The interference produces spectral modulation (scalloping) – minima (spectral nulls) at fundamental
and multiple frequencies (Cohen, 1970; Kulhanek, 1971; Bakun and Johnson, 1973; Heyburn et al.,
2013; Husebye et al., 2013). The spectral null f01 of the fundamental mode is represented as follows:
f01 = Vp/[2cos (i)*h], f0n = n f01 , n = 1, 2, … (1)
@seismicisolation
@seismicisolation
nulls of the spectra averaged from the respective array traces (Fig.3 and Table 2).
Fig. 3. The blue, red and black arrows indicate spectral nulls of pP+P, sP+P and pPcP+PcP. The sP phases are observed
at NVAR for the 2006 and 2013 tests and at PDAR for the 2009 and 2013 tests. The PcP phase appears only at PDAR
with a distance of 81°.
@seismicisolation
@seismicisolation
Fig. 4. The P-wave spectral nulls are determined using pP+P wave spectra at ASAR and WRA. The same spectral nulls
of 1.10 Hz at ASAR, while spectral nulls of 1.35 Hz, 1.25 Hz and 1.25 Hz at WRA are estimated for the 2006, 2009 and
2013 nuclear tests of North Korea.
Fig. 5. The open blue arrows and red solid arrows indicate spectral nulls of sP-P at around 1.1 Hz for the 2009 and 2013
and spectral nulls of pP-P at around 1.6 Hz for the 2006, 2009 and 2013 nuclear tests at FINES. The delay times of pP-
P are much shorter than those of other arrays which may be due to the high-velocity lower crust through the stable
Fennoscandian Shield with high Q.
@seismicisolation
@seismicisolation
In case of a small take-RIIDQJOHRIS3IRUDORQJUDQJHHSLFHQWUDOGLVWDQFH ¨ ZHPD\WDNHDQ

approximate equation as K§WpP-P*Vp/2 from equation (1), which was utilized by many researchers
(Kulhanek, 1971; Douglas et al., 1972; Gitterman et al., 2015). Ideally, the pP-P time would be
obtained from the time difference between the onsets of P and depth phase pP in the time domain.
However, those onsets are difficult to pick so an alternative is to use spectral nulls of the average
spectra because the depth phases are always 180° out of phase with the incoming P-wave arrivals.
i.e. we measured the time difference between the maximum positive amplitude P-wave arrival and
the maximum negative amplitude of pP assuming P and pP have an identical shape with negative
polarity and interference between the two pulses (Douglas et al., 1987). The absence of depth phase
is the result of scattering and defocusing of P waves by lateral variation of medium above the source
(Douglas et al., 1990). However we can minimize the nonlinear topographic effects of the
mountainous source and the azimuthal variations of pP waveforms (Burdick et al., 1984) using the
average spectra obtained from a broad and good azimuthal distribution of teleseismic arrays (Fig.
1a). We used spectral nulls of the average spectra of multi-pathing arrays and utilized the full depth
formula with take-off angles assuming the average overburden P-wave velocity to be Vp=5.1 km/sec
(Bonner et al., 2008) at the DPRK nuclear test site. All spectra of single array show similar ‘spectral
nulls’ indicating that the pseudo-1D radiation in one direction is constant and neglecting the
contribution of different source components. So data from a test in the middle of the mountain area
will be look alike in all directions since an onset time of P-wave arrivals at each element of an array
follows Fermat Principle and Snell’s Law which indicate that a raypath corresponds to a minimum
of travel time and that an incident angle (take-off angle) is always the same as a reflected angle of a
single raypath within a limited time window. It should be noted that there is no elevation difference
between a source and a reflector with a shallow dip as the closest approximation possible on average
in the present state of evidence (Table 2). Consequently we do not count on the unknown and
complicated terrain of the Mt Mantap area (Fig. 1b) and neglected the trade-off between our applied
simplifications and the claimed accuracy of the results from the nonlinear topographic effects
considering almost uniform azimuthal coverage by means of the global seismic arrays from the
source.
From the relationship between ray paths and travel times, W[ VLQ i)/v0 =1/vz for a flat earth,
where i= a take-off angle for the reflected pP wave, v0= P-wave velocity at the focal depth, vz = the
bottoming average Vp velocity (vav) where the ray returns at i=90° where 1/Vz =1/vapp= (travel-time
curve slope). Using the take-off angle i, where i= arcsin (v0/vav) (Richter, 1958), the time delay for
pP-P with 1/cos (i) from equation (2) and P-wave velocity at the focal depth, v0 as 5.1 km/sec, we
estimated the average depths of burial for the 2006, 2009 and 2013 at 2.10 km, 2.09 km and 2.09
km, respectively (Figs 3, 4, 5 and Table 3). Lay (1991) stated that the source depth estimated by pP-
P delay times may be overestimated biasing effects such as delay times of depth phase propagation
for very short lag times with limited bandwidth. However, we offset these effects by using the
average spectra of signals from different azimuthal propagation of seismic arrays and take-off
angles.
h = (pP-P)Vp/2cos (i) (2)
The estimated depth using sP-P delay times is derived using Snell’s law as follows (Fig. 1b):
h = (sP-P)VpVs/[Vp cos (j) + Vs cos (i)] (3)
NOA, NVAR, PDAR and broadband vertical components from the WRA and YKA. We used pP+P
depth phases from all arrays, but we also used some sP+P phases in NVAR for the 2006 and 2013
tests and PDAR for the 2009 tests and 2013 which particularly appear in Fig. 3.
@seismicisolation
@seismicisolation
Source depth can be estimated for a simple single-layer over a half space model from the time delay
between Pn and pPn using the following equations. The source depth by means of utilizing pP-Pn
can be derived considering the take-off angle for the Moho discontinuity with its velocity Vn and
the overburden average velocity Vp as follows:
h= (pPn-Pn)Vp/2cos(Į) (4)
h=(sPn-Pn)/F (Vp, Vs, Vn, Įȕ) (5)
ZKHUH) 9S9V9QĮȕ >FRVĮ 9S9Q-9S9VVLQȕ FRVȕ 9V9Q-9S9VVLQĮ @
FRVĮFRVȕ9S9V9Q(see the deriving procedure from APPENDIX A in Chapter Twenty-one).
where Į and ȕ are take-off angles for pPn and sPn phases. Į = arcsin(Vp/Vn) and ȕ =arcsin[Vs/Vp
sin (Į)] according to Snell’s law. We used the average spectra from spectral nulls for destructive
interference pPn + Pn since the ray paths should pass through the heterogeneous and anisotropic
crustal structures to reach each station of the array. We can determine burial depths for the near-
field nuclear explosions using equations (4) and (5) assuming that Vp and Vs are 5.1 km/sec and
Vs=2.9 km/sec and that Vn=7.8 km/sec from the refracted P-wave velocity of ray paths through the
central part of the Chugaryeong Rift Zone and the subcontinental crust of the East Sea (Chung, 1995;
Kim and Bae, 2006; Phillips et al., 2007), but for MDJ data Vn=8.0 km/sec. Thus the source depths
are roughly estimated at 2.11 km, 2.01 km and 2.00 km for the 2006, 2009 and 2013 tests,
respectively, using data from local stations (Table 4). The CHC station (Chunchon, South Korea)
for the 2013 test was slightly moved (approximately 20 m) from the previous position occupied for
the 2009 test and this site movement was reflected in spectrum of the CHC station for the 2013 test
(Figs 6 and 7).
@seismicisolation
@seismicisolation
Fig. 6. Spectral nulls from depth-phase interference of pPn +Pn spectra for the CHC, KSRS, MDJ, SEO and SES. The
spectral nulls of pPn + Pn which are revealed in the waveforms are smaller than those of spectral nulls of pP+ P for the
teleseismic P waves. The source spectra and spectral nulls for the 2009 and 2013 are almost identical for data from the
KSRS, MDJ, SEO and SES. Data from for CHC station were slightly different because the station position was slightly
changed. MDJ is a pure-continental raypath station showing source characterization of each explosion.
@seismicisolation
@seismicisolation
Fig.7. The spectral characteristics for the 2006, 2009 and 2013 nuclear tests from local stations. The spectra for the 2009
and 2013 tests are identical compared to the 2006 test, indicating that the raypath from source to station for the 2006
test is quite different from the 2009 and 2013 tests even if the source parameters and depths are almost identical for all
three tests. A slightly different shape of the spectra for the Chunchon (CHC) station is due to a slight change in the site
position of the seismometer. This indicates that the 2006 test may be different from source mechanism and/or depth
compared with the 2009 and 2013 tests
DEPTH INTERPRETATION USING pP-P AND pPn -Pn DELAY TIMES

We estimated spectral nulls to determine depths taking average spectra with the band-pass filtered
between 0.5-3.0 Hz taking into account the fundamental-mode alone. Here we presented the only
spectra of ASAR, FINES, NVAR, PDAR and WRA as representative teleseismic arrays because of
their specific locations (Figs 3, 4 and 5). Overall, the average spectral nulls of pP+P destructive
interference for the 2006, 2009 and 2013 nuclear tests were estimated at 1.29 Hz, 1.28 Hz and 1.27
Hz selecting the spectral holes of the fundamental-mode spectra, respectively (Table 2). There are
some holes other than what we expect to pick up at NVAR and PDAR for the 2009 and 2013 tests
in Fig. 3. The spectral nulls at lower frequencies than pP+P at NVAR and may be those of sP+P, but
those at PDAR may be mixed with pPcP+ PcP phase (see Fig. 3). The higher frequencies after pP+P
may be related to multiples due to ambient wind noise from seasonal variation and slowness-azimuth
errors (Wang et al., 1999) of the high elevation arrays. The small discrepancies of null frequencies
between different arrays are attributed to the site effects of the array location and the effects of the
propagation path for the depth phase. In general, there is not a much greater variation in delay time
estimates for a deep source as shown in the Amchitka tests (Lay, 1991). The depth phase method is
most successful when P-wave arrival has a signal-to-noise ratio (SNR) greater than 4-7 and the depth
phase pP exhibits a SNR greater than ~2 (Bonner et al., 2002). In this study, a S/N is good enough
not to consider any noise spectra. Nonetheless, the delay times of the teleseismic arrays are very
sensitive to the propagation path effects of the depth phase pP. The low frequency of spectral nulls
at the ASAR (Fig. 4) may be more closely related to the low velocity layer of the large aquifers
beneath Alice Springs area in the upper crust and possibly including the low P-wave velocity layer
@seismicisolation
@seismicisolation
in the upper mantle of the Late Palaeozoic Alice Springs Orogeny beneath the array according to
the tomography study (Van Der Hilst et al., 1998), whereas the anomalously high frequencies at the
FINES array (Fig. 5) may be attributed to a high-velocity lower crust through the Archean-
Proterozoic of the Fennoscandian Shield beneath the FINES array with a small aperture of 2 km
(Yliniemi et al., 2004). We also found abnormally higher or lower spectral nulls due to the bottom
of the 660-km discontinuity of the Mantle Transition Zone (410 km-660 km) at a depth around 600
km at FINES and ASAR as compared with at PDAR in Fig.8. These phenomena are attributed to
higher crustal velocity beneath the FINES seismic array and the lower crustal velocity beneath the
ASAR seismic array which is located in the Great Artesian Basin in the Central Australia. As a
result, we also found the abnormally higher or lower spectral nulls to be related to the site effects of
observing stations. We presumed that the deep-focus earthquake occurred at a depth of around 3 km
from the bottom of the 660-km Discontinuity.
explosion depth and the P-wave velocity in the material above the shot point (Douglas and Hudson,
1990) because the corresponding pP arrival would be expected to be reduced in amplitude and
possibly delayed relatively to what would be expected for true pP in case of the effect of a zone of
damage above an explosion. The reflected signals containing the fracture and crushing associated
with the explosion would be reduced and delayed resulting in a slapdown phase as a secondary
source. From a known firing depth of 0.463 km in granite and a P-wave velocity of 4.8 km/sec, they
found it impossible to see the predicted delay times of 0.2 sec from pP-P, because the pP-P delay
times were found to be 0.8 sec (corresponding to a depth at about 2 km) from the Pile Driver
explosion (Kulhanek, 1971; Douglas and Rivers, 1988) since it was conducted at a shallow depth
(around 500 m) of granite resulting in damage of fracture and crushing from a spalling. However in
this study, it should be noted that North Korean nuclear explosions were conducted in the deep
homogeneous granite without the surface spallation indicating that the spectral nulls of observations
are reasonably consistent with those of synthetic spectra for local and teleseismic waveforms. We
also neglect the diffraction effects of depth phases from the edge of a spalling in this study (Douglas
and Hudson, 1990). There are some anomalous examples of a deep-depth estimate using the depth
phase pP-P for a deep atomic explosion (Douglas et al., 1972). On May 21, 1968 a 47-kiloton (mb
=5.6) atomic explosion (detonated in salt) was fired at a depth of 2.45 km near Bukhara, USSR to
seal an oil well. The geology in the vicinity of the explosion is a sedimentary basin in which an
overburden velocity of 3.0 km/sec (Douglas et al., 1972). It was observed by pP-P delay times of
1791 m of a depth below the free surface (HOB= –1700 m) (Johnston, 2005) corresponding to the
observed spectral nulls of 0.94 Hz (1.06 s) by pP+P spectra using single channels at GBA, EKA and
YKA stations from the Cannikin explosion of the Amchitka Island explosions (mb=6.8, yield=5 Mt).
The low spectral null of 0.94 Hz or a longer pP-P delay time (1.06 s) also attests to overestimation
by an overshoot of pP pulses (Douglas et al., 1987). It is due to the slapdown phase from spallation
at a certain depth of subsurface as well as surface reducing P-wave velocity in the crushed rock
around the explosion in the island site.
Table 2. Spectral nulls (Hz) of pP+P spectra for the 2006, 2009 and 2013 tests using teleseismic
arrays.
Year ASAR FINES ISN KURK NOA NVAR PDAR WRA YKA Mean
2006 1.10 1.66 1.19 1.25 1.25 1.35 1.24 1.29±0.17
2009 1.10 1.63 1.25 1.23 1.17 1.35 1.25 1.25 1.27 1.28±0.14
2013 1.10 1.63 1.22 1.23 1.17 1.25 1.28 1.25 1.30 1.27±0.14
* ISN, KURK and NOA (sub-array of NORSAR) data were obtained from another study (Gitterman et al., 2015).
@seismicisolation
@seismicisolation
The depths for the 2006, 2009 and 2013 tests determined via sP - P delay times of equation (3) were
used only for NVAR and PDAR since greater uncertainty and a few observations exist in the use of
sP-P delay times due to ambiguous spectral nulls in the spectra from weak S-wave generation. For
sources where the depth phases have been accurately identified, the error in the estimated source
depth will be much less.
a b
c d
Fig. 8. a, b and c indicate seismograms and spectra of P waves reflected at the bottom of 660-km Discontinuity. The
spectral nulls at ASAR, FINES, and PDAR are due to the reflection from the 660-km Discontinuity by a deep-focus
earthquake which occurred at a depth of around 600 km in the NE China on January 2, 2016 (depth=585.5 km, M=5.8).
We found spectral nulls at ASAR, FINES and PDAR to be 1.48, 1.88 and 1.64 Hz respectively. The low spectral null
at ASAR are attributed to the Great Artesian Basin beneath the seismic array which includes a large aquifer with water-
bearing formation, whereas the high spectral nulls at FINES are due to the higher crustal velocity of the Fennoscandian
Shield beneath the seismic array resulting in the fast P-wave arrivals with high Q and low attenuation. d shows a seismic
tomography near the Mantle Transition Zone (410-660 km) with a hypocenter of the deep-focus earthquake (white star).
@seismicisolation
@seismicisolation
Table 3. Depth determination using spectral nulls and take-off angles. Azimuth (AZ) direction
to the station and angular distance (ǻ are measured in degrees.
Array names 2006 2009 2013 Take-off
$=ǻ (km) (km) (km) Angle (°)
ASAR (175.1, 64.8) 2.43 2.43 2.43 17.57
FINES (327.4, 60.3) 1.62 1.65 1.65 18.62
ISN (295.6, 73.5)* 2.12 2.18 16.13
KURK (302.8, 35.7) 2.33 2.25 2.25 22.86
NOA (332.0, 66.2) 2.28 2.28 17.36
NVAR (47.5, 79.7) 2.11 (2.05) 1.95 2.11 (1.98) 14.62
PDAR (39.5, 81.0) 2.11 2.11 (1.83) 2.06 (1.85) 14.33
WRA (174.3, 61.1) 1.99 2.15 2.15 18.40
YKA (27.4, 64.7) 2.16 2.11 2.06 17.57
Average depth 2.10±0.05 2.09±0.05 2.09±0.04 17.50±2.38
* indicates data from the representative station of PRNI in the Israel Seismic Network (ISN).
Depth values in the brackets refer to sP-P delay times.
Table 4. Source depths (km) using pPn-Pn delay times from fixed local stations
Year CHC MDJ KSRS, BHZ SEO SES Mean depth

