high-frequency spectra of regional phases from earthquakes...

Bulletin of the Seismological Society of America, Vol. 84, No. 5, pp. 1365-1386, October 1994

High-Frequency Spectra of Regional Phases from Earthquakes and Chemical Explosions

by W. Y. K i m , D. W. S i m p s o n , 1 and P. G. R ichards

Abstract We analyze the high-frequency (1 to 50 Hz) spectra of chemical explosions and earthquakes at local and regional distances in the northeastern United States and in Norway to understand the seismic signal characteristics of single explosions, multiple-hole instantaneous explosions, ripple-fired quarry blasts, and earthquakes. Our purpose is to evaluate practical discriminants, and to obtain a physical understanding of their successes and failures.

High-frequency spectra f rom tipple-fired blasts usually show clear time- independent frequency bands due to the repetitive nature of the source and are distinctively different from the spectra of instantaneous blasts or earthquakes. However, like other discriminators based on spectral estimates, the spectrogram method requires data with high signal-to-noise ratios at high frequencies for unambiguous discrimination. In addition, banding is not seen in spectrograms for shots with small delay times (less than 8 msec) and short total durations.

We have successfully modeled the observed high-frequency spectral bands up to about 45 Hz of the regional signals f rom quarry blasts in New York and adjacent states. Using information on shot-hole patterns and charge distribution, we find that ripple firing results in an enrichment of high-frequency S waves and efficient excitation of the Rg phase. There is an azimuthal dependence of P-wave amplitude associated with orientation of the path with respect to local topography (ridges, benches) in which the shots are emplaced.

To discriminate instantaneous explosions f rom earthquakes, we find the P/S spectral amplitude ratio at high frequencies is complementary to the use of spectrogram methods. A high P/S spectral ratio above 10 Hz is a stable characteristic of instantaneous explosions.

Introduction

The high-frequency seismic spectra of characteristic crustal phases, such as Pn, Pg, and Lg, on regional records provide important data for determining crustal structure and for studying seismic source properties. During the Advanced Research Project Agency's project "VELA UNIFORM" in the early 1960s, many researchers studied seismic records from chemical explosions and earthquakes (e.g., Frantti, 1963; Pollack, 1963; Willis, 1963). Most of these earlier works were based on seismic signal below 10 Hz.

Recently, high-frequency spectra (up to 30 to 40 Hz) from regional events have been advocated as a crucial tool in seismic verification of low-yield coupled and fully

1Present address: Incorporated Research Institutions for Seismol- ogy, 1616 N. Fort Myer Drive, Suite 1050, Arlington, Virginia 22209.

decoupled underground nuclear explosions (e.g., Evem- den et al., 1986). The possibility of new limitations on nuclear testing, which would require monitoring for nuclear explosions small enough to be comparable with large industrial explosions, has renewed interest among seis- mologists in the seismic signals from chemical explosions, particularly from ripple-fired explosions. Data from recently installed high-quality, high-frequency seismo- graph stations and networks now make it possible to investigate whether high frequencies can improve the ability to discriminate between different types of explosions and earthquakes.

We present spectrograms of ground velocity recorded at regional distances from ripple-fired quarry blasts, from single-hole explosions, and from earthquakes. Our goal is to contribute to the evaluation of spectrograms as a basis for discriminating between these three types of seismic sources, as well as to the improved understand-

1365

1366 W.Y. Kim, D. W. Simpson, and P. G. Richards

ing of high-frequency seismic signals from these types of sources.

We recognize that a number of special studies have pointed out the utility of spectra and spectrograms for discrimination, including the occurrence of time-independent bands in spectrograms (e.g., Greenhalgh, 1980; Baumgardt and Ziegler, 1988; Smith, 1989; Hedlin et al., 1989, 1990). We have sought to evaluate problems that arise in applying spectrogram methods of discrimination to seismic data routinely acquired in Norway and in the northeastern United States. In this way, we are able to report on a variety of practical experiences and can point out where spectrograms can provide effective discriminants, and when problems may arise. Thus, we are interested in (1) the effect of different signal-to-noise ratios; (2) the effect of different delay times (for ripple firing), compared to the sampling interval (for recording seismic motion); and (3) consequences of different geological conditions, especially in the vicinity of the source. An advantage of spectrogram methods over other tech- niques is the use of the complete regional seismogram trace rather than isolated phases, such as Pn, Pg, and Lg. However, like other discriminators based on spectral estimates, the method requires data with high signal-to- noise ratios, especially at high frequencies.

We find that fairly good discrimination capability can be demonstrated in several different regions, with a variety of delay time patterns and diverse geological settings. However, for purposes of evaluating the method and as an essential part of delineating where it works well, it is necessary to pay particular attention to failures. In some cases, these failures will provide a guide to what new types of data may be needed for more successful discrimination.

In sections that follow, we describe applications of the spectrogram technique to regional seismic waves recorded from chemical explosions in two different parts of the world: northeastern United States and Norway. For the northeastern United States, we show spectrograms of ripple-fired and single-hole explosions, and earthquakes, obtained from standard recordings acquired by the New York State Seismic Network. The discrimination power of P/S spectral amplitude ratios is examined for classifying earthquakes from chemical explosions and for complementing the spectrogram method. A final section discusses the underlying causes of suc- cess and failure of the spectrogram method of discrimination.

A Note on Terminology. We use P wave to denote all first-arrival P waves on the records with a group velocity of greater than about 4 km/sec without further classi- fication, and likewise S wave is used to denote all S waves arriving with group velocities of about 3.6 km/sec. Fre- quency content is used to indicate that there is substantial energy above background level in the frequency band of

interest. The term time-dependent spectral peaks is used to describe the typical earthquake speclrogram with peaks of high amplitude, limited in temporal extent and associated with the arrival of characteristic phases such as P and S. Time-independent spectral bands are prominent spectral bands, limited in frequency content, that extend throughout the duration of the seismogram.

Chemical explosions may be classified into single explosions, multiple-hole instantaneous explosions, and ripple-fired explosions, depending upon the time delays and shooting patterns used. Multiple-hole instantaneous explosions are distributed single explosions designed to be detonated within a very short time interval (an 8 msec interval is one standard used for regulatory and practical purposes in the mining industry). Ripple-fired explosions most typically consist of 20 to 50 such instantaneous explosions with separate delays between them. Almost all chemical explosions above about 1 ton are ripple fired (Richards et al., 1992).

Data Analysis

Frequency-time displays (spectrograms) of seismograms are useful tools to study the frequency content of entire seismic waveforms observed at local and regional distances (e.g., Hedlin et al., 1989). They are especially helpful when contrasting the time-dependent peaks in the spectrograms typical of most earthquakes with the presence of time-independent spectral bands observed in spectrograms from many ripple-fired mining blasts.

In this study, the following steps were taken to cal- culate spectrograms of regional seismograms:

1. Record segments were selected starting from about 10 sec prior to the first-arrival P waves in order to obtain estimates of background noise.

2. Data were demeaned and band-pass filtered using a third-order Butterworth filter over the passband where the instrument response is not more than 6 dB below the peak response.

3. Spectral estimates were calculated for each time window (usually about 4 sec) by applying the adaptive multi-taper spectral estimation method of Thomson (1977, 1982).

4. Spectral estimates for the whole seismogram were achieved by moving the time window with offset of about 0.75 of the window length between each suc- cessive time window.

5. Final spectral estimates of all time windows were displayed in time-frequency space using a continuous curvature surface gridding algorithm (Smith and Wessel, 1990).

High-Frequency Spectra of Regional Phases from Earthquakes and Chemical Explosions 1367

All spectrograms shown in this article correspond to the signal velocity spectra. Spectral estimates are not corrected for the instrument response, since all the seismograms analyzed in this study were obtained by in- struments with a response nearly flat to the ground velocity over a wide frequency band. We tried several methods for presenting the three-dimensional (time, frequency, amplitude) spectral data, including spectra of short time segments, color-shaded contours of spectral amplitude, and wire-line (perspective) views. No single visualization draws out the relevant features in all cases. To provide a consistent presentation of all analyses, we present all spectrograms as wire-line diagrams in this article. In several cases, summed spectra l in which spectral amplitude at each frequency is obtained by summing up contributions of all the time windows on the spectro- g ram-a re used to display overall characteristics of the several spectrograms together.

Explosions and Earthquakes in New York and Adjacent States

The New York State Seismic Network (NYSSN) has been operated by Lamont-Doherty Earth Observatory since the early 1970s (Sbar and Sykes, 1977). The NYSSN consists of about 25 short-period, high-gain seismographs, including three-component sensors at some stations (Fig. 1). Data are telemetered in analog mode and

~ L e v

N e w York a n d Ad jacen t S ta tes pT.. Loz. ~U~

WNYI R2 I HBVT

• " ~ i ~ 0 1 RI

~ . ~ I I ElmNWC ; ¢

• / , o = o = : , ° , . . / • N Y S S N stat ions i 1 i i i

O N Y N E X shots / / 75" " 74" " 71" " 72' • Ear thquakes

+ Quar r i es /

/ / ) ~ . ' ~ E6.? j , i , , , , , , , , 41"

02°0 i o o~ 44"N ~ ~o . ~ i

=DHN

40"N 0 100 200krr " 0 20 40 60 80km

8o:w " ~ ' 7 ~ ' " 7a' " 7~" 7 o : w " 76' 7~" 7a" 73"

Figure 1. Locations of New York State Seis- mic Network (NYSSN) stations (filled squares) and events used in this study. Open circles denote NY- NEX shot points and closed circles indicate epicenters of earthquakes. Locations of quarries are indicated by crosses. (a) Adirondacks and western Vermont area and (b) southern New York and New Jersey.

recorded digitally. The NYSSN spans the main structural provinces of the eastern United States, including the Pre- cambrian Grenville North American shield ( -1 .2 b.y.) exposed in the Adirondack Mountains, the St. Lawrence rift in northern New York, the Paleozoic Appalachian platform in the western and central part of the state, the Appalachian front and crystalline overthrust sheets in eastern New York and western Vermont, the Newark Mesozoic rift basin and the Cretaceous-Cenozoic Coastal Plain of northeastern New Jersey (Stanley and Ratcliffe, 1985). The area covered by the NYSSN has numerous active quarries and mines and a moderate level of natural seismicity. Thus, seismic data from the NYSSN provide an excellent opportunity to study high-frequency regional seismic wave propagation in diverse geologic settings and to test several discriminators between rip- pie-fired quarry blasts, instantaneous (single-hole or multiple-hole) chemical explosions, and earthquakes.

