Content uploaded by Xiaoning Yang
Author content
All content in this area was uploaded by Xiaoning Yang on Mar 27, 2015
Content may be subject to copyright.
Identi®cation of Mining Blasts at Mid- to Far-regional Distances
Using Low Frequency Seismic Signals
MICHAEL A. H. HEDLIN,
1
BRIAN W. STUMP,
2,3
D. CRAIG PEARSON,
3
and XIAONING YANG
3
Abstract Ð This paper reports results from two recent monitoring experiments in Wyoming.
Broadband seismic recordings of kiloton class delay-®red cast blasts and instantaneous calibration shots
in the Black Thunder coal mine were made at four azimuths at ranges from 1°to 2°. The primary focus of
this experiment was to observe and to explain low-frequency signals that can be seen at all azimuths and
should routinely propagate above noise to mid-regional distances where most events will be recorded by
International Monitoring System (IMS) stations.
The recordings clearly demonstrate that large millisecond delay-®red cast blasts routinely produce seismic
signals that have signi®cant spectral modulations below 10 Hz. These modulations are independent of time,
the azimuth from the source and the orientation of the sensor. Low-frequency modulations below 5 Hz are
seen beyond 9°. The modulations are not due to resonance as they are not produced by the calibration shots.
Linear elastic modeling of the blasts that is guided by mine-blast reports fails to reproduce the ®ne detail of
these modulations but clearly indicates that the enhanced ``spectral roughness'' is due to long interrow
delays and source ®niteness. The mismatch between the data and the synthetics is likely due to source
processes, such as nonlinear interactions between shots, that are poorly understood and to other eects,
such as variations of shot time and yield from planned values, that are known to be omnipresent but cannot
be described accurately. A variant of the Automated Time-Frequency Discriminant (HEDLIN, 1998b), which
uses low-frequency spectral modulations, eectively separates these events from the calibration shots.
The experiment also provided evidence that kiloton class cast blasts consistently yield energetic 2±10
second surface waves. The surface waves are strongly dependent on azimuth but are seen beyond 9°.
Physical modeling of these events indicates that the surface waves are due mainly to the extended source
duration and to a lesser extent to the slap-down of spalled material. The directionality is largely a path eect.
A discriminant that is based on the partitioning of energy between surface and body waves routinely
separates these events from the calibration shots.
The Powder River Basin has essentially no natural seismic activity. How these mining events compare to
earthquake observations remains to be determined.
Key words: Spectral modulations, attenuation, surface waves, remote source characterization.
1. Introduction
The recent Comprehensive Nuclear-Test-Ban Treaty (CTBT) is unlike any earlier
test ban accord, such as the Threshold Test-Ban Treaty (TTBT) which prohibits tests
1
IGPP-University of California, San Diego, La Jolla, CA, 92093-0225, U.S.A.
2
Southern Methodist University, U.S.A.
3
Los Alamos National Laboratory, U.S.A.
Pure appl. geophys. 159 (2002) 831±863
0033 ± 4553/02/040831±33 $ 1.50 + 0.20/0
ÓBirkha
Èuser Verlag, Basel, 2002
Pure and Applied Geophysics
above 150 kilotons, as it bans nuclear explosions of any yield. The exclusive nature of
this treaty and the fact that a 1 kiloton contained explosion in hard rock is well above
magnitude 4 motivates interest in the accurate characterization of small (m
b
< 4.0)
seismic events (MURPHY, 1995). Detonating a nuclear explosion in a cavity can
further reduce the magnitude of the seismic waves (U.S. CONGRESS, OTA REPORT,
1988). Interest in small events has increased both the numbers of events that must be
considered and the types. It is estimated that, globally, 21,000 events with m
b
above
3.5 occur per year (U.S. CONGRESS OTA REPORT; p. 78). Some of these small events
will not be associated with natural seismic activity but are due to commercial blasting
which occurs globally. The blasting technique favored worldwide is delay ®ring
(LANGEFORS and KIHLSTRO
ÈM, 1978) in which a number of charges are arranged in a
spatial grid and detonated in sequence. This technique is favored as it yields ecient
fracturing of rock while minimizing damaging seismic and acoustic signals in areas
proximal to the mine. Commercial blasting is common but usage varies widely
(LEITH, 1994). KHALTURIN et al. (1997) surveyed over 30 regions worldwide and
found that several hundred industrial blasts each year have a magnitude greater than
3.5. Large blasts, often associated with construction, were relatively common in the
Former Soviet Union and in China however the current blasting practice is not yet
well known (Bill Leith, personal communication). In Wyoming, very large (>1 kt)
surface coal mine blasts are common. The Black Thunder coal mine, one of several
mines in the Powder River Basin in NE Wyoming, typically detonates 1 to 2 blasts of
this magnitude each month (PEARSON et al., 1995; STUMP, 1995) with a few in the
magnitude range of 3.5 to above 4.0.
The Reviewed Event Bulletin (REB), published by the prototype International
Data Centre (PIDC), indicates that large mining explosions are commonly detected
by IMS seismic stations at all regional distances; some are seen teleseismically.
Although there has been very promising progress in identifying mine blasts using
correlation techniques (HARRIS, 1991; ISRAELSON, 1991; RIVIERE-BARBIER and
GRANT, 1993), signi®cant changes in how mine blasts are detonated at an individual
mine are known to occur (MARTIN et al., 1997). This gives us the impetus for
exploring other, complimentary, methods. Large mining events are problematic not
just because these events will trigger the IMS, but these large, controlled, explosions
oer a means to obscure nuclear tests. The ``hide-in-quarry blast'' evasion scenario
(BARKER and DAY, 1990; BARKER et al., 1994; RICHARDS and ZAVALES, 1990; SMITH,
1993), in which a nuclear test is colocated with a mining blast, might be troubling
because blasting anomalies, in which a large part of a mining explosion shot grid
detonates simultaneously, are not uncommon (MARTIN et al., 1997). A nuclear test
colocated with an industrial explosion might be entirely hidden or mistaken for a
detonation anomaly.
The CTBT calls for an International Monitoring System (IMS) which will
comprise four networks of sensors (seismic, infrasound, radionuclide and hydroa-
coustic). The seismic network will consist of 50 primary stations and over 100
832 Michael A. H. Hedlin et al. Pure appl. geophys.,
secondary stations (Fig. 1). This network will place a station within local distance
(1°) of 2% of the Earth's landmass. The near-regional (within 5°) coverage will be
34%. The mid-regional coverage (within 15°of the source) is nearly complete at 89%
(Fig. 2). If multiple recordings are required for accurate source characterization, the
coverage is more limited. Just 22% of the Earth's landmass is within 10°of 3 stations;
74% is within 20°. Some prominent mining regions (e.g., the Kuzbass/Abakan
mining region in Russia near the IMS station ZAL) will be monitored at near-
regional range (Fig. 2) and a full suite of high- and low-frequency characterization
techniques can be brought to bear on suspicious events (STUMP et al., 1999a). In
many regions, however, monitoring of man-made and natural seismic activity will
rely on recordings made at mid- to far-regional range. Events in these regions will
require a more limited set of seismic discriminants: those that operate at low
frequencies. Although millisecond delay-®red industrial explosions will produce
diagnostic spectral peaks at high frequency, because of interference between shots,
the industry standard intershot time delay is 35 msec and the resulting spectral peak
is at 30 Hz (the inverse of the intershot time delay). As the standard sampling rate
for IMS stations is 40 samples/sec (sps), this energy will lie beyond the Nyquist
frequency of most recordings that will be made under the CTBT.
Quantitative detection capability of the IMS involves many factors, including
spatial decay rates, event size, noise levels, instrument type (e.g., array, single
station), and phase (e.g., Pand S). There have been many studies quantifying these
into the detection capability of the IMS (e.g., WUSTER et al., 2000). The analysis in
Figure 1
Distribution of current and proposed primary (red) and auxiliary (yellow) IMS seismic stations. Array sites
and three-component stations are represented as circles and triangles, respectively. Additional stations (in
black) are in the IRIS GSN. A regional network considered in this study was deployed in Wyoming and is
represented by the oval.
Vol. 159, 2002 Identi®cation of Mining Blasts 833
this case is to emphasize the distance range at which observations of mining
explosions are most likely to be made.
