S11B-1730
Nuclear Explosion Monitoring History and Research and Development
Within a year after the nuclear detonations over Hiroshima and Nagasaki the Baruch Plan was presented to the newly formed United Nations Atomic Energy Commission (June 14, 1946) to establish nuclear disarmament and international control over all nuclear activities. These controls would allow only the peaceful use of atomic energy. The plan was rejected through a Security Council veto primarily because of the resistance to unlimited inspections. Since that time there have been many multilateral, and bilateral agreements, and unilateral declarations to limit or eliminate nuclear detonations. Almost all of theses agreements (i.e. treaties) call for some type of monitoring. We will review a timeline showing the history of nuclear testing and the more important treaties. We will also describe testing operations, containment, phenomenology, and observations. The Comprehensive Nuclear Test Ban Treaty (CTBT) which has been signed by 179 countries (ratified by 144) established the International Monitoring System global verification regime which employs seismic, infrasound, hydroacoustic and radionuclide monitoring techniques. The CTBT also includes on-site inspection to clarify whether a nuclear explosion has been carried out in violation of the Treaty. The US Department of Energy (DOE) through its National Nuclear Security Agency's Ground-Based Nuclear Explosion Monitoring R&D Program supports research by US National Laboratories, and universities and industry internationally to detect, locate, and identify nuclear detonations. This research program builds on the broad base of monitoring expertise developed over several decades. Annually the DOE and the US Department of Defense jointly solicit monitoring research proposals. Areas of research include: seismic regional characterization and wave propagation, seismic event detection and location, seismic identification and source characterization, hydroacoustic monitoring, radionuclide monitoring, infrasound monitoring, and data processing and analysis. Reports from the selected research projects are published in the proceedings of the annual Monitoring Research Review conference.
S11B-1731
Infrasound Sensor and Porous-Hose Filter Characterization Results
The Ground-Based Nuclear Explosion Monitoring Research and Development (GNEM R&D) program at Sandia National Laboratories (SNL) is regarded as the primary center for unbiased expertise in testing and evaluation of geophysical sensors and instrumentation for nuclear explosion monitoring. Over the past year much of our work has focused in the area of infrasound sensor characterization through the continuing development of an infrasound sensor characterization test-bed. Our main areas of focus have been in new sensor characterization and understanding the effects of porous-hose filters for reducing acoustic background signals. Three infrasound sensors were evaluated for characteristics of instrument response, linearity and self-noise. The sensors tested were Chaparral Physics model 2.5 low-gain, New Mexico Tech All-Sensor and the Inter-Mountain Labs model SS avalanche sensor. For the infrasound sensors tested, the test results allow us to conclude that two of the three sensors had sufficiently quiet noise floor to be at or below the Acoustic low-noise model from 0.1 to 7 Hz, which make those sensors suitable to explosion monitoring. The other area of focus has been to understand the characteristics of porous-hose filters used at some monitoring sites. For this, an experiment was designed in which two infrasound sensors were co- located. One sensor was connected to a typical porous-hose spatial filter consisting of eight individual hoses covering a 30m aperture and the second sensor was left open to unimpeded acoustic input. Data were collected for several days, power spectrum computed for two-hour windows and the relative gain of the porous-hose filters were estimated by dividing the power spectrum. The porous-hose filter appears to attenuate less than 3 dB (rel 1 Pa**2/Hz) below 0.1 Hz and as much as 25 dB at 1 Hz and between 20 to 10 dB above 10 Hz. Several more experiments will be designed to address the effects of different characteristics of the individual porous-hoses, such as length, number and geometric arrangement. This work directly impacts the Ground-Based Nuclear Explosion Monitoring mission by providing a facility, equipment, and personnel to give the operational monitoring agencies confidence in deployed instrumentation and capability for mission success.