The stations of the NYSSN have either 1- or 2-Hz seismometers (HS-10) and their response to ground velocity is flat from the seismometer natural period to 25 Hz (6 dB level). Data are recorded at a sampling rate of 100 samples/sec and provide useful information up to at least 25 Hz.

NYNEX Explosions

We start our analysis using data recorded on the NYSSN for controlled explosions from the Ontario-New York-New England Seismic Refraction Experiment (NYNEX) conducted by the USGS, the Air Force Geo- physics Laboratory, and the Geological Survey of Can- ada during September 1988 (Mangino and Cipar, 1990). Our purpose here is to investigate spectrogram properties for single-hole shots, which we would expect to be sim- pler than for ripple-fired shots and earthquakes. A total of 35 single-hole shots were detonated at 23 shot points almost equally spaced on a 640-km-long profile trending roughly east-west (Fig. 1). For most shots, the explo- sives were loaded into a single 0.2-m-diameter drill hole cased to bedrock that varied in depth from 49 to 55 m. A few shots were detonated in water and in sediments. Ammonium nitrate and fuel oil (ANFO) was used as the explosive and the charge size in each shot ranged from 270 to 2100 kg, with the majority of shots near 1000 kg. Most of the shots were fairly well recorded by NYSSN stations in the Adirondack Mountains and adjacent western Vermont. Two of the largest shots (the largest shot of 2100 kg in hard rock and an especially well-coupled shot of 1360 kg in water) were also recorded at most of the more distant stations of the NYSSN.

The NYSSN seismic record section from NYNEX single-hole shot number 20 is displayed in Figure 2. Note that Figure 2 is a group-velocity section, in which seismograms are plotted as a function of group velocities (distance/time), instead of more conventional time sec-


tions. Thus, time scales, which remain linear for each seismogram, are increasingly compressed with distance, giving the impression of higher frequency content at greater distances. Such a section emphasizes seismic energy traveling with constant group velocities, and re- veals the local variation of seismic velocities across the whole network more clearly than a conventional time section. Over the distance range 150 to 210 km, major first P energy arrives with group velocities of about 6.1 to 6.3 km/sec. For the distance range between 280 and 380 km, the group velocity increases to 6.5 km/sec. At distances less than 400 km, Lg waves arrive with group velocities between 3.5 and 3.7 km/sec across the whole section. At TBR (A = 462 km) and other stations in southern New York (Fig. 1), waves traveling in the top of the mantle, Pn and Sn, arrive with group velocities of 7.2 and 4.3 km/sec, and the relative amplitude of Lg in the wave train has decreased substantially compared with earlier arrival phases between Sn and Lg. These stations are in southern New York-New Jersey, suggesting that Lg propagation is disrupted by the Appalachian platform lying along the path. Thus, the section shows ef-

Group Velocity (km/sec) NYNEX shot #20 g 8 7 6 5 4 3 2

140 - L

"MEDY "az.203

- • ~ " , ' ,'~- ,~ ,~, - DHN ,y"" . . . . . i' r . . . . . .

v , ' - . . i ] ~ 7 . . , E ~ ~.2 . . . . . ::" " -MSNY

az=74

1 7 X 7 ...... .~, . . . . . . . . . . . . . . . . . ~;Ir& ~;:5,2'. . . . . . . . . . . . . . -,,.~o,

3 0 0 ~ . . . . . . A , , ~ L , a ~ ~ . (~ .az=91

o ........... a ~ l . a . ,, : ~v

; ~ ] M I r , d ~ l l l t ~ .~ , , . . . ~ . .... .az=9r ~-~ I" . . . . . . . . . . ~ r 'T'q~r~ V' . . . . . . . . FLET az=8,4

Pg 4 0 0 - L g

5 0 0

P n S n Lg I i

I i t , . . ,~ , t ~ . . . . . . . . , . J l ~ l J ~ . , . ~ . ,

5 4 0 -~.,T..,v...F ,,... i... , i , . •, ] , , , , i . . . . I . . . . ~ . . . . I 9 8 7 6 5 4 3

FBR ~z=142

CRNY az~135

-PRIN :az .151

Figure 2. Group-velocity seismic section from NYNEX shot number 20 recorded on the vertical components of the short-period seismographs of the NYSSN. Each seismic trace is plotted with group velocities between 9 and 2 km/sec. Note that the horizontal axis is group velocity, not the usual time axis (see text for details). Station code and azimuth are given at the end of each trace.

fects caused by lateral velocity and structural variations associated with major geologic provinces in the region.

Single-Hole Shots Fired in Competent Bedrock. Spectrograms observed at HBVT and WNY from a single-hole shot in competent bedrock (NYNEX shot number 7) are displayed in Figure 3a. For this and other records from shots in competent rock, the spectrograms are characterized by time-dependent spectral energy distribution (i.e., the seismic signals show strong energies associated with the arrivals of P and S waves, otherwise the seismic energy is distributed fairly randomly in both frequency and time). P waves have higher amplitude at higher frequencies (above 10 Hz) than S waves at all stations in the distance range 10 to 200 km. At distances greater than 100 km, spectrograms at some stations show limited spectral bands, but these are not uniform through the whole seismogram trace and are not consistent at all stations.

Single-Hole Shots in Water-Filled Quarry Sites. Spectrograms at stations MEDY and WNY for shot number 20 are shown in Figure 3b. Although this explosion is a single shot, the spectrograms in Figure 3b show long- lasting spectral bands centered at about 5 and 7.5 Hz and a weaker band at 11 Hz. These spectral bands are present at all stations. This shot was detonated in a water-filled quarry (Mangino and Cipar, 1990) and so the spectral bands are likely due to a combination of an odd harmonic series with fundamental f0 = v/4h (where v = speed of sound in the water, h = water depth), an odd harmonic series with fundamental f l = v/4d, (d = detonation depth), and the complete harmonic series with fundamental f2 = 1/t , where t = first bubble pulse period (Weinstein, 1968). For this shot, h ~- d ~-- 195 m (J. H. Luetgert, personal comm, May 1991), which gives a predicted f0 of 1.9 Hz, assuming v = 1.509 km/sec at 25°C (Press, 1966). If the observed spectral bands at about 5, 7.5, and 11 Hz are related to the third, fifth, and sev- enth harmonic series, respectively, then f0 = fl ~ 1.6 Hz. The detonation depth, d, is estimated to be about 236 m, which is about 20% higher than the known depth. It is noted that higher-order harmonics are not apparent in the observed spectrogram, mainly because of increas- ing attenuation at the higher frequencies. Based on the banding revealed in the spectrograms, this event could be identified as a multiple shot source; however, our knowledge of the source shows that the multiplicity results from reverberations in a water column, rather than a ripple-fired explosion.

The observed amplitudes from this shot are an order of magnitude higher than the amplitudes from other shots with similar charge size at similar distances, showing the much more efficient coupling of seismic energy for underwater shots compared to boreholes. It is also inter- esting to note that the frequency content of S waves (1


(a)

o ~

shot #7 HBVF Z

%

shot #7 WNY Z t~=132 km c13,=281"

(b) shot .2o shot #20 WNY Z A=303 km ~9o'

(c)

<,

shot #10 WNY Z A--5353, km

N ; 1

Figure 3. (a) Spectrograms at HBVT and WNY from the single-hole explosion, NYNEX shot number 7 (competent bedrock site). Notice the strong P waves with high frequency content relative to S waves; (b) Spectrograms at MEDY and WNY from NYNEX shot number 20 (water-f'dled quarry site). Notice clear spectral bands at about 5 and 7.5 Hz, and a weak band at about l 1 Hz (e.g., MEDY) due to reverberation in the water; (c) Spectrograms at WNY and FIN from NYNEX shot number 10 (in sediments). Notice strong spectral peaks at about 1.5 Hz due to the Rg phase, and stronger S waves than P.

shot #10 PTN Z t~=140 km $=295"


to 15 Hz) at most of the distant stations (A > 200 km) from this shot is higher than or comparable to P waves, whereas the opposite is observed on records from the largest NYNEX explosion, shot number 1 (Table 1) at similar ranges.

Single-Hole Explosion in a Sedimentary Layer. The presence of a strong Rg phase with dominant periods 0.5 to 2 sec on regional seismograms has been claimed to be indicative of a shallow focal depth (e.g., Langston, 1987; Kafka, 1990). Events at such shallow depths are usually presumed to be explosions. NYNEX shot number 10 was detonated in sediment at the southern end of the Lake Champlain Valley (Fig. 1; Table 1). Seismograms from this shot are characterized by the presence of a strong Rg phase with group velocity of about 2.9 km/sec out to distances of at least 80 km. Typical spectrograms are displayed in Figure 3c, P waves at most stations have a spectral peak at about 5 to 10 Hz, whereas the S waves have a broader frequency content of between 2 and 15 Hz. Rg phases are confined to a frequency band below about 2 Hz.