It is apparent that large blasts emit seismic energy which can be used for source
characterization from local to far-regional distances. Spectral modulations below
10 Hz have been observed by several groups (incl., BAUMGARDT and ZIEGLER, 1988)
and are usually attributed to long intershot, or interrow, delays. GITTERMAN and VAN
ECK (1993) attributed a low frequency spectral notch in recordings of mine blasts to
source ®niteness. Unusual time-domain characteristics also have been observed.
ANANDAKRISHNAN et al. (1997) and STUMP and PEARSON (1997) observed signi®cant
surface waves caused by cast explosions in Wyoming. GITTERMAN et al. (1997)
observed surface waves in recordings of quarry blasts in Israel. ANANDAKRISHNAN
Figure 2
Coverage of the Earth's landmass permitted by the current and proposed IMS primary and secondary
seismic networks is indicated by the solid curves. Single-station coverage (N1) is nearly complete within
10°of the source. If multiple recordings of an event are required for adequate source characterization
(N> 1), most of the observations will be made at mid- to far-regional or teleseismic range. Coverage of
three prominent mining regions is indicated by shaded symbols. Krasnogorsky, the main Kuzbass mine in
Russia, at 53.6°N, 87.8°E; Black Thunder in Wyoming at 43.7°N, 105.25°W and Kursk, Russia at 51.8°N,
36.5°E are represented by diamonds, triangles and stars, respectively. These mining regions will be
monitored somewhat more closely than the average. For example, three stations are within 8°of the
Kuzbass mine. Just 8% of the Earth's landmass has better coverage. The dashed curve is coverage given by
the IMS networks (current and proposed) and existing stations in the GSN. Many stations in the GSN that
are not already in the IMS are located on oceanic islands and so the improvement is limited.
834 Michael A. H. Hedlin et al. Pure appl. geophys.,
et al. (1997) modeled the Wyoming blasts and concluded that the surface waves are
due to long source duration and signi®cant spall, including material cast into the pit.
This paper presents further evidence for signi®cant seismic signals produced by
large mining blasts and not by instantaneous explosions. The paper presents evidence
for spectral modulations below 10 Hz produced by documented (``ground truthed'')
cast blasts in Wyoming. Low-frequency modulations, below 5 Hz, are seen at an
IMS station at a range of 9°. The paper identi®es observations of signi®cant surface
waves also produced just by these events. Source modeling is used to understand
these observations and to gauge the sensitivity of these signals to changes in blasting
parameters at the source. We assess the usefulness of these signals, and low-frequency
spectral modulations, for characterizing mining explosions using the IMS.
2. Regional Monitoring Experiments in Wyoming
To investigate low-frequency seismic signals produced by large mining blasts,
researchers from the University of California, San Diego (UCSD), in collaboration
with researchers from the Los Alamos National Laboratory (LANL), Southern
Methodist University (SMU) and the Air Force Technical Applications Center
(AFTAC), conducted two regional monitoring experiments in Wyoming in 1996 and
1997. Five broadband three-component seismic stations (STS-2's; ¯at response from
0.0083 to 40 Hz) were deployed within near-regional range of the Black Thunder mine
(Fig. 3). Four of the sensors were deployed in a ring at 200-km range to study the
azimuthal dependence of the seismic signals. The ®fth station was deployed (in 1996)
at a range of 100 km to the north of the mine along the azimuth to station CUST to
allow examination of the range dependence. A three-element, 100-m aperture
infrasound array was deployed at station MNTA (Fig. 3). The temporary deploy-
ments complemented permanent, nearby, deployments at PDAR, RSSD and more
distant stations in the IMS and the IRIS Global Seismographic Network (GSN).
The seismic activity in this region is almost entirely man-made and is
concentrated in the Powder River Basin coal mining trend (Fig. 3). Two types of
blasting are most common in this trend. The largest blasts are used to cast
overburden to the side to expose the coal seam (MARTIN and KING, 1995). Smaller
shots are used to fracture the coal seam to facilitate recovery (ATLAS POWDER
COMPANY, 1997). Event clusters correlate well with known mine locations (Fig. 3).
The events of greatest interest were the large cast blasts and calibration shots, located
in the Black Thunder coal mine (PEARSON et al., 1995; STUMP, 1995), as these were
closely monitored with video and acoustic equipment by the LANL team and were
carefully documented by personnel at the mine. During the two experiments, four
large cast blasts were detonated at Black Thunder (Fig. 4; Table 1). These blasts
ranged from 2.5 million pounds (Aug. 1, 1996) to 7 million pounds (Aug. 14, 1997)
and were used to move overburden. Three of four blasts occurred in the south pit of
Vol. 159, 2002 Identi®cation of Mining Blasts 835
Black Thunder where overburden is cast to the north. Typical cast blasts include
several (typically 7) rows of shots. Intershot time spacing is 35 msec, rows are
spaced by 200 to 300 msec. For reasons as yet not fully understood, one cast blast
(Aug. 1, 1996) detonated, in part, nearly simultaneously. Six calibration shots
(ranging from 5000 to 16,000 pounds) were detonated in the mine in 1997 (Fig. 4;
Table 2). The ®rst four calibration shots (yields 5000 to 5500 pounds) consisted of a
single cylindrical borehole. The ®fth and sixth calibration shots (yields 12,000 and
16,000 pounds) consisted of 3 and 4 boreholes, respectively. The boreholes in the
larger shots were spaced 20 to 30 m apart and detonated simultaneously. In the
Figure 3
Two regional experiments were conducted in Wyoming by LANL, SMU, AFTAC and UCSD in 1996 and
1997. Four broadband seismic stations (three-component STS2's) were deployed in a ring surrounding the
Powder River Basin (PRB) at a range of 200 km from the Black Thunder coal mine. A ®fth station
(KRET) was deployed at a range of 100 km. Permanent seismic stations are located at PDAR and RSSD.
Infrasound sensors were deployed at MNTA. Mining explosions detected by the temporary seismic
stations during 49 days in 1996 and 1997 are plotted and correlate well with known mines. The slight
northward bias is likely due to the 1-D model. The oval indicates approximate limits of the PRB.
836 Michael A. H. Hedlin et al. Pure appl. geophys.,
Figure 4
A satellite photo of the Black Thunder coal mine in Wyoming. The symbols give approximate locations of
the major events considered in this paper. The 1996 and 1997 experiments produced recordings of four
signi®cant cast blasts in the Black Thunder coal mine. The July 19, Aug. 1, 1996 and Aug. 14, 1997 blasts
occurred at the west end of the south pit. The west end of the Aug. 1 blast is believed to have detonated
simultaneously. In the July 19, Aug. 1, 2 (all 1996) and Aug. 14 (1997) cast blasts 4.5, 2.5, 2.8 and 7.0
million pounds of ANFO were detonated. The two largest calibration tests involved 12,000 and 16,000
pounds of ANFO.
Table 1
Major blasts at Black Thunder
Date Pit/®ring direction # Rows/# holes Intershot/interrow delays
(m
s
)
Total explosive yield
(pounds)
July 19, 1996 South/west 7/620 (decked) 35/200±275 4,549,366
Aug. 1, 1996 South/west 7/341 35/200±275 2,460,730
Aug. 2, 1996 NE/south 7/422 35/200±275 2,784,540
Aug. 14, 1997 South/west 7/702 35/200±400 5,958,010
Vol. 159, 2002 Identi®cation of Mining Blasts 837
discussion that follows, we consider the time and frequency domain characteristics of
all large Black Thunder events that occurred during the two experiments. However,
we will focus on two events; the 4.5 million pound cast blast that occurred on July 19,
1996 and the 16,000 pound calibration shot.
3. Observations of Low-frequency Seismic Signals from Mining Blasts
Time Domain Energy Partitioning
Amplitudes of body waves produced by the 4.5 million pound July 19 cast blast
are comparable to those produced by the 16,000 pound calibration shot (Fig. 5). This
is expected as the July 19 blast yield is spread out over 620 shots which were arranged
in 7 rows and ripple-®red over 4.8 s. Each shot hole was decked ± that is two
explosive charges were detonated with a small delay in each hole. Although the two
events produced similar body waves, just the cast shot produced signi®cant surface
waves (Fig. 5). Un®ltered recordings of the July 19 blast (Fig. 6) show the
progression from high-frequency body waves to low-frequency surface waves at
four azimuths. The surface-wave amplitudes exhibit a clear dependence on azimuth.