S11B-1732
Directional microphone arrays: Reducing wind noise without killing your signal or filling up your disk
The bane of infrasound signal detection is the noise generated by the wind. While the physics of the noise is
still a subject of investigation, it is clear that sampling pressure at many points over a length scale larger
than the spatial coherence length of wind turbulence attenuates the noise. A dense array of microphones
can exploit this approach, but this presents different challenges. Two mechanical wind filters using this
approach are commonly employed by the nuclear monitoring community (rosette pipe and porous-hoses
networks) and attach to a central microphone. To get large wind noise reduction and a low signal detection
threshold in the frequency band of interest, these filters require large apertures. However, these wind filters
with such large apertures have a poor omnidirectional instrument response for typical infrasound signals
because the pressure signal propagates at the speed of sound through the pipes/hoses to the central
microphone. A simple, but improved averaging approach would be to instantaneously sample a long length
of the infrasound signal wavefront. Optical fiber infrasound sensors (OFIS) are an implementation of this
idea. These sensors are compliant sealed tubes wrapped with two optical fibers that integrate pressure
change instantaneously along the length of the tube with laser interferometery.
Infrasound arrays typically consist of several microbarometers with wind filters separated by distances that
provide predictable signal time separations, forming the basis for processing techniques that estimate the
phase velocity direction. An analogous approach is to form an array of OFIS arms. The OFIS instrument
response is a predictable function of the orientation of the arm with respect to the signal wavefront. An array
of many OFIS arms with different azimuths permits at least one OFIS to record any signal without the signal
attenuation inherent in equivalently-sized onmi-directional mechanical filters. OFIS arms that are wavefront-
orthogonal have a very similar instrument response to that of the mechanical filters of the same size.
We show that although an OFIS array is an improved alternative to an individual rosette pipe filter, one can
also use a single OFIS array as an infrasound station where space is at a premium. Synthetic and real
infrasound data recorded by an OFIS array and a collocated pipe array at the Piñon Flat Observatory
indicate that the instrument response of an OFIS can be exploited with multiple arms to determine the signal
direction. Three types of OFIS instrument-response-dependent beamforming and an array deconvolution
technique are evaluated. The results suggest that an array of five radial OFIS arms with equal azimuthal
separation and an aperture that depends on the frequency band of interest provides excellent directional
resolution while requiring less space than an equally effective array of five microphones with rosette wind
filters. A five-arm OFIS array with an aperture of ~150 m is being compared with a collocated porous
hose array at our Camp Elliott field site in San Diego county.
http://sail.ucsd.edu/~ofis/
S11B-1733
Infrasound monitoring in Kazakhstan: source localization and characterization
The IMS infrasound station (IS31, Aktyubinsk) was installed in 2001 years ago in the northwest of the Republic of Kazakhstan. An automatic detection procedure has been implemented for the last three years. PMCC is the basis of the detection algorithm used which is able to routinely extract infrasound signals across the array within non-coherent noise. Sources of signals have been identified but can not be located using one single station. To do so, seismo-acoustic studies were carried out using the Akbulak seismic array. They allow a firm identification of some regional sources being detected on both arrays. The analyses of satellite views surrounding the station and the characteristics of the detected signals suggest that one prominent source is the Zhanazhol oil and gas field. With the support of PTS/CTBTO and CEA/DASE, a temporary experimental infrasound array, collocated with the Akbulak seismic array, has been installed in 2007. This experiment confirmed the location of this source. Moreover, the analysis of the automatic processing results shows clear seasonal trends in the detected backazimuth and slowness. Sine variations of the azimuth deviation are recorded within a range of 15°. Such variations are explained using a ray tracing method combined with a climatological atmospheric model. Other regional sources are also localized using both IS31 and the Akbulak infrasound arrays, such oil and gas fields, open mines, industrial explosions or earthquakes. Such studies would help to build up experience and discriminate between environmental noise and events of interest.