Sedimentary layers along the path produce apparent time-dependent arrivals of packets of P and S waves; however, there is no discernable time-independent spectral banding. The seismograms from this shot show that P waves have lower frequency content than the other shots detonated in more competent rock (cf., shot number 7 in Fig. 3a), and that S waves have frequency content higher than P waves at most of the stations. The strong S waves with higher frequency content may be due to the more efficient excitation of S for an explosive

source in the sediment, because much of the P-wave energy is trapped in the sedimentary layer and progres- sively converts to S as the wave propagates from the source to receiver.

Quarry Blasts

There are several active mining areas in upstate New York and western Vermont (see Table 2; Fig. 1). We analyzed seismograms recorded at the NYSSN from about 100 quarry blasts in the area and found them signifi- cantly different from single-hole shots of the NYNEX experiment.

Comparison between Single-Hole Shots and Quarry Blasts. Seismogram data from a granite quarry R1, in Barre, Vermont (Fig. 1), which is close to the NYNEX shot number 8, provided the opportunity to compare quarry shots with single-hole shots for almost identical source receiver paths. Example spectrograms for the blasts on 06 /17/91 and 07 /26 /91 , for which we obtained a copy of blast reports, show that there are weak but clear spectral bands with equal spacing of about 3 Hz (Fig. 4). These spectral bands are clearly observed at all stations recording these and nine additional events from R1.

Overall spectral shapes at several stations for these two ripple-fired shots can be better displayed using summed spectra, in which spectral amplitude at each frequency is obtained by summing up contributions of all the time windows of the spectrogram. Summed spectra at four stations in the distance range 53 to 115 km and azimuth range 254 ° to 312 ° from the blasts on 0 6 / 1 7 / 91 and 07/26/91 are plotted in Figures 4c and 4d, to-

Table 1 NYNEX Shots and Their Locations

Date Origin Time Latitude Longitude Charge Weight Shot ld (m/d/yr) (hr:min:sec) (ON) (~V) (kg) Remark

1 0 9 / 1 7 / 8 8 06:04:01 44.590 69.746 2100 bedrock 7 0 9 / 1 7 / 8 8 04: 04:00 44.179 72.237 1225 bedrock 8 0 9 / 2 4 / 8 8 04:00:00 44.151 72.577 910 bedrock

10 0 9 / 1 7 / 8 8 08:04:00 44.054 73.386 1360 sediment 13 0 9 / 2 4 / 8 8 06:04:00 43.968 74.262 1 040 bedrock 20 0 9 / 3 0 / 8 8 04:00:00 44.477 77.658 900 water filled

Table 2 Quarries in New York, New Jersey, and Vermont*

Latitude Longitude Id Quarry Location (°N) (°W)

R1 Rock-of-ages Barrc, Vermont 44.14 72.48 R2 NYCO Lewis, New York 44.32 73.64 R3 Vermont Asbestos Eden, Vermont 44.77 72.53 S1 Mt. Hope New Jersey 40.94 74.54

*Locations from the U.S. Department of the Interior, Bureau of Mines (1984).


(a) Quarry R, (b) 06/17/91 HBVT Z

~9 ° ~ 6

Quarry R1 07/26/91

~ ~ ~ ~ = pI-IBVT Z b,=52.6 km

98"

(C) 06/17/91 (d) 07/26/91

. ~

.: " - s ' E

Z 0 , , , , , , , ~ , ,

0 1 0 2 0 3 0

Frequency (Hz)

(e) shot #8 ~, .-.- ~.,.. ...'.... ..-:

-"~ ";- ,i -~- i~! "" • " -.-." ~ ~:" .-% . ~ ~ - -~ :." ~

. - : - • ~:<-.~.. - ~ ".- -. ~! :'.'...:~:-...

"~N ~ 0 " 5 \ .',~ i.."....'-~ ~ ""-

Z 0 0 10 20 30

Frequency (Hz)

0 10 20 30

(f) ID "0

2-

E1 < 11)

• ~ 0

rr" -1.

q-2 0

Frequency (Hz)

.................. i ........................ i .................... i ....................................................

' ; 0 20 30 40 5'0 Frequency (Hz)

F i g u r e 4 . A comparison of spectrograms from two different blasts from the same quarry, R1 (Table 2; Fig. 1) at HBVT. (a) Spectrogram from a ripple-fired quarry blast on 0 6 / 1 7 / 9 1 ; (b) spectrogram from a blast on 0 7 / 2 6 / 9 1 . Notice the higher frequency contents of signal from this blast than blast on 0 6 / 1 7 / 9 1 ; (c) summed spectra at four stations for blast on 0 6 / 1 7 / 9 1 ; (d) summed spectra for blast on 0 7 / 2 6 / 9 1 . Notice the different spectral shape compared with (c); (e) summed spectra at eight stations from a single-hole explosion (NYNEX shot number 8). Notice the lack of coherent spectral peaks as compared with those in (c) and (d); (f) predicted source spectra for the two blasts shown in (a) and (b). Overall observed spectral shapes are predictable from their source spectra.


gether with their mean spectra. Spectral shapes of the observed records over the wide distance and azimuth range show remarkably similar characteristics. In contrast, the spectra from the single-hole shot (shot number 8, Fig. 4e) do not show any clear spectral peaks in common.

A comparison of the quarry blasts at R1 with single- hole shots (numbers 7 and 8) in the distance range 50 to 220 km indicates that P waves from the single-hole shots are much stronger than S at frequencies above 10 Hz (Fig. 3a), whereas for the quarry blasts, the S waves have substantial energy up to 25 Hz comparable to P waves (Fig. 4).

The overall spectral characteristics of signals from these quarry blasts can be explained by their predicted source spectra calculated with their known spatio-temporal shot-hole pattern. The predicted source spectra are calculated using the approach given in Chapman et al. (1992). For example, the blast on 07 /26/91 consisted of a total of 63 shot holes that were arranged in seven linear rows in a rectangular pattern. The distance between shot holes (spacing) in each row was 1.52 m and the distance between rows (burden) was 1.83 m. Shot holes of 7.62-cm diameter are drilled to a depth of 14 m. Each shot hole is loaded with 3.2 kg of high-speed (5486 m/sec) explosive at the bottom of the hole and 42 kg of ANFO derivatives are filled as blasting agents above the bottom charge. The top 1.5 m of the hole is filled with drill cuttings (stemming) to contain the explosion. The shot is initiated at the end of the first row closest

to the free face of the blast bench and progressed to the adjacent holes with time interval of 25 msec. The rows are sequentially wired with 10 msec time delay to be added from the previous row. Total duration of the blast was 385 msec. Other pertinent information is listed in Table 3.

The predicted amplitude spectrum (Fig. 4f) includes spectral minima due to the total duration of the blast, n/T , which gives minima at 2.6-Hz intervals for T = 385 msec (blast on 07/26/91) . Several minima at low frequency can be identified as due to this total duration (see Fig. 4f and Chapman et al., 1992), whereas spectral maxima are associated with delay times of 25 msec between shot holes and of 10 msec between the rows, as well as a combination of these two delay times. The first harmonic of primary spectral peaks due to the dominant delay time of 10 and 25 msec should occur at frequencies centered at 100 and 40 Hz. However, fine details of the source spectra of these blasts cannot be fit, because of a large number of delay periods employed (Ta- ble 3) and probably because of some scatter of delay times and charge sizes of subshots.

More details of the signals from a ripple-fired quarry blast may be examined by synthesizing a quarry blast record. Figure 5 shows a linear synthesis of the ripple- fired quan'y blast record for the blast on 07/26/91 . The source time series incorporates the planned spatio-temporal shot-hole pattern of the multiple-hole ripple-fired blast, as well as the azimuth of the receiver. The record

Table 3 Blast Information for Quarries in New York, New Jersey, and Vermont

Total Number Number Max Charge Number Total Delay Row* Date Origin Time Charge of of Per Delay of Delay Duration Time AZ

(m/d/yr) (hr:min) (kg) Rows Holes (kg) Period (msec) (msec) (°)

06/17/91 20:24 3486 10 07/26/91 20:41 2823 7

06/24/92 19:47 6954 4 07/30/92 20:16 5682 5

R1, Barre, Vermont

96 109 90 290 10, 25 63 90 62 385 10, 25

R2, Lewis, New York

54 771 13 325 25, 50 40 567 16 400 25, 50

R3, Eden, Vermont

06/06/91 15:34 4654 1 17 4654 1 0 0 350]" 06/21/91 14:33 1545 2 10 1545 1 0 0 75 07/03/91 16:29 7309 2 23 3655 2 9 9 355t 07/22/91 17:00 4441 2 14 2223 2 18 18 288 08/02/91 15:19 5914 2 20 3545 2 18 18 5 08/02/91 19:06 3300 1 10 3300 1 0 0 45]" 08/21/91 14:20 2882 1 20 2882 1 0 0 330 08/21/91 18:40 6168 2 20 3086 2 9 9 190t

S1, Mt. Hope, New Jersey

06/03/92 18:18 38591 3 118 982 26 716 25, 50, 58 07/08/92 16:16 12409 3 35 1064 11 241 25, 58

225 255

*Row AZ is the direction of the row(s) measured from the north where the free face is at the right-hand side. tRows are curved.