Figure 7 indicates that although the low-frequency waveforms are highly dependent
on azimuth, these signals do not seem to depend strongly on the ®ne details of the
source if the blasts are in the same pit.
ANANDAKRISHNAN et al. (1997) and STUMP and PEARSON (1997) have published
PDAR recordings of signi®cant 4 to 12 second surface waves which are produced by
cast blasts in Wyoming. ANANDAKRISHNAN et al. (1997) used linear source modeling
to argue that the surface waves are the result of the long source duration and spall
impacts. The 1996 and 1997 regional experiments indicate that these surface waves
are routinely produced by the cast explosions and are readily detected at near-
regional range.
In Figure 8, peak amplitudes in two pass bands (1 to 10 Hz; centered at Ponset
and 2 to 10 seconds centered on the surface wave) are estimated from noise-
corrected envelopes of recordings made by the stations in the 200-km ring (Fig. 3).
The cast blasts and calibration explosions are easily separated despite the strong
Table 2
Calibration shots at Black Thunder in 1997
Date Yield (pounds)
Aug. 14, 1997 5500
Aug. 14, 1997 5500
Aug. 14, 1997 5500
Aug. 14, 1997 6000
Aug. 15, 1997 12,000
Aug. 15, 1997 16,000
838 Michael A. H. Hedlin et al. Pure appl. geophys.,
dependence of the peak surface-wave amplitudes on azimuth. Although small
calibration shots yield Pwaves that are comparable in strength to those produced
by the much larger, but not concentrated, cast blasts, just the latter events produce
signi®cant surface waves. As indicated in this ®gure, the calibration shots yielded
no surface-wave energy above noise. The lone outlier, among the population of cast
blasts, is the Aug. 1, 1996 event. A signi®cant simultaneous detonation, which
occurred within this blast, signi®cantly boosted peak P-wave amplitudes. The
network average surface-wave amplitudes are almost identical, despite the
signi®cant dierences in how these blasts were detonated. Despite the boosted P-
wave amplitudes observed in the Aug. 1, 1996 recordings, the amplitude ratio
comparing surface waves to body waves still separates this event from the single
shots.
Figure 5
Un®ltered vertical component recordings of a 16,000 pound calibration shot (top) and a 4.5 million pound
cast shot made at CUST. Both recordings are plotted at the same scale. The station was located 200 km to
the north of the events (Fig. 3) which occurred in the Black Thunder coal mine. The tiny calibration shot
rivals the immense cast shot as a source of Pwaves but is an insigni®cant source of surface waves. The
dissimilarity of un®ltered vertical component recordings made at CUST is consistent with previous ®ndings
(KIM et al., 1994; ANANDAKRISHNAN et al., 1997) and suggests that a regional variant of the M
s
:m
b
discriminant could be eective for separating large mine blasts from instantaneous explosions.
Vol. 159, 2002 Identi®cation of Mining Blasts 839
Frequency Domain Modulations
Previous studies (incl. STUMP and PEARSON, 1997) have shown that the large cast
explosions in Wyoming do not yield obvious spectral modulations above 10 Hz. The
intershot delays of 35 msec concentrate energy at multiples of 29 Hz (1/0.035 s).
Close-in data show a spectral increase that corresponds to the 35 ms delays although
the peak is relatively broad, re¯ective of the dierent spatial locations of the
individual charges and possibly variance in their individual detonation time. Rapid
attenuation in this area is an additional cause of the faintness of the high-frequency
modulations. These explosions do, however, produce signi®cant modulations below
Figure 6
Un®ltered vertical component seismograms from the azimuthal network show the progression from high-
frequency body waves to substantial surface waves. The July 19, 1996 blast used 4.5 million pounds of
ANFO which detonated as planned. The arrow indicates the direction of shooting. The waveforms are
highly dependent on azimuth. Propagation to the east across the Black Hills Pluton resulted in relatively
little surface-wave energy.
840 Michael A. H. Hedlin et al. Pure appl. geophys.,
10 Hz. Spectral estimates taken from recordings of the July 19, 1996 cast blast made
by the 5 station regional network are shown in Figures 9 and 10. These spectra
exhibit only a modest dependence on time, the recording direction, and the azimuth
from the mine. No organized spectral modulations are seen in the recordings of any
of the calibration shots. In this ®gure we display spectra from the 16,000 pound shot
(dashed curves). As will be seen later, the low-frequency modulations below 5 Hz are
seen out to 9°.
4. Waveform Synthesis
To give these basic observations a physical basis, we turn to synthetics. The
synthesis of extraordinarily complex mining explosions has become relatively easy
given the early work of BARKER and DAY (1990), BARKER et al. (1993) and
MCLAUGHLIN et al. (1994) and recent work by X. Yang who has modi®ed the linear
elastic algorithm of ANANDAKRISHNAN et al. (1997) and packaged it into an
interactive MATLAB package (MineSeis; YANG, 1998). The algorithm assumes the
linear superposition of signals from identical single-shot sources composed of
Figure 7
A comparison of band-pass ®ltered regional seismograms from three dierent cast blasts at ®ve dierent
stations at varying azimuths from the mine (see Fig. 3). All of these blasts occurred in the south pit of the
Black Thunder coal mine. All shots were detonated in the south pit of the Black Thunder coal mine
(Fig. 4). Low-frequency seismic signals from the south pit events are robust although highly dependent on
azimuth.
Vol. 159, 2002 Identi®cation of Mining Blasts 841
isotropic and spall components. Both shooting delays and location dierences among
individual shots are taken into account in calculating delays of the superposition,
although the Green's functions are assumed to change slowly so that a common
Green's function is used for all the single shots. We used a re¯ectivity method to
calculate the Green's functions. A one-dimensional velocity model was used
(PRODEHL, 1979; ANANDAKRISHNAN et al., 1997).
Figure 8
A comparison of 2 to 10 second surface wave and 1 to 10 Hz P-wave peak amplitudes using recordings
made at a range of 200 km by MNTA, CUST, LBOH & SHNR (Fig. 3). All events occurred in the Black
Thunder coal mine. Each trace is ®ltered, converted to an envelope and adjusted downward by an amount
determined by pre-onset noise. Above are displayed the logarithms of the individual station peak
amplitudes. The large symbols represent the network average for each event. Each labeled curve indicates a
constant ratio of surface wave to P-wave amplitude [i.e. log(2 to 10 s peak)/log(1 to 10 Hz peak)]. As
expected the calibration shots yield essentially no surface-wave energy above noise. The downward arrows
indicate that the maximum P-wave amplitude is well constrained but the surface-wave amplitude lies below
noise. For this reason, no network averages are displayed for these events. The Aug. 1, 1996 cast shot
appears as being somewhat explosion-like due to the sympathetic detonation. The sympathetic detonation
greatly boosted P-wave amplitudes but left the surface waves untouched. Unadjusted amplitudes are
plotted as all stations are at the same range from the mine. We also display peak amplitudes from a
synthetic version of the July 19 cast blast and from a synthetic 16,000 pound calibration shot. The energy
partitioning from the synthetic events is in agreement with observations.
842 Michael A. H. Hedlin et al. Pure appl. geophys.,
To model the July 19 cast blast we used a blast report issued by the Black
Thunder mine. The blast consisted of 7 rows and a total of 620 decked shots with a
total yield of 4.5 million pounds (Table 1). In a decked shot, more than one charge is
detonated in the same hole. In this event, each hole contained two charges separated
by 50 to 200 msec. Although the number of shots in each row varied from 85 to 93,
for simplicity we assumed each row had 89 shots and that each decked shot had a
total yield of 0.0033 kt. We assumed that all rows were spaced by 9.1 m and that all
adjacent shots in the same row were 10.4 m apart. Intershot delays were 35 msec,
interrow delays ranged from 200 to 275 msec. SOBEL (1978) estimated that 9:6109
kg of material is spalled by each kt detonated. In our experiment, each decked shot
Figure 9
Spectral estimates taken from pre-event noise and signal recorded at CUST. In the upper panel, we display
spectra taken from the recording of the July 19, 1996 cast blast (see Fig. 5). In the lower panel we display
spectra taken from the CUST recording of the 16,000 pound calibration shot. The noise spectral estimates
were untapered. Both P-wave spectra have been convolved with a boxcar function spanning 0.04 Hz to
give a clearer view of the spectral modulations in the upper panel.