2006 2.217 1.938 2.166 2.134 2.11±0.01
2009 2.059 1.940 2.022 1.911 2.134 2.01±0.08
2013 2.059 1.940 1.957 1.911 2.134 2.00±0.08
Using regional stations (KMA), the source depth can be estimated by using individual pPn-Pn delay
times (Fig. 6) by means of selecting a short time window. The spectral analysis is performed using
a time window of 5 seconds which were band-pass filtered between 0.5 Hz and 3.0 Hz taking into
account the first arrival Pn and pPn phases which are within 1 second after P-wave arrival excluding
Pg and P* phases. It should be noted that the spectral nulls of spectra using most of individual
Korean stations are not constant due to the complicated raypaths through the East Sea and sub-
continent (Fig. 6). Nonetheless, we found similarities in spectral shape for the 2009 and 2013 nuclear
tests indicating that the source depths and source mechanisms were almost identical for both 2009
and 2013 tests (Figs 6 and 7). In particular MDJ is a good station to find source characterization
because of the pure-continental raypaths. We found similarities of spectra for the 2009 and 2013
tests implying that not only the raypaths from source to station but also mechanism and depth should
be identical for those tests as compared to the 2006 test.
@seismicisolation
@seismicisolation
@seismicisolation
@seismicisolation
Fig.9. The spectral characteristics for the 2006, 2009 and 2013 nuclear tests from KSRS Array.
The spectra for the 2009 and 2013 tests are identical compared to the 2006 test, indicating that the
raypath from source to station for the 2006 test is quite different from the 2009 and 2013 tests even
if the source parameters and depths are almost identical for all three tests. A slightly different shape
of the spectra for the Chunchon (CHC) station is due to a slight change in the site position of the
seismometer. This indicates that the 2006 test may be different from source mechanism and/or
location compared with the 2009 and 2013 tests
In local stations there is strong variation in velocity-sensitive parameters such as the ray paths and
crustal structures at the near-field phases resulting in incongruity with the estimates of pPn-Pn delay
times (Table 4). It is very worthwhile to compare the spectral nulls of the individual regional P-wave
spectra with those of the average spectra from the seismic arrays. We found the evident and dominant
spectral nulls of the average spectra from the local seismic array of KSRS whereas we could hardly
see the evident spectral nulls for the spectra of the individual local stations (Fig. 6). Therefore, the
source depths are more reliably estimated using a local seismic array, KSRS (Korea Seismological
Research Station, Wonju, South Korea) which is located in the stable region with high Q in the
Archean-early Proterozoic Massif of the central Korean Peninsula. The use of KSRS can enhance a
good SNR (signal-to-noise ratio) by superposing each channel signal, resulting in the evident
spectral nulls of the average spectra for 19 traces from KSRS array indicating that predominant
spectral nulls may present the identical depth delay times from depth phases (Fig. 9). We neglected
the trade-off between nonlinear topographic effects of the mountain and applied simplifications
taking into account a source in the middle of the mountain area neglecting a difference between a
source elevation and an elevation at each reflector with a shallower dip. The pseudo-1D radiation in
one direction is constant and will be even lookalike in spectral nulls of the average spectra for 19
channels. We analyzed the P-wave spectra of waveform data recorded at the KSRS which consists
of 19 short-period vertical components with a 5-second window excluding P* phase as well as Pg
phase arrivals. We also selected the frequency range 1.0 - 3.0 Hz in which we are interested in this
study. Belyashova et al., (2001) compiled travel-time phases of the crust and mantle using Russian
@seismicisolation
@seismicisolation
nuclear explosions showing Pn, Sn and Pg in the range 0 to 740 km and found that the arrival time
difference between Pg and Pn is greater than 6 seconds at an epicentral distance of 440 km.
According to the conspicuous spectral holes of KSRS spectra (Fig. 9), the depths for the 2006, 2009
and 2013 tests were estimated at 2.12 km, 1.95 km and 1.95 km by pPn-Pn, respectively, whereas
those by sPn-Pn were estimated at 2.17 km for all three tests from the spectral nulls of the average
spectra from 19 channels at the KSRS (Fig. 9). We can conclude that source depths of three nuclear
tests obtained from the KSRS are consistent with depths estimated by pP-P. The reason why the
three tests show the same depth using sPn-Pn is that the delay time difference of sPn–Pn is almost
the same for three tests owing to the small delay time difference of slow Vs compared to fast Vp for
the near-field array. Nonetheless, it is important to note that the delay time of pPn-Pn of the 2006 is
slightly greater than that of the 2009 or 2013 test implying that the 2006 test must be deeper than
the 2009 and 2013 tests. We also found that the onset of P-wave arrival (pPn) for the 2006 test at
the KSRS was small and weak compared with the onsets for the 2009 and 2013 tests implying that
the 2006 explosion in addition to being a smaller size, may have occurred in a large cavity with
decoupling effects.
Fig.10. We show synthetic seismograms and a spectral null at 1.75 Hz for the epicentral distance of 440 km and at 1.25
Hz for the epicentral distance of 81° in case of depth at 2.15 km assuming that the synthetics was calculated based on
the flat Earth model for simplifications
uncertainties to other parameters (Stein and Wiens, 1986; Douglas and Hudson, 1990). The earth
structure models for synthetics are used by Korea Model (Kim and Bae, 2006; Herrmann 2016) in
the near-field and AK135 (Kennett et al., 1995; Montagner and Kennett, 1996) in the far-field for
the teleseismic arrays taking a short span of the first P-wave arrivals for 5-6 seconds. The preferred
pP-P and pPn-Pn delay times are in very close agreement with observed results from Fig. 10. The
synthetic seismograms and spectra are produced using Hudson96 and Herrmann programs
(Herrmann, 2016). The synthetic spectral nulls of short-period P waves for the near-field and those
of long-period P waves for the far-field (Fig. 10) are considerably consistent with the observed
spectral nulls (Burdick et al.,1984; Douglas and Hudson, 1990).
DEPTH ESTIMATION FROM SPECTRAL NULLS OF THE FUNDAMENTAL-MODE
RAYLEIGH WAVE SPECTRA
It is worth noting that Relationship between a spectral null of fundamental-mode Rayleigh wave and
depth in a half-space is very useful to estimate depth for DPRK’s nuclear tests. It should be noted
that the particle motion of the fundamental-mode Rayleigh waves becomes prograde below a depth
of about at z = - Ȝ Ȝ ZDYHOHQJWK 7KHUHLVDYDOXHIRUWKHKRUL]RQWDOGLVSODFHPHQWX1 is null,
z = -ȜZKHUHDVWKHYHUWLFDOGLVSODFHPHQWX3 is never null. The spectral null in Rayleigh wave
@seismicisolation
@seismicisolation
excitation is due to a zero-crossing of the horizontal displacement eigenfunction. This special feature
has been used for estimating source depths of underground explosions as well as earthquakes (Tsai
and Aki, 1971; Fox et al., 2012). At the depth where u1 is null, its amplitude changes sign and below
the depth the particle motion is prograde. With increasing depth, the amplitudes of u1 and u3 decrease
exponentially, with u3 always larger than u1. The purpose of this section is to find the spectral null
of Rg-wave spectra due to the zero-crossing of the fundamental-mode Rayleigh wave controlled by
CLVD source.
)RU5D\OHLJKZDYHVRIIUHTXHQF\ȦWKDWSURSDJDWHLQWKH[GLUHFWLRQZLWKYHORFLW\FSRWHQWLDOĳ
SXUHGLODWDWLRQDO DQGȥ SXUHURWDWLRQDO LQKRPRJHQHRXVDQGLVRWURSLFKDlf–space,
ĳ I ] H[S>LN [-ct)]
ȥ J ] H[S>LN [-ct)] (6)
where f(z) and g(z) present an amplitude, i, k and c indicate ξ(െ1) ZDYH QXPEHU ʌIF DQG
Rayleigh wave velocity, respectively. %\LQVHUWLQJĳDQGȥLQWRZDYHHTXDWLRQVDVEHORZ
2f]2 + kr2f(z) = 0
2J]2 + ks2g(z) = 0 (7)
where r=(c2/Vp2-1)1/2, s=(c2/Vs2-1)1/2
The solutions of equation (7)
f(z)=A exp (ikrz) + D exp (-ikrz)
g(z)=B exp(iksz) + E exp (-iksz) (8)
D=0 and E=0 since the values of f(z) and g(z) decrease with depth.
The equation (6) can be rewritten as follows:
ĳ $H[S>LNU]LN [-ct)]
ȥ %H[S>LNV]LN [- ct)] (9)
Using boundary conditions at free surface (P31=P32=P33=0), numerals for r=0.8475i, s=0.3933i and
B=-1.4675iA the horizontal displacement eigenfunction u1 and vertical displacement eigenfunction
u3 are as follows:
u1 = ĳ[– ȥ]
u3 = ĳ]ȥ[ (10)
from equation (9)
u1 = Aik[exp(-0.8475kz – 0.5773exp (-0.3933kz)]exp[ik(x-ct)]
u3 = Aik[0.8475exp(-0.8475kz – 1.4679exp(-0.3933kz)]exp[ik(x-ct)] (11)
Considering taking real parts, the horizontal and vertical displacement eigenfunctions for the
fundamental-mode Rayleigh waves are as follows:
u1 = -Aksin (kx- ȦW >H[S -0.8475kz) – 0.5773exp (-0.3933kz)] (12)
u3= -Akcos (kx- ȦW >H[S -0.8475kz) – 1.4679exp (-0.3933kz)] (13)
ZKHUH$ DPSOLWXGHȦ ʌIDQGN ʌIKFZKHQ] K
@seismicisolation
@seismicisolation
From equation 12, at the depth where u1is null, its amplitude changes sign.
1-0.5733exp(0.4542kz) = 0 (14)
Therefore, source depth h= 0.1923c/fnull (15)
where c=Rg-wave velocity, 1.46 km/sec from Rayleigh Equation (Udias, 1999) and fnull is a spectral
null of the fundamental-mode Rayleigh wave amplitude spectra.
corroborate the estimated depths extrapolated by body waves. Rayleigh wave excitation is sensitive
to source depth, especially at intermediate and short periods due to the approximate exponential
decay of Rayleigh wave displacement with depth. The frequency-dependent “spectral null” in the
Rayleigh wave amplitude spectra are most pronounced for vertical dip-slip and 45° dip-slip focal
mechanisms (Tsai and Aki, 1971). The spectral null of the fundamental-mode Rayleigh wave
spectrum is dependent on source mechanism, depth and source-receiver azimuth but does not vary
much with the shot medium and the shot yield (Tsai and Aki, 1971; Langston, 1980; Massé, 1981;
Okal, 1992; Fox et al., 2012; Heyburn et al., 2013). However, in case of North Korean nuclear
explosions which were detonated in the same velocity structure of granite a spectral null strongly
depends on the path rather than mechanism because we assume that the general mechanism would
be 45° dip-slip focal mechanisms (Tsai and Aki, 1971; Massé, 1981). So it is more possible to
determine a source depth using the spectral null of the fundamental-mode Rayleigh wave amplitude
spectra because the strongest signal is more produced by a source placed beneath the mountain
(Viecelli, 1973). However, it will not be precise enough to really pin down the depths of the tests to
the level of accuracy required because of lack of geological structure and topography. We found
spectral nulls at the fundamental-mode Rayleigh wave spectra of the MDJ and DSN (Dongbei
Seismic Network) through the pure continental-path data from the 2006 test in Fig. 11. If depth and
azimuth (source-receiver) are the same for three explosions recorded at the pure-continental path of
the MDJ and dsn stations, the same spectral nulls for all three tests should appear in the amplitude
spectra of the vertical component contingent on being via the same constraint of CLVD source,
However, we observed spectral nulls at 0.13 Hz and 0.134 Hz using spectra of MDJ and dsn for the
2006 test but none for the 2009 and 2013 tests (Fig. 11), indicating that the 2006 raypath was
different from the 2009 and 2013 tests. The average spectral null at 0.132 Hz is caused by an
excitation null for short-period fundamental-mode Rayleigh waves, termed Rg waves inducing a
CLVD (compensated linear vector dipole) source by normal mode theory (Massé, 1981; Tsai and
Aki, 1971; Fox et al., 2012; Heyburn et al., 2013; Kim et al., 2015) which can be manifested by
showing 0.14 Hz in the theoretical spectrum of the fundamental-mode of Rayleigh waves in Fig.
11 . We selected the frequency range up to 0.2 Hz to use the high frequency Rayleigh waves which
have less than 8 s (0.25 Hz) of period that occurred less than a mean depth of 3 km for North Korean
nuclear tests. Bonner et al. (2003) have examined the performance of MS scales on 7-sec Rayleigh
waves recorded distances less than 500 km from Nevada Test Site. This shows that the CLVD depth
(Langston, 1980) correlates with a source depth derived from pP-P and pPn-Pn delay times. The
conical dip-slip reverse faulting may accompany a deep-seated tensile failure occurring at a depth
above shot point (Massé, 1981; Patton and Taylor, 2008) generating the strong Rg wave radiation
from a vertically oriented CLVD source (Rodgers et al., 2010) for the 2006 test in the mountainous
region. In particular the 2006 test was a spallation-like explosion but it is not associated with surface
spallation due to the absence of spall as a consequence of test being conducted in a strong material
such as granite. The estimated centroid depth via the CLVD model disambiguates the estimated
source depth for the 2006 nuclear test.
However we cannot exclude some poor S/N signals for the 2006 test because of the small event.
@seismicisolation
@seismicisolation
Fig. 11. a) The fundamental-mode Rayleigh waves and displacement spectra for the BHZ band-pass filtered (0.02-0.1
Hz) to isolate the fundamental-mode (Fox et al., 2012; Heyburn et al., 2013). The 2006 nuclear test (black line) shows
a spectral null at 0.13 Hz but no nulls for the 2009 (blue line) and 2013 (red line) tests using Rg waves from MDJ station.
b) We present spectral nulls at 0.134 Hz from DSN (Dongbei Seismic Networ) data only for the 2006 but no spectral
nulls for the 2009 test data with amplitude in count
There are more elliptical motions of Rg and the ‘retrograde‘ reverses ‘prograde’ at a depth of
d=0.192 wavelength which represents a spectral null (hole) at the spectra. This source mechanism
of CLVD source could be assumed to be a pure dip-slip (reverse-faulting), generating various
synthetic seismograms and spectra depending on different depths and mechanisms in Fig. 12. The
synthetic seismograms were produced based on Green function and the modified AK135 Model
(Kennett et al., 1995; Montagner and Kennett, 1996) considering the specified nuclear test site using
a narrow band-pass filter between 0.02 and 0.1 Hz for Rg waves. We utilized the only forward model
with known parameters to draw synthetic seismograms and spectra of Rayleigh waves skipping over
the detailed explanations in this study. Various researchers (Ford et al., 2009; Shin et al., 2010;
Barth, 2014; 9DYU\þXN and Kim, 2014; Cho et al., 2016) have found different source mechanisms
from moment tensor inversion to be some uncertainties for the 2006, 2009 and 2013 North Korean
@seismicisolation
@seismicisolation
nuclear explosions. In particular there are some trade-offs between spectral nulls and moment tensor
inversion for underground nuclear explosions owing to a shallow-depth event with a little DC and
some CLVD components. We found a spectral null to be 0.14 Hz at a depth of around 2 km for the
2006 test in Fig.11a, but the spectral null will be shifted from one place to another according to a
strike angle which is theoretically absent in case of a pure explosion mechanism. Nonetheless, we
selected the optimal strike angle 30 degrees to match the estimated depth based on pP-P and sP-P
depth phases taking into account the tectonic stress (ENE) in this region (9DYU\þXNDQG.LP .
As you see, we can hardly see spectral nulls for the 2009 and 2013 tests in Fig 11 because they may
be due to the different raypath through the complicated crustal structures rather than mechanisms.
The burial depth of the 2006 nuclear test could be determined by the spectral null which is due to
the destructive interference of reflections associated with P and SV from the free surface controlled
by shock waves of an explosion and a CLVD source. As a result, we estimated a source depth at
2.12 km using a spectral null of Rg-wave spectrum since from h=–0.192* wavelength for a Poisson
solid (Poisson’s ratio=0.25), i. e., h= 0.192*7.5758*1.46 km/sec where 1.46 km/sec is a Rg-wave
velocity obtained from Rayleigh equation (Udias, 1999). Consequently the final depths for the 2006,
2009 and 2013 nuclear tests were estimated at 2.12 km, 2.06 km and 2.05 km from the free surface,
respectively (Table 5), which reasonably agree with the previous study by Gitterman et al. (2015).
We also estimate the source depth for the 2006 test at 2.12 km, using spectral nulls and average
spectral holes of the fundamental-mode Rg-wave amplitude spectra from MDJ and DSN.
Table 5. The final average depths (km) for the 2006, 2009 and 2013 tests using spectra of
teleseismic arrays, local stations and the KSRS including Rayleigh wave spectra of the
Mudanjang (MDJ) station and DSN (Dongbei Seismic Network) and KSRS array by 19 short-
period records are in a good agreement with those by the broadband of KSRS records.
Test pP-P & sP-P pPn-Pn average spectral hole of Rg- KSRS array Mean depth
year km km wave spectra MDJ & DSN, Wonju, km km
km
2006 2.10 2.11 2.12 2.12 (2.17) 2.12±0.02
2009 2.09 2.01 1.95 (2.17) 2.06±0.01
2013 2.09 2.00 1.95 (2.17) 2.05±0.01
The depth for MDJ and DSN refer to depths from estimates using the average of the fundamental mode-Rayleigh wave
by equation (15) and the parentheses indicate depth using sPn-Pn from body waves.
The synthetic seismograms were produced based on Green function and the modified AK135 Model
(Kennett et al., 1995; Montagner and Kennett, 1996) on condition that depth = 2 km, rake= 90° and,
dip = 45° but strike varies with every 15° using a narrow band-pass filter between 0.02 and 0.1 Hz
for Rg waves in Fig. 12. Various researchers (9DYU\þXN and Kim, 2014; Cho et al., 2016; Ford et
al., 2009; Shin et al., 2010; Barth, 2014) have found different source mechanisms from moment
tensor inversion with some uncertainties for the North Korean nuclear explosions. In particular there
are some trade-offs between spectral nulls and moment tensor inversion for underground nuclear
explosions owing to DC and CLVD components in North Korean nuclear tests (Kim et al., 2015;
9DYU\þXN and Kim, 2014; Cho et al., 2016; Shin et al., 2010). We found a spectral null to be 0.14
Hz at a depth of around 2 km in Fig. 12 from the synthetic seismograms and spectra.
@seismicisolation
@seismicisolation
Fig. 12. The synthetic seismograms and spectra of Rg-waves are band-pass filtered between 0.02-0.1 Hz to obtain the
fundamental-mode based on the Korea model considering a general mechanism for an explosion with dip 45° and rake
90° varying with every 15° of strike to 90°. The higher amplitude at an azimuth of 90° attributes the large radiation
pattern of Rayleigh wave and it decreases again up to 180°. We found the most fitting depth (2 km) at 0.14-0.15 Hz
from an azimuth of 30°which corresponds to the observations of tectonic stress (ENE) and source mechanism study by
moment tensor inversion (9DYU\þXNDQG.LP .
COMPARISON WITH PREVIOUS STUDIES