S11B-1734
Dispersed Infrasound Signals in the "Zone of Silence"
During the last few years the use of seismo-acoustic recordings have become increasingly important; however, infrasound analysis methods lag behind seismic methods. The current paper discusses infrasonic signals in the so called 'Zone of Silence', at a distance up to 300 km from the source. During controlled source experiments in 2006 and 2007, tropospheric, stratospheric and thermospheric signals were recorded at a suite of temporary infrasound arrays; some of the tropospheric arrivals exhibit dispersion. The most common types of infrasonic signals observed during our studies beyond 76 km are stratospheric. These signals are not predicted by atmospheric modeling (raytracing and PE calculations) using the Naval Research Laboratory Ground to Space (G2S) model. The G2S model does not explain the observed tropospheric arrivals. Meteorological data from balloon launched rawinsondes obtained in the path of the propagating signals are able to predict the tropospheric propagation if the data closest to the detonation time is used. In a previous research study in the China Sea, a suite of dispersed infrasound signals were successfully interpreted as propagating in a low velocity waveguide. We also observed dispersed signals from an explosion at White Sands Missile Range recorded at TXIAR at a distance of 546 km. The signals exhibit dispersion between 0.2-1 Hz, and can be successfully modeled as propagating in a low velocity layer 1.2 km thick. However, dispersed signals recorded in Nevada suggest propagation in a waveguide consisting of a single low velocity layer is too simplistic. Effects due to lateral changes in effective sound speed, topography, turbulence and multiple layering can significantly affect the dispersion.
S11B-1735
Misty Picture: A Unique Experiment for the Interpretation of the Infrasound Propagation from Large Explosive Sources
In the framework of the Comprehensive Nuclear-Test-Ban Treaty, the International Monitoring System develops a 60 micro-barometric stations network. These stations, which records infrasound, detect various powerful natural and artificial sources like long range explosions, oceanic swell, and volcano eruptions. The Misty Picture experiment is a high explosive event (4685 Tons of ANFO) realized in 1987 in New Mexico (US). Infrasound waves were recorded by an amount of 22 sensors installed by the Sandia National Laboratories (J.W. Reed et al., 1977, SAND--87-2978C), the Los Alamos National Laboratories (R.W. Whitaker et al. 1990, 4th LRSP) and the CEA (E. Blanc, 1998, CEA). Multi-reflected tropospheric, stratospheric and thermospheric phases are detected until a distance of 1000 km in a quiet background noise condition. Signals recorded near the source (1 km away) and observed in the geometrical shadow zone (between 150 km and 250 km) are of particular interest. This reference experiment is used to improve our understanding of the atmospheric propagation of infrasound as well as to evaluate our models. Using various methods such as ray tracing and parabolic equation, we investigate effects of the wind, atmospheric absorption, nonlinearity, refraction and scattering by small atmospheric scales on observed phase kinds, their travel time and their waveform.
S11B-1736
Infrasound monitoring, acoustic-gravity waves and global atmospheric dynamics
For the verification of the Comprehensive nuclear Test Ban Treaty, the International Monitoring System has been developed. As part of this system, the infrasound network provides an unique opportunity to monitor continuously pressure waves in the atmosphere. Such infrasonic waves propagate in the channel formed by the temperature and wind gradients of the atmosphere. Long term observations provide information about the evolution of the propagation conditions and then of atmospheric parameters. The monitoring of continuous sources, as ocean swell, gives the characteristics of the stratospheric wave channel submitted to stratospheric warming effects. Large scale gravity waves, which are also observed by the network, produce a forcing of the stratosphere at low and middle latitudes and long-lived changes in the stratospheric circulation towards high latitudes, leading to fluctuations in the strength of the polar vortex. These fluctuations move down to the lower stratosphere with possible effects on the tropospheric temperature. Gravity wave monitoring in Antarctica reveals a gravity wave system probably related to the wind effect over mountains. At mid latitudes an additional main sources of disturbances is the thunderstorm activity. The infrasound monitoring system allows a better knowledge of the atmospheric wave systems and of the dynamics of the atmosphere. In return this better knowledge of the wave systems allow a better identification of the possible explosion signals in the background of the atmospheric waves and then to improve the discrimination methods
S11B-1737
Detection, Location, and Characterization of Hydroacoustic Signals Using Seafloor Cable Networks Offshore Japan
The hydroacoustic monitoring by the International Monitoring System for CTBT (Comprehensive Nuclear- Test-Ban Treaty) verification system utilizes hydrophone stations (6) and seismic stations (5 and called T- phase stations) for worldwide detection. Some conspicuous signals of natural origin include those from earthquakes, volcanic eruptions, or whale calls. Among artificial sources are non-nuclear explosions and airgun shots. It is important for the IMS system to detect and locate hydroacoustic events with sufficient accuracy and correctly characterize the signals and identify the source. As there are a number of seafloor cable networks operated offshore Japanese islands basically facing the Pacific Ocean for monitoring regional seismicity, the data from these stations (pressure and seismic sensors) may be utilized to increase the capability of IMS. We use these data to compare some selected event parameters with those by IMS. In particular, there have been several unconventional acoustic signals in the western Pacific,which were also captured by IMS hydrophones across the Pacific in the time period of 2007-present. These anomalous examples and also dynamite shots used for seismic crustal structure studies and other natural sources will be presented in order to help improve the IMS verification capabilities for detection, location and characterization of anomalous signals.