Vol. 159, 2002 Identi®cation of Mining Blasts 843
Figure 10
Obvious time-independent modulations exist below 10 Hz in the spectra of recordings of the July 19 cast
blast (upper solid lines). At each station three curves, each representing the log of the spectral amplitude
from a single component, are plotted. These are similar to the high-frequency modulations observed by
HEDLIN et al. (1989) although these are likely due to source ®niteness and interrow delays. Both these
spectra and those observed by HEDLIN et al. (1989) are independent of the recording direction. A very
slight dependence of the modulations on azimuth can be seen. The 16,000 pound calibration shot (dashed
lines) produced no discernable spectral modulations. Each detrended multitaper spectral estimate was
taken from 125 seconds of Pand Scoda. All components, from each three-component station, are plotted.
844 Michael A. H. Hedlin et al. Pure appl. geophys.,
had a total yield of 0.0033 kt and thus the Sobel relation gives 31 kt of spalled
material. For the July 19 Black Thunder event we found this ®gure yielded surface
waves that were more energetic than those that were recorded and so we reduced the
spall ®gure to 20 kt. As will be discussed later, this discrepancy can be due to
the incomplete conversion of spalled kinetic energy into seismic or to inadequacy of
the velocity model. We assumed that the spalled material was cast at an angle 10°
above the horizontal and fell 20 m.
Some vertical component synthetics are shown in Figure 11. This ®gure illustrates
the relatively energetic surface waves that can be expected from cast blasts. The
16,000 pound point synthetic source, modeled as an isotropic source, produces a
weak surface wave. The ®gure also illustrates the slow attenuation of the surface
wave produced by the synthetic cast blast.
The peak amplitudes of the point source and the July 19, 1996 cast blast
synthetics at a range of 200 km have been calculated for the four outer stations in the
regional network (Fig. 3) and are displayed in Figure 8 with the observed amplitudes
from the recorded events. Despite the assumptions listed above, we found the
surface- and body-wave amplitudes produced by the synthetic event were consistent
with the recorded waves at MNTA and CUST. The other two stations (LBOH and
SHNR) yielded broadly dispersed surface waves which had lower peak amplitudes in
the frequency band from 2 to 10 seconds. This pronounced mismatch is not due to
unmodeled source eects but results from the propagation of the energy through a
crust that diers signi®cantly from the one-dimensional Prodehl model. Any
comments at this time on what these dierences imply about the crustal velocity
model would be imprecise. A surface-wave inversion is currently being conducted by
Rongmao Zhou and Brian Stump at Southern Methodist University. Although a
signi®cant azimuthal dependence of the surface-wave amplitudes is observed, the
surface waves produced by the cast blasts are signi®cantly more energetic at all
azimuths than those produced by the single shots. The relatively minor azimuthal
dependence that is seen in the synthetics is due to source directivity.
Experiments in Varying Source Parameters
Although we have reproduced the relative peak amplitudes of body and surface
waves generated by the July 19 cast blast, several important questions remain
unanswered. We have yet to determine which of the many source parameters included
in the model play a leading role in generating the energetic surface waves. We need to
gauge the sensitivity of the observed signals to changes in source parameters and to
assess the utility of these signals for source characterization. ANANDAKRISHNAN et al.
(1997) considered the eect of changes in several source parameters on regional
waveforms. In this paper, we conduct a similar exercise; however, we are most
interested in how these changes aect the partitioning of energy between the surface
and body waves. We focus on the frequency bands considered in Figure 8. We
Vol. 159, 2002 Identi®cation of Mining Blasts 845
Figure 11
Simulations of two events. A single 16,000 lb shot is displayed on top, the July 19, 1996 cast shot is
displayed on the bottom. The single shot is an insigni®cant source of surface waves. The surface waves
excited by the cast blast decay relatively slowly and dominate the waveform of the cast shot at all ranges
from 200 to 1000 km. The single shot synthetics are magni®ed for clarity. The range dependence of all
synthetics was reduced by simply scaling all amplitudes by the range.
846 Michael A. H. Hedlin et al. Pure appl. geophys.,
consider a broad suite of mine blasts. The blasts we synthesize are all the same as the
presumed correct one reviewed in the previous section except one source parameter is
allowed to vary while the others are held ®xed. In turn, we vary the spalled mass,
individual shot yield, and the duration of the blast. In a separate experiment we also
allowed shot times to deviate from the planned 35 msec delays. All synthetics are
calculated for CUST, the station located 200 km to the north of the mine. The results
for the other three stations are essentially the same and are not plotted.
An example is shown in Figure 12 which displays synthetics for a suite of cast
blasts in which the total yield of each shot hole is varied from 20% of the actual value
to 200% (from 1468 to 14,676 pounds). The synthetics suggest that the body-wave
amplitudes will scale rapidly with individual shot yield however the surface waves
show a weaker dependence. This is consistent with ANANDAKRISHNAN et al. (1997)
who concluded that the surface waves are mostly due to the dierent yield scaling in
the explosion and spall source models.
Figure 12
The dependence of vertical component waveforms on shot parameters is easily simulated using the
MineSeis Matlab package. These plots show the dependence of the waveform on the yield of the individual
shots. The standard cast, with individual shots of 7338 lb, is the third trace from the top. The scaling of the
shot yield is indicated by the text to the right of each trace. The body-wave amplitudes depend strongly on
shot yield. The surface waves show a weaker dependence.
Vol. 159, 2002 Identi®cation of Mining Blasts 847
A more general illustration of source eects on body and surface-wave
amplitudes is given in Figure 13. In the upper panel of this ®gure we consider peak
amplitudes in two ®ltered versions of the synthetic traces. We ®lter the synthetic
traces between 2 and 10 s, and between 1 and 10 Hz and calculate the log of the peak
amplitude in each trace. In Figure 13, we display the ratio of the peak amplitude in
the the low-frequency trace to that in the high-frequency trace. In the upper panel, we
Figure 13
In the upper panel we display the surface-body wave ratios for a suite of blasts in which a single source
parameter is varied while all others are held ®xed at levels believed to be appropriate for the July 19, 1996
cast blast. The ®lled circles represent sources in which individual shot yield is varied from 182 to 36,300 lb.
The ®lled triangles represent blasts in which the spalled mass per shot is varied from 0.5 kt to 80 kt. The
squares represent blasts grids that range from a single row with three shots to a 10 row blast with 200 shots
in each row. Varying blast duration clearly has the most signi®cant impact on energy partitioning. Only the
severely restricted grid partitions energy like the synthetic 16,000 pound shot (represented by the horizontal
line at a ratio of 0.61). All of these simulations assumed the subshots detonated exactly when planned. The
curve in the lower panel represents blasts in which the shot times are distributed normally about the
planned times. The shot scatter considered ranges up a variance of 30% of the planned intershot delay of
35 msec.
848 Michael A. H. Hedlin et al. Pure appl. geophys.,
display the ratios obtained from synthetic events in which a signi®cant source
parameter has been varied between 2.5% and 500% of the standard value. As seen in
Figure 12, the surface wave to body-wave amplitude ratio decreases with increasing
shot yield. This eect is represented by the ®lled circles in the upper panel of
Figure 13. This eect is seen to be relatively minor as varying the yield over more
than three orders of magnitude (from 182 to 36,300 pounds) changes the body-
surface ratio by 10%. As seen in Figure 12, this ratio change results from changes
in both the surface- and body-wave amplitudes.
Varying the spalled mass has the opposite eect. When the spalled mass is varied
from 0.5 kt to 80 kt (from 2.5% to 400% of the standard) the surface-body wave
ratio changes from 0.9 to 1.13 (®lled triangles in the upper panel of Fig. 13). The
change in this ratio results from the eect spalled mass has on the surface waves as
the spalled mass is predominantly a source of low-frequency energy. Changing the
spalled mass had little eect on the body-wave amplitudes. Although the production
of surface waves is reduced when essentially no material is spalled into the pit, the
restricted blast still produces surface waves that are signi®cantly more energetic than
those produced by the single shot.