Our depth estimations were quite different from other studies by Murphy et al. (2013) and Zhang
and Wen (2013) which concluded depth estimations of 200 m (2006 test)/550 m (2009 test) by
Murphy et al. (2013) and 610 m (2009 test)/430 m (2013 test) by Zhang and Wen (2013). Murphy
et al. (2013) used the spectral ratios of two explosions at common regional stations to be compared
with the theoretical spectral ratios (Mueller/Murphy) assuming that both explosions were detonated
at either a depth of 200 m or 800 m. They then compared these ratios with the theoretical Mueller–
Murphy source spectral model and estimated the yield for the 2006 test as 0.9 kt if detonated at a
depth of 200 m, and 4.6 kt for the 2009 test if detonated at a depth of 550 m. In their study, observed
spectral ratio data were more consistent with the hypotheses that the 2006 test was conducted at a
depth of 200 m and the 2009 test was conducted at a depth of 550 m. Murphy et al. (2013) failed in
finding appropriate spectral nulls of around 5 Hz for the 2009 test and those of around 10 Hz for the
2006 test (Gitterman et al., 2015). Zhang and Wen (2013) estimated depths based on the relative
method using the first 2006 test as a reference point (master event) subtracting the elevation
difference of the reference level and the respective surface of the source locations of the other tests
via data gathered through satellite images. However, we are not sure that the master event of the
2006 test is located at the same level as the West Portal (see Fig. 1b). Rougier et al. (2011) also
estimated the minimum depth of burial and yield for the 2009 North Korean test at 375 m and 5.7
kt, respectively based on constraints from hydrodynamic simulation which is mostly unknown. On
the other hand, the fore-mentioned locations are considerably inconsistent with those by the absolute
methods of IDC (International Data Centre, CTBTO) and National Earthquake Information Service
(NEIC, USGS) using global seismic stations. It seems that the location by IDC is more reliable than
that by USGS (NEIC) because IDC used more observations of 170 stations which are located
beneath solid rock consisting in seismic arrays with many sensors, utilizing updated various
@seismicisolation
@seismicisolation
monitoring technologies in order to detect, locate and verify the only global nuclear explosions.
Nonetheless, IDC did not determine and announce the source depths of the North Korean nuclear
explosions.
As a result, the absolute method is the best fitting approach to estimate depth of North Korean
nuclear tests taking into account neglecting the nonlinear effects due to a shallow sloping topography
and non-surface spallation owing to the granite source rock for North Korean nuclear tests (Kim et
al., 2017a).
scaled depth of burial should not be less than 122Y1/3 (Glasstone and Dolan, 1977) where depth is
measured in meters and yield in kilotons (kt). Consequently, the optimum minimum depths are found
to be 145 m, 255 m and 373 m by the yield of Murphy (1981) and 129 m, 193 m and 291 m by the
yield of Bowers et al. (2001) for the 2006, 2009 and 2013 tests respectively. Nevertheless a depth
of about 2000 m would allow containment of an underground nuclear explosive event having a yield
of up 4.4 Mt. However, this scaled depth concept does not fit the North Korean nuclear tests because
the maximum body wave magnitude was estimated at 5.1.
Gitterman et al. (2015) estimated source depths at 2.01 km, 2.13 km and 2.11 km for the 2006, 2009
and 2013 tests, respectively, using only pP-P delay times without considering take-off angles
whereas we calculated source depths at 2.12 km, 2.06 km and 2.05 km from the free surface for the
2006, 2009, and 2013 tests, respectively, using spectral nulls of Rayleigh wave spectra as well as
those of local and teleseismic body wave spectra taking into account take-off angles for further
corroboration.
The striking similarity of the particle motions of the 2009 and 2013 tests (Fig. 13) highlights similar
raypaths of Rg-waves rather than mechanism with identical depths (Table 5). The generation of non-
isotropic radiation shows horizontal and transverse motion of Love waves on a tangential component
and Rayleigh waves on a radial component, indicating that the source parameters of the 2009 and
2013 tests are fairly identical in particle motions as shown in Fig. 13 The deviation of an ellipticity
of the particle motion is measure of the effects of noise and lateral heterogeneity. Generally we
should be aware that the complete separation of Love waves and Rayleigh waves in a homogeneous
and isotropic half-space is not absolutely true in the real Earth because of heterogeneity and
anisotropy. Even if the stations used for particle motions are randomly distributed, the receiver–
source path for the 2006 test is quite different form that for the 2009 and 2013 tests in the very near
field implying that the particle motions are mainly dependent on the raypath owing to the constraints
of the same depth and mechanism. It turns out that the particle motions of surface waves for the
2009 and 2013 are identical compared with the 2006 test implying that source mechanisms for the
2009 and 2013 tests could be all but identical (9DYU\þXNDQG.LP on the basis of the same
source-receiver path taking into account a depth of about 2 km for three North Korean nuclear tests.
@seismicisolation
@seismicisolation
@seismicisolation
@seismicisolation
Fig. 13. A) T-R particle motions of surface waves of the 2006, 2009 and 2013 explosions at the MDJ, BJT, SES2, SEO
and CHC2 stations. The velocity records were band-pass filtered between 0.04-0.1 Hz. b). For the subduction zone-path
stations such as YSS, ERM and MAJO show good generation of Love waves whereas the pure-continental path stations
such as MDJ, BJT, SEO, SES2 and CHC2 reveal significantly Rayleigh waves.
SUMMARY AND CONCLUSIONS

We estimated depths of burial at 2.12 km, 2.06 km and 2.05 km from the free surface for the 2006,
2009 and 2013 nuclear tests using body and Rg wave spectral nulls. We can take into account the
trade-off between our applied simplifications and the non-linear topographic effects of the mountain
assuming that each reflection point is constant on the pseudo-1D radiation in one direction. There is
nothing about the contribution of different source components according to Fermat principle and
Snell’s law indicating that a raypath is along a minimum of travel time and that a reflected angle is
always the same as an incident angle (take-off angle). The data from a test in the middle of a
mountain will be even lookalike in terms of spectral nulls of the average spectra from all directions
of seismic arrays. Nonetheless we cannot help but rule out trade-offs between the predetermined
depth before the test and the calculated depth from observations. In order to justify our method, it is
worthwhile to apply our technology for the depth estimation of the Pakistan nuclear test for which
the overburden depth of about 700 m (HOB=–700 m) was predetermined before the nuclear test
(Azam, 2000). The Pakistan nuclear explosion (mb 4.88) was conducted in the Ras Koh Hills of the
Chagai district on May 28, 1998 (Gupta et al., 1999; Bowers, et al., 2002) after long preparation.
The site of the Pakistan nuclear test is very similar to that of the North Korean nuclear test in terms
@seismicisolation
@seismicisolation
of granite source rock and the high mountain. We estimated the average depth at 772 m via both pP-
P and sP-P using KSRS, WRA and YKA (see Table 6 and Fig.14). We cannot exclude that the
additional depth of around 70-100 m may be due to the surface elevation difference between the
source and the reflector point owing to the mountain slope. As a result, our determination
substantially agree with the predicted depth if we consider some spallation for a shallow explosion
including the source configuration as the vertical shaft (collapse chimney) sealed with averting
materials like pebble and cement up to about 50 m from the horizontal tunnel. We also approved
that the depth estimate of 10 km by pP-P was well coherent with that by the conventional method
for the Gyeongju Earthquake on September 12, 2016 (Kim et al., 2017b).
Fig. 14. The average spectral nulls of spectra using WRA and YKA for the Pakistan nuclear test on May 28, 1998. The
black, blue and red arrow marks represent depth phase of pPcP+PcP, sP+P and pP+P, respectively
Table 6. Depth parameters used for the Pakistan nuclear test of May 28, 1998 (mb=4.88)
Array $=ǻ pP+P Depth sP+P Depth Take-off Take-off
names (Hz) (m) (Hz) (m) angle angle
pP (°) sP (°)
KSRS 63.2/52.54 3.75 766.2 2.50 720.7 19.34 12.22