(a) ,,L

Green's _ j ~ A j i a L , , | j , . , i . ~ . , , i j l ~ S h o t #8 - _ _ 45.6 km

function W~l~,m~.~r'~|~lllqlm~-I,rr'r ! V " " ~ . . . . ' . . . . . " az= 301

~c

Source _ _ ~ pulse S Rg

Synthetic ~ ~ . ~ ~ i ~ ~ ~ ~ ~

/ |

blast " .... ~ ] ~ li lt ~ [ ~ I I11"V ~r?~l~l'~[ tii az= 29e

I ' i , , I | ' ' | I ' ' i , I ' ' ' ' I

5 10 15 20 25

time (sec)

(b)

.o /

(c) shot #8 Synthetic Blast HBVTZ ~ HBVTZ ~=4s.6 ~n ~ a=4s.~ km

L r _,j ~,~ ° o.

Figure 5. (a) Comparison of a synthetic and the observed ripple-fired quarry blast record at HBVT (A = 52.6 kin). The record of the single-hole shot (NYNEX shot number 8) used as a Green's function, and the source time series of the ripple-fired blast, are also plotted. Time axis for the source pulse is plotted with an enlargement of × 10 and * indicates convolution; (b) spectrogram of the observed Green's function; (c) spectrogram of the synthetic ripple-fired quarry blast, which shows banding much like Figure 4b.


from a single-hole NYNEX shot number 8 (Fig. 5b) is used as a Green's function, because the shot point is located close to the quarry R1 (see Fig. 1). Although this single shot had double the shot-hole diameter and about 10 times greater maximum charge weight per delay than the quarry blast, the resulting synthetic record is reason- able within the frequency band of interest up to about 25 Hz. Both the synthetic and the observed blast have strong Rg, because of strong spectral amplitude reinforcement at frequencies below 5 Hz (Fig. 4f) from rip- pie firing associated with the total duration of the blast (T -- 0.385 sec in this case). S waves are also enhanced at high frequencies by ripple firing. Finally here, we note the excellent agreement between the spectrogram of the synthetic (Fig. 5c) and of the observed blast (Fig. 4b), suggesting that linear superposition may be a valid as- sumption in the frequency band of interest (see, e.g., Stump and Reinke, 1988).

Quarry Blasts with Strong Rg Excitation. Figures 6a and 6b show spectrograms from station MIV (A = 28.6 km) for two of the 12 blasts we studied from mining area R2 in Lewis, New York (Table 2; Fig. 1). Seismograms from the blasts in this area show a strong Rg spectral peak below 2 Hz. All the spectrograms show clear banding with frequency spacing of 4.5 to 5 Hz in the frequency between 10 and 30 Hz. While these events are all from the same quarry, there are striking differences in the character of the waveforms recorded at the same station from one event to another. These differences in waveforms and spectral bandings are due to the spatio- temporal pattern of each blast.

Blast reports at this quarry indicate that shots consist of three to five rows of shot holes and an average of about 50 shot holes. The nominal delay times between shot holes is 25 msec, except for those at both ends of a row, where 50 msec delays are used (Table 3). For blasts on 06 /24 /92 and 07 /30 /92 , time series of the source pulses are shown in Figures 6c and 6d and predicted source spectra are shown in Figure 6e. The higher frequency content of the signals from the blast on 0 6 / 24/92 can be explained by the predicted source spectrum, which has broader spectral maxima at 12 to 28 Hz compared with the blast on 07 /30 /92 (Fig. 6e). Figure 6f shows a comparison between observed P-wave amplitude spectra and predicted source spectra at two stations, MIV (A = 28.5 km) and HBVT (A = 45.6 km), from the blast on 07 /30 /92 . The predicted source spectra are calculated with pertinent information retrieved from the blast report. A prominent spectral peak centered at 40 Hz is due to the primary delay time of 25 msec between shot holes. Observed P-wave amplitude spectra are corrected for the instrument response and for the attenuation along the paths using an average Q of 115 (Saikia et al., 1990). Figure 6f shows a fairly good agreement between the predicted source spectra and the

observed P-wave spectra in the frequency band between about 15 and 45 Hz.

Quarry Blasts with Apparent Single-Hole Nature. Twenty-four quarry blasts from the area R3 in Eden, Vermont were analyzed (Table 2; Fig. 1). In the spectrogram shown in Figure 7a, there is no clear spectral banding and the spectral shape and waveform are similar to those observed from the single-hole NYNEX shots (see, e.g. , Fig. 3a). Seismograms from these events are characterized by very strong high-frequency P waves and relatively weak S waves at frequencies higher than 10 Hz. Summed spectra at 10 stations for the blast show the similarity of the blast with those of NYNEX single-hole shots (cf. Figs. 4e and 7b). We suspected that these blasts were multiple-hole instantaneous shots or shots with very short time delays between subshots. There is little doubt that these are quarry blasts (based on their location and origin times and since all 24 events that occurred over a 2-yr period had similar characteristics), but they cannot be identified as such based on spectral character alone.

These events were later identified as mining blasts (Vermont Asbestos) and we obtained copies of blast reports for eight of them. These blasts consisted of one or two linear or curved rows of shot holes (10 to 23 holes). Four blasts were detonated without time delays between shot holes or rows and another four blasts were fired with 9- or 18-msec (i.e., quite short) delay times (Table 3). Thus, these shots were all effectively multiple-hole instantaneous blasts. It is observed that P-wave amplitudes vary as much as a factor of about 3 among the stations from the same blast. This strong amplitude variation appears to be due to a directivity effect, as shown by the following discussion.

The amplitude of P-wave signals from quarry blasts depends principally upon the maximum charge per delay. The peak velocity of P waves, Vpeak, at distance A, from a blast can be predicted by a relation (Devine and Duvall, 1963; Nicholls et al., 1971)

Vr~ak = K [A/N/'-~ -t3,

where W = maximum charge weight per delay (delay intervals of 9 msec or larger), A / V W = scaled distance, K = constant for each blast site and shooting procedure, /3 = constant (regression exponent or slope). We determined constants K and/3 using peak velocities observed at four stations from eight blasts. Regression of the peak velocities on the scaled distance yields K = 49.8 + 1.1 and /3 = 1.61 +-- 0.17 (Fig. 7c). The exponent of the scaled distance obtained is close to an empirical value, /3 = 1.58, determined for short distance ranges (a few tens of meters to 1000 m) given by Nicholls et al. (1971) and suggests that the above relation may also be valid for larger distances. The observed peak velocities have

High-Frequency Spectra o f Regional Phases f rom Earthquakes and Chemical Explosions 1375

(a) oo~,~o~ (b) MIV Z &=28.6 km

=163"

Q u a r r y R2 07/30192 M I V Z & = 2 8 . 6 km

63 '

,~

cr~

v

t -

t -

O

(c) 7 6 8 -

512

256

0 L! 0

! i! 100 200

time (msec)

i I I

300

(d)

512 -

2 5 6

o! 0

,!,!l 1 O0 200 300

time (msec)

!!, 400

"o

e~ E

(]) r r 0-

o _.1

(e)

- 1 i i

0 10 20 30 40 50 Frequency (Hz)

c - (D

E o

r~

a

.>__

r r

o ._1

(f) 3 -

4-.......M!Y i . 7"" 2 - ~ ...... ~ ........................ [ ......................... i .................... ' " ' " ~ - . ..................... ~,. - . ~ . - . , / % ' . 2 ~ .,-~:-',.!' / . I • . : ' ~ ~ ........... _ . . . . . . ~ 1 - ~ / " " : ~ : r ...... -" ' / " ~ v

0 . . . . . . . . . . . . . . . . . . . . . . . . . ~ . . . . . . . . . . . . . . . . . . . . . . ~ . . . . . . . . . . . . . . . . . . . . . i . . . . . . . . . . . . . . . . . . . . . . . . . i ~s -0u ree / | : , HBVT ! r . ~ o b s ! ~ /

...:~ ................... : ....................... ~ . . . . . . . . . . i ................ . ........ -1 t ~.-.,, ,-. ,,.~ , ~ : . i ; . , ~ ~ : -

t :-.-.-- .......... i .................. "~i ~ ," ." • i ~! i'i

- 3 / , i , i , , , i , ,

0 10 20 30 40 50 Frequency (Hz)

Figure 6. A comparison of spectrograms from different blasts from the same quarry, R2 (Table 2; Fig. 1) at MIV. (a) Spectrogram at MIV from a ripple-fired blast on 0 6 / 2 4 / 9 2 ; (b) same as (a) but from a blast on 0 7 / 3 0 / 9 2 . Notice the lower frequency content of signal from this blast than blast on 0 6 / 2 4 / 9 2 ; (c) t ime series calculated from the blast pattem of a blast on 0 6 / 2 4 / 9 2 for a station azimuth 163 °. Vertical axis corresponds to charge weight in kilograms; (d) same as (c) but calculated for the blast on 0 7 / 3 0 / 9 2 ; (e) predicted source spectra for the two blasts shown in (a) to (d); (f) detailed comparison between observed and predicted P-wave spectra at MIV (28.6 km) and HBVT (45.6 km). Notice the fit of the 40-Hz spectral peak at both stations.

1376 W . Y . Kim, D. W. Simpson, and P. G. Richards

(a) Quarry R3 (b) 08121/91

A=62.3 km ~ "~ : " ~ i

,:/!" :": ' ' : ; ";t,\~"i

0 10 20 30 Frequency (Hz)

(c) ._.. 200-

t - "

£ Ioo O

E 50

30- O o 20- >

10-

(3. 5 0.5

"''"%..........,.... _o ....... ............

t~

""'"... 0 ""-. . 0 "'. .... " 0 . . . . . . . . . . . O o ~ _.-

"'O... O E '-....