The most signi®cant changes in the waveform result from variations in the total
duration of the blast. To vary the blast duration we scaled both the number of rows
and the number of shots in each row. The standard blast had 7 rows and 89 shots in
each row and lasted a total of 4.8 s. The largest blast grid we considered had 10
rows, each with 200 shots. The extended blast spanned 9.5 s ± 200% of the actual
duration and had a total yield of 14 million pounds. The reduced blast grids consisted
of 5 rows, 50 shots per row; 4 rows, 40 shots per row; 3 rows with 30 shots per row; 2
rows with 20 shots per row and 1 row with 10 shots. To reduce the blast further we
decreased the number of shots in the ®nal row to 5 shots and then to 3. The duration
of these reduced blasts ranged from 3.1 s (for the blast with 5 rows and 50 shots per
row) to 70 msec (for the smallest blast). As we see in Figure 13, extending the blast
beyond the standard one that was used on July 19 yielded essentially no change in the
surface-body wave amplitude ratio. However reducing the scale of the grid has a
signi®cant eect on the partitioning of energy between surface and body waves.
Reducing the scale of the blast has a modest eect on body-wave amplitudes but the
surface waves are strongly dependent on this source attribute. The ratio drops most
rapidly when the blast duration is reduced below 2/10 of the actual duration (down
to 0.5 s from 4.8 s). The production of 2±10 s surface waves is reduced signi®cantly
when the blast duration is reduced below 2 seconds. When the blast duration extends
much beyond 5 seconds, production of these surface waves is not increased
substantially. Of all the blasts considered, just the brief blast resembled the single
shot (which is represented by the horizontal line at a ratio of 0.61).
All synthetics considered thus far assumed a perfect temporal and spatial
adherence of the actual shot grid to the design grid. Introducing shot time scatter,
which is believed to be omnipresent (STUMP and REAMER, 1988; STUMP et al., 1994,
Vol. 159, 2002 Identi®cation of Mining Blasts 849
1996), is predicted to have no eect on the surface-wave amplitudes but could increase
the body-wave peak amplitudes signi®cantly (Fig. 13; lower panel). Shot scatter
increases the likelihood that shots will detonate simultaneously. An extreme example
of shot scatter is the simultaneous detonation of a portion of the shot grid. This
occurred in the August 1, 1996 blast and, as predicted by the synthetics, the surface
waves were not aected (Fig. 8). The network averaged peak surface-wave amplitudes
for the three mine blasts shown in this ®gure are very similar, despite the wide range of
explosive yields. It appears that surface waves are just generated by temporally
extensive mine blasts without regard for exactly how the blasts are detonated, and
whether the blast sequence includes any signi®cant detonation anomalies. The
anomalous event was assigned a body wave magnitude of 4.0 (REB Bulletin).
5. Synthesis and Automated Recognition of Spectral Modulations
Synthesis
Seismic signals produced by delay-®red sources have spectral modulations at a
wide range of frequencies. High-frequency modulations result directly from intershot
delays. These delays are typically 35 msec and the modulations occur at multiples of
1/35 msec (30 Hz). Interrow delays are typically longer. The cast blasts we consider
in this paper have interrow delays of 200 to 300 msec which give modulations every
3 to 5 Hz. Source ®niteness will also cause subtle modulations spaced at the inverse
of the duration of the event (e.g GITTERMAN and VAN ECK, 1993). Spectral
modulations can also be acquired during propagation to the receiver (HEDLIN et al.,
1989). To illustrate this problem, and to illustrate the modulations that can result
from source ®niteness, we consider two synthetic sources ± a single shot and a simple,
1 row, 25 shot delay-®red source. As shown in Figure 14, a single shot yields spectra
with subtle modulations. These modulations are due to resonance in the near-surface
layer of the model. This layer has a two-way travel time of 0.4 s and thus produces
modulations spaced at 2.5 Hz. Much more signi®cant modulations below 10 Hz
are produced by the delay-®red event. The simple delay-®red event consisted of a
single row of 25 shots. Intershot spacing of 35 msec produces a high-frequency
modulation starting at 29 Hz (1/0.035 s). The shot sequence lasts for 0.875 s. The
source ®niteness causes a spectral modulation at the inverse of the source duration
(peaks every 1.14 Hz).
To understand better the modulations produced by the July 19 event, we contrast
signals from three dierent blasts with those produced by an instantaneous shot.
Starting at the top of Figure 15, the upper three black traces are modulations
between 1 and 10 Hz predicted for 1, 4 and 7 row cast blasts, respectively. These
synthetic blasts are modeled after the July 19 decked blast, we have just altered the
number of rows. The dashed spectra were taken from a synthetic 16,000 pound
850 Michael A. H. Hedlin et al. Pure appl. geophys.,
instantaneous shot which was located at the same point. In the single row delay-®red
event we see broad modulations spaced at 2.5 Hz and ®ne-scale modulations
spaced at 0.3 Hz. The ®ne scale modulations are due to the ®niteness of the source
(which spanned 3 s). The broader modulations are seen in the spectra from both
sources and are due to resonance. Strong modulations spaced at 0.75 Hz appear in
the spectrum of the 7 row blast. These are not continuous across the band from 0 to
10 Hz but appear to be strongest at 3 Hz. These modulations are not present in the
spectra of the single-row blast and are due to the combined eect of the interrow and
interdeck delays (which range from 0 to 300 msec) and source ®niteness. Although
the 7 row event had a total duration of 5 s, the rate at which explosives were
detonated was strongly dependent on time. The ®rst shot in the ®nal row detonated
Figure 14
A synthetic 16,000 lb single shot at a range of 200 km is shown on the left. The faint spectral modulations
seen in the spectra are due to resonance in near-surface low-velocity strata. A simple cast blast consisting of
1 row of 25 shots spaced at 35 msec is shown on the right. This simulation is for a station at a range of
200 km. The prominant modulations seen in the untapered spectral estimates are due to source ®niteness.
This shot lasted 0.875 s and produced modulations at the inverse of the duration (1.14 Hz). The ®rst peak
due directly to the intershot delays lies at 29 Hz.
Vol. 159, 2002 Identi®cation of Mining Blasts 851
1.9 s after the shot sequence began. The ®nal shot in the ®rst row detonated 3.2 s
into the sequence. For 1.3 s in the middle of the shot grid, all 7 rows were being
detonated and the explosive yield per time delay was at a peak (25,000 pounds per
8 msec delay period). The spectral modulations produced by this trapezoidal source
time function are not spaced at the inverse of the total shot duration (every 0.2 Hz)
but rather are more broadly spaced at 0.75 Hz (the inverse of the 1.3 s period
during which explosive yield is at a peak). These ®ne scale modulations are not
continuous across the band from 0 to 10 Hz because of destructive interference with
long interrow and interdeck delays.
The fourth spectrum displayed in Figure 15 was taken from the vertical
component recording of the July 19 event made at CUST. At the bottom of this
®gure we display a spectrum taken from the CUST recording of the 16,000 pound
calibration shot. Strong peaks are observed in the cast blast spectrum at 2, 4, 6 and
10 Hz (see also Fig. 10). Smaller modulations are seen across the band from 1 to
10 Hz at a spacing of 0.75 Hz. These modulations are not seen in the spectrum of
the calibration shot and are clearly a source eect. These modulations require source
Figure 15
A comparison of spectra from synthetic and real events. The upper three black traces are vertical
component spectra from synthetic cast blasts that are based on the July, 1996 south pit blast. The third
spectrum is calculated from the full 7 row shot pattern. The middle and upper traces were calculated by
using the front 4 and 1 rows, respectively. The grey curve plotted with each spectrum is from a synthetic
16,000 pound calibration shot. All synthetics were calculated for a receiver located 200 km to the north of
the mine at the CUST station. The fourth spectrum is from the CUST vertical component of the July 19
event. The lowest trace is from the recorded 16,000 pound calibration shot. Broad modulations seen in all
synthetic spectra result from resonance in the one-dimensional model. The ®ne-scale modulations are a
source eect. Although the synthetics do not reproduce the ®ne detail of the low-frequency spectra
recorded by the regional network, they do reproduce the general periodicity and variance of the
modulations. In this ®gure, all spectral estimates are untapered.
852 Michael A. H. Hedlin et al. Pure appl. geophys.,
delays of >1 s to 500 msec. The synthetic test displayed in the upper part of this
®gure suggests that these modulations are due to the combined eects of source
®niteness and the complex interference between the 7 rows and 2 decks at this source;
however, the ®ne spectral details are not reproduced.