WRA 117.2/82.59 3.29 793.3 2.33 803.9 13.87 7.83
YKA 359.8/88.88 3.27 803.2 2.34 746.5 12.31 6.96
Average 787.6±19.2 757.0±42.6
We concluded that the DPRK nuclear tests were conducted at a depth of at least 2 km beneath the
vicinity of Mt. Mantap with an elevation of 2205 m. A depth of about 2000 m would allow
containment of an underground nuclear explosion having a yield of up 4.4 megatons. However this
scaled depth cannot be applied for the North Korean nuclear tests since we estimated a maximum
body-wave magnitude at 5.1. Concerning source characterization, we also infer that the first 2006
nuclear test was a vertically-distributed source in the underground shaft (Viecelli, 1973) generating
mostly strong Rayleigh waves. However the 2009 and 2013 tests may have been vertically-
distributed sources in the mutually and horizontally connected-underground tunnels resulting in SH
and Love wDYHV 1XWWOL7RNVऺ]DQG.HKUHU/D\DQG+HOPEHUJHU LQGXFHGIURP
tectonic stress and/or new cracks (blocks) in addition to Rayleigh waves. The radioisotope was also
not detected in the atmosphere for the 2009 and 2013 nuclear tests (VavryþXNDQG.LP%DUWK
2014) because of well-containment by blocking radionuclides through the long winding
underground tunnels whereas it was released for the first 2006 nuclear test in the vertical shaft
(Koper et al., 2008).
@seismicisolation
@seismicisolation
ACKNOWLEDGMENTS
criticism. We would like to thank Paul Richards for his providing DSN (Dongbei Seismic Network)
data and Rick Benson (IRIS) and David Jepson (Geoscience Australia) for the teleseismic data and
KMA (Korea Meteorological Administration) for providing local waveform data. We also
appreciate Albert Brouwer as well as the Executive Secretary of CTBTO, Lassina Zerbo and
Director of IDC (International Data Centre), CTBTO for their collaboration in IDC data collection
for this study. One of author (SGK) would greatly appreciate CTBTO’s funding to participate in
SnT2015, SnT2017and SnT2019, CTBTO, Vienna, Austria. We acknowledge Director of ISC
(International Seismological Centre), Dmitry Storchak for providing phase reading data for every
North Korean nuclear test.
REFERENCES
Azam, RMS, 2000, When mountains move-the story of Chagai, defence note, Defence Journal June
2000 1-10.
Bakun, J. H. Johnson L. R., 1973.The convolution of teleseismic P waves from and Cannikin,
Geophys. J. R. astr. Soc., 34, 321-342.
Barth, A., 2014.Significant release of shear energy of the North Korean nuclear test on February 12,
2013. J of Seismology, 18 (3), 605-615.
Belyashova, M. N. Shacilov, V. I., Mikhailova, M. N,, Konarov, I. I., Sinyova, Z. I. Belyashov, V.
Malakhova, M, N, 2001,.On the use of calibration explosions at the former Semipalatinsk test
site for compling a travel-time model of the crust and upper mantle, Pure. Appl. Geophys. 158
(1-2),193-209.
Bonner, J. L., Reiter, D. T., Shumway, R. H., 2002, Application of a ceptral F statistic for improved
depth estimation. Bull. Seism. Soc. Am., 92 (5), 1675-1693.
Bonner, J. L., Herrmann, R. B., Harkrider., Pasyanos,. M. 2008, The surface wave magnitude for
the October 2006 North Korea Nuclear Explosion. Bull. Seism. Soc. Am., 98 (5), 2498-2506.
Bonner, J. L., Harkrider, D. G., Herrin. E. T., Shumway, R. H., Russell, S. A., Tibuleac, I, M., 2003,
Evaluation of short-period, near-regional MS scales for the Nevada Test Site. Bull. Seism. Soc.
Am., 93(4), 1773-1791.
Bowers, D., Marshall, P. D., Douglas, A.,2001, The level deterrence provided by data from the
SPITTS seismometer array to possible violations of the Comprehensive Test Ban in Novaya
Zemlya Region. Geophys. J. Int., 46, 425-438.
Bowers, D., Douglas, A., Selby, N. D., Marshall, P.D., Porter. D.,Willis, N. J., 2002 Seismological
identification of the 1998 May 28 Pakistan nuclear test, Geophys. J. Int.. 150 153-161.
Burdick, L. J., Wallace, T., Lay, T., 1984, Modeling near-field and teleseismic observations from
the Amchitka test site, J.. Geophys. Res,, 89 (B6) 4373-4388.
Cho, C., Shin. J. S., Kim, G., 2016, Comparison of results of relative location methods and moment
Tensor inversion for the nuclear explosions experimented in North Korea, 2016AGU Fall
Meeting, San Francisco, 12-16 December 2016 S31A-2722.
Chung, T. W. 1995, A quantitative study on the crustal structure of the Korean Peninsula based on
the earthquakes from 1991 to 1994. J. Kor. Earth Science Soc. 16 (2), 152-157.
Cohen, T. J., 1970 Source-depth determination using spectral, pseudo-autocorrelation and cepstral
Analysis. Geophys. J. R. astr..Soc., 20, 223-231.
Douglas, A., Corbishley, D. J., Blamey, C., Marshall. P. D., 1972. Estimating the firing depth of
underground explosions, Nature 237, 26-28.
Douglas, A., Marshall, P. D., Young, J. B., 1987. The P waves from the Amchitka Island explosions.
Geophys. J. R. astr. Soc., 90, 101-117.
Douglas, A., Rivers, D. W.,1988. An explosion that looks like an earthquake. Bull. Seism.Soc. Am.
78 (2), 1011-1019.
@seismicisolation
@seismicisolation
Douglas, A., Richardson, L., Hutchins, M., 1990. Surface reflections and S-to-P conversions on P
Seismograms. Geophys. J. Int., 100, 303-314.
Douglas, A., Hudson, J. A., 1990. The effect on the teleseismic P of the zone of damage created by
an explosion. Geophys. J. Int. 103, 111-133.
Ford, S. R., Douglas, D. S., Walter, W. R., 2009. Source analysis of the Memorial Day Explosion
Kimchaek North Korea. Geophys. Res. Lett., 36 (21), L21304.
Fox, D. B., Selby. D. N., Heyburn, R., Woodhouse, J. H., 2012 Shallow seismic source parameter
determination using intermediate-period surface wave amplitude spectra. Geophys. J. Int., 191,
601-615.
Glasstone, S., Dolan, P. J., (compiled & edited), 1977. The Effects of Nuclear Weapons, United
States Department of Defense and the Energy Research and Development Administration, 657
pp.
Gitterman, Y., Kim, S. G., Hofstetter, R., 2015. Spectral modulation effect in teleseismic P-Waves
from North Korean nuclear tests recorded in broad azimuthal range and possible source depth
estimation. Pure Applied Geophys. DOI: 10.1007/s00024-015-1169-8.
Gupta, H. K., Bhattacharya, S. N., Kumar, M. R., Sarkar, D., 1999. Spectral characteristics of the
11 May 1998 Pokhran and 28 May 1998 Chaghai nuclear explosions, Current Science, 76 (8),
1117-1121.
Harjes, H-P, Henger N., Muller, G., Wilhelm, H., 1985. A system design for the gradual
imporovement of seismic monitering and verification capabiliuies for a Comprehensive Nuclear
Test Ban. Conference on Disarmament, CD/624, GE. 85-63293, 26 July 1985 Federal Republic
of Germany 68 pp.
Herrmann, R. B., 2016. Computer Programs in Seismology. Current Version 3.30 is dated May 8,
2016 (NP330.May-08-2016.tgz), Saint Louis University, Saint Louis, MO. 186 pp.
Heyburn, R., Neil, D., Fox, B., 2013. Estimating earthquake source depths by combining surface
wave amplitude spectra and teleseismic depth phase observations. Geophys. J. Int., 94 (2), 1000-
1010.
Husebye, E.S., Matveeva, T., Fedorenko, Y. V., 2013. Focal-depth estimation using Pn-coda Phases
including pP, sP, and PmP. Bull. Seism. Soc. Am., 103 (3), 1771-1783.
Johnston, W. R., 2005. Database of nuclear tests, United States: Part 2, 1964-1972, complied and
last modified 19, June 2005
Kennett, B.L.N., Engdahl, E. R., Buland. R., 1995. Constraint on seismic velocities in the Eart from
Traveltimes. Geophys. J. Int., 221, 108-124.
Kim, S. G., Bae. H., 2006. Investigation of post-sites using local seismic tomography in the Korean
Peninsula, Kr. Soc. Econ. Environ. Geol 39 (2), 111-128.
Kim, S. G., Gitterman, Y., Lee, S., Vavrycuk, V., Kim, M., 2015. Estimating depth and source
characteristics of nuclear tests by the Democratic Peoples’s Republic Korea in 2006, 2009 and
2013 using regional and teleseismic networks. SnT2015 (CTBTO) Vienna, Austria, 22-26, June
2015.
Kim, S. G., Gitterman, Y., Lee, S., Vavrycuk, V., 2017a. Accurate depth determination and Source
characteristics of the DPRK nuclear tests [2006, 2009, 2013, 2016J (01/06/2016) and 2016S
(09/09/2016)] using regional and teleseismic arrays. SnT2017 CTBTO Vienna, Austria, 26-30,
June 2017.
Kim, S. G., Gitterman, Y. and Lee, S-K., 2017b. On the computation of depth and source
characteristics of the DPRK;s nuclear tests (2006, 2009, 2013, 2016J, 2016S and 2017) taking
into account 3D topography of the epicenter area, Korea Seismological Institute Research Report,
KSI 2017--02
Koper, K. D., Herrmann, R. B., Benz, H. M., 2008. Overview of open seismic data from the North
Korea event of 9 October 2006. Seism. Res. Lett., 78,178-185.
Kulhanek. O., 1971. P-wave amplitude spectra of Nevada underground nuclear explosions. Pageoph
88, 121-136.
@seismicisolation
@seismicisolation
Langston, C. A., 1980. A note on spectral nulls in Rayleigh waves. Bull. Seism. Soc. Am., 70 (4),
409-1414.
Lay, T., Wallace, T. C., Helmberger, D. V., 1984. The effects of tectonic release on short-period P
waves from NTS explosions. Bull.Seism. Soc. Am., 74 (3), 819-842.
Lay, T., 1991. The teleseismic manifestation of pP: problems and paradoxes. In Explosion Source
Phenomenology, Geophysical Monograph 65, 109-125.
Massé, R. P., 1981. Review of seismic source models for underground nuclear explosions. Bull.
Seism. Soc. Am., 71 (4),1249-1268.
Montagner, J. P., Kennett, B. L. N., 1996. How to reconcile body-wave and normal mode reference
Earth models?. Geophys. J. Int., 125, 229-248.
Murphy, J. R.,1981. P wave coupling of underground explosions in various geologic media. In
pp. 201-205.
Murphy, J. R., Kitov, I., Rimer, N., Barker, B., 1996. Further studies of the seismic characteristics
of Russian explosions in cavities: implication for cavity decoupling of underground nuclear
explosions PL- TR-96-2017 Phillips Laboratory Hanscom AFB MA 66 pp.
Murphy, J. R., Stevens, J. L., Kohl, B. C., Bennett, T, J. 2013. Advanced seismic analyses of the
source characteristics of the 2006 and 2009 North Korean Nuclear Tests. Bull. Seism. Soc. Am.,
103 (3),1640-1661.
Nuttli, O. W., 1969. Travel times and amplitudes of S waves from nuclear explosions from Nevada.
Bull. Seism. Soc. Am., 59,385-398.
Okal, E. A., 1992. A student’s guide to teleseismic body wave amplitudes. Seism. Res. Lett., 63 (2),
169-180.
Pabian, F., Hecker. S.,2012. Contemplating a third nuclear test in North Korea, Bull Atomic Sci
Agust.
http://www.thebulletin.org/web-edition/features/contemplating-third-nuclear-test- north-korea.
Pabian, F., Coblentz, D., 2017. North Korea’s Punggye-ri Nuclear Test Site: Analysis Reveals Its
potential for Additional Testing with Significantly Higher Yields, . 38 NORTH (03/16/2017).
http://38north.org/2017/03/punggye031017/
Patton, H. J., Taylor, S. R.,2008. Effects of shock-induced tensible failure on mb - MS discrimination:
contrasts between historic nuclear explosions and the North Korean test of 9 October 2006.
Geophys. Res. Lett., 35 (14). DOI: 10.1029/2008GL034211
Phillips W. S., Begnaud, M. L., Rowe, C. A., Steck, L. K., Myers, S. C., Pasyanos, M. E., Ballard,
S. 2007. accounting for lateral variations of the upper mantle gradient in Pn tomography studies.
Geophys. Res. Lett., 35 (14). DOI: 10.1029/2008GL034211.
Richter, C. F., 1958. Elementary Seismology. WH Freeman and Company Inc. San Francisco and
London.
Rogers, A. J., Petersson, N. A., Sjogreen. B., 2010. Simulation of topographic effects on seismic
waves from shallow explosions near the North Korean nuclear test site with emphasis on shear
wave generation. J. Geophys. Res., 115 (B11309),1-27 (2010). doi:10.1029/2010JB007707.
Rougier, E., Patton, H. W., Knight, E. E., Bradley, C. R., 2011. Constraints on burial depth and yield
of the 25 May 2009 North Korea test from hydrodynamic simulation in a granite medium.
Geophys. Res. Lett., 38 (16). DOI:10.1029/2011GL048269.
Science Academy Press. 1964. The Geological structure and underground resources of NE Korea
and Southern Primorsky Krai of the Maritime Province, USSR (in Korean), The Geological
Institute, Pyongyang, DPRK and The Far East Geological Institute, Vladivostok, Siberia Branch,
AS, USSR.
Shin, J. S., Sheen, D., Kim. G., 2010. Regional observation of the second North Korean nuclear test
on 2009 May 25. Geophys. J. Int., 160, 243-250.
Stein, S. Wiens, D., 1986, Depth determination for shallow teleseismic earthquakes : methods and
results, Reviews Geophys., 24 (4), 806-832
@seismicisolation
@seismicisolation
Toksoz, M. N., Kehrer, H. H.,1971. Underground nuclear explosion: tectonic utility and dangers.
Science 173, 230-233.
Tsai, Y-B., Aki, K., 1971. Amplitude spectra of surface waves from small earthquakes and
underground nuclear explosions. J. Geophys. Res., 76 (17), 3940-3952.
Udias, A., 1999. Principles of Seismology. Cambridge University Press, Cambridge, UK
9DYU\þXN 9 .LP 6 * 1RQ-isotropic radiation of the 2013 North Korean nuclear
explosion. Geophys. Res. Lett. 41 (20),7048-7056. DOI: 10.1002/2014GL06126.
Van Der Hilst, R. D., Kennett, B. L. N., Shibutani, T., 1998. Upper mantle structure beneath
Australia from portable array deployments. Geodynamics 26, 39-57.
Viecelli, J. A., 1973. Topography and Rayleigh wave generating efficiency of buried explosive
sources. J. Geophys. Res., 78 (17). 3334-3339.
Wang, J., Israelsson, H., McLaughlin, K., 1999. Signal detection and estimation at the new
International monitoring ystem array at Mina, 21th Sesimic Research Symposium, sponsored by
U. S. Department of Defense, Defense Threat Reduction Agency, Contract DTRA01-99-C-0025.
Yliniemi, J., Kozlovskaya, E. Hjeltb, S-E., Komminacho, K., Ushakov A., 2004. Structure of the
crust and uppermost mantle rbeneath southern Finland revealed by analysis of local events
registered by the SVEKALAPKO seismic array SVEKALAPKO Seismic Tomography Working
Group. Tectonophysics 394 (1-2). 41-67.
Zhang, M., Wen, L., 2013. High-precision location and yield of North Korea’s 2013 nuclear test.
Geophys. Res. Lett., 40, 2941-2946.
@seismicisolation
@seismicisolation
APPENDIX: Principles of Nuclear Weapons

HEU Atomic Bomb Dropped over Hiroshima
Nuclear detonation-Hiroshima type. On August 6, 1945, USA B29 dropped a HEU atomic bomb
aka “Little Boy” equivalent to 15 kt and burst at a height of 500 m of Hiroshima for the first time in
the world history causing about 79,000 casualties.
Plutonium Atomic Bomb Dropped over Nagasaki
Implosion detonation-Nagasaki type. On August 9, 1945, USA B29 dropped a plutonium atomic
bomb aka “Fat Man” equivalent to 21 kt and burst at a height of 600 m of Nagasaki for the second
time in the world history causing about 74,000 casualties.
____________
Take a look at the below youtube for war nuclear bomb explosions
https://www.youtube.com/watch?v=hkTZmY3e8rk
@seismicisolation
@seismicisolation
Calculations of Nuclear Bomb Explosion Parameters for Seismological Calibration 2

SUNG-TACK SHIN
Abstract:
It is worth noting here the characteristics of earthquakes and explosions at the source parameters.
The most important calculation of nuclear bomb explosion parameters for seismological calibration
is the fission and fusion reaction energy release. When a nucleus of a uranium 235 atom captures a
neutron, it has a high probability of fission into two fragments of unequal mass. The missing mass_
has been converted into energy. The Einstein relation E=mc2 tells us that energy will be released.
The energy release is approximately 200MeV per fission of each nucleus. Nuclear energy can also
be released by the fusion of two nuclei. This is possible because the fusion products for nuclei of
small atomic number have less mass than the combined mass of the original nuclei that are fused. In
the nuclear reaction of a deuterium and a tritium nuclei fusion, l 7.6 MeV is the energy released.
How many generations occur in an atomic bomb explosion? This calculation shows the total energy
Most of this energy remains in the bomb and serves to heat it. Now 200 MeV is not a very large
number, by macroscopic standards. It is only 3.2 x 10-11 joule. But when there is a very large number
of fissions, the resulting energy is large.
How long does this process take? Most of energy comes out in the last few generations. The neutrons
released travel at speeds of about 10 million meters per second, or roughly 3 percent of the speed of
light. The characteristic time for a generation is the time it takes a neutron to cross the diameter of
the ball of U235. The total energy release and the process time are so important to seismological
calibration that the physicists could a discrimination of earthquakes and explosions.
Introduction to Nuclear Explosion Parameters
An explosion is far easier to define than an explosive. Berthelot, in 1883, defined an explosion as
the sudden expansion of gases into a volume much greater than their initial one, accompanied by
noise and violent mechanical effects. Explosions are common in nature. Such as physical explosions
are nevertheless inherently destructive and capable of causing damage by air blast and by the
propulsion of fragments at high velocity.
Nuclear explosions are a second class of explosions. These result in the sudden release of enormous
quantities of heat by fission or fusion processes. The heat causes the rapid expansion of the air
surrounding the device and also vaporizes material in the close vicinity which adds to the effect. The
radioactive elements are not themselves explosives in the conventional sense but the device is
triggered by the use of conventional explosives.
The third class of explosion is called a chemical explosion. This is caused by the extremely rapid
reaction of a chemical system to produce gas and heat. The reaction is usually a combustion process
and is often accompanied by the production of smoke and flame.
The most important calculation of underground nuclear explosion parameters for seismological
calibration is the fission and fusion reaction energy release. The energy parameter is the basic
information for seismic wave propagation. The yield of nuclear weapons is measured by the energy
released in the explosion. The energy release of nuclear weapons is measured in units of the energy
released from 1000 tons, or 1 kiloton(kt), of TNT (4.2 x 1012 joules)[l]. For thermonuclear bombs
the yield is often measured in megatons (Mt) of TNT. One megaton is equal to 1 million tons or
1000 kilotons. The energy released by the Hiroshima bomb was about 13 kilotons. The largest bomb
ever tested (a Soviet weapon) released about 60 megatons. The explosion time for a fission and
fusion reaction energy release is another important parameter for seismological calibration. Yet just
2
Calculation of nuclear bomb explosion parameters for seismological calibration,73-83 in Modern Seismology, edited
by So Gu Kim (1996). @seismicisolation
@seismicisolation
as significant as the energy released in the fission reaction is the fact that several neutrons are also
produced in the reaction.
These neutrons can be used to propagate a fission chain reaction. Most of these fission neutrons
appear essentially instantaneously (within 10 -14 sec) of the fission event. However, these prompt
neutrons live until across to bomb core (about 0.1 meter). The time it will take for average neutron
to cross the sphere will be calculated the following chapter.
Energy Release from Nuclear Reaction
The origin of the energy released by nuclear bombs is contained in Einstein's famous relation E =
mc2, where mis mass and c is the speed of light. One gram of matter, if converted entirely into
energy, would give an energy release equivalent to about 20 kilotons of TNT, which was the yield
of the bomb that destroyed Nagasaki.
Fission of uranium or plutonium nuclei provides a practical means by which the potential in
Einstein's relation can be realized. In a fission explosion a small fraction (about 0.1 percent) of the
uranium or plutonium mass is converted into energy.
Fusion of tritium and deuterium nuclei to form helium also produces nuclear energy. This is the
reaction of the thermonuclear bomb.
The energy release in a Hiroshima-size (13-kiloton) bomb is roughly equal to the energy that would
be released if all the automobiles in the United States (over 100 million of them) were accelerated
to a speed of 60 miles per hour and simutaneously crashed into each other. A 1-megaton explosion
would release 75 times more energy.
Energy Release from Nuclear Fission
A typical energy from nuclear fission reaction such as
1
on + 92U235 ĺ 38Sr95 + 54e139 + 2on1 + 199 MeV
Efission ¨PF2 a @;0H9DPX§ 200MeV
releases throughout a variety of reaction products, including the fissioned nuclei or fission products
and several neutrons as well as numerous gammas, betas, and neutrinos as in the following Table 1.
Of this 200MeV, some 168MeV appears as the kinetic energy of the fission fragments. These
fragments come to rest within about 0.01cm of the fission site so that all of their energy is converted
into thermal blast.
Table 1. Energy Release in Nuclear Fission [2]
Fonn Erelease (Me V) Range Time Delay
Ekinctic of fission fragments 168 FP instantaneous
Fast neutrons 5 10-100 cm instantaneous
Fission gamma energy 7 100 cm instantaneous
Fission product beta decay 8 short delayed
Neutrinos 12 nonrecoverable delayed
Neutron capture reactions 10 100 cm delayed
The basic concept behind obtaining energy from nuclear fission is