S11B-1738
Implications for mb-Ms Discrimination Based on a New Long-Period Explosion Source Model with Shock-Induced Tensile Failure
A major challenge of low-magnitude nuclear test monitoring is the sheer numbers of events to be identified,
requiring the most reliable discriminants possible. mb-Ms has proven to be one of the best at teleseismic
distances, and regional-distance extensions are being pursued vigorously for applications to small events.
Nevertheless, conundrums have surrounded the discriminant for many years, the departure of mb-Ms
observations from 1-to-1 scaling being one example for which explosion source models fail to explain. We
present a new model involving tensile failure at depth, in contrast to classical spallation at the Earth's
surface. Shock-induced, deep-seated tensile failure is manifested as vertical block motions in extension due
to synergy between the dynamics of cavity rebound and gravitational unloading of spalled surface layers.
The body-force representation of this source is assumed to be a compensated linear vector dipole (CLVD),
its relative strength given by an index K defined as a ratio between diagonal elements of the moment tensor,
2Mzz/(Mxx+Myy). Rayleigh waves excited by a CLVD with vertical axis of symmetry destructively interfere with
waves from the explosion monopole, reducing Ms for 1 S11B-1739 Analyses and Simulation of 3D Scattering due to Heterogeneous Crustal Structure and Surface Topography on Regional Phases; Seismic Source Discrimination
The main purpose of this study is the understanding of scattering and its effects through step-by-step
numerical experiments and waveform modeling, with the goal of providing useful insights into ongoing
research for development of simple empirical models of scattering that can be used in reducing the scatter in
measures of Lg and coda wave magnitude, and for discrimination purpose as well. We are performing
anelastic 3D finite-difference simulations of wave propagation in highly heterogeneous media for a range of
source depths, receiver distances and source types. In a first stage of our study we investigated the effect of
small-scale crustal heterogeneities on wave propagation scattering. During the second stage we
investigated the effect of surface topography combined with crustal heterogeneities on wave propagation
scattering. In our simulations of regional wave propagation we used a finite difference computer program
using our computer cluster. The seismograms were calculated up to 3.5 Hz for regional distances up to 300
km. The topography elevation was simulated using correlated random variations along the free surface. Our
simulations show that the surface topography increases the wave-path scattering effects. The combined
effects of crustal heterogeneities and surface topography produce Lg, P and S coda waves with significant
energy even for explosion sources. The energy of Lg coda waves depends on the source depth. Numerical
experiments with explosion-type sources and earthquake-type sources show that P/Lg ratios estimated at
different frequencies indicate that this ratio could be a good discriminant between explosions and
earthquakes when calculated at high frequencies. The wave-path scattering effects were also investigated by
simulating observed high frequency source directivity effects, as well as fault zone wave trapping in highly
fractured conditions from aftershocks of the Big Bear earthquake sequence.