We lack the necessary ground-truth information to reach de®nitive conclusions
regarding probable causes of the mismatch between the observed and synthesized
modulations; however, unmodeled source eects such as shot time and yield scatter
seem most likely. The synthetic events we considered in this experiment consisted of
shots that had exactly the same yield and detonated exactly when they were supposed
to. Any deviations from uniform spacing of identical shots will change the manner in
which signals from the dierent source processes (®niteness, interrow, interdeck and
intershot delays) interfere with one another and will cause the recorded modulations
to dier substantially from those predicted by synthetics. The synthetics suggest that
most modulations observed in the recorded cast blast are due to long source delays
and to source ®niteness; however, this observation remains tentative as the ®ne
details of the spectral modulations are not matched.
Automated Recognition
HEDLIN (1997, 1998a, b) used a variant of cepstral analysis to quantify time-
independent spectral modulations at high frequencies (up to 40 Hz). The standard
cepstrum is the Fourier transform of the log of a single spectrum. Hedlin used the
two-dimensional Fourier transform of sonograms to isolate spectral energy that is
periodic in frequency and independent of time. The same processing technique can be
used for the low-frequency modulations observed here. Figure 16 shows discrimi-
nation parameters output by the Automated Time Frequency Discriminant (ATFD)
for individual stations. The three parameters displayed are the autocorrelation,
which measures the independence of spectral modulations with time; the cross
correlation, which measures the independence from recording component; and
cepstral extreme, which indicates the strength of time-independent spectral modu-
lations. Although there were too few ground-truthed events to put the output to a
statistical test, we tentatively conclude that the network averaged parameters
separate single explosions from cast explosions.
6. IMS Recordings of the July 19, 1996 Cast Blast
The regional experiment provided evidence that large mining events will routinely
yield signi®cant low-frequency seismic signals. This network was deployed within 2°
of the Powder River Basin so this test is rather unrealistic. Under a foreseeable
monitoring scenario, where the bulk of the data comes from IMS stations, these
events will be detected from mid-regional range. Detection statistics from the PIDC
Vol. 159, 2002 Identi®cation of Mining Blasts 853
854 Michael A. H. Hedlin et al. Pure appl. geophys.,
show that many of the PRB explosions are seen at far-regional to teleseismic range
but can these attenuated signals be used for source characterization?
IMS recordings of the Wyoming events can give some indication of the range
from which these signals might be used for source characterization. As shown in
Figure 17, a low-frequency surface wave packet from the July 19, 1996 Black
Thunder blast are seen out to ULM at a range of 9.1°. Figure 18 shows that the low-
frequency spectral modulations also survive to this range; however, those above 6 Hz
have fallen below noise.
7. Discussion and Conclusions
Mining Explosions and the CTBT
Under a CTBT, mining explosions will be problematic. As these events occur
worldwide, many produce signi®cant seismic signals which can be detected at mid- to
far-regional and to teleseismic distances. Mining blasts are exceedingly complicated
events. The complexity in the cast blasts considered in this paper stems primarily
from the interaction of delay-®red explosives, rock fracture, and spall. The challenge
mining events pose for the monitoring community is heightened because, through
time, mine operators will occasionally experiment with new shot patterns. Further-
more, mine blasts will typically not detonate exactly as planned. Most deviations
from the planned shot grid, such as shot scatter (e.g., STUMP and REAMER, 1988), are
not highly signi®cant. Other anomalies, such as the sympathetic detonations
discussed in the introduction, are relatively infrequent but can be signi®cant. A
constraint is also placed on characterization methods by the IMS stations which
sample the signal at 40 sps. Most diagnostic signals due to intershot delays are
beyond the recording Nyquist frequency and are likely to be attenuated. For these
reasons, we believe that it is important that a suite of approaches are developed to
characterize these events. Under the CTBT, monitoring of mining blasts at any range
will have to rely heavily on low-frequency signals.
Figure 16
Discrimination parameters calculated by the Automated Time-Frequency Discriminant (ATFD, are
described in detail in HEDLIN (1998). Each panel shows the results of applying a single operator to the time-
frequency expansions of the data. The autocorrelation operator (top panel) assesses the dependance of the
spectral modulation pattern on time. The cross correlation (middle panel) is a measure of the independence
of the modulation pattern from the recording component. The cepstral extreme (bottom panel) is a
measure of the amount of energy in the coda that is periodic in frequency and independant of time. Each
three-component recording gives rise to nine parameters ± three from each operator. Two calibration shots
(left) and four cast blasts (right) were considered. The large symbols represent unweighted network
averages.
b
Vol. 159, 2002 Identi®cation of Mining Blasts 855
The Relative Merits of the Discriminants
Methods based on spectral modulations are potentially useful. For this approach
to have value, it is necessary to separate modulations acquired at the source from
those acquired during propagation and from those present in the background noise.
Techniques that are dependent on high frequency modulations are typically limited
to near-regional ranges and settings where the propagation Qis high (e.g., stable
cratons). For this reason, this paper has focused on the causes of modulations below
10 Hz. It is apparent that complex mining blasts produce spectral modulations at a
broad range of frequencies. The Wyoming ®eld experiments have provided evidence
that signi®cant mining explosions will produce spectral modulations below 10 Hz
that are not single-explosion like. Low-frequency modulations below 5 Hz will be
detected beyond near-regional range. Just how common these signals are beyond
near-regional range, and whether they are noticeable at lower blast yields is implied
by our analysis of synthetics and by the recordings at ULM but will require further
study of recorded events. The path from Wyoming to ULM probably has a relatively
Figure 17
IMS/WYnet recordings of the Jul. 19, 1996 cast blast. The low-frequency bandpassed recordings exhibit an
energetic dispersed surface wavetrain out to the IMS station ULM at 9.1°.
856 Michael A. H. Hedlin et al. Pure appl. geophys.,
Figure 18
Multitaper spectral estimates taken from IMS and WYnet recordings the July 19 cast blast. The dashed
portion of the ULM spectrum is noise. This robust low-frequency spectral roughness is not produced by
small single explosions (Fig. 10).
Vol. 159, 2002 Identi®cation of Mining Blasts 857
high propagation Q. As a result, this observation cannot be taken as representative
most regions but is the kind of scenario one might expect in a stable continental
region. One region in which large mining blasts are not uncommon and propagation
Qis relatively high is the Kuzbass/Abakan mining region in Kazakhstan. In regions
where propagation Qis low, such as Wyoming, the high-frequency modulations due
to intershot delays are quickly lost, although those due to longer duration source
processes remain.
Long-period surface waves from large mining blasts in Wyoming are readily seen
at mid-regional distances. The expenditure of a lot of energy spread out in time
results in body waves that are relatively inenergetic and surface waves that are large.
Our experiment indicates that the energy partitioning of mining blasts is not
explosion-like and might be useful at mid-regional distances. It is clear that in
geologically complex regions like Wyoming, surface waves are highly dependent on
the path (Figs. 6±8). Even in this region, however, our study indicates that peak
amplitudes along paths where surface waves are rapidly dispersed, are high.
Our analysis of a small number of events indicates that the observables (spectral
roughness, energy partitioning) can be reduced to simple discrimination parameters.
This result indicates that the ATFD approach developed by HEDLIN (1997, 1998a, b)
can be simply scaled to lower frequencies where necessary. We have not yet
attempted to de®ne similar discrimination parameters for the body-surface amplitude
ratios. In large part, this is due to a shortage of events included in the analysis and a
shortage of recordings at ranges other than 200 km. A number of researchers (incl.
ANANDAKRISHNAN et al., 1997; STUMP and PEARSON, 1997) have pointed out
anomalous surface waves. If this observation appears to be routine from studies of
events from other regions the next logical step is to take source-receiver range into
account and de®ne a M
b
:m
s
relationship for mining blasts.
The potential for misidentifying a mining blast by using any one of these
individual tools is rather high. It will, no doubt, be necessary to regionalize these
techniques by examining, in detail, blasting practices and propagation in each area of
interest. These tools will, perhaps, some day be used alongside other tools, such as
cross correlation, that have been proven to be powerful by a number of researchers.
Simulations
The state-of-the-art in simulating mine blasts is not able to match wiggle for
wiggle the seismic waves from these very complex events in the time or in the
frequency domain. However, general features of these events (energy partitioning
between body and surface waves, enhanced spectral roughness below 10 Hz) have
been reproduced. We have used the synthetics as an interpretation tool to reproduce
the general character of these events to better understand which source processes are
important, but an exact replication might never be realized.
