A UNITED STATES DEPARTMENT OF COMMERCE PUBLICATION c S^.\S '. tKL \TU-B^L°l ESSA TR ERL 176-ESL 9 ESSA Technical Report ERL 176-ESL 9 U.S. DEPARTMENT OF COMMERCE Environmental Science Services Administration Research Laboratories Norsar Microseisms JAMES N. MURDOCK JOHN H. PFLUKE • BOULDER, COLO. JULY 1970 ESSA RESEARCH LABORATORIES The mission of the Research Laboratories is to study the oceans, inland waters, the lower and upper atmosphere, the space environment, and the earth, in search of the under- standing needed to provide more useful services in improving man's prospects for survival as influenced by the physical environment. Laboratories contributing to these studies are: Earth Sciences Laboratories: Geomagnetism, seismology, geodesy, and related earth sciences; earthquake processes, internal structure and accurate figure of the Earth, and distribution of the Earth's mass. Atlantic Oceanographic and Meteorological Laboratories: Oceanography, with emphasis on the geology and geophysics of ocean basins, oceanic processes, sea-air interactions, hurricane research, and weather modification (Miami, Florida). Pacific Oceanographic Laboratories: Oceanography; geology and geophysics of the Pacific Basin and margins; oceanic processes and dynamics; tsunami generation, propaga- tion, modification, detection, and monitoring (Seattle, Washington). Atmospheric Physics and Chemistry Laboratory: Cloud physics and precipitation; chem- ical composition and nucleating substances in the lower atmosphere; and laboratory and field experiments toward developing feasible methods of weather modification. Air Resources Laboratories: Diffusion, transport, and dissipation of atmospheric con- taminants; development of methods for prediction and control of atmospheric pollution (Silver Spring, Maryland). Geophysical Fluid Dynamics Laboratory: Dynamics and physics of geophysical fluid systems; development of a theoretical basis, through mathematical modeling and computer simulation, for the behavior and properties of the atmosphere and the oceans (Princeton, New Jersey). National Severe Storms Laboratory: Tornadoes, squall lines, thunderstorms, and other severe local convective phenomena toward achieving improved methods of forecasting, detecting, and providing advance warnings (Norman, Oklahoma). Space Disturbances Laboratory: Nature, behavior, and mechanisms of space disturb- ances; development and use of techniques for continuous monitoring and early detection and reporting of important disturbances. Aeronomy Laboratory: Theoretical, laboratory, rocket, and satellite studies of the physical and chemical processes controlling the ionosphere and exosphere of the earth and other planets. Wave Propagation Laboratory: Development of new methods for remote sensing of the geophysical environment; special emphasis on propagation of sound waves, and electro- magnetic waves at millimeter, infrared, and optical frequencies. Institute for Telecommunication Sciences: Central federal agency for research and services in propagation of radio waves, radio properties of the earth and its atmosphere, nature of radio noise and interference, information transmission and antennas, and meth- ods for the more effective use of the radio spectrum for telecommunications. Research Flight Facility: Outfits and operates aircraft specially instrumented for re- search; and meets needs of ESSA and other groups for environmental measurements for aircraft (Miami, Florida). ENVIRONMENTAL SCIENCE SERVICES ADMINISTRATION BOULDER, COLORADO 80302 ^fiVCf St«^ tVS U. S. DEPARTMENT OF COMMERCE Maurice H. Stans, Secretary ENVIRONMENTAL SCIENCE SERVICES ADMINISTRATION Robert M. White, Administrator RESEARCH LABORATORIES Wilmot N. Hess, Director ESSA TECHNICAL REPORT ERL 176-ESL 9 Norsar Microseisms JAMES N. MURDOCK JOHN H. PFLUKE This work was sponsored by the Advanced Research Projects Agency under ARPA Order No. 800, Amendment No. 11. EARTH SCIENCES LABORATORIES BOULDER, COLORADO July 1970 For sale by the Superintendent of Documents, U. S. Government Printing Office, Washington, D. C. 20402 Price $1.75 The Research Laboratories of the Environmental Science Services Administration do not approve, recommend, or endorse any product, and the results reported in this document shall not be used in ad- vertising or sales promotion, or in any manner to indicate, either implicitly or explicitly, endorsement by the United States Govern- ment of any specific product or manufacturer. XI CONTENTS 1. INTRODUCTION 