0 Z i i i

1 2 3

S c a l e d D i s t a n c e (km/k~/2 )

(d) 2

o ® o

e e e e e~ .............................................................................. e . .ee~:e ......... ® o o ~ o ° o

o

0 I I -90 -60 -30 0 3'0 6'0 9'0

Angle from Row to Station (*)

(e) 07/22/91 17:00 ( f ) 08/02/91 19:06

MDV 1.75

HBVT 1.05

PNY

0.75

~ AZ=76 ° MDV ~ ¢ ~ I ] ~ i ~ j L ~ A Z = _ 1 3 " 0.73 " -w l ' i , v "" I q - " 1 0 0 . 1 k m

~ AZ=64 ° HBVT 0.54

~ ~ AZ=I 2"

I I t l l l l l ~ l l l l l l l l l l l t l l , l l l l l l I

0 1 2 3 4 5 6

time (sec)

PNY 1.66

~ ~ AZ= - 1 ° 62.3 km

~ J ~ ~ AZ= 50 °

I I r I , l ' - 8 1 . 9 km

0 1 2 3 4 5 6

time (sec)

Figure 7. (a) Spectrogram at HBVT from quarry R3 (Table 2; Fig. 1). Notice the strong P waves with high frequency content and no clear spectral bands; (b) summed spectra at 12 stations for blast on 08/21/91. Notice the similarity between this and shot number 8 (Fig. 4e); (c) observed peak velocity of P waves from eight blasts are plotted with scaled distances log-log scale. The solid line is the regression line and dotted lines indicate one standard deviation; (d) normalized peak velocities for eight blasts are plotted with respect to the angles from row to station (angle convention: 0 ° = station parallel to the row, - 9 0 ° = station perpendicular to the row and facing the free face, +90 ° = station perpendicular to the row and behind the free face); (e) first 5 sec of P waves are plotted for the two blasts with large amplitude variations.


a dominant frequency of about 15 Hz and are corrected for the attenuation using Q = 115 (Saikia et al., 1990).

To examine the cause of this amplitude variation, the peak velocity of P waves at each station is normalized by using scaled distance with the constants K and /3. The normalized peak velocities at each station are plotted against the angle from row to station in Figure 7d. Ideally, the normalized peak velocities should be close to unity. Though there is some scatter, the peak velocities at stations lying behind the free face of the blast bench (positive angle) show about a factor of 1.5 greater amplitude than at stations facing the free face of the bench (negative angle). The stations with high amplitudes are lying close to the direction perpendicular to the blast bench, whereas the stations with small amplitudes are located along the azimuth of the row. The P waves from the two blasts at three stations shown in Figures 7e and 7f demonstrate the significant amplitude variation among these stations with respect to their azimuths. For the blast on 07 /22 /91 , the row azimuth is 288 °, and station MDV (A =- 100.1 km, 4) = 212°), located about 76 ° from the row azimuth, has an amplitude greater by about a factor of 2.3 than PNY (A = 81.9 km, ~b = 276°), which is at about 12 ° from the row azimuth (Fig. 7e).

The main difference between these blasts is the azimuth of the ridge from which the rock is quarried. This leads to a difference in the azimuth of the system of benches and of shot-hole rows used to move the rock: the azimuth of rows is determined by topography. The strong apparent azimuthal dependence of P-wave amplitude appears to be due to interaction of the wave field with the topography immediately surrounding the shot holes.

Earthquakes in the Adirondack Mountains

We have shown that many ripple-fired quarry blasts can be identified by using the presence of spectral bands as a discriminator. However, discrimination between earthquakes and instantaneous (single- or multiple-hole) shots is not possible using this method alone, since nei- ther type of seismic source would be expected to show spectral banding. In the following sections, we find that P/S spectral ratios at high frequency provide a reliable discriminant for earthquakes and explosions in the same region.

Comparison of Single-Hole Shots and Earthquakes. NYNEX shot number 13 is within a few kilometers of the epicenters of several aftershocks of the October 1983 Goodnow earthquake (mb = 5.2, h = 7.5 km). Although the source depths are different, their paths to most of the more distant stations are nearly identical. A comparison of the NYNEX shot with one of the Goodnow aftershocks is shown in Figure 8. The NYNEX records are characterized by initial strong P waves with high-frequency content (5 to 25 Hz) followed by S waves with slightly

lower-frequency content (between 1 and 20 Hz) at most of the stations in the range 20 to 135 km (Fig. 8b). The aftershock shows a weak initial P wave compared to S, and the S waves have much higher frequency content than P across a broad frequency range (about 1 to 25 Hz) at most of the stations (Fig. 8a). Though strong S waves may be partly due to the thrust mechanism for the aftershock (Nabelek and Suarez, 1989), which generates strong S and weak P waves (see, e.g., Kim, 1987; Lil- wall, 1988), most of the earthquakes in the region show S waves with higher-frequency content than P.

The difference between the earthquake and explosion is most pronounced at frequencies above 10 Hz. In this band, the P/S spectral ratio is higher for the explosion than for the earthquake. Therefore, the P/S spectral amplitude ratio in the frequency band 10 to 25 Hz can discriminate these aftershocks from single-hole shots. Other earthquakes and explosions in the region can be discriminated by their P/S spectral amplitude ratios at high frequencies (Kim et al., 1993).

Quarry Blasts and Earthquakes in Southem New York and New Jersey

Earthquakes and numerous quarry blasts in southern New York and northern New Jersey provide an important test of the spectrogram method in seismic discrimination, since a large part of the area is covered by the thick sediments of the Newark Basin, raising the possibility that time-independent spectral bands can be acquired during propagation through highly reverberating, shallow, low-velocity horizons.

Quarry Blasts in Southern New York and New Jersey. Among numerous quarry blasts in this area, we studied seven blasts from the quarry S 1, Mt Hope in New Jersey (Table 2; Fig. 1). According to blast reports we obtained, these blasts had total charge sizes ranging from 12.3 to 38 tons and blasted with 35 to 118 shot holes. The average charge used in each hole ranges from 181 to 345 kg. All the blasts are detonated with a delay time of 25 msec between the holes in a row and each row is delayed 58 msec from the previous row. We show ex- amples from the largest blast (06/03/92) and the smallest blast (07/08/92) among the seven (Table 3; Fig. 9). Spectrograms from the largest blast show clear frequency bands at 9, 14, and 15.5 Hz and weak bands at 5, 19, 22, and 26 Hz, whereas the spectrograms from the smallest blast show relatively simple and clear bands at 12.5 and 18 Hz. A strong spectral peak at about 1.5 Hz is present at stations out to 53 km from both blasts because of a strong Rg phase.

Summed spectra at six stations in the distance range 23 to 93 km and azimuthal coverage 49 to 232 ° are plotted in Figures 9c and 9d. Similarity of spectra at stations dispersed over a wide range of azimuths indicates the absence of significant directivity effects. The reason is


that the spatial dimension of the blast pattern (50.3 by 9.1 m) is too small to exhibit significant directivity in the frequency band used. Again, the summed spectra show remarkable similarities in spectral character among stations in each blast, as well as a distinct source signature of each ripple-fired blast. The largest blast had 118 shot holes with 26 delay periods, whereas the smallest blast had 35 shot holes and 11 delays (Table 3). The differences in the spatio-temporal blast patterns are seen in their predicted source spectra (Fig. 9e).

A comparison between the observed and predicted P-wave spectra at station TBR (A = 39 km) from the blast on 07 /08 /92 is shown in Figure 9f. There is fairly good agreement between the two spectra at 10 to 20 Hz and at 35 to 45 Hz, whereas the spectral fit is poor at frequencies 20 to 33 Hz and below 10 Hz. The strongest spectral peak at 35 to 40 Hz is primarily due to 25-msec

delay times between the shot holes. A spectral peak at around 18 Hz is the first harmonic due to 58-msec delay times between the rows, and a peak at 12.5 Hz may be related to combined delay times of 25 and 58 msec. Note that the second harmonics of 58-msec delays as well as third harmonics of 83-msec delays contribute to making the peak at 35 to 40 Hz broad and prominent. The observed spectrum is corrected for ittenuation using an average Q = 115 (Hough and Seeber, 1991). Comparisons of observed and predicted spectra for the largest blast on 06 /03 /92 were poor, probably because of increased variance of both the delay times from their nominal settings and the distributions of charges. These events show that high-frequency spectral bands can still be recog- nized in the spectrograms from ripple-fired quarry blasts even when the paths cross a large portion of a reverberating sedimentary basin.

(a)

<,

Eq E1 M=2.0 PTN Z A=91.2 km ~=321

9,- f.

(b)

g

shot #13 PTN Z A=88.5. km ~=320

~' - '~ :"~ , .'~'-!~.,t~ "~.... !

<

Z 0 0

0 10 20 30 0 10 20

Frequency (Hz) Frequency (Hz) Figure 8. A comparison of spectrograms from an earthquake with those from a single-hole explosion (NYNEX shot number 13, Fig. 1) at PTN. Notice clear differences in P-wave spectral amplitudes and frequency content.

30


(a) (b) Mt Hope Quarry O6/O3/92

b = TBR Z ~34 .9 km

Mt Hope Quarry 07108/92 TBR Z b.=34.9 km $=49'

(c) 1 ~ ..-" 4 1

< 0.5 i i \ ,, i'i~ • ""~ 0.5

Z 0 0


(d)

ii'i ?], & ii ii i ~t i!

' ' ' ' I ' ' ' ' I ' ' ' ' I


(e) (f) 1 ] . ,'-,.j, source

I?-°~'~ ~ i , ; , , . .