S11B-1740 On the Robust Estimate of Lg Source Spectral Parameters
It is well known that the widely used stochastic modeling of Lg spectra
contains error owing to its oversimplification. We have suggested in the
past that, with such errors, a simultaneous estimation of Lg source
spectral and path Q parameters using a formal inverse algorithm can be
subject to significant errors and parameter trade-off (Xie, 1998, BSSA).
We proposed a Bayesian algorithm to reduce the errors and constrain the
trade-off, which essentially constrains the inversion by using a priori
knowledge of path Q and source spectra from two-station and empirical
Green's function (EGF) analyses. We have found recently that the
modeling error is more substantial and complex in the observed Lg
spectra. For example, Lg radiated from the source seems to vary with
path, resulting not only from a radiation pattern, but from the source
finiteness that is not well captured by the Brune model. We have tested
the robustness of Lg source spectral estimates with some Eurasian
earthquakes for which we possess a priori knowledge on both source
corner frequencies and lateral variations of Lg Q. A simultaneous
inversion of source spectral and path-variable Q parameters typically
fails unless negative Q values are allowed. On the other hand, we always
obtain reasonable estimates of Lg source spectral parameters if a priori
knowledge on path-averaged 1 Hz Lg Q is used, leaving only averaged
frequency dependence of Lg Q as the unknown path parameter, together
with the unknown source parameters. The a priori knowledge on source
corner frequency from the EGF method is not required for these
estimates, although it can always be used to verify the corner frequency
estimates. We are analyzing more data to confirm this finding, and to
explore how the a priori tomographic Q model can be improved by using
many path-averaged Q measurements.
S11B-1741 Tomographic Inversion for Regional Phase Attenuation and Source Parameters
Our ability to monitor seismic events that are too small to be observed at teleseismic distances depends
critically on the accurate characterization of the effects of source, path and site on regional phases.
Magnitude, yield and event identification procedures rely on path and site corrections, while identification
schemes such as MDAC (magnitude and distance amplitude correction) additionally rely on source correction
via a regionally appropriate scaling model. Independently determined Mw can drive the source correction.
Here, we focus on the use of Lg amplitudes to obtain a laterally varying attenuation model with power law
frequency dependence, site amplification terms, and source scaling parameters such as apparent stress,
using data from the USArray deployment. We collected amplitudes from broadband vertical channels of 605
USArray stations for 986 western US PDE events through June 2008. The IRIS DMC provided waveform data
and instrument response information. We measured RMS amplitudes in eleven overlapping octave width
bands between 0.25 and 16 Hz, for windows defined by group velocities 3.6 to 3.0 km/s. Over 320,000
amplitudes passed a pre-phase signal-to-noise threshold of 2. Berkeley moments were used to set absolute
levels. We found 1-Hz attenuation (Qo) to be high in stable regions and across batholiths, and low in areas of
recent volcanism. The power law exponent (eta) varied from 0.4 to 0.9 and was generally lower in the high Q
regions. We explore methods that break the tradeoff between attenuation and stress by damping mean Qo
and eta to values determined by inverting for frequency dependent source terms, and by applying ground
truth scaling information from relative coda studies for particular events. Tests that evaluate regionalization of
apparent stress are underway, although we acknowledge that the limited bandwidth and deployment period
are not ideal for this purpose.