858 Michael A. H. Hedlin et al. Pure appl. geophys.,
There are several likely causes of the mis®t. The code makes a number of
assumptions about these sources that limit how well we can ®t the data. Nonlinear
interactions between shots are not taken into account and are likely important
(MINSTER and DAY, 1986). The model assumes 100% of the spall kinetic energy
converts into seismic. It is known that much of this energy is used to compact the
spalled mass by collapsing voids and fracturing the rock, more energy is irretrievably
lost to friction; however, these losses are not well understood at this time and are not
taken into account in this code. As a remedy, this loss of energy can be modeled to
some degree by assigning a smaller amount of spalled mass. The spalled mass is
directly proportional to the seismic energy the spall generates. All shots in the blast
are assumed to be identical. However it is well known that explosive yield is highly
variable. STUMP et al. (1999b, c) analyzed single-®red shots ranging in explosive
weight from 5500 to 50,000 pounds. They found amplitudes from several 5000 pound
shots varied by up to 2 orders of magnitude. The calibration shots considered by
STUMP et al. (1999b, c) were, 30 m apart at the same vertical depth and so it is
unlikely that changes in the physical properties of the medium are the cause of this
variable performance. Velocity of detonation measurements in the holes were not
made although video footage indicates a lot less borehole response in terms of
motion and permanent displacement around the borehole for the event producing the
smaller motions. These arguments support degraded explosive performance.
Other processes associated with each shot (e.g., spall tonnage and throw
direction) are also assumed to be identical. Decorrelation between shots is also
unavoidable (STUMP et al., 1999b, c) and will degrade the constructive interference
(BAUMGARDT, 1995). The analysis of single shots conducted by STUMP et al. (1999b,
c) indicates that for the shot separations commonly used in these mining blasts
(10 m) we should expect good correlation below 4±5 Hz. Decorrelation is not
taken into account in the current code as all sources are assigned a common Green's
function. Shot timing is also an issue. We have concluded that low-frequency
modulations result from long source delays and source ®niteness; however, the ®ne
details of these modulations are dependent on the interplay, or interference, between
these processes. Marked changes in the modulation patterns can be caused by
making seemingly minor adjustments in the shot pattern. Shot scatter, or the
mismatch between intended and actual shot detonation times, is ever present (STUMP
and REAMER, 1988) and is not taken into account in our simulations.
Despite these limitations, synthetics reproduce the general character of the
spectral modulations observed and indicate that, in the absence of strong crustal
resonance which will also yield spectral modulations, this trait can be useful for
separating delay-®red blasts from instantaneous explosions. The synthetics under-
score the need for taking into account seismic resonance. In the one-dimensional
model, the low-velocity surface layer is continuous between the source and receiver.
In practice, seismic resonance will be observed if the layer is discontinuous and
present only at the source or at the receiver.
Vol. 159, 2002 Identi®cation of Mining Blasts 859
The energy partitioning between body and surface waves is more easily
reproduced. The synthetic test clearly indicates that source duration has the most
signi®cant eect on surface wave amplitudes. Spall is a second-order eect. This
successful simulation gives a physical basis to the surface waves observed. The source
parameter tests are a ®rst step in using synthetics to gauge the sensitivity of surface
waves to changes at the source.
Outstanding Issues
This empirically based study has identi®ed a number of characteristics of seismic
waves from large-scale cast blasting that can be used for identifying the source.
Within the context of the CTBT, the extent to which this observational experience
can be used to assess discrimination techniques in other regions where propagation
path eects are dierent and blasting practices may vary needs to be assessed. In
some of these other areas there may be little or no access to information concerning
blasting practices and so it is hoped that these detailed studies may provide the
foundation for the interpretation of the observations after consideration of
propagation path eects.
A number of outstanding issues associated with the generation of regional
waveforms from mining explosions remain independent of speci®c information on
propagation paths. Mechanisms for the generation of regional surface waves have
been demonstrated but need further investigation and constraint. The impact of a
range of blasting practices from long-time duration cast blasting in coal mines to
short duration rock fragmentation blasts in surface coal mines needs to be explored.
The modeling and data analysis have assumed that design blast parameters are those
that are implemented. For purposes of assessing the reliability of discriminants, the
quanti®cation of anomalies in blasting and their implications for regional seismo-
grams needs additional exploration. Finally this study has focused on observations
from mining and single-®red explosions but has not compared these observations to
those from earthquakes along similar propagation paths. Studies which include
earthquake and mining explosion sources along comparable propagation paths will
allow the assessment of the proposed discriminats for separating earthquake and
mining explosion populations.
This study has focused on seismic observations from mining explosions. There is
increasing evidence that infrasonic observations may help in the identi®cation of
surface mining explosions (SORRELLS et al., 1997). It is possible that the colocation of
seismic and infrasonic instrumentation surrounding mining regions may signi®cantly
add to the identi®cation capability of mining explosions.
Mining events are controlled and thus might provide an opportunity for
clandestine tests as part of an advanced nuclear weapons program. There is a strong
incentive for learning how to use the IMS to distinguish anomalies from clandestine
nuclear tests and avoid On-Site Inspections.
860 Michael A. H. Hedlin et al. Pure appl. geophys.,
Unannounced mining events that include signi®cant simultaneous explosive
energy releases are problematic for several reasons. Such blasts are unwanted by the
mining community as these events have both reduced rock-fracturing eciency and
increased seismic eciency. Some of these blasts will trigger the IMS and cause some
to question whether the anomalous energy release was chemical in nature. Eective
characterization techniques, which rely on remote IMS observations, will reduce the
need for on-site inspections. Our preliminary analysis suggests that such events can
be identi®ed using spectral modulations and the relative strength of surface- and
body-wave seismic energy. A fuller analysis of errant blasts will be the subject of a
forthcoming paper.
Acknowledgments
The combination of regional and close-in measurements for solving problems
related to mining explosions could not have been made without the close collaboration
with Vindell Hsu at the Air Force Technical Applications Center (AFTAC). Bob
Martin, David Gross, Al Blakeman and Terry Walsh at the Black Thunder mine
provided essential support. Portable deployments were made possible by CL Edwards,
Diane Baker and Roy Boyd (LANL) and Adam Edelman, Aaron Geddins and John
Unwin. The authors would like to thank Doug Baumgardt and an anonymous
reviewer for helpful comments. Funding and equipment provided by LANL (under
contracts 1973USML6-8F and F5310-0017-8F) and the eorts of Frank Vernon and
the IGPP north lab permitted us to deploy the azimuthal network. Data processing
and analysis by MAHH was funded by DTRA under contracts DTRA01-97-C-0153,
and DTRA01-00-C-0115 and LANL under contract F5310-0017-8F.
R
EFERENCES
ANANDAKRISHNAN, S., TAYLOR, S. R., and STUMP, B. W. (1997), Quanti®cation and Characterization of
Regional Seismic Signals from Cast Blasting in Mines: A Linear Elastic Model, Geophys. J. Internat. 131,
45±60.
ATLAS POWDER COMPANY (1997), Explosives and Rock Blasting, Dallas, TX.
BARKER, T. G. and DAY, S. M. (1990), A Simple Physical Model for Spall from Nuclear Explosions Based
upon Two-dimensional Nonlinear Numerical Simulations, PL Report SSS-TR-93-13859.
BARKER, T. G., MCLAUGHLIN, K. L., STEVENS, J. L., and DAY, S. M. (1993), Numerical Models of Quarry
Blast Sources: The Eects of the Bench, PL Semiannual Report, SSS-TR-93-13915.
BARKER, T. G., MCLAUGHLIN, K. L., STEVENS, J. L., and DAY, S. M. (1994), Numerical Simulation of
Quarry Blasts Part 2: Implications for Discrimination, Seismol. Res. Lett. 65, 71.
BAUMGARDT, D. R. (1995), Case Studies of Seismic Discrimination Problems and Regional Discriminant
Transportability, Phillips Lab Report PL-TR-95-2106.
BAUMGARDT, D. R. and ZIEGLER, K. A. (1988), Spectral Evidence for Source Multiplicity in Explosions:
Application to Regional Discrimination of Earthquakes and Explosions, Bull. Seismol. Soc. Am. 78, 1773±
1795.
Vol. 159, 2002 Identi®cation of Mining Blasts 861
GITTERMAN, Y. and VAN ECK, T. (1993), Spectra of Quarry Blasts and Microearthquakes Recorded at Local
Distances in Israel, Bull. Seismol. Soc. Am. 83, 1799±1812.