1 2. NORWAY OPERATIONS AND PRECISION OF THE DATA 2 2.1 Data Acquisition 2 2.2 Precision of the Data 5 2.3 Production of Seismograms in Norway 9 3. DIFFERENCES BETWEEN THE SURFACE AND SUBSURFACE SEISMIC FIELDS 12 3.1 Microseisms 12 3.2 Seismic Signals 22 3.3 Other Observations 22 U. PARAMETERS OF THE TIME- VARYING MICROSEISMIC FIELD 2k U.l Estimates by Analog Techniques 2h U.2 Estimates by Digital Techniques 2J> It. 3 Statistical Properties of the Data 28 5- COHERENCE OF MICROSEISMS 33 5.1 Data Reduction and Construction of Displays 33 5.2 Observed Sample Coherence 3h 5.3 Model of the Coherence Variations I46 $.k Comparison of Models with Observations 52 $,$ Time Parameters of the Model 56 6. MICROSEISM AND METEOROLOGICAL CONDITIONS $9 6.1 Correlation of Meteorological Activity with Time-Varying Microseismic Trends %9 6.2 Discussion 7 4 7. OBSERVED EARTHQUAKE SIGNALS 75 7.1 Data Reduction Techniques 75 7.2 NORSAR Detection Threshold for Teleseisms 76 7.3 Average Signal Frequencies of Teleseisms 79 7 • U Discussion 70 8. SUMMARY 83 9. ACKNOWLEDGMENTS 84 10. REFERENCES 86 APPENDIX 87 111 ABSTRACT Spectral density and coherence estimates, depicting seismic noise recorded at the Norway Seismic Array by the Earthquake Mechanism Labora- tory, show the following for the P signal band: (1) the difference be- tween the surface microseism field and that sensed in a 60-m well is normally less than 1 dB, even at times of high wind velocities j (2) the coherence between adjacent subarray sensors is quite small. The coherence of microseisms, particularly at frequencies of 0.1 to 0.6 Hz, shows a strong azimuthal dependency. At lower frequencies, the estimated coherence is greatest in the direction of apparent propagation, perhaps indicating that the noise field is a propagating interference pattern. A mathematical model of the coherence of the postulated inter- ference process agrees qualitatively with the observations. Analog and digital estimates, produced to describe the time-varying microseism trends, demonstrate that the noise field is approximately space stationary at the eight center subarrays. There is a weak correla- tion between time-varying barometric gradients and time- varying microseisms for the bands near 0.3 and 1.0 Hz. For epicentral distances of 20 to 90°, the single seismometer P de- tection threshold ranges from approximately U.3 to 5.U i%. Earthquakes perceived showed P waves predominantly in the 0.8- to 2.0- Hz band. Signal frequency increases as a function of decreasing magnitude. Key Ifords: coherence of microseisms, interference patterns, microseisms and atmospheric conditions, NORSAR, Norway Seismic Array, detection threshold for teleseisms, spectral analyses of interference processes, spectral estimates of microseisms, surface and subsurface seismic fields. IV NORSAR MICROSEISMS James N. Murdock and John H. Pfluke 1. INTRODUCTION The Earthquake Mechanism Laboratory (EML), an Environmental Science Services Administration (ESSA) research facility, was requested by the Advanced Research Projects Agency (ARPA) to perform tests on instrumenta- tion installed in Norway during the winter of 1968-196 Q (the Norway Seis- mic Array Project). The EML participation is monitored by the Electronics System Division (ESD) of the United States Air Force. Primary objectives of our study program were: (1) to describe the difference between seismic activity sensed in a shallow (60 m) well and at the surface directly above, and (2) to describe the coherence of microssisms recorded by sensors em- placed 2 km to 6 km apart. Secondary objectives included descriptions of the time-varying microseism field and its correlation with meteorological conditions . This paper describes the EML field operations in Norway and presents interpretations of data gathered at the Norway Seismic Array (NORSAR) facility. In section 2 we show map positions of the field installations, describe the chronology of operations in Norway, and comment on the pre- cision of the data* Section 3 presents differences in the seismic field observed from simultaneous recordings of surface and subsurface activity) section U describes time- varying microseism trends j section 5 presents both microseism coherence estimates and our hypothetical noise model. Section 6 gives a description of meteorological conditions correlated with microseisra amplitudes. Section 7 describes the earth- quake signal frequencies observed and gives an estimate of detection threshold for a single NORSAR