Total energy = energy release per fission x large number of fissions
# of fissions =[ m(U 235)gr /235 gr/mole]x NAvogadro
@seismicisolation
@seismicisolation
Suppose that a bomb contains 5 kg of 92U235 (100% enrichment) and that 10 percent of it fissions in
a nuclear explosion. How much energy will be released? In order to calculate this we need to know
how many uranium atoms there are in 5 kg of uranium.
The key to calculating the number of atoms is found in the mass number. In U235 there are 235
protons and neutrons. If we know the mass of an individual proton and neutron, then the number of
protons and neutrons can be found by dividing their individual masses into the total mass of the
element or isotope. This kind of calculation is done very often in science. To make it simple we use
the concept of a mole. A mole of a substance is the mass (in grams) corresponding numerically to
the mass number of the element. Thus the mole weight of U235 is 235 gr, and the mole weight of
U238 is 238 gr. The advantage of the mole-weight concept is that 1 mole weight (often called gram-
molecular weight) contains the same number of nuclei, regardless of the mass number.
There are just 6.023 x 1023 nuclei contained in 1 mole of any substance. This number is so important
and so often used that it is given a special name· such as Avogadro's number. One mole of uranium-
235 has a mass of 235 gr. Hence 235 gr of uraniurn-235 contains 6.023 x 1023 uranium atoms. In
our example we assume:
5 kg of U235 & fission 10% of it = 5 kg x 0.01 = 500 gr
500 gr / 235 gr/mole = 2.13 moles
2.13 moles x (6.023 x 1023) = 1.3 x 1024 fissioned U235 nuclei
If each nucleus of the atom releases 200MeV when it fissions, the total energy release is
(1.3 x 1024fissions) x 200MeV/fission) x (1.6 x 1013J/M eV) = 4.1 x 1013J
Since there are 4.2 x 1012 J/kt, the energy released in this example is approximately equivalent
to 10 kt of TNT.
Calculation of the Energy Released in a Fusion Bomb
Nuclear energy can also be released by the fusion of two nuclei. This is possible because the fusion
products for nuclei of small atomic number have less mass than the combined mass of the original
nuclei that are fused. The fusion of deuterium and
tritium to form helium( a-particle) and a neutron with a release of energy may be written:
1H2 + 1H3 ĺ2He4 + on1 + 17.6 MeV
{[m(1H2) + m(1H3)] - [ m(2He4) + m(on1 @`[60H9DPX§0H9
Suppose that a fusion bomb is composed of I00 kg of deuterium (2gr /mo le) and I50 kg of
tritium(3gr /mole), and that 10% of the material in the bomb fuses in the explosion. What is the
energy release? The total mass involved in the fusion reactions is 0.1 x (I00 kg) of deuterium and
0.1 x (150 kg) of tritium. If we use the deuterium to measure the number of reactions, we have
(10 kg) x (l000 gr / kg ) x (6.023 x 1023nuclei/mole) x (0.5 mole/g r)
= 3.01 x 1027 fusion reactions
Each reaction releases l7.6 MeV of energy. Since there are 1.6 x 10-131/MeV, each reaction releases
l7.6 MeV x 1.6 x 10-13J/MeV = 2.82 x 10-12J. The total energy released is then the product of the
energy released by the individual reaction multiplied by the number of reactions:
@seismicisolation
@seismicisolation
Total energy released = (2.82 x 10-12J/reaction) x (3.01 x 1027reactions)

= 8.5 X 1015 J /4.2 X 1012J/kt
= 2 Megatons
Energy Release from High Explosives
High explosives is called a chemical explosion in the nuclear weapon process. Both 'A Gun-Type
Bomb (Little Boy)' and 'An Implosion Bomb (Fat Man)' are surrounded by very carefully arranged
layers of high explosives. At the proper time the explosive is symmetrically detonated. The force of
the resulting shock wave compresses the nuclear material into a much smaller volume. As the total
surface area decreases, there is less area for neutrons to escape from. The surface to mass ratio
decreases and the material becomes supercritical. This process can be represented as the steady-state
diffusion equation: [3]
The requirements for the high explosives used in implosion bombs are critical. The explosive must
produce an extremely precise shock wave which will collapse and compress the nuclear material
symmetrically so as to achieve a critical mass and sustain it for enough generations
for_the_design_energy to be released. Symmetrical compression is accomplished using shaped
charges and explosive lenses.
Performance of Explosives
The effectiveness of an explosive depends on two factors. The first is the amount of in the explosive
and secondly, the rate of release of the available energy when the explosion occurs. To measure the
effectiveness of different explosives a variety of performance parameters may be used such as:
‫ ٲ‬Heat of explosion
‫ٲ‬Pressure of explosion
‫ ٲ‬Rate of burning
‫ ٲ‬Detonation pressure.
‫ ٲ‬Temperature of explosion
‫ ٲ‬Power index
‫ ٲ‬Detonation velocity
+HDWRI([SORVLRQ
When an explosive is initiated either to rapid burning or to detonation, its energy is released in the
form of heat. The heat so released under . adiabatic conditions is called the Heat of Explosion and
determines the work capacity of the explosive.
@seismicisolation
@seismicisolation
Table 2. Calculated Heats of Explosion for a Nuclear Bomb HE [4]

leading Hes for A-bombs joules per gramme
TNT(C2H5N3O6) 4080
RDX(C3H4N6O6) 5130
HMX(C4H8N8O8) 5130
RDX/TNT 60:40 4500
2[\JHQ%DODQFH
7KHR[\JHQEDODQFH RIDQH[SORVLYHLVGHILQHGDVWKHSHUFHQWDJHE\ZHLJKWRIR[\JHQpositive
or negative, remaining after explosion, assuming that all the carbon and hydrogen atoms in the
explosive are converted into CO2 and H2O. All the explosive classes contain only the elements
carbon hydrogen oxygen and nitrogen and are called CHNO explosives having the general formula:
Ca Hb Nc Od
Simple algebra shows that the oxygen balance is given by the formula:
G- 2a - b /2) x1600/M (M=the relative molecular mass of the HE)
When there is exactly enough oxygen in the explosive to fully oxidize the carbon and hydrogen to
FDUERQGLR[LGHDQGZDWHU LH WKHQWKHKHDWRIH[SORVLRQZLOOEHRSWLPDOWKDWLV KDVWKH
highest heat, either far positive or negative has a lower heat of explosion.
Table 3. Oxygen balance of a Nuclear Bomb HE
Oxygen Balance per cent
Nitroglycerine + 3.5
TNT(C7H5N3O6) - 74
RDX(C3H6N6O6) - 22
HMX(C4H8N8O8) - 21
Detonation Pressure (p)

The peak dynamic pressure in the shock front is called the detonation pressure of the explosive. An
empirical method of calculating it is due to Cook, as follows:
P(kbar) ¨D2 x 2.50 x 10 -6
:KHUH ¨ LV WKH FKDUJH GHQVLW\ LQ JUFP3. Dis the velocity of detonation (m / sec) TABLE 4.
Detonation Parameters for Some Military Explosives
@seismicisolation
@seismicisolation
Type D/m s-1 · -3

AW¨/ gr cm p / kilobar
TNT 6950 1.57 190

RDX 8440 1.70 300
HMX 9110 1.89 392
RDX/TNT 60/40 7900 1.72 268
The Concept of Equivalent Megatons (EMT)

In evaluating the destructive power of a nuclear weapon system it is usual to use the concept of
equivalent megatons. Equivalent megatonnage is defined as the actual megatonnage raised to the
two-thirds power:
EMT = Y2/3 where Y is the yield in megatons
This relation stems from the fact that the destructive power of a bomb does not vary linearly with
the yield. A given blast pressure occurs at a distance from ground zero which varies as the cube root
of the weapon yield. The area destroyed by blast is proportional to the square of this. Because the
destructive power of a given megatonnage depends nonlinearly on the yield, division of
megatonnage among a number of warheads increases the total amount of destruction possible.
Table 5. Destructive power of lMt distributed in several ways [5]
No. of bombs EMT Destroyed

and yield per bomb area, square miles
1 X l Mt 1 80
8 X 125 kt 2 160
20 X 50 kt 2.7 216
100 X10 kt 4.6 368
How Long Does an Atomic Bomb Explosion Take?

The speed of a neutron released in the fission process can be estimated as follows. The average
neutron energy is about lMeV. The neutron mass is 1.6749544 x 10-2 kg. Using the expression
relating kinetic energy to speed, we have:
KE = ms2/2
KE = (1MeV) X (1.6 X 10-13J/MeV) = 1.6 X 10-13 J
S=(2KE/m)1/2=[2x(1.6x10-13J)/1.67x10-27kg]1/2 = 1.4x107 m/sec
How long does this nuclear reaction process take? The neutrons released travel at speeds of about
10 million meters per second, or roughly 3% of the speed of light. The characteristic time for a
generation is, roughly, the time it takes a neutron to cross the diameter of the ball of U235 (If neutrons
escape most of the time, there will be no sustained reaction; if no neutrons escape, there is extra
uranium present and the bomb can be made smaller.) Suppose we have a ball of uranium that is
.
roughly base ball-sized, with a diameter of about 10 centimeters. The time t it will take for a neutron
to cross the sphere is
@seismicisolation
@seismicisolation
t= 0.1 meter/1.4x107 meters/sec =

7.1429x10-9 sec =7 nanoseconds
that is, about a hundred-millionth of a second. The complete process of bomb explosion, from the
release of the first neutron to the total energy release, is about 80 times this number, or roughly 100
hundred-millionths of a second, or a microsecond. Since 99.9% of the energy is released in the final
10 generations, the important macroscopic energy-release processes take place in about one-tenth
of a microsecond. (it was called 'shake ', means "as fast as the shake of a lamb's tail")
Producing Neutrons and Assembly Speed (U235 vs. Pu239)
One simple initiator design uses a radioactive source of alpha particles surrounded by beryllium
powder is a spherical shell of thin metal such as 0.001inch aluminum. When the plutonium is
compressed by the high explosive surrounding it, the metallic shell ruptures, allowing the alpha
particles to bombard the beryllium powder. The following nuclear reaction is an efficient means for
producing neutrons in a controlled manner:
4
2H ( ࢻ-emitter)+ 4Be9 ĺ6C12+ 0n1(seed neutron)
Figure l. A schematic representation of the initiator mechanism of an implosion bomb. [6]
Let us compare Pu239 and U235, with the goal of estimating the required bomb assembly speed.
Consider a 10 kg mass of material. (This is substantially more material than would be required for
a bomb – 4.4 kg is the critical mass in a configuration in which the plutonium is surrounded by a
natural uranium reflector.) The spontaneous neutron emission rates are as follows.
Pu239: (0.03 neutron / gr. sec) x (1 x 10 gr)
= 300 neutrons / sec
Where the mean time between neutrons is 1 / 300 sec = 3300 μ sec.
U235 : (4x10-4 neutron/gr.sec)x(1x104 gr) = 5 neutrons/sec
where the mean time between neutrons is 1 / 5 sec = 200,000 μ sec.
If the material is not pure, then the spontaneous neutron production rates may be much higher. The
need to keep spontaneous neutron emission rates low is one reason why weapons-grade plutonium
must contain at most small quantities of isotopes other than Pu239 (The other important reason has
to do with absorption of neutrons - a major reason why weapons-grade uranium must be highly
enriched in U235
ACKNOWLEDGMENTS
This paper was originally reproduced from the Proceedings of the Korea-China International
) Joint
Seminar and Seismological Workshop, Seoul, Korea on January 29-31, 1996. We appreciate the
support by the Seismological Institute, Hanyang University, Institute of Geophysics, State
Seismological Bureau, China and Center for International Science and Technology Cooperation,
@seismicisolation
@seismicisolation
STEPI, Republic of Korea.

REFERENCES
1. Paul P. Craig, John A. Jungerman, Nuclear Arms Race Technology and Society, McGraw-Hill
Book Company, (1986)
2. John R. Lamarsh, Introduction to Nuclear Engineering 2nd. Edition, Addison Wesley Publishing
Company, (1983)
3. James J. Duderstadt, Louis J. Hamilton, Nuclear Reactor Analysis, John Wiley & Sons, Inc.,
(1976)
4. A. Bailey, S. G. Murray, Explosives, Propellants and Pyrotechnics, Brassey' s (UK), London,
(1989)
5. Dietrich Schroeer, Science, Technology, and The Nuclear Arms Race, John Wiley & Sons, (1984)
6. G. E. Strickfaden, Nuclear Energy and Proliferation Intelligence Workshop, Part 8: Weaponization
and Manufacturing, USA. Los Alamos National Laboratory Non-official Textbook Printing for
International Intelligence Officer, (1993 June)
@seismicisolation
@seismicisolation
CHAPTER TWENTY-THREE
ON COMPUTATION OF DEPTH AND SOURCE CHARACTERISTICS OF THE

DPRK NUCLEAR TESTS (2006, 2009, 2013, 2016J, 2016S AND 2017)
USING DEPTH PHASES AND WAVEFORM STACKING
SO GU KIM, YEFIM GITTERMAN, SEOUNG-KYU LEE, HYUNG-SUB BAE

AND GILL JAE LEE
Topography near nuclear test site for the North Korean nuclear explosions determined by CTBTO
(closed stars) and USGS (open stars)
ABSTRACT
nuclear explosions on October 9, 2006 (M 4.3), May 25, 2009 (M 4.7), February 12, 2013 (M 5.1),
January 6, 2016 (M 5.1), September 9, 2016 (M 5.3) and September 3, 2017 (M 6.3). The principal
objective of this paper is to estimate absolute source depths for the North Korean nuclear tests
using spectral nulls of body and Rayleigh wave spectra. Source depths were calculated from the
spectral nulls (minima) of Rg wave as well as P-wave spectra. The source depths for the 2006,
2009, 2013, 2016J, 2016S and 2017 nuclear tests were thus estimated at 2.12, 2.06, 2.05, 2.06,
,2.05 and 1.97 km respectively taking into account the minimum nonlinear effects of topography
and damage by surface spallation. It should be noted that our over-burial determination is
significantly greater than expected from the standard experiment practice of nuclear nations. In
particular we found the Rg-wave spectral@seismicisolation
nulls in North Korean nuclear tests to appear due to a
@seismicisolation
500 Chapter Twenty-Three
most general source type of pure dip-slip motion with about 45°dip. We assume that the North
Korean nuclear test source may be an inverted conical volume dip-dip slip motion which could be
a vertically distributed source in vertically contained shafts through mutually connected-deep
tunnels. The over-burial determination would affect the interpretation of MS-mb discriminants and
the estimation of the seismic yield as well.
Keywords: CTBT, depth phase, spectral null, Love wave, Mt. Mantap
INTRODUCTION
Source depth is pertinent for ensuring enforcement of the Comprehensive Nuclear-Test-Ban-
Treaty (CTBT) as nuclear explosions are unlikely to be fired at depths of more than a few
kilometers1, 2, 3. Two basic techniques of spectral minima for body and Rayleigh wave spectra are
utilized for depth analysis. Rayleigh wave spectra can be used to determine depths, as mode
excitation reflects the behavior of the eigen functions with depth. The primary limitation on this
study arises from lateral heterogeneity of velocity structure and mechanism. The body wave
spectra are also used to estimate depths by the time separation between the direct P-wave arrivals
and near-source surface reflections. These studies can be limited by a partial trade-off between
depth and source time function duration, but the trade-off with source time function duration is
found not to generally produce major difficulties for small events like nuclear explosions4
especially North Korean nuclear explosions within the rigid granite source. The DPRK conducted
the first nuclear test (M 4.3) on October 9, 2006, the second one (M 4.7) on May 25, 2009 and the
third one (M 5.1) on February 12, 2013, the fourth one (M5.1) on January 6, 2016, the fifth one
(M5.3) on September 9, 2016 beneath the vicinity of Mt. Mantap, but sixth one (6.3) on September
3, 2017 far away NW from the Mt. Mantap in the Punggye-ri region of northeastern DPRK (Table
1).
Although depth determination through the measured time differences between the direct P-wave
and the pPand/or sP depth phase is difficult for the nonlinear effects of mountain topography and
shallow spallation in this region, the average spectra of global seismic arrays can be used by the
identification of spectral nulls due to the interference between the direct and surface reflected
arrivals including showing the 3D topography map. To provide robustness of the result is to
consider teleseismic and regional arrays at a wide range of better azimuthal distribution. For a
simple structural model, simple formulas are developed to estimate the source depth from
identified spectral nulls.
Table 1. Source parameters of North Korean nuclear tests. Origin times, locations and
magnitudes were extracted from the United States Geological Survey (USGS). Seismic yields
calculated from methods utilized by Murphy (1981) and Bowers et al. (2001) in the brackets,
respectively. Depths and magnitudes in the brackets refer to this study and IDC/CTBTO
determination, respectively.
Date Origin time* Location (N, E) mb Seismic yield (kt) Depth (km)
10/09/2006 01:35:28 41.294, 129.094 4.3 (4.1) 1.67 (1.17) 0 (2.12)
05/25/2009 00:54:43 41.306, 129.029 4.7 (4.5) 9.18 (3.98) 0 (2.06)
02/12/2013 02:57:51 41.308, 129.076 5.1 (4.9) 28.63 (13.59) 0 (2.05)
01/06/2016 01:31:01 41.326, 129.010 5.1 (4.8) 28.63 (13.59) 0 (2.06)
09/09/2016 00:30:01 41.302, 129.043 5.3 (5.1) 50.55 (25.12) 0 (2.05)
09/03/2017 03:30:01 41.333, 129.056 6.3 (6.1) 867.50 (505.48) 1.0** (1.97)
*UTC (Universal Time Coordinated). and ** represents GEOFON, GFZ Germany estimate. The epicenter for the
2017 is estimated at 41.3296º N and 129.0267º E by IDC which are preferred in this study.
The strength of a seismic signal and the body-wave magnitude (mb) depends strongly on the
medium in which the explosion is detonated. Explosions in hard rock, like granite or tuff give
considerably stronger amplitudes than explosions in unconsolidated rock such as alluvium.
@seismicisolation
@seismicisolation
On Computation of Depth and Source Characteristics of the DPRK Nuclear Tests 501
Seismic signal strength depends not only on seismic yield and the medium but also on the wave
transmission properties of the Earth. These properties vary considerably for different source-
receiver paths. The range of the yield would be a factor of about 10. Taking this uncertainty into
account, we correlate magnitude 5.0 roughly with 100 kt in dry unconsolidated rock and with 10 kt
for wet hard rock5. These crude relations can be considered as indicating an upper and lower yield
limit for a given magnitude. Furthermore, various researchers derived different magnitude-yield
relations depending on the detonation site (Table 1). For example, using mb=3.92 + 0.81 logY for
wet hard rock6, it resulted in a 29 kt yield for the 2013 nuclear test while it alternatively estimated
at a 14 kt yield when using mb = 4.25 + b logY for a fully coupled explosion where Y is the
explosive yield in kilotons (kt), with b=0.75 for < and b=1.0 for Y<1 which is suitable for
conservatively estimating the yield threshold of a potential violation of the CTBT in the Novaya
Zemlya region2. Seismic signals from an explosion can also be reduced by detonating an explosion
in a large underground cavity due to decoupling effects5,7. Because emplacement conditions are
unknown for North Korean nuclear explosions, it is most difficult to estimate seismic yields of
underground nuclear explosions conducted surreptitiously. The seismic yield, depth of burial and
medium around the emplaced nuclear device including configuration play significant roles in
verifying the observed seismic signals of underground nuclear explosions.
DEPTH ESTIMATION USING PP-P AND SP-P DELAY TIMES