GITTERMAN, Y., PINSKY, V., and SHAPIRA A. (1997), Application of Spectral Semblance and Ratio
Discriminants to Regional and Teleseismic Events Recorded by ISN and NORESS, AFTAC/DOE/DSWA
Seismol. Res. Symp. 19, 369±378.
HARRIS, D. B. (1991), A Waveform Correlation Method for Identifying Quarry Explosions, Bull. Seismol.
Soc. Am. 80, 2177±2418.
HEDLIN, M. A. H. (1997), A Global Test of a Time-frequency Small Event Discriminant, AFTAC/DOE/
DSWA Seism. Res. Symp. 19, 390±399.
HEDLIN, M. A. H. (1998a), Identi®cation of Mining Blasts at all Regional Distances using Low-frequency
Seismic Signals, 20th Ann. Seis. Res. Symp. 20, 335±344.
HEDLIN, M. A. H. (1998b), A Global Test of a Time-frequency Small-event Discriminant, Bull. Seismol. Soc.
Am. 88, 973±988.
HEDLIN, M. A. H., MINSTER, J.-B., and ORCUTT, J. A. (1989), The Time-frequency Characteristics of
Quarry Blasts and Calibration Explosions Recorded in Kazakhstan,U.S.S.R., Geophys. J. Int. 99, 109±
121.
ISRAELSSON, H. (1991), Correlation of Waveforms from Closely Spaced Regional Events, Bull. Seismol. Soc.
Am. 80, 6, 2177±2193.
KHALTURIN, V. I., RAUTIAN, T. G., RICHARDS, P. G., and KIM, W. Y. (1997), Evaluation of Chemical
Explosions and Methods of Discrimination for Practical Seismic Monitoring of a CTBT, Phillips Lab.
Final Report.
KIM, W. Y., SIMPSON, D. W., and RICHARDS, P. G. (1994), High-frequency Spectra of Regional Phases from
Earthquakes and Chemical Explosions,Bull. Seismol. Soc. Am. 84, 1365±1386.
LANGEFORS, U. and KIHLSTRO
ÈM,B.The Modern Technique of Rock Blasting (Halsted Press, Wiley, New
York. 1978).
LEITH, W. (1994), Large Chemical Explosions in the Former Soviet Union and Blasting Estimates for
Countries of Nuclear Proliferation Concern, Arms Control and Nonproliferation Technologies, 1st
quarter 1994, 25.
MARTIN, R., GROSS, D., PEARSON, C., STUMP, B., and ANDERSON, D. (1997), Black Thunder Coal Mine and Los
Alamos Experimental Study of Seismic Energy Generated by Large-scale Mine Blasting, Proc. Twenty-third
Ann. Conf. on Explosives and BlastingTechnique, Las Vegas, Nevada, Feb. 2±5, 1997, International Society
of Explosive Engineers, Cleveland, Ohio, pp. 1±10.
MARTIN, R. L., and KING, M. G. (1995), The Eciency of Cast Blasting in Wide Pits, Proc. Twenty-®rst
Conf. on Explosives and Blasting Technique, Nashville, Tennessee, Feb. 5±9, 1995, International Society
of Explosive Engineers, Cleveland, Ohio, pp. 176±186.
MCLAUGHLIN, K. L., BARKER, T. G., STEVENS, J. L., and DAY, S. M. (1994), Numerical Simulation of
Quarry Blast Sources, PL Final Report SSS-FR-94-14418.
MINSTER, J.-B. and DAY, S. (1986), Decay of Wave®elds near an Explosive Source due to High-strain,
Nonlinear Attenuation, J. Geophys. Res. 91, 2113±2122.
MURPHY, J. R. (1995), Types of Seismic Events and their Source Description,Monitoring a Comprehensive
Test Ban Treaty, NATO ASI Series E: Applied Sciences, Vol. 303, (E. Husebye and A. Dainty ed.) 225±
245.
PEARSON, D. C., STUMP, B. W., BAKER, D. F., and EDWARDS, C. L. (1995), The LANL/LLNL/AFTAC
Black Thunder Mine Regional Mining Blast Experiment, 17th Annual PL/AFOSR/AFTAC/DOE Seism.
Res. Symp., 562±571.
PRODEHL, C. (1979), Crustal Structure of the Western United States, US. Geological Survey Professional
Paper, 1034.
RICHARDS, P. G. and ZAVALES, J. (1990), Seismic Discrimination of Nuclear Explosions, Annu. Rev. Earth
Planet. Sci. 18, 257±286.
RIVIERE-BARBIER, F. and GRANT, L. T. (1993). Identi®cation and Location of Closely Spaced Mining
Events, Bull. Seismol. Soc. Am. 83, 1527±1546.
SMITH, A. T. (1993), Discrimination of Explosions from Simultaneous Mining Blasts, Bull. Seismol. Soc.
Am. 83, 160±179.
862 Michael A. H. Hedlin et al. Pure appl. geophys.,
SOBEL, P. A. (1978), The Eect of Spall on m
b
and M
s
, Teledyne Geotech Rept., SDAC-TR-77-12, Dallas,
TX.
SORRELLS, G. G., HERRIN, E. T., and BONNER, J. L. (1997), Construction of Regional Ground-truth
Databases Using Seismic and Infrasound Data, Seism. Res. Lett. 68, 743±752.
STUMP, B. W. (1995), Practical Observations of U.S. Mining Practices and Implications for CTBT
Monitoring, Phillips Lab Report, PL-TR-95-2108.
STUMP, B. W. and REAMER, S. K. (1988), Temporal and Spatial Source Eects from Near-surface
Explosions, 10th Annual AFGS/DARPA Seism. Res. Symp., 95±113.
STUMP, G. W., RIVIERE-BARBIER, F., CHERNOBY, I., and KOCH, K. (1994), Monitoring a Test-ban Treaty
Presents Scienti®c Challenges, EOS, Trans. of the Am. Geophys. Union 75, 265.
STUMP, B. W., ANDERSON, D. P., and PEARSON, D. C. (1996), Physical Constraints on Mining Explosions:
Synergy of Seismic and Video Data with 3-D Models,Seismol. Res. Lett. 67, 9±24.
STUMP, B. W. and PEARSON, D. C. (1997), Comparison of Single-®red and Delay-®red Explosions at
Regional and Local Distances, 19th Ann. Seism. Res. Symp., 668±677.
STUMP, B. W., HEDLIN, M. A. H., PEARSON, D. C., and HSU, V. (1999a), Characterization of Mining
Explosions at Regional Distances, Rev. Geophys., in Review.
STUMP, B. W., PEARSON, D. C., and HSU, V. (1999b), Empirical Scaling Relations for Contained Single-®red
Chemical Explosions and Delay-®red Mining Explosions at Regional Distances, 1999 SSA Meeting.
STUMP, B. W., PEARSON, D. C., and HSU, V. (1999c), Empirical Scaling Relations for Contained Single-®red
Chemical Explosions and Delay-®red Mining Explosions at Regional Distances, 21st Ann Seism. Res.
Symp., 764±772.
STUMP, B. W. and REAMER, S. K. (1988), Temporal and Spatial Source Eects from Near-surface
Explosions, 10th Ann. AFGS/DARPA Seism. Res. Symp. 95, 113.
U.S. CONGRESS,OFFICE OFTECHNOLOGY ASSESSMENT (1988), Seismic Veri®cation of Nuclear Testing
Treaties, OTA-ISC-361 (Washington, DC: U.S. Government Printing Oce, May, 1988).
WUSTER, J., RIVIERE, F., CRUSEM, R., PLANTET, J. -L., MASSINON, B., and CARISTAN, Y. (2000), GSETT-
3: Evaluation of the Detection and Location Capabilities of an Experimental Global Seismic Monitoring
System, Bull. Seism. Soc. Am. 90, 166±186.
YANG, X. (1998), MineSeis ± A Matlab GUI Program to Calculate Synthetic Seismograms from a Linear,
Multi-Shot Blast Source Model, 20th Ann. Seism. Res. Symp. 755±764.
(Received June 15, 1999, revised June 4, 2000, accepted June 15, 2000)
To access this journal online:
http://www.birkhauser.ch
Vol. 159, 2002 Identi®cation of Mining Blasts 863