S11B-1742 Synthetic Seismograms for CLVD and Tensile Cracks Using Green's Functions Due to Single Force, Force Dipole and Single-Couple Sources
In this paper, the frequency-wavenumber integration and surface-wave generation algorithms are modified to
compute Green's functions from single force, single dipole and single-couple sources. Using source
coefficients for these elementary point sources from Haskell (1964), a modified source matrix Σ is
constructed where Σ = E S [see Haskell (1964) for explicit expresssions of E and S matrices] to
propagate the source wavefield discontinuity through a multilayered earth model. These Green's functions
are then used to generate waveforms from other source such as the compensated linear vector dipole
(CLVD) and tensile cracks. In an orthogonal co-ordinate system, since a CLVD source represents a
compression dipole of moment 2 along one axis and two dilatational dipoles of unit moment along the two
remaining axes (Ben-Menahem and Singh, 1980) and likewise representing the tensile cracks based on
defintion from Day and McLaughlin (1991), these Green's function are used to generate synthetic
seismograms for these individual sources. These synthetic waveforms have been compared with those
synthetic seismograms generated for the three orthogonal dipoles in the moment-tensor matrix using source
coefficients of a double-couple point source without moment at various depths. The CLVD, tensile crack,
double-couple and compressional synthetic seismograms are also used to invert regional seismograms to
estimate contributions of these indivdual mechanisms, thereby establishing the complexity in some known
seismic sources. Singular value decomposition (SVD) is implemented while inverting for these source
contributions, and includes estimations of errors in the solution vector at each depth for a given earth model.
S11B-1743 Advances in Three-Dimensional Modeling and High-Performance Computing to Improve Nuclear Explosion Phenomenology and Monitoring
Detailed modeling of the physical effects of nuclear explosions has been a subject of intense investigation at
the U.S. national laboratories since their inception. Recent advances in numerical methods for wave
propagation and the inexorable growth in computational power offer new opportunities to improve the fidelity
of simulations of seismic wave phenomena for the purposes of improving physics-based understanding and
monitoring of nuclear explosions. This presentation will summarize new modeling capabilities developed to
improve understanding of the seismic waves emerging from an explosion. These capabilities allow
representation of three-dimensional (3D) variations in physical properties of the earth and rely on the world-
class computational resources available at LLNL. Specifically we are working in three thrust areas: 1)
computation of regional distance intermediate-period (>5 seconds) synthetic seismograms in 3D earth
models to assess the ability of these models to predict observed seismograms; 2) coupling of non-linear
hydrodynamic simulations of explosion shock waves with an anelastic finite difference code for modeling the
dependence of seismic wave observables on explosion emplacement conditions and near-source
heterogeneity; and 3) simulation of surface topography in our anelastic finite difference code to include
scattering and mode-conversion due to a non-planar free surface.
S11B-1744 A Global Model and Methods for Accurate Real-Time Calculation of Regional-Phase Travel Times
We improve on existing methods of operational, regional travel time calculation by capturing 1st-order effects
of 3-dimensional earth structure. A seamless global tessellation of nodes with spacing of approximately
1° provides global support. Three-dimensional structure is rendered by interpolation of velocity profiles
at each node. The velocity profile at each node is a 7-layer crust, mantle velocity at the Moho, and a linear
velocity gradient below the Moho. The linear gradient parameterization in the mantle leads to an analytical
approximation for the diving Pn/Sn ray that results in ~1 millisecond computation time for travel times out to
far regional distance. We improve prediction across Eurasia and North Africa using a tomographic
formulation that adjusts the average crustal velocity, Pn/Sn velocity, and the mantle gradient at each node.
Pn tomography reduces travel-time variance by 78%. In location tests, epicenter accuracy is improved by an
average of 30% for regional network geometries with large azimuthal gaps. Regional travel time residuals
relative to our model are zero mean when event origin times are constrained by teleseismic (ak135) P-wave
arrivals, so our model improves location accuracy when regional and teleseismic data are combined.
Prepared by LLNL under Contract DE-AC52-07NA27344.