The usual way to estimate the depth of an event using teleseismic waves is to measure the time
difference between the arrival time of the direct P wave and the arrival time of the surface
reflections pP and sP. pP and sP are signals that travel upward from the focus as P and S waves8-10
and then travel steeply downwards following the same path as to the receiving stations of the P
waves. The relative amplitudes and polarities of these phases depend on the source mechanism
while the time delay depends on the velocity structure and the ray paths. For shallow sources,
identification of individual arrival times is difficult in the time domain because the arrival pulses
interfere with each other. However, this interference has a strong effect on the amplitude spectra
that may be exploited to estimate the depth.
To provide waveforms for spectral analysis, we used stacked P-wave signals of the vertical
components from nine telesemic seismic arrays (Figure 1). Such stacking increases the S/N of the
P-wave signals by suppressing the effects of local noise. We performed spectral analyses taking 10
or 30 second time- windows of short-period vertical components for the ASAR (Alice Springs,
Australia), FINES (FINESS, Finland), KURK (Kurchatov, Kazakhstan) NOA (Norway), NVAR
(Mina, USA), PDAR (Pinedale, USA) including South Korean local stations and KSRS (Korea
Seismological Research Station, Wonju, South Korea), USRK (Ussuriysk, Russia) and broadband
records for WRA (Warramunga, Australia) and YKA (Yellowknife, Canada), ISN (Israel Seismic
Network) and DSN (Dongbei Seismic Network) along the border between northeastern China and
North Korea (Figure 1). The effects of absorption and the recording system can be removed by
utilizing a source in the same high velocity layer and the same array stations. The spectral nulls
(minima) from destructive interference of pP+P and sP+P spectra are equivalent to a reciprocal of
pP-P and sP-P lag time so that the pP delay times estimated from the spectral analysis are
compatible with the pP phase arrivals in the time domain. The high quality factor Q in the granite
source rock (1000) from amplitude A(t) = A (0)exp(-ʌIW4 and low attenuation of source in the
high velocity layer clearly reveal high frequency arrivals of pP and sP in the teleseismic
seismograms of the seismic arrays which fit to detect underground nuclear tests.
@seismicisolation
@seismicisolation
Figure 1. Closed squares represent global arrays of the ASAR, FINES, ISN (PRNI), KURK, NOA, NVAR, PDAR,
WRA and YKA in better azimuthal coverage and regional arrays KSRS and USRK including MDJ. Closed triangles
represent DSN (Dongbei Seismic Network) deployed near the border between China and North Korea including South
Korean stations used in this study. The red star indicates the underground nuclear test site.
The interference produces spectral modulation (scalloping) – minima (spectral nulls) at

fundamental and multiple frequencies1, 11-13. The spectral null f01 of the fundamental mode is
represented as follows:
f01 = Vp/2cos (i)*d, f0n = n* f01 , n = 1, 2, … (1)
nulls of the stacked spectra from the respective array traces to enhance seismic signals in noise
(Figure 2).
In case of a small take-off angle of pP for a long range epicentral distance ¨ 75.0°), we may
approximate equation (2) as d § pP-P*Vp/2 which was utilized by many researchers 8, 12, 15.
Ideally, the pP- P time would be obtained from the time difference between the onset of P and that
of the downswing depth phase of pP in the time domain. However, those onsets are difficult to
pick so an alternative is to use spectral nulls of the average spectra because the depth phases are
always 180°out of phase with the incoming P-wave arrivals. i.e. we measured the time difference
between the maximum positive amplitude P- wave arrival and the maximum negative amplitude of
pP assuming P and pP have an identical shape with negative polarity and interference between the
two pulses 16. The absence of depth phase is the result of scattering and defocusing of P waves by
lateral variation of medium above the source17. However, we can minimize the nonlinear
topographic effects of the mountainous topography and the azimuthal variations of pP
waveforms18 using the average spectra obtained from a broad and uniformly azimuthal distribution
of teleseismic arrays (Figure 1). We used spectral nulls of the average spectra of multi-pathing
arrays and utilized the full depth formula with take-off angles assuming the average overburden P-
wave velocity to be Vp=5.1 km/sec19 at the DPRK nuclear test site.
From the relationship between ray paths and travel times, W[ = sin (i)/v0 =1/vz for flat earth,
where i= a take-off angle for the reflected pP wave, v0= P-wave velocity at the focal depth, vz =
the bottoming average Vp velocity (vav) where the ray returns at i=90° where 1/Vz=1/vapp = ߲ܶ/߲¨
(travel-time curve slope). Using the take-off angle i, where i= arcsin (v0/vav), the time delay for pP-
@seismicisolation
@seismicisolation
P from the spectral null (a reciprocal of spectral null), and P-wave velocity at the focal depth, v0 as
5.1 km/sec, we can determine depth from equation (2). Lay20 stated that the source depth estimated
by pP-P delay times may be overestimated by biasing effects such as delay times of depth phase
propagation for very short lag times with limited bandwidth. However, we assume that the
nonlinear effects due to surface spallation or damage do not influence the homogeneous source
rock of granite for most of North Korean nuclear tests. We also offset the nonlinear topographic
effects by using the average spectra of stacked signals from different azimuthal propagation of
seismic arrays and take-off angles.
From Figure 2a,
d = (pP-P)Vp/2cos (i) (2)
In case of a small take-off angle of pP for a long range epicentral distance ¨ 75.0°, we may
approximate equation (2) as d § pP-P*Vp/2 which was utilized by many researchers 11, 12. 15. The
estimated depth using sP-P delay times is derived using Snell’s law as follows (Figure 2a):
d = (sP-P)VpVs/[Vp cos (j) + Vs cos (i)] (3)
NOA, NVAR, PDAR and broadband vertical components from the WRA and YKA.
Source depth can be also estimated for a simple single-layer over a half-space model from the time
delay between Pn and pPn using the following equation. The source depth by means of utilizing
pP-Pn can be derived considering the take-off angle for the Moho discontinuity with its velocity
Vn and the overburden average velocity Vp as follows:
From Figure 2b,
d= (pPn-Pn)Vp/2cos(Į) (4)
depth d =(sPn-Pn)/F(Vp, Vs, Vn, Į ȕ) (5)
where
F(Vp, Vs, Vn, Į ȕ = 9S9QFRVĮ + 9V9QFRVȕ – 9S9VFRVĮVLQ ȕ – 9S9VFRVȕVLQ
Į 9S9V9QFRVĮFRV ȕ
= >FRVĮ (VpVn-VpVs VLQȕ cos ȕ (VsVn-VpVs sin Į @FRV ĮFRVȕ9S9V9Q (6)
where Į and ȕ are take-off angles for pPn and sPn phases. Į = arcsin (Vp/Vn) and ȕ =arcsin
[Vs/Vp sin (Į)] according to Snell’s law. We used the average spectra from spectral nulls for
destructive interference pPn + Pn and sPn+Pn since the ray paths should pass through the
heterogeneous and anisotropic crustal structures to reach each station of the array. Using equations
(4) and (5), Vp=5.1 km/sec, Vs=2.9 km/sec and Vn=7.8 km/sec from the refracted P-wave velocity
of ray paths through the central part of the Chugaryeong Rift Zone and the subcontinental crust of
the East Sea21-23 whereas Vn=8.0 km for USRK array and MDJ.
@seismicisolation
@seismicisolation
a b
Figure 2. a) b) Ray trajectories of P, pP and sP at source of the nuclear test. i and j are take-off angles for pP and sP
ray-paths. G, G', R and R' indicate the contact points on the surface and on the ray-path by pP and sP. S and O
represent seismic stations and the center of the Earth, respectively. b) Depth can be calculated using travel time
difference using pPn-Pn and sPn-Pn from the ray paths for pPn, sPn and Pn.
DEPTH INTERPRETATION USING PP-P AND PPN-PN DELAY TIMES

The spectral nulls are used to determine depths taking an average spectra with the band-pass
filtered between 0.5 Hz-3.0 or 0.5 Hz-5.0 Hz. Here the only spectra of WRA are presented as a
representative teleseismic array for the spectral analysis.
The depth phase method is most successful when P-wave arrival has a signal-to-noise ratio (SNR)
greater than 4 -7 and the depth phase pP exhibits a SNR (signal-to-noise ratio) 24 greater than ~2.
The most of nuclear tests were valid for these conditions, but the 2006 test was hampered by weak
signals and noise (Figure 3). It is often not certain if the difference in spectral nulls is due to
intrinsic source differences or because of its path within the limited time window for the near-
field. Nonetheless, the delay times of the teleseismic arrays are very sensitive to the propagation
path effects of the depth phase pP.
explosion depth and the P-wave velocity in the material above the shot point25 because the
corresponding pP arrivals would be expected to be reduced in amplitude and possibly delayed
relatively to what would be expected for true pP in case of the damage effect zone above an
explosion. The reflected signals containing the fracture and crushing associated with the explosion
would be also reduced and delayed resulting in the slapdown phase as a secondary source.
However, these phenomena of depth delay times would be excluded in North Korean nuclear
detonation at a source of rigid granite even if there may be some subsurface spallation at not above
shot point for large explosions such as 2016S and 2017 test. We show the spectral analyses for five
North Korean nuclear tests using WRA in Figure 3 and spectral nulls for the 2016J test in Table 2
for examples. The low frequency of spectral nulls at ASAR (Table 2) may be more closely related
to the low velocity layer in the Great Artesian Basin beneath the seismic array which is a large
aquifer with unwater-bearing formation including the low velocity zone in the upper mantle of the
Late Palaeozoic Alice Springs Orogeny beneath the array according to the tomography study25.
From a known firing depth of 0.463 km in granite and a P-wave velocity of 4.8 km/sec, they found
it impossible to see the predicted delay times of 0.2 sec from pP-P, because the pP-P delay times
were found to be 0.8 sec (corresponding to a depth at about 2 km) from the Pile Driver explosion12,
26
since it was conducted at a shallow depth (around 500 m) of granite resulting in the surface
@seismicisolation
@seismicisolation
damage of fracture and crushing from a spalling. However, in this study it should be noted that
North Korean nuclear explosions were conducted in the deep homogeneous granite without the
surface spallation27. There is another example of a deep-depth estimate using the depth phase pP-P
for a deep atomic explosion28. On May 21, 1968 a 47-kiloton (mb =5.6) atomic explosion
(detonated in salt) was fired at a depth of 2.45 km near Bukhara, USSR to seal an oil well. It was
observed by pP-P delay times of 1.7 seconds at YKA (79°) indicating that source depth was
estimated at about 2.6 km which was in very good agreement with the actual depth.
Figure 3. Representative spectral nulls to determine the detonation depth by using WRA array for the 2006, 2009,
2013, 2016J and 2016S tests. The source depth is estimated by pP-P delay times from the destructive interference
(pP+P) of P and the surface reflected P wave (pP).
All spectra of single array show similar ‘spectral nulls’ indicating that the pseudo-1D radiation in
one direction is constant and neglecting the contribution of different source components. So data
from a test in the middle of a mountain will be look alike in all directions since an onset time of P-
wave arrivals at each element of an array follows Fermat Principle and Snell’s Law that a raypath
corresponds to a minimum of travel time and that an incident angle (take-off angle) is always the
same as a reflected angle of a single raypath within a limited time window. It should be noted that
there is almost no elevation difference between a source and a reflector with a shallow dip as the
closest approximation possible on average.
Consequently we do not count on the unknown and complicated terrain of the test sites and
neglected the trade-off between our applied simplifications and the claimed accuracy of the results
from the nonlinear topographic effects considering almost uniformly azimuthal coverage by means
of global seismic arrays from the source.
In local stations there is strong variation in velocity-sensitive parameters such as the ray paths and
crustal structures with different phases resulting in incongruity with the estimates of pPn-Pn delay
times. However, the source depths are more reliably estimated using a local seismic array which is
@seismicisolation
@seismicisolation
located in the stable region with high Q in the Korean Peninsula and can enhance a good SNR by
superposing signals, resulting in strong average spectral nulls of regional P-wave spectra in KSRS
array (Wonju, Republic of Korea) with a low Moho discontinuity velocity of Pn (7.8 km/sec) (see
Figure 4), whereas a low velocity layer at USRK (Ussuriysk, Russia) array lying beneath the
Pacific slab of a high Moho discontinuity velocity of Pn (8.0 km/sec) (Figures 5). Consequently,
the irregularly dipping mountainous topographic effects do not significantly influence the depth
estimation.
Table 2. The source depth of the 2016J nuclear test was determined via pP-P delay times
with take-off angles using 10 seismic arrays. distance and AZ (direction to the station) are
measured in degrees. The asterisk (*) symbol indicates the representative station PRNI of the
Israel Seismic Network.
(°) (Hz) (°) (km)
ASAR (64.8, 175.1) 1.1 17.57 2.43
ARCES(56.4, 335.8) 1.25 18.91 2.16
EKA (75.5, 334.1) 1.35 15.56 1.96
ISN (73.5, 295.6)* 1.28 16.13 2.07
KURK (35.6, 302.8) 1.22 22.86 2.27
NORESS (66.2,331.9) 1.25 17.15 2.14
NVAR (79.7, 47.5) 1.25 (0.98) 14.62 (8.25) 2.11 (1.92)
PDAR (81.0, 39.5 ) 1.25 (1.02) 14.33 (8.09) 2.11 (1.85)
WRA (61.1, 174.3) 1.25 18.40 2.15
YKA (64.7, 27.4) 1.25 17.57 2.14
Values in the brackets refer to sP+P spectral null and depth including a take-off angle of sP phase
@seismicisolation
@seismicisolation
Figure 4. a) Comparison of the spectral power density of the Pn-wave observed for the nuclear explosions and two
earthquakes recorded at KSRS (KS31) was produced by German researchers (Hartmann et al., 2017) 29 indicating that
the spectral nulls of pPn-Pn and sPn-Pn are consistent with those of our study. The mean spectral amplitudes decrease
between 2 and 8 Hz approximated by a least-square fit is marked. b) The average source depths were estimated at
2.13, 2.05, 2.05, 2.03 and 2.03 km respectively using pPn-Pn and sPn-Pn for the 2006, 2009, 2013, 2016J and 2016S
tests from KSRS. These estimates are fairly consistent with those of pP-P estimates. b)
@seismicisolation
@seismicisolation
Figure 5. The spectral analyses of regional P waves to determine the detonation depth for the 2009, 2013. 2016J and
2016S using USRK. The source depth is estimated by the delay times of pPn-Pn (spectral nulls of pPn+Pn) from the
destructive interference P and pPn refracted from the Moho discontinuity. The spectral nulls for 2009, 2013, 2016J
and 2016S tests are estimated at 1.57 Hz. 1.56 Hz, 1.50 (1.42) Hz and 1.50 Hz whereas the spectral null at 1.46 (1.50)
Hz for the 2006 test in the lower figure.
The distinction at the Pn signals between nuclear explosions and earthquakes can also be verified
by the frequency characteristics of the decrease of the amplitudes at frequencies between 2 and 8
Hz in Figure 4b. The gradient of the decrease for earthquakes exhibits a larger decrease with
higher frequencies than the explosions.
The spectral nulls for the 2009, 2013, 2016J and 2016S tests were found to be 1.57 Hz, 1.56Hz,
1.50 Hz and 1.50 Hz indicating that depths should be 2.11, 2.12, 2.20 and 2.20 km respectively
assuming that the Pn velocity is 8.0 km/sec for USRK in the Sikhote Alin region (Figure 5), but
the low velocity layer overlying the Pacific slab which reduces the velocity beneath the USRK
seismic array. Furthermore, the spectral null for the 2006 test is much lower than other tests.
Consequently, the actual spectral nulls should be higher than those of observations so that depth
@seismicisolation
@seismicisolation
estimate for USRK is unreliable.