seismometer. Observations and interpre- tations are summarized in section 8. The appendix describes instrumen- tation used, outlines digital data reduction techniques, and presents seismograms recorded during times of high wind. 2. NORWAY OPERATIONS AND PRECISION OF THE DATA 2.1 Data Acquisition "When completed, the NORSAR instrument will consist of two concen- tric rings of subarrays (named B and C) centered on the subarray 01A. Seismometers, amplifiers, and signal cables for the subarrays 01A and 01B to 07B (see fig. 2.1) were emplaced by NORSAR contractors during the fall of 1968 and the winter of 1968-1969. Each subarray consists of a central terminal vault (CTV), a long-period seismometer vault (LPV), and six permanent short-period (1.0 Hz free-period) vertical seismometer installations. The CTV, LPV, and a short-period seismometer emplaced in a 60-m well occupy a common position in each subarray. A seventh short-period vertical seismometer was emplaced in the LPV to allow simultaneous descriptions of the surface and subsurface seismic fields. The EML signal processing and recording instruments were emplaced in the central terminal vaults (see fig. 2.2 and app.) to acquire informa- tion from the seismometer positioned in the well, in the LPV, and from four remote subarray seismometers. Because high winds might produce noise at the surface that would be attenuated at depth, anemometers 03B LEGEND /K SP station, 60m hole ^ SP - station, deep hole 23 SP - station, shallow hole Figure 2.1. The NORSAR subarrays 01A and 01B to 07B. Hamar is approximately ISO km north of Oslo, Norway K- '50m- EML DATA CONDITIONING AND RECORDING CONSOLES LPV SURFACE SEISMOMETER WELL-HEAD VAULT SNOW, ~0.5-1.5m SUBSURFACE SEISMOMETER m WELL Figure 2.2. Artist ' s concept of the CTV-LPV-well area. ■were installed (fig. 2.2) to provide continuous wind velocity informa- tion. During the first week of January 1969, EML occupied positions 01A, 01B, and OIlB (fig. 2.1). In mid- January, EML in addition occupied posi- tions 02B and 03B. In late February, the instruments at 01A, 01B, and QiixB were moved to positions 05>B, 06B, and 07B. We operated EML instru- ments at positions in the B ring until early April 1969. Figure 2.3 depicts the history of operations and also indicates periods during which significant malfunctions occurred. During the operational period, one NORSAR seismic data amplifier and one NORSAR seismometer failed. The EML instrumental malfunctions we judge significant included two tape recorder stoppages, several instances of tape recorder spiking, and EML amplifier gain drift at position 06B. k 01A 10 - O) JO 20 - 9) *" 30 H tt 40 < u 50 H >- u. ; o 70 - 80 - 90 - 100 01B 01B 02B POSITION 03B 04B 05 B 06B Z^/V, ^^ 05B 06B ■00 -10 -20 -30 -40 -50 -60 -70 -80 -90 100 03B 04B V/SXA gpikas ■HHI rtcordir stopped iiiimomttir or amplifier out Figure 2.3. History of operations and intervals of malfunctions 2.2 Precision of the Data Ihe following calibrations and tests were normally conducted by the field party for each element being monitored (see app. for more details); (1) seismometer frequency response calibrations (one calibration per 10-day magnetic tape), (2) seismic data amplifier frequency response calibrations (one calibration per 10-day magnetic tape), (3) electrical analog calibrations of the anemometer channel (one for each EML instrument package), and (U) system noise tests (one at the beginning and one at the end of each 10-day magnetic tape). Referenced to the first applied calibration signal, from field re- cordings and field logs, we estimate the NORSAR seismic data amplifiers to have normally drifted less than $ percent (at 1.0 Hz) during a UO-day recording period. From laboratory playback displays, we estimate the entire NORSAR system plus the entire EML system to have normally drifted less than 8 percent at a typical subarray. Drifts of 30 percent were observed for the EML amplifier positioned at 06B; this amplifier was re- placed on day 089. From the field anemometer channel calibrations, and from a wind tunnel test conducted at the manufacturer's laboratory, we believe most of our estimated average wind velocities are accurate to within several miles per hour. Ws used two different digital data reduction techniques: one pri- marily for spectral density estimates, and another for coherence esti- mates; both are described in the appendix. For the spectral density estimates, a typical system noise sample is compared with a typical