0 1 ................ ~ ..... 5~'°"', ,., ~ , , , . . , , , ~ :~, , . 1 l I , ... t' ~-i~',:~i/'~""/~]'~[l:'\k:" .......... i '~.' i{~,/-\~i i / J ~ ~ . i-.

~ o . . . . . . . . . . . . . . . . . . . . .

0 _ , ~ ._1-2

0 10 20 30 40 50 0 10 20 30 40 50 Frequency (Hz) F requency (Hz)

Figure 9. A comparison of spectrograms at TBR from two quarry blasts at the same quarry (S1; Table 2, Fig. 1). (a) Spectrogram from the blast on 06/03/92; (b) spectrograms from the blast on 07/08/92; (c) summed spectra at seven stations from the largest blast on 06/03/92; (d) summed spectra at seven stations from the smallest blast on 07/08/92; (e) predicted source spectra for both blasts. Notice the complexity of the source spectrum of the 06/03/92 blast; (f) fit between the predicted and the observed P-wave spectra at TBR for the blast on 07/ 08/92.


Earthquakes in Southern New York and New Jersey. Although a large part of the region is covered with thick sediments, seismograms from earthquakes in this region do not show clear Rg phases, probably because of the deeper depth of the earthquakes. Seismograms from earthquakes that occurred in four different locations were analyzed (Table 4; Fig. 1). There are no clear spectral bands in the spectrograms and the energy is more or less randomly distributed in time and frequency. Figure 10 shows spectrograms from two earthquakes at station PRIN in the epicentral distance range 87 to 124 km. There are weak spectral bands associated with P- and S-wave arrivals, but these are discontinuous and do not extend throughout the seismogram. This apparent spectral banding may be due to source-receiver paths that lie mostly in the Newark Basin. P waves from these earthquakes are much weaker than S waves over a broad frequency range (5 to 25 Hz). The P/S spectral amplitude ratio in the high-frequency band (10 to 25 Hz) successfully dis- criminates all events as earthquakes rather than instantaneous explosions.

Ripple-Fired Mining Explos ions in Norway

As a final example of the use of the spectrogram method, we test the ability to identify spectral banding in a very poor signal-to-noise ratio environment, but where we have basic information on the explosions, number of delays, and total charge size, and hence on the expected spectral banding characteristics. Seismograms recorded at the Norwegian seismic array (NORESS) have been used extensively by many researchers for detecting and discriminating earthquakes and chemical explosions (e.g., Baumgardt and Ziegler, 1988; Hedlin et al., 1990). Pre- vious studies were mainly conducted for quarry blasts from two large mining areas, Titania and Bl~sj¢ in southern Norway. These studies used data recorded on the conventional NORESS short-period channels with sampling rate of 40 sample/sec, thus they were able to interpret spectral modulation only up to 20 Hz.

We use data recorded on the special three-component high-frequency seismic element at NORESS, which has nearly flat response between 10 and 55 Hz and a

Table 4 Earthquakes in New York and New Jersey*

Date Origin Time Latitude Longitude Depth Event (m/d/yr) (hr:min:sec) (°N) (°W) (kin) Magnitude Area

E1 10/07/83 10:48:39 43.938 74.258 7.5 2.0 Goodnow aftershock E4 01/22/89 08:27:15.9 40.884 73.942 - - 2.0 Englewood Cliffs, New Jersey E6 04/12/91 11:12:12 41.136 73.654 - - 2.0 Westchester Co., New York

*Location, origin time, depth, and magnitude are from Quarterly Seismicity Bulletin of the New York State Seismic Network, Lamont- Doherty Earth Observatory.

Eq E4 m=2.0 01/22/89 08:27

. • PRIN Z &=87.1 km, h=6 km baz=48.6 °

Eq E6 m-2.1

PRIN Z &=124 kin, h=10 km

.1

• ' ~ o £ ~

S'q.~ @

Figure 10. A comparison of spectrograms from two earthquakes in southern New York and New Jersey (Table 4, Fig. 1) at station PRIN. Note that P waves from all earthquakes are much weaker than S waves over a broad frequency range (5 to 25 Hz).


sample rate of 125 samples/sec (Ringdal et al., 1986). Detailed information on the events analyzed are listed in Table 5 and locations are depicted in Figure 11. Since all of the delay times are 45 msec, the frequency spacing of the primary spectral peaks should be about 22 Hz, so that primary spectral banding cannot be observed using the conventional NORESS short-period seismogram data. All of the explosions are at large distances from NORESS and the signal-to-noise ratios are very low. The major portion of seismic energy arrives in the frequency band between 2 and 10 Hz and shows weak time-dependent spectral peaks associated with P, S, and Lg arrivals.

Explosions N1, N2, and N3 are all at the same location. The smallest explosion, N2, does not show any consistent spectral banding above 15 Hz. But the larger

blasts, N1 and N3, showed a clear spectral band close to the expected 22.5-Hz value. Figure 11 shows the spectrogram for N3, with a spectral band at 23 to 25 Hz. The spectral peak near 25 to 28 Hz on the spectrogram is apparently a noise signal, since it exists prior to the expected arrival of the explosion signal. More information, such as charge per delay, was available to us for a ripple-fired explosion at Aheim about 353 km NW of NORESS (N4; Fig. 11). Different amounts of explosive were used in each subshot, with the largest subshot of 2.1 ton (Table 5). The signal-to-noise ratio for this event is even lower than the spectrogram in Figure 11, but a broad spectral band at about 23 Hz is observed on all three components.

This analysis of ripple-fired explosions whose basic

Table 5 Mining Explosions in Norway Recorded at NORESS

Total Max Charge Number Total Delay Date Origin Time Latitude Longitude Charge Per Delay of Duration Times

Id (m/d/yr) (hr:min) (°N) (°E) (ton) (ton) Delays (msec) (msec)

N1 01/28/86 10:18 66.24 14.35 149.0 6.7* 22 N2 05/09/86 17:14 66.24 14.35 102.2 3.9* 26 N3 12/01/87 12:55 66.24 14.35 217.6 9.9* 22 N4 05/24/89 13:35 62.04 5.52 9.4 2.1 8

945 45 1125 45 945 45 315 45

*Estimates of mean max charge weight per delay (total charge/number of subshots).

70"N

65"

60"

55* 5" 10" 15" 20"E

Figure 11. (Upper panel) Locations of NORESS and the ,~daeim and Stor- forshei mines in Norway (Table 5); (lower panel) spectrogram at NORESS from a 218-ton ripple-fired quarry blast (N3, Table 5) in northern Norway. Expected spectral band at 22 Hz is indicated by an arrow at the upper right-hand edge of the spectrogram. Note that a spectral band with constant amplitude at about 30 Hz is due to noise.

Z


detonation characteristics were known a priori shows that, even when the signal-to-noise ratio is very low, it is possible to observe spectral bands at a distance of several hundred kilometers, because of source multiplicity re- suiting from ripple firing. Care must be taken, however, first to identify background noise, from electronic or cul- tural sources, that may also produce spectral banding.

Discussion

Spectral Characteristics of Regional Records. Ob- servations of high-frequency (1 to 35 Hz) regional signals from earthquakes and explosions in New York and adjacent states may be summarized as follows.

Instantaneous explosions (single- or multiple-hole) in competent bedrock produce strong P waves with higher frequency content than S waves (NYNEX shot numbers 7 and 13, Figs. 3a and 8b; quarry R3, Fig. 7), whereas the single shot in sediments (NYNEX shot number 10, Fig. 3c) generates P waves with much lower frequency content but strong S waves. The single shot in sediments also generates a strong Rg phase out to about 100 km. The excitation of Rg and an enhancement of S waves relative to P waves are closely associated with the presence of low-velocity surface layers along the path. Sed- imentary layers appear to have a stronger effect when they are present near the source rather than along the path or close to the receiver. This can be explained as efficient P to S conversion occurring in the sedimentary layer. The conversion to S is more efficient near to the source simply because more convertible P energy is available near the source region than for later parts of the path with longer elapsed time. Thus, the decrease in frequency content of the P wave is much more substantial than for the S wave.

S waves from earthquakes are stronger than P waves by a factor of about 2 over a broad frequency band (1 to 25 Hz). P and S waves from ripple-fired quarry blasts are rich in high frequency because of spectral modulation caused by ripple firing of subshots. The records from ripple-fired quarry blasts often show a strong Rg phase out to about 100 km because of spectral amplitude reinforcement at Rg dominant frequencies (0.5 to 2 Hz) resulting from extended source duration. The mean source duration of the ripple-fired surface quarry blasts in New York and adjacent states examined in this study is about 0.4 sec. The short-period Rg waves are essentially confined to depths of about 1 to 5 km, and as such, Rg signals are likely to experience significant attenuation re- suiting from low Q materials in the shallow crust. Hence, Rg phases are observed only at near-regional distance ranges (up to 200 km).

Ripple-Fired Quarry Blast and Spectral Banding. The observations of clear high-frequency spectral banding from

the ripple-fired quarry blasts, from an instantaneous explosion in a water-filled quarry site in New York and adjacent states (Figs. 3b, 4, 6, and 9), and in Norway (Fig. 11), suggest that spectral banding is a unique feature of the repetitive seismic sources. From some of our observations, however, it is clear that spectral banding is not a universal feature of all quarry blasts (e.g., Fig. 7). If the delay times are very short (within --8 msec), the blast becomes similar to an instantaneous explosion and the first primary spectral band (= inverse of delay times) is likely to lie above the passband of typical regional seismographs.