S11B-1745 The 9 October 2006 North Korean event: full moment tensor solution and seismic discrimination
On 9 October 2006 an underground nuclear explosion took place at North Korea's Chik-Tong nuclear test
site, in the Hamgyeong Province. The explosive signatures of this seismic event were captured over a broad
frequency band at near-regional and regional distances (150 - 350 km). We present a full moment tensor
solution for this nuclear test and compare its signatures with those produced by natural earthquakes and
chemical explosions. The velocity model we used for the moment tensor inversion is a modified version of
what had been used by the Jilin Provincial Seismological Bureau for location purposes. The modifications
were based on minimization of the travel-time residuals. We used a total of 10 stations and tested different
frequency bands for the moment tensor inversion. We achieved the highest variance reduction of 62.2% at
the frequency band of 0.06 to 0.1 Hz. At frequencies below 0.06 Hz, low-frequency noise overwhelms the
signal. The inverted Mw is 3.8, which is at the lower end of the estimated magnitude range of 3.8 to 4.3. With
72% isotropic component, but only 22% for the compensated linear vector dipole, and 6% for the double
couple, the moment tensor solution is highly consistent with that of an underground explosion. For
comparison, we also studied a similar-sized earthquake (16 December 2004), which was well recorded by our
stations. Of the two, the earthquake is much closer to our seismic network than is the North Korea nuclear
event. The full moment tensor solution with the greatest variance reduction (66.7% with five available
stations) is also achieved in the 0.06-0.10 Hz frequency band. The Mw in this case is 3.9. The solution, being
characterized by 37% double couple, 24% compensated linear vector dipole, and 39% isotropic, is
ostensibly not that of a typical earthquake. This might be related with the fact that this earthquake is located
near the Changbaishan Volcano.
S11B-1746 A Tale of Two Countries, Three Discriminants, and Many Mining Explosions
With the advent of not only the IMS, but the contribution of high-quality stations and networks throughout the
globe, nuclear explosion monitoring strategies have shifted from teleseismic to regional scales. At regional
distances, the event detection threshold is greatly decreased, meaning mining explosions must be included in
the event identification process. The addition of this event class leads to several questions which much be
addressed in the discrimination process: (1) can delay-fired mining explosions be discriminated from
earthquakes; (2) can the understanding of mining discriminants illuminate the explosion discrimination
process in regions where there are no known nuclear explosions, and (3) are mining discriminants regionally
dependent, and are there corrections to account for this dependence?
To address these questions, we have assembled a database of ~2500 mining explosions and ~1300
earthquakes recorded at a number of regional stations, comprising a total of ~150,000 waveforms. The
database is focused on two regions, the Western United States (WUS) and the Altai-Sayan (AS) region of
Russia, which are both areas of prolific mining activity. Due to extensive collaboration with the largest coal
mine in Wyoming and the nation, we have detailed shot information for ~1000 mining events, classified into
six distinct blast types. We have limited information for events in the AS based upon contacts with the AS
Seismological Expedition.
We have applied 3 discriminants to data from 11 stations and 1 array in the WUS. The first discriminant, time-
of-day (TOD), assesses the event origin time. Mining events occur between 9 am and 6 pm, while
earthquakes are randomly distributed in time. The second discriminant, spectral ratios (SR), exploits
differences between regional phases due to source type. Results are highly station centric, although the
largest mining events separate from earthquakes that are < 250 km from the mine; as the earthquake
dataset expands spatially, discrimination performance degrades. 1D path corrections provide improvement,
but additional calibrations are necessary to optimize this discriminant. The third discriminant, time-frequency
(TF), capitalizes on the unique spectral signature of delay-fired mining events as a function of time. This
discriminant separates the larger types of blasts with the longest source duration at all stations. Smaller
blasts do not discriminate because of the shorter shot durations. We use classification trees to combine
these discriminants, which yield an intuitive decision grid for classifying events based on the discriminant
value for each of the 3 discriminants.
In the AS region, we calculated these same discriminants at for ~260 earthquakes and ~850 mining events.
TOD results are similar to the WUS in that presumed mining events fall within working hours. The SR
discriminant shows significant overlap of the earthquake and mining populations. Certain events do separate,
but we have no information on what makes these events unique. Similar results are seen for the TF
discriminant. We do not know if the discriminant itself fails, or if the majority of our data points are from
smaller shots that have shorter time durations. These unanswered questions illustrate the need for detailed
ground-truth information. Future studies of mining discrimination, particularly where large datasets are to be
acquired, should involve cooperation with mine operators in order to address ambiguities such as those
identified in the AS study.