We found the spectral nulls for the 2017 test (Figures 6, 7 and 8) to be low and complicated to read
on the spectra indicating that the source estimate was influenced by a sub-surface spallation at a
certain depth near the source. The source is located beneath the middle of the top rugged
topographic mountain with the reflector located on the rugged mountain. As a result, we observe
more surface waves such as Love waves as well as Rayleigh waves (Figure 9) including the
secondary source of a slapdown phase (see Table 3 and Figure 7). The slapdown may not be
related to a surface spall above a shot point.
Figure 6. Red and blue arrows indicate spectral nulls of pP + P and sP + P using teleseismic arrays for the 2017 test.
The spectral nulls for the 2017 test present low frequencies and do not appear at some arrays like ASAR and NVAR.
The slapdown phases appear at 0.57 Hz (after 1.75 seconds from the onset) from the spectra of ARCES and EKA
records. The vertical black, red, blue and brown bars in ARCES indicate P-wave, pP, sP and slapdown phase arrivals.
We found the spectral nulls for the 2017 test (Figures 6 and 7) to be low and complicated to read
on the spectra indicating that the source estimate was influenced by some sub-surface spallation
near the source. The source is located in the middle of the top rugged topographic mountain with
the reflector located on the rugged mountain resulting in generation of more surface waves in Love
waves as well as Rayleigh waves including the secondary source of a slapdown phase (see Table 3
and Figure 7). The slapdown may not be related to a surface spall above a shot point.
@seismicisolation
@seismicisolation
(°) (Hz) (°) (km)
ASAR (64.8, 175.2) none 17.57 none
ARCES (56.4, 335.8) 1.47 (0.95) 18.91 (10.62) 1.83 (2.01)
EKA (75.5, 334.1) 1.40 15.56 1.89
FINES (60.3, 327.4) 1.57 (0.97) 18.62 (10.46) 1.71 (1.95)
NORESS (66.2, 332.0) 1.35 17.14 1.98
PDAR (81.0, 39.6) 1.33 14.33 1.98
WRA (61.2, 174.4) 1.43 18.40 1.88
YKA (64.7, 27.4) 1.24 17.57 2.05
Average depth , km 1.92± 0.10
Note: The numbers in the brackets refer to spectral nulls and depths via sP-P delay times including a take-off angle of
sP phase. NORESS is a sub-array of NOA.
a b
Figure 7. The source depth for the 2017 test was estimated at 2.04 km using spectral nulls of pP +Pn and sP+Pn from
KSRS (a) and USRK (b) arrays. The slapdown phase appears at 0.57 Hz, but no spectral null for sP+P at USRK
spectra.
@seismicisolation
@seismicisolation
a b
Figure 8. Synthetic seismograms and spectra of the vertical component for regional and teleseismic data to account for
a) a spectral null (minimum) due to the destructive interference at 1.75 Hz showing pPn + Pn for the near-field (441
km) and b) at a spectral null at 1.25 Hz showing pP + P for the far-field (distance 81°) in case of a depth at 2.0 km
assuming the flat Earth model
We also found that spectral nulls of the spectra for synthetic seismograms are well consistent with
those of observations in Figure 9. The data are best fit by a depth near 2.0 km since the depth
determinations by spectral nulls of body waves are robust to uncertainties to other parameters4,26.
The earth structure models for synthetics used Korea Model29 for the local stations and AK135 for
the teleseismic arrays. The effect of anelastic attenuation using t* which is the ratio of travel time
and a specific quality factor Q is assumed to be independent of frequency. Several researchers18,26
found that the short-period P wave synthetic seismograms showing two troughs include the effect
of anelastic attenuation computed using t* whereas no effects of anelastic attenuation are included
in the long-period P wave synthetic seismograms showing a trough in in Figure 8. The synthetic
seismograms and spectra are produced using Hudson96 and Herrmann programs30 The spectral
nulls of P-wave spectra due to the destructive interference by the direct and reflected P waves from
the surface are reasonably consistent with those of our observations for the near-field and the far-
field.
DEPTH ESTIMATION USING SPECTRAL NULLS OF RAYLEIGH WAVES
exponential decay of Rayleigh wave displacements with depth. The frequency-dependent “spectral
null” in the Rayleigh wave amplitude spectra are most pronounced for vertical dip-slip and pure
dip-slip with 45° dip (reverse faulting) mechanisms. The spectral null of the fundamental-mode
Rayleigh wave spectrum is dependent on source mechanism, depth and source-receiver azimuth
but does not vary much with the shot medium and the shot yield8,10,14,31,32,33. The conical dip-slip
reverse faulting may accompany a deep-seated tensile failure occurring at a depth above shot
point9,34 generating the strong Rg wave radiation from a vertically oriented CLVD* source35. We
also estimated source depths for North Korean nuclear tests using spectral nulls of the
fundamental-mode amplitude spectra of Rayleigh waves. We found spectral nulls at 0.138 Hz
using Dongbei Seismic Network (DSN) through the pure continental-path data only for the 2006
test (Figure 9a). The spectral minima appear at 0.146 Hz for the 2017 from INCN and R720B to a
reverse faulting dipping at about 45° dip which is consistent with findings from other
researchers8,10,14,31,32,33 whereas no spectral nulls were found in the mechanisms for neither pure
@seismicisolation
@seismicisolation
strike-slip motions nor reverse faulting mechanisms with about 45° dip, possibly oblique-reverse
faulting mechanisms37,38. The source mechanism for the North Korean nuclear explosions may be
assumed to be the conical dip-slip volume accompanying a reverse faulting motion as a vertically-
distributed source in a shaft. We estimated a spectral null at 0.138 Hz for the 2006 nuclear test, but
no spectral nulls for the 2009 test from lack of Rg waves in the shorter distance less than 200 km,
including the azimuth change from the source to stations implying change of an azimuth of the
nodal line (strike) of mechanism. A spectral null (minima) is caused by an excitation null for short-
period fundamental-mode Rayleigh waves, termed Rg waves as a CLVD (compensated linear
vector dipole) source which correlates with normal mode theory in the form of resonant
frequency9,14,32,33. Bonner et al.3 have examined the performance of MS scales on 7-sec Rayleigh
waves recorded distances less than 500 km from Nevada Test Site. The spectral null (minimum)
from the CLVD depth38 (9DYU\þXN and Kim, 2014) correlates with source depths derived from pP-
P and pPn-Pn delay times. The conical dip-slip reverse faulting may accompany a deep-seated
tensile failure occurring at a depth above shot point9,34 generating the strong Rg wave radiation
from a vertically oriented CLVD source37,39 which is a kind of a spallation-like source function,
but it is not associated with the surface spall because it was conducted at granite in the deep
tunnels. The estimated depth via the CLVD model also disambiguates the estimated source depth
by body waves (Figure 9).
The conical dip-slip faulting may accompany a deep-seated tensile failure occurring at a depth
above the shot point as a CLVD source generating Rg waves. Rg-waves resulting in the distinctive
interference by P an SV waves reflected from the surface is induced from the tensile failure from a
vertically oriented CLVD source showing spectral nulls of Rg wave spectra. There are more
elliptical motions of Rg and the ‘retrograde’ reverses ‘prograde’ at a depth of d = 0.192
wavelength which represents a spectral null (hole) at the spectra. Therefore, a source depth is
estimated at 2.09 km (d = 0.192 * 7.4627 s * 1.46 km/sec = 2.03 km) using spectral nulls at 0.138
Hz from DSN for the 2006 test. We found 1.97 km (d = 0.192 *1/0.146* 1.46 km/sec = 1.97 km)
using 0.146 Hz from INCN and R720B data for the 2007 test (Figure 10b). Rg-wave velocity of
1.46 km/sec is obtained from Rayleigh wave equation assuming a source rock to be a Poisson
solid39. This source mechanism of CLVD source could be assumed to be a pure dip-slip (reverse-
faulting), and the absence of spectral nulls or slight change may be due to an azimuth change and
site effects from the different raypath through the complicated crustal structures in the very shorter
distance.
______________
*There are source changes in volume (isotropic), shear fracture (double-couple) and sudden changes in rigidity (shear
modulus) in moment tensor inversion. The sudden change in shear modulus is called a CLVD source
@seismicisolation
@seismicisolation
@seismicisolation
@seismicisolation
Figure 9. The fundamental-mode Rayleigh waves and displacement spectra for the BHZ band-pass filtered (0.02-0.1
Hz) with amplitude in counts. a) A spectral null at 0.134 Hz was found at Dongbei Seismic Network (DSN) data only
for the 2006 but no spectral null for the 2009 test. A spectral null of 0.14 Hz at USRK was observed for the 2016S
test. b) Spectral nulls (minima) at 0.146 Hz were estimated at INCN and R720B for the North Korean nuclear test on
September 3, 2017 using Rg-wave spectra from IRIS station at Incheon and Raspberry Shake Seismograph station
which is located on the hill of Mt Gobong in the northwestern part of South Korea.
The synthetic seismograms were produced based on Green function and the modified AK135
Model taking into account the average depth of 2 km calculated from Rg-wave spectra and a
general nuclear explosion mechanism of a reverse faulting with 45°dip, rake = 90° and, dip = 45°
with a depth = 2 km, and strike varies with every 15° using a narrow band-pass filter between 0.02
and 0.1 Hz (Figure 10) increasing spectral nulls with azimuth (from source to station). We
calculated the most-fitting depth 2 km at 0.14-0.15 Hz from an azimuth of 30° which corresponds
to the tectonic stress (ENE) in this region38.
@seismicisolation
@seismicisolation
Figure 10. The synthetic seismograms and spectra of Rg-waves are band-pass filtered between 0.02-0.1 Hz to obtain
the fundamental-mode based on the Korea model considering a general mechanism for an explosion with dip 45° and
rake 90° varying with every 15° of strike to 90°. The higher amplitude at an azimuth of 90° is attributed to the large
radiation pattern of Rayleigh wave and it decreases again up to 180°. We found the most fitting depth (2 km) at 0.14-
0.15 Hz from an azimuth of 30°which corresponds to the tectonic stress (ENE) in this region38.
COMMON DEPTH POINT STACKS (CDPS) FOR REGIONAL AND TELESEISMIC