Overall spectral shapes and frequency content of the observed quarry blast signals can be explained by predicted source spectra calculated with the spatio-temporal shot-hole pattern (e.g., Smith 1989; Chapman et al., 1992). However, predicted source spectra do not repro- duce all details of the spectral minima and maxima of the observed ripple-fired quarry blasts spectra, because of the variations in spatio-temporal shot-hole patterns from nominal settings, uneven charge distributions, and the influence of paths.

Even in those cases where the observed complex spectral bands do not lead to clear correlation with the spectral bands because of the primary delay times (between the shot holes or between the rows), we still find scalloping that is observed coherently in spectra at stations over wide distance ranges and azimuths. The fact that observed spectral scalloping of a ripple-fired blast is independent of paths and station site responses is shown in the summed spectra (Figs. 4 and 9). Chapman et al. (1992) also reported similar observations of the spectral modulations that are largely independent of paths and receiver sites from mining explosions in Kentucky, at distance ranges of 180 to 400 km. The spectral scalloping can be distinctly different for different ripple-fired shots, even for blasts from the same quarry.

Discrimination of Earthquakes and Explosions using the Spectrogram Method. The spectrograms from earthquakes and instantaneous explosions are characterized by a time-dependent spectral energy distribution that is associated with the arrivals of P and S waves, plus seismic energy that is distributed fairly randomly in both frequency and time. The observations of clear time- independent high-frequency spectral bands from the rip- pie-fired quarry blasts in diverse geologic settings indicate that this feature can be used to identify ripple-fired quarry blasts. The spectrogram method successfully de- ciphers multiplicity of the source (e.g., Fig. 3b). How- ever, spectral banding can also develop from factors other than a multiple source [e.g., resonances along path, as shown by Sereno and Orcutt (1985), or near the recording site; electronic noise in the recording system]. It is important that the ambient spectral characteristics of the recording site and the source-receiver path be well known.


In attempting to use spectral banding in the identification of ripple-fired explosions, care must be taken to isolate other sources of time-independent spectral bands by careful analysis of background noise.

Discrimination of Earthquakes and Explosions Using the P/S Spectral Ratio. Spectrograms shown in the previous sections indicate that the P/S (=Pg/Lg) spectral amplitude ratio of records from earthquakes, ripple-fired quarry blasts, and single-hole explosions can be a useful discriminant. In the frequency band 5 to 25 Hz, P/S spectral amplitude ratios of vertical-component records from explosions and earthquakes used in this study and additional data from earthquakes that occurred in the eastern United States recorded at NYSSN may be summarized as showing that: (1) P waves from instantaneous explosions have higher frequency content than S waves with a mean P/S spectral amplitude ratio of about 1.25; (2) P waves from earthquakes are weaker than S waves with a mean P/S spectral amplitude ratio of about 0.5; (3) P and S waves from ripple-fired quarry blasts have fairly equal frequency contents with a mean P/S spectral amplitude ratio of about 1.0. Single-hole instantaneous explosions and ripple-fired quarry blasts have somewhat overlapping P/S spectral ratios and so these events cannot be discriminated with P/S ratios. But these two types of explosions as a group are distinctly different from earthquakes.

We performed linear discriminant function analysis for the data from the 30 known explosions (230 records) and 30 earthquakes (255 records) by using log,o(Pg/Lg) spectral ratios at frequencies 5, 10, 15, 20, and 25 Hz as variables (see e.g. , Seber, 1984). Noise analysis sug- gested that signal-to-noise ratio is poor at frequencies higher than about 25 Hz for several records from distant events or from weak events. Assuming equal prior probabilities for two groups, we find that about 15% of earthquake records and about 15% of explosion records are classified incorrectly. The discriminant scores for each record in two populations are plotted in Figure 12a with respect to mean loglo(Pg/Lg) spectral amplitude ratio of each record. Note that vertical lines in the figure denoted as Eq and Ex are a projection of the multivariate mean of records from the earthquake and explosion popula- tion, respectively. More details of the method used are reported in Kim et al. (1993).

If the network-averaged mean discriminant score for each event is used to classify the explosions from earthquakes, then all explosions and earthquakes are correctly classified, except one ripple-fired explosion from quarry R2 (Fig. 12b). This result suggests that the P/S spectral amplitude ratio provides very good discrimination power with a total misclassification probability of only about 1%. Single-record P/S ratios shown in Figure 12a indicate that the frequency content of regional P and S waves is somewhat influenced by propagation paths as

well as by local receiver site responses. These effects can be averaged out with network data, for purposes of identifying spectral properties of the source.

It is difficult to compare directly our results of P/S ratio with other studies, since in most earlier studies, data available for discrimination analyses were limited to frequencies below 10 Hz and much previous work on P/S amplitude ratios was based on time domain observations at dominant frequencies near 1 Hz (Pomeroy et al., 1982). Yet, we note that similar observations are reported in the western United States, but with much higher misclassification probabilities ranging from about 16 to 25% (e.g., Taylor et al., 1989). Some observations in the western United States have even been reported as having Pg/Lg higher for earthquakes than explosions (e.g., Bennett and Murphy, 1986), which are contrary to our observations.

Chan et al. (1990) reported observations similar to

(a)

0.5 ml

13_ v

O

~:~ o- o

_J t - -

t~

- 0 . 5 -

/.,

A

-1

-10 -8

Ex Do Eq

o ~

0 o

0

I I I I I

-6 -4 -2 0 2 4 6 8

Discriminant Score

(b)

I

-4

ZX A A N k A /X

I I I I I I i

-3 -2 -1 1 2 3 4 Mean Discriminant Score

Figure 12. (a) Discriminant scores of earthquake (circles) and explosion (triangles) records are plotted with respect to the mean loglo(Pg/Lg) ratios of each record. Vertical lines denoted as Eq and Ex are the projection of the multivariate mean of the earthquake and explosion populations, respectively. The vertical line, Do, serves to classify the events when the a priori probability of the two populations is the same; (b) mean discriminant scores of earthquakes (circles) and explosions (triangles) are plotted: Overlapping data points are plotted with vertical offset for clear display.

I

10


ours for underground explosions and earthquakes in central Asia using P/Lg ratios in the frequency band 1 to 20 Hz, and Deneva et al. (1989) reported a successful discrimination of small earthquakes and explosions using P/S amplitude ratio at short ranges (A < 50 km) in Bul- garia. These and our results of P/S ratios suggest that though the regional discriminants must be evaluated on a regional basis, the P/S spectral ratio at high frequencies (5 to 25 Hz) is an adequate discriminant for explosions from earthquakes in the eastern United States.

In the western United States, spectral ratios Lg (0.5 to 2 Hz)/Lg(2 to 8 Hz) have been reported to perform better than P/S ratios to discriminate underground explosions from earthquakes (Murphy and Bennett, 1982; Pomeroy et al., 1982; Taylor et al., 1988, 1989). How- ever, this method would work poorly for discriminating the ripple-fired explosions from other types of events because of spectral reinforcement resulting from the ripple firing.

The presence of Rg, with a dominant period of about 0.5 to 1.5 sec, on regional records suggests that the source is most likely very shallow. The Rg phase from near- surface explosions and shallow earthquakes has been successfully used as a source depth discriminant. For example, Bath (1975), Langston (1987), and Kafka (1990) used Rg/S or Rg/Lg ratios to estimate the source depths of mining rock burst and quarry blasts in Sweden, local earthquakes in Australia, and quarry blasts in New En- gland, respectively. However, the lack of Rg waves is not necessarily indicative of a deep source, since many shallow instantaneous explosions do not generate discernable Rg. The Rg/S ratio is a useful depth discriminant, but it does not necessarily discriminate different source types. Shorter distance ranges of Rg observation make the Rg phase less useful than Lg for regional seismic discrimination.

Conclusions

The diagnostic features of regional waveforms that can be used to identify regional earthquakes, instantaneous explosions, and ripple-fired quarry blasts include temporal variations in the spectra, high-frequency spectral banding, excitation of Rg, and the P/S spectral amplitude ratio.

The observation of regular spectral banding at high frequencies is the most reliable discriminant of industrial chemical blasting. The spectral banding (or scalloping) is the unique source signature of a ripple-fired blast and it is independent of paths and station site responses. The spectral scalloping of individual ripple-fired blasts may differ for different blasts from the same quarry. Scal- loping with an expected peak was seen even at distances of several hundred kilometers from chemical explosions in Norway--but only if the data were recorded at sample rates much higher than typically used by NORESS.

We observed a class of chemical explosions that do not show the usual banding. They were shots without delay times between different holes, or with delays so short that scalloping would not be expected except at frequencies above 50 Hz, or shots with short total duration. In practice, we find that the predominant spectral band due to the primary delay time of tipple-fired quarry blasts cannot be easily discerned for a blast with delay times shorter than about four times the sampling interval of the recording system (one factor of 2 or so for the usual passband of the regional seismographs and another factor of 2 for the Nyquist frequency). The observed variation of P-wave peak velocity with azimuth in these shots appears to be caused by the local topography surrounding the shot holes. Topography controls the orientation of the free face of the blast bench and hence the azimuth of rows.

The presence of short-period Rg on regional seismograms is diagnostic of a shallow source. This usually implies an explosive source, but very shallow earthquakes can also produce Rg. A strong Rg phase is often generated from explosions in areas with surface low-velocity layers. Rg is more efficiently excited by ripple- fired quarry blasts than by instantaneous explosions, because the extended duration of the ripple-fired shots often lead to spectral reinforcement at the Rg dominant frequencies.

The P/S spectral amplitude ratio at high frequencies provides a complementary tool to discriminate between explosions and earthquakes--a tool that works very well for instantaneous explosions, when spectral banding fails. In the northeastern United States, a high P/S spectral ratio above 10 Hz is a stable characteristic of instantaneous explosions, distinguishing them from earthquakes. Regional discriminants must be evaluated on a regional basis; however, we presume the P/S ratio reported here would work to discriminate small nuclear explosions from small earthquakes, in regions that support high-frequency signals. The key to our P/S method and conclusions is the observability of frequencies up to about 20 Hz at regional distances in the eastern United States.

Finally, we note that the discriminants we found most effective all required acquisition of regional data at high frequency--preferably up to about 40 Hz. This conclu- sion has significant implications for the design of regional networks.

Acknowledgments

We thank A. Lerner-Lam who helped us in the early stages of this work. R. Such and D. Johnson provided us with much of NYSSN data used. We thank R. Hansen, A. Dahle, and F. Ringdal of NORSAR for providing us with NORESS high-frequency element seismogram data. M. Hedlin reviewed the manuscript and provided helpful com- ments. This research was sponsored by the Defense Advanced Re- search Projects Agency and monitored by the Air Force Phillips Lab- oratory under Contract Number F19628-88-K-0041, and by the USGS

High-Frequency Spectra o f Regional Phases f r o m Earthquakes and Chemical Explosions 1385

under Grant Number 1434-92-A-0969. Lamont-Doherty Earth Ob- servatory Contribution Number 5179.

References

Bfith, M. (1975). Short-period Rayleigh waves from near-surface events, Phys. Earth Planet. Interiors | 0 , 369-376.

Baumgardt, D. R. and K. A. Ziegler (1988). Spectral evidence for source multiplicity in explosions: application to regional discrimination of earthquakes and explosions, Bull. Seism. Soc. Am. 78, 1773-1795.

Bennett, T. J. and J. R. Murphy (1986). Analysis of seismic discrimination capabilities using regional data from western United States events, Bull. Seism. Soc. Am. 76, 1069-1086.

Chan, W. W., R. Baumstark, and R. K. Cessaro (1990). Spectral Discrimination between Explosions and Earthquakes in Central Eurasia, Scientific Report, GL-TR-90-2017, Geophys. Lab., Hanscom AFB, Massachusetts, 38 pp.

Chapman, M. C., G. A. Bollinger, and M. S. Sibol (1992). Modeling delay-fired explosion spectra at regional distances, Bull. Seism. Soc. Am. 82, 2430-2447.

Deneva, D., L. Khristoskov, V. Babachkova, N. Dotsey, and K. Marinov (1989). Detection of industrial explosions and weak earthquakes with local seismological networks, Izvestiya Earth Phys. 25, no. 9, 750-753.

Devine, J. F. and W. I. Duvall (1963). Effect of charge weight on vibration levels for millisecond delayed quarry blasts, Earth- quake Notes 34, no. 2, 17-24.

Evernden, J. F., C. B. Archambeau, and E. Cranswick (1986). An evaluation of seismic decoupling and underground nuclear test monitoring using high-frequency seismic data, Rev. Geophys. 24, 143-215.

Frantti, G. E. (1963). Spectral energy density for quarry explosions, Bull. Seism. Soc. Am. 53, 989-996.

Greenhalgh, S. A. (1980). Effects of delay shooting on the nature of P-wave seismograms, Bull. Seism. Soc. Am. 70, 2037-2050.

Hedlin, M. A., J. B. Minster, and J. A. Orcutt (1989). The time- frequency characteristics of quarry blasts and calibration explosions recorded in Kazakhstan, USSR, Geophys. J. Int. 99, 109- 121.

Hedlin, M. A., J. B. Minster, and J. A. Oreutt (1990). An automatic means to discriminate between earthquakes and quarry blasts, Bull. Seism. Soc. Am. 80, 2143-2160.

Hough, S. E. and L. Seeber (1991). Seismological constraints on source properties of the m~ = 4.0, 1985 Ardsley, New York, earthquake: a characteristic rupture? J. Geophys. Res. 96, 18183- 18195.

Kafka, A. L. (1990). Rg as a depth discriminant for earthquakes and explosions: a case study in New England, Bull. Seism. Soc. Am. 80, 373-394.

Kim, W. Y. (1987). Modelling short-period crustal phases at regional distances for seismic source parameter inversion, Phys. Earth Planet. Interiors 47, 159-178.

Kim, W. Y., D. W. Simpson, and P. G. Richards (1993). Discrim- ination of earthquakes and explosions in the eastern United States using regional high-frequency data, Geophys. Res. Lett. 20, 1507- 1510.

Langston, C. A. (1987). Depth of faulting during the 1968 Mecker- ing, Australia, earthquake sequence determined from waveform analysis of local seismograms, J. Geophys. Res. 92, 11561-11574.

Lilwall, R. C. (1988). Regional mb:Ms, Pg/Lg amplitude ratios and Lg spectral ratios as criteria for distinguishing between earthquakes and explosions: a theoretical study, Geophys. J. 93, 137- 147.

Mangino, S. and J. Cipar (1990). Data report for the 1988 Ontario- New York-New England seismic refraction experiment: three-

component profiles, Environmental Research Papers Number 1054, GL-TR-90-0039, Geophysics Laboratory, Hanscom AFB, Massachusetts, 143 pp.

Murphy, J. R. and T. J. Bennett (1982). A discrimination analysis of short-period regional data at Tonto Forest Observatory, Bull. Seism. Soc. Am. 72, 1351-1366.

Nabelek, J. and G. Suarez (1989). The 1983 Goodnow earthquake in the central Adirondacks, New York: rupture of a simple, circular crack, Bull. Seism. Soc. Am. 79, 1762-1777.

Nicholls, H. R., C. F. Johnson, and W. I. Duvall (1971). Blasting vibrations and their effects on structures, Bureau Mines Bull. 656, U.S. Dept. of the Interior, Washington, D.C., 105 pp.

Pollack, H. N. (1963). Effect of delay time and number of delays on the spectra of ripple-fired shots, Earthquake Notes 34, no. 1, 1- 12.

Pomeroy, P. W., W. J. Best, and T. V. McEvilly (1982). Test ban lleaty verification with regional data--a review, Bull. Seism. Soc. Am. 72, $89-$129.

Press, F. (1966). Seismic velocities, in Handbook of Physical Con- stants, S. P. Clark, Jr. (Editor), Memoir 97, Geological Society of America, New York, 587 pp.

Richards, P. G., D. A. Anderson, and D. W. Simpson (1992). A survey of blasting activity in the United States, Bull. Seism. Soc. Am. 82, 1416-1433.

Ringdal, F., B. K. Hokland, and T. Kvaerna (1986). Initial results from the NORESS high frequency seismic element (HFSE), Semi- annual Technical Summary, NORSAR Scientific Report Number 2-85/86, Kjeller, Norway, 31-39.

Saikia, C. K., A. L. Kafka, S. C. Gnewuch, and J. W. McTigue (1990). Shear velocity and intrinsic Q structure of the shallow crust in Southeastern New England from Rg wave dispersion, J. Geophys. Res. 95, 8527-8541.

Sbar, M. L. and L. R. Sykes (1977). Seismicity and lithospheric stress in New York and adjacent areaS, J. Geophys. Res. 82, 5771- 5786.

Seber, G. A. F. (1984). Multivariate Observations, John Wiley, New York, 686 pp.

Sereno, T. J., Jr. and J. A. Orcutt (1985). Synthesis of realistic oceanic Pn wave trains, J. Geophys. Res. 90, 12755-12776.

Smith, A. T. (1989). High-frequency seismic observations and models of chemical explosions: implications for the discrimination of ripple-fired mining blasts, Bull. Seism. Soc. Am. 79, 1089-1110.

Smith, W. H. F. and P. Wessel (1990). Gridding with continuous curvature splines in tension, Geophysics, 55, 293-305.

Stanley, R. S. and N. M. Ratcliffe (1985). Tectonic synthesis of the Taconian orogeny in western New England, Geol. Soc. Am. Bull. 96, 1227-1250.

Stump, B. W. and S. K. Reinke (1988). Experimental confirmation of superposition from small-scale explosions, Bull. Seism. Soc. Am. 78, 1059-1073.

Taylor, S. R., N. W. Sherman, and M. D. Denny (1988). Spectral discrimination between NTS explosions and western United States earthquakes at regional distances, Bull. Seism. Soc. Am. 78, 1563- 1579.

Taylor, S. R., M. D. Denny, E. S. Vergino, and R. E. Glaser (1989). Regional discrimination between NTS explosions and western U.S. earthquakes, Bull. Seism. Soc. Am. 79, 1142-1176.

Thomson, D. J. (1977). Spectrum estimation technique for charac- terization and development of WT4 wavegnide-I, Bell Sys. Tech. J. 56, 1769-1815.

Thomson, D. J. (1982). Spectrum estimation and harmonic analysis, Proc. IEEE 70, 1055-1096.

U.S. Department of the Interior, Bureau of Mines (1984). Minerals Available System Domestic Deposit Listing.

Weinstein, M. S. (1968). Spectra of acoustic and seismic signals generated by underwater explosions during Chase experiment, J. Geophys. Res. 73, 5473-5476.

1386 W . Y . Kim, D. W. Simpson, and P. G. Richards

Willis, D. E. (1963). Comparison of seismic waves generated by different types of source, Bull. Seism. Soc. Am. 53, 965-978.

Lamont-Doherty Earth Observatory of Columbia University Palisades, New York 10964

(W.Y.K., D.W.S., P.G.R.)

Department of Geological Sciences Columbia University Palisades, New York 10964

(P.G.R.)

Manuscript received 12 August 1993.

high-frequency spectra of regional phases from earthquakes...

Documents