ARRAYS
The nulls correlate with the delay times between the onset of the upswing P-wave arrival and that
of the downswing of depth phases (pP, sP, pPn and sPn) as shown in Figures 11a and 11b which
are produced by SUMSTACK via the SAC program. We clearly identified depth phases (pP, sP,
pPn and sPn) showing the reverse polarity of the depth phases in the frequency domain as well as
in the time domain. The depth phases of pP, sP, pPn and sPn are clearly presented in the
waveforms produced by superposition with respective channels of the array which are 180 º out of
phase relatively to the first P- wave arrivals9,39.
. The delay times for depth phases present the difference between the onsets of P-wave arrivals and
those of the downswing depth phases of pP (pPn, sPn) in the time domain. We can show the
common depth point stack using the trace gather for the moveout corrections of the multiple
coverage for the near-field and teleseismic arrays. In the frequency domain (Figures 11 and 12),
the delay time for the depth phase produces a sharp scalloping (hole) of the amplitude spectrum,
yielding minima at periods Tn=T/n for n=1, 2, … where T is the delay time for pP-P or sP-P (pPn-
Pn, sPn-Pn)4,18,41. The onsets of pPn-wave arrivals appear on the zero lines in KSRS arrays
(Figures 11a and 12a) whereas those of pP-wave arrivals appear at the peak in the teleseismic
arrays (Figures 11b and 12b) because of the overshoot of the teleseismic P-wave propagation path
whereas there are no overshoots at CDPS on the regional seismograms of KSRS.
a b
@seismicisolation
@seismicisolation
Figure 11. a).The common depth point stacks were produced through the trace gather for the moveout corrections of
the multiple coverage of the KSRS array (19 channels) for the 2006, 2009, 2013, 2016J (01/06/2016) and 2016S
(09/09/2016) nuclear tests of North Korea showing Pn, pPn and sPn phases including an unknown phase (?) shortly
after pPn for the 2006 test. b).The common depth point stacks were produced through the trace gather for the moveout
corrections of the multiple coverage using teleseismic arrays. The onsets of P-wave arrival and depth phase (pP) are
clearly shown from the normalized superposition of the oncoming teleseismic body waves into ASAR, NVAR, PDAR,
WRA2016J (01/06/2016), WRA2016S (09/09/2016) and YKA arrays for the 2016 tests.
@seismicisolation
@seismicisolation
a b
Fig. 12. a) The arrows represent spectral nulls at 1.75 Hz for the 2009, 2013, 2016J (01/06/2016) and 2016S
(09/09/2016) for pPn+Pn, but at 1.62 Hz for the 2006 nuclear test at KSRS. The spectral nulls for sPn+Pn are found to
be identically 1.12 Hz for the 2006, 2009, 2013, 2016J and 2016S nuclear tests using the KSRS records. b) The
spectral nulls using pP+P were identically estimated at 1.25 Hz for the 2016 nuclear test using NVAR, PDAR, WRA-J
(01/06/2016), WRA-S (09/09/2016) and YKA arrays except for ASAR whose spectral null was estimated at 1.1 Hz
due to the low velocity layer beneath the stations.
Our depth estimations were quite different from other studies by42,43 which concluded depth
estimations of 200 m (2006 test)/550 m (2009 test) by42 and 610 m (2009 test)/430 m (2013 test)
by 43. Murphy et al.44 used the spectral ratios of two explosions at common regional stations to be
compared with the theoretical spectral ratios (Murphy et al.44) assuming that both explosions were
detonated either at a depth of 200 m or 800 m. They then compared these ratios with the
theoretical Mueller–Murphy source spectral model and estimated the yield for the 2006 test as 0.9
kt if detonated at a depth of 200 m, and 4.6 kt for the 2009 test if detonated at a depth of 550 m. In
their study, observed spectral ratio data were more consistent with the hypotheses that the 2006
test was conducted at a depth of 200 m and the 2009 test was conducted at a depth of 550 m.
Murphy et al.43 determined the burial depths of the 2009 and 2013 tests to be about 610 m and 430
m, respectively, based on elevation differences between the first 2006 test location used as a
reference point and the respective source locations of the other tests via data gathered through
satellite images. Furthermore, the source locations by IDC and NEIC are quite different from those
by the relative method. Rouguier et al.45 also estimated the minimum depth of burial and yield for
the 2009 North Korean test at 375 m and 5.7 kt, respectively based on constraints from
hydrodynamic simulation.
scaled depth of burial should not be less than 122Y1/3 where depth is measured in meters and yield
in kilotons (kt)45. Therefore, a depth of about 2000 m would allow containment of an underground
@seismicisolation
@seismicisolation
nuclear explosive event having a yield of up 4.4 megatons. However, this scaled depth concept
does not fit the North Korean nuclear tests in the light of the observed magnitude. As a result, it
seems sure that the over- burial depths of the North Korean nuclear tests were made beyond the
economic cost and the ordinary tunneling technology taking into account the geology and depth
(Figure 13a).
a b
Figure 13. a) Geological map of the nuclear test site and surrounding area (reproduced from the Geological Society of
Korea, 1987). 1. The Mesozoic intrusive granite rock in Kwanmo Massif, 2. Quaternary volcanic rocks in Kwanmo
Massif, 3. Kilju-Myongchon Graben (Cenozoic sedimentary rocks), 4. Machollyong supergroup (Proterozoic rock), 5.
Rangnim Massif (Proterozoic-Archaeozoic), 6. Triassic intrusives, 7. Hayang group (Cretaceous), 8. Daedong
supergroup (Jurassic-Triassic), 9. Sangwon system (Proterozoic-Archeozoic). The star indicates the site of the nuclear
detonation. b) Satellite image of Mt. Mantap (2205 m) and surrounding area near the test site was produced by French
Satellite ‘Spot-5’ on May 14, 2009 before the second test. The scale is 1:25,000. The star indicates the site of the
nuclear detonation.
Three white spots at the bottom are the tunnel entrances AKA West, South and East portals46,47,
but they were built before the 2009 and 2013 nuclear tests implying pre-existed portals (Figure
13b). Very recently48 described rather a large piled spoil in front of west and north portals
implying that a large nuclear explosion is impending. However, the piled spoil from the satellite
image is nothing to do with the actual source locations as well as the seismic yield because the
exact source location is undisclosed.
@seismicisolation
@seismicisolation
Figure 14. The map is reproduced from the original satellite digital data (2014 KOMPSAT-2) of National Geographic
Information Institute of Korea showing the contour map surrounding nuclear test sites. According to IDC locations
which are more reliable in this study, especially 2017REB (Revised Event Bulletin) is located at 41.3296º N and
129.0267º E by IDC/CTBTO issued on 5 September 2017. The elevation difference between a source and a reflector is
not significantly large because the slope angles surrounding sites are not fast, but slow in the range of 7-8º indicating
that the nonlinear effects of topography may be negligible.
@seismicisolation
@seismicisolation
Figure 15. Topography map using 2D and 3D around the test site area is constructed using the numerical geology
configuration data (1:25000) from National Geographic Information Institute. Closed stars. open stars and a triangle
represent epicenters with event year determined by IDC (CTBTO), NEIC (USGS) and Mantapsan. The 3D and 2D
topography maps around the North Korean nuclear tests show shallow slope (almost flat) near the test sites, especially
epicenters determined by IDC/CTBTO (Figures 14 and 15).
@seismicisolation
@seismicisolation
CONCLUSIONS
The source depths for the 2006, 2009, 2013, 2016J, 2016S and 2017 nuclear tests were estimated
at 2.12 km, 2.06 km, 2.05 km, 2.06 km, 2.05 km and 1.97 km, respectively using body and
Rayleigh wave spectra. We did not use the relative method based on ostensible satellite data and
unclear scaled depth in relation to depth estimate. There are certain investigators who have some
strong controversies concerning the source depth against our over-burial depth estimate.
Nevertheless, our determination is the best method as far as given data and seismological methods
are concerned. So it should be clearly noted in this study that DPRK conducted nuclear tests at the
over-burials at around 2 km beneath the vicinity of Mantapsan (Mt. Mantap) (2,205 m) neglecting
the nonlinear effects of rugged topography taking into account the epicenter locations with shallow
slopes (Figures 14 and 15) and surface spallation owing to the granite source rock. In particular the
2D (Figure 14) and 3D (Figure 15) topography maps around the North Korean nuclear tests show
shallow slope (almost flat) near the test sites, especially epicenters for 2006, 2016J, 2016S and
2017 tests determined by IDC/CTBTO which seem to be located at high elevation with shallow
slope whereas the elevation for 2009 and 2013 tests may be lower than that of around the area
resulting in a slightly deeper depth (100-200 m) than surrounding high elevation. As a result, the
non-linear topographic effects in use of depth phase may be de minimus in this best estimated
study assuming that the elevation of reflector point is almost the same as that of the source
resulting in almost flat Earth near the test sites. However, we cannot exclude some errors to
estimate depth by depth phase because of the non-linear topographic effects which may be at most
a few hundred meters. Nonetheless, the spectral nulls of Rayleigh wave spectra using many
observations were very useful to compensate for the poor depth phase estimate. It was also
inferred from spectral minima of Rayleigh waves that most of the nuclear tests may be related to a
CLVD source13 which can be generated from a vertically- distributed source49 in the shaft beneath
mutually connected-underground tunnels. Besides it is worthwhile to note that we also observed
SH and Love waves14 induced from tectonic stress and/or new cracks50,51,52 in addition to Rayleigh
waves.
ACKNOWLEDGMENTS
criticism. We would like to thank Paul Richards for his providing DSN (Dongbei Seismic
Network) data and Rick Benson (IRIS) and David Jepson (Geoscience Australia) for the
teleseismic data and Korea Meteorological Administration and National Geographic Information
Institute of Korea for providing local waveform data and satellite digital map, respectively. We
also appreciate the Executive Secretary of CTBTO, Lassina Zerbo and Director of International
Data Centre (IDC), CTBTO for their assistance in retrieving IDC data. One of author (SGK)
would greatly appreciate CTBTO’s funding to participate in SnT2015, SnT2017 and SnT2019,
CTBTO, Vienna, Austria including permission to use IMS data.
REFERENCES
1. Bakun, B. H. and Johnson, L. R., Geophys. J. R. astr. Soc., 1973, 34, 321-342.
2. Bowers, D, Marshall, P.D. and Douglas, A., Geophys. J. Int., 2001, 46, 425-438.
3. Bonner, J. L., Harkrider, D. G. Herrin, E. T., Shumway, R. H., Russell, R. A. and Tibuleac,
I. M. Bull. Seism. Soc. Am..2003, 93 (4), 1773-1791.
4. Stein, S. and Wiens, D., Reviews Geophys., 1986, 24 (4), 806-832.
5. Harjes, H-P, Henger, M., Muller, G. W. and Wilhelm, H., Conference on Disarmament,
&' GE. 85-63293, 26 July 1985 Federal Republic of Germany 68.
6. Murphy, J. R., In Identification Seismic Sources, ed. Huseby E. S. and Mykkeltveit, S D.,
Reidel Publishing Company, 1981, 201- 205. B., PL- TR-96-2017 Phillips Laboratory
Hanscom AFB MA 66.
7. Sykes, L. R., Dealing with Decoupled Nuclear Explosions under a Compresensive Test Ban
@seismicisolation
@seismicisolation
Treaty, 247-293, in Monitoring a CTBT, NATO Science Series, 1996, 25, ed. by E. Y.
Husebye and A. M. Dainty.
8. Douglas, A., Corbishley, D. J., Blamey, C. and Marshall, P. D., Nature, 1972, 237, 26-28.
9. Massé, R. P., Bull. Seism. Soc. Am., 1981 71 (4), 1249-1268.
10. Okal, E. A., Seism. Res. Lett., 1992, 63 (2),169-180.
11. Cohen, T. J., Geophys. J. R. astr. Soc., 1970, 20, 223-231.
12. Kulhanek, O., pageoph, 1971, 88, 121-136.
13. Husebye, E. S., Matveeva, T. and Fedorenko, Y., Bull. Seism. Soc. Am., 2013, 103 (3), 1771-
1783. 14. Heyburn, R., Neil, D. and Fox, D. B., Geophys.J. Int., 2013, 94 (2), 1000-1010.
15. Gitterman, Y. Kim, S. G. and Hofstetter, R., Pure Applied Geophysics. ,2015
DOI:10.1007/s00024- 015-1169-8.
16. Douglas, A., Marshall,P. D. and Young. J. B., Geophys. J. R. astr. Soc.,1987. 90. 101- 117.
17. Douglas, A., Richardson, L. and Hutchins, M., Geophys. J. Int., 1990, 100, 303-314.
18. Burdick, L. J., Wallace, T. and Lay, T., J. Geophys. Res., 1984, 89 (B6), 4373-4388.
19. Bonner, J. L., Herrmann, R. B., Harkrider, D. and Pasyanos, M., Bull. Seism. Soc. Am., 2008,
98 (5). 2498- 2506.
20. Lay, T., Geophysical Monograph 65, 1991,109-125.
21. Chung, T. W., J. Kor. Earth Science Soc.,16 (2),1995,152-157.
22. Kim, S. G. and Bae, H., Kr. Soc. Econ. Environ.Geol., 39 (2), 2006, 111-128.
23. Phillips, W. S. Begnaud, M. L., Rowe, C. A. Steck, L. K, Myers, S. C., Pasyanos, M. E. and
Balla, S., Geophys. Res. Lett., 35 (14), 2007. DOI: .1029/2008GL034211.
24. Bonner, J. L., Reiter, D. L. and Shumway, R. H., Bull. Seism. Soc. Am., 92 (2), 2002, 1675-
1693.
25. Van Der Hilst, R. D., Kennett, B. L. N. and Shibutani, T., Geodynamics, 26. 1998, 39-57.
26. Douglas, A. and Rivers, D. W., Bull. Seism. Soc. Am.,78 (2), 1988, 1011-1019
27. Douglas, A. and Hudson, J. A., Geophys. J. Int., 103, 1990, 111-133.
28. Bolt, B. A., Nuclear Explosions and Earthquakes. The Parted Veil. W. H. Freeman and
Company. 1976, San Francisco CA.
29. Hartmann, G., Barth, A., J. O. Ross, Grunberg and Frei, M. (2017). Verification of the North
Korean nuclear explosions 2006, 2009, 2013, and 2016.137-165– Pilger, Ceranna and
Boennemann (eds): Monitoring Compliance with the Comprehensive Nuclear-Test-Ban
Treaty (CTBT), Federal Institute of Geosciences and Natural Resources (BGR), Hannover,
Germany, 328pp
30. Kennett, B. L. N., Engdahl, E. R. and Buland, R., Geophys. J. Int., 221, 1995, 108-124.
31. Herrmann, R. B. Computer Programs in Seismology. Current Version 3.30 is dated May 8,
2016 (NP330.May-08-2016.tgz), Saint Louis University, Saint Louis, MO. 186.
32. Tsai, Y-B and Aki, K., J. Geophys. Res.,1971. 76 (17), 3940-3952.
33. Fox, D. B., Selby, D. N., Heyburn, R. and Woodhouse, J. H., Geophys. J. Int., 2012, 191,
601-615.
34. Langston, C. A., Bull. Seism. Soc. Am., 1980,70 (4).1409-1414.
35. Patton, H. J. and Taylor, S. R., Bull. Seism. Soc. Am., 1995, 85, 220-226.
36. Rodgers, A. J., Petersson, N. A. and Sjogreen, B., J. Geophys. Res., 2010, 115(B11309), 1-
27. doi:10.1029/2010JB007707.
37. Kim. S. G., Gitterman, Y., Lee, S. and Vavrycuk, V., Final Paper Number: S34A-02, AGU
Fall Meeting, San Francisco, December 12-16, 2016.
38. 9DYU\þXN V, and Kim, S. G., Geophys. Res. Lett., 2014, 41 (20), 7048-7056. DOI:
10.1002/2014GL06126.
39. Cho, C., Shin, J. S., and Kim, G., S31A-2722, 2016AGU Fall Meeting, San Francisco,
December 10-12, 2016
40. Udias, A. Principles of Seismology, 1999, Cambridge University Press, Cambridge, UK.
41. Montagner, J. P. and Kennett, B. L. N., Geophys. J. Int.,1996, 125, 229-248.
42. Stead, R. J. and Helmberger, D. V., PAGEOPH, 1998, 128 (1/2), 157-193.
@seismicisolation
@seismicisolation
43. Zhang, M. and Wen, L., Geophys. Res. Lett., 2013, 40, 2941-2946.
44. Murphy, J. R., Stevens, L., Kohl, B. C. and Bennett, T. J., Bull. Seism. Soc. Am., 2013, 103
(3),1640-1661.
45. Rougier, E., Patton, H. W., Knight, E. E. and Bradley, C. R., Geophys. Res. Lett., 2011,38
(16). DOI:10.1029/2011GL048269.
46. Glasstone, S. and Dolan, P. J., (compiled & edited) The Effects of Nuclear Weapons, 1977,
United States Department of Defense and the Energy Research and Development
Administration, 657.
47. Pabian, F. and Hecker, S., Bull. Atomic Sci.Aug.,
http://www.thebulletin.org/web-edition/features/contemplating-third-nuclear-test-north-
korea.
48. Israelsson, H., Technical Note 2016-01, 2016, Seismic-Infra Research, Washington DC, 17.
49. Pabian, F. and Coblentz, D. http://38north.org/2017/03/punggye031017/.
50. Viecelli, J. A., J. Geophys. Res., 1973, 78 (17), 3334-3339.
51. Nuttli, O. W., Bull. Seism. Soc. Am., 1969, 59, 385-398.
52. Viecelli, J. A., J. Geophys. Res., 1973, 78 (17), 3334-3339.
53. Kim, S. G., Gitterman, Y., Lee, S. and Vavrycuk, V., SnT2017 (CTBTO) Vienna, Austria,
26-30, June 2017.
@seismicisolation
@seismicisolation

Forensic Explosion Seismology Technologies and Applications Kim

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Forensic Explosion Seismology Technologies and Applications Kim

Uploaded by

Copyright:

Available Formats

Forensic Explosion Seismology

Technologies and Applications

So Gu Kim and Yefim Gitterman

Edited by So Gu Kim and Yefim Gitterman

This book first published 2020

Cambridge Scholars Publishing

Lady Stephenson Library, Newcastle upon Tyne, NE6 2PA, UK

British Library Cataloguing in Publication Data

Copyright © 2020 by So Gu Kim, Yefim Gitterman and contributors

ISBN (10): 1-5275-4474-5

Prologue .................................................................................................................................. viii

Chapter One ................................................................................................................................1

Chapter Two ..............................................................................................................................24

Chapter Three ............................................................................................................................56

Chapter Four .............................................................................................................................78

Chapter Five ............................................................................................................................108

Chapter Six ..............................................................................................................................120

Chapter Seven .........................................................................................................................135

Chapter Eight ..........................................................................................................................155

Chapter Nine ...........................................................................................................................174

Chapter Ten .............................................................................................................................207

Chapter Eleven ........................................................................................................................230

Chapter Twelve .......................................................................................................................252

Chapter Thirteen .....................................................................................................................270

Chapter Fourteen .....................................................................................................................285

Chapter Fifteen ........................................................................................................................326

Chapter Sixteen .......................................................................................................................342

Chapter Eighteen .....................................................................................................................377

Chapter Twenty .......................................................................................................................406

Chapter Twenty-One ...............................................................................................................434

Chapter Twenty-Two ..............................................................................................................461

Chapter Twenty-Three ............................................................................................................499

SURFACE-WAVE MAGNITUDES OF EURASIAN

OTTO W. NUTTLI, SO GU KIM AND HUI-YUIN WEN

Table 1. Dates, times and locations of soviet explosions studied

Table 2. Magnitudes and locations of events studied

a An “E” after the date indicates a presumed underground explosion.

Figure 2. Seismograms showing 20-sec period waves at small distances.

DISCRIMINATION OF ANOMALOUS EARTHQUAKES

OTTO W. NUTTLI, SO GU KIM, HUI-YUIN WEN AND JOHN A. WAGNER

PUBLICATIONS RESULTING FROM RESEARCH PERFORMED UNDER CONTRACT

Table 1. Hypocentral coordinates of events studied

Table 2. Body-wave (mb) and surface-wave (ms) magnitudes of events studied

Event No.a ms(NEIC)b ms(ISC)b mbb,c MS(NEIC)b MS,RZb,c MS,Lb,c

Table 3. Earthquakes for which mb (isc)-ms 1.2

Date Origin Time Lat. ( °N) Long. (°E) mb୉MS Location

Table 4. Hypocentral coordinates of foreshocks, mainshocks and aftershocks studied

Region Date Origin Time Lat.(°N) Long.(°E) Depth(km) mb MS

Table 1 Hypocentral coordinates and magnitudes of earthquakes studied

Szechwan 1 16 Aug 71 04-58-00.3 28.9 103.7 19 5.7 4.8

2 16 Aug 71 18-53-54.7 28.9 103.6 17 5.4 5.4

3 16 Aug 71 22-37-33.6 28.8 103.6 13 5.3 4.7

Iraq-Iran 1 12 Jun 72 13-34-00.7 33.1 46.3 16 5.3 4.9

2 13 Jun 72 00-55-37.3 33.1 46.3 12 5.1 4.5

3 14 Jun 72 04-34-28.1 33.0 46.1 15 5.3 4.1

Tibet 1 14 Jul 73 04-51-21.0 35.2 86.5 19 6.0 6.6

2 14 Jul 73 13-39-30.0 35.3 86.6 22 6.0 5.4

2 23 Jan 72 21-17-52.1 -13.2 166.4 28 5.9 6.8

3 24 Jan 72 03-55-42.5 -13.0 166.4 26 5.9 5.9

Table 2 Comparison of ೗ms and log ೗a·values

Region Earthquake Pair ᶭMS log ᶭA

No. 1 - No. 3 0.1 0.49 (n = 6)

Table 3. Earthquakes for which mb (isc)-ms 1.2