microseism spectral density estimate in figure 2.U. For the coherence estimates, figure 2.5> shows the spectral density estimate of a system noise sample, believed to depict the worst-case system noise, compared with a typical microseism spectral density estimate. In figure 2.U, peaks in system noise approach the microseismic level at 2.8 Hzj this noise peak may have been caused by the EML tape recorder. The peak at 0.6 Hz was introduced primarily in the digitizing process. From re- dundant processing of microseism samples, we have found errors as high as 3 dB caused by the 0.6- Hz noise peak, a noise common to all magnetic 40 £ 30 q < 20 N X "a. E 10 > 35 |-20 tr w S o o- -30 -40 TYPICAL MICROSEISM SPECTRUM ^ 1 M -^v TYPICAL SYSTEM NOISE SPECTRUM 0.1 1.0 FREQUENCY (Hz) 5.0 Figure 2.4. Typical spectral density estimate of a micro- seism sample compared with a spectral density estimate of system noise. 40 30 2 20 < N X £ 10 a. E g -10 z UJ Q CC UJ g-20 -30 -40 -^T TYPICAL MICROSEISM SPECTRUM T- — - SYSTEM NOISE SPECTRUM \ \ 0.1 1.0 4.0 FREQUENCY (Hz) Figure 2.5. Typical spectral density estimate of a microseism sample used for coherence estimates compared with spectral density estimate of system noise. tape channels. Figure 2.5 shows a signal-to-noise ratio of more than 30 dB for the 0.1- to 1.0-Hz band, and more than 1$ dB for the 1.0- to 2.0- Hz band. Our method of comparing two different time series, described in the appendix, can produce significant bias if large delays exist between the series (see Jenkins and Watts, 1968, p,399). To estimate the bias in our computations, caused by time delays, we selected a pair of seismo- grams whose cross- correlation function showed a peak at 1.2 s (repre- sentative of the larger offsets observed in our data). One of the time series was translated 1.2 s to make the cross-correlation peak occur at the zero lag position. Coherence estimates for the translated and un- translated sets, compared in figure 2.6, show that the two agree to within 0.1. 2.3 Production of Seisraograms in Norway Professor Markvard Sellevoll, Director of the seisraological observ- atory at the University of Bergen, kindly made playback facilities avail- able for our use at Bergen. At the playback facility (fig. 2.7) we per- formed the following: (1) Produced paper seisraograms from our field magnetic tape re- cordings at 7-hour real time intervals (see app. for frequency response of the EML-University of Bergen system). (2) Monitored the filtered output of a surface seismometer channel with two root-mean- square voltmeters ; one meter 1.0 0.8 0.6 ui O z Ui cc ui O 0.4 o 0.2 0.0 n NOT TRANSLATED ""-^---^. 5 / /i TRANSLATED o / '' i >■ m — V "s ° A« / \ \ \ m I x. v, / / » O / i\ i 1 / i V \ \ o \ | / / 1/A >> -A \ \ \ \ v/\y v v v \ V \ I \ 1 \ CO - d tau (s) 1-5 +5 \ i i i i 1 i i i i 1 i i i i I i i i i L \ ■ \\ FROM \\ l> TRANSLATED \\ SERIES \\ \\ \ 1 \ A L i * v i \ / 1 \ /7 \ 1 7 \ lv 1 / 1 if \ lU A^ V 1 ' 1 H 1 if r JL- li V» Ju y l y \0W 0.1 0.1 3.0 FREQUENCY. (Hz) Figure 2.6. Bias in coherence estimates caused by off-centered correlation function. The bias is the difference between estimate of the centered (translated) series and that of the of f -centered (not translated) series. 10 displayed the output bandpass filtered at 1.0 Hz real time; the second, at 0.3 Hz. Logs were made of the meter readings at 7-hour real-time intervals. (3) Monitored the anemometer channel at 7-hour intervals by using a counter to display the frequency of the anemometer output. The anemometer produces a signal -whose frequency was proportional to wind velocity. Paper sei sinograms made at the University of Bergen will be referred to in this report as "sei sinograms having a time base of 3 mm/s" or words to this effect. DATA FROM 5 .SEISMOMETERS ^/^ AND TIME BASE FREQUENCY RMS RMS COUNTER METER METER Figure 2.7. Artist's aonaept of playback operations at the University of Bergen. 11 3. DIFFERENCES BETWEEN THE SURFACE AND SUBSURFACE SEISMIC FIELDS 3.1 Microseisms To describe the differences between the surface and subsurface microseismic fields, we computed spectral density estimates from seismic data recorded simultaneously by the subsurface and surface sensors (see fig. 2.2 for their relative positions). Data were chosen for spectral analyses to include samples of (1) recordings made during times of both high and low wind velocity and (2) recordings made during times of both high and low microseismic activity (see fig. 3.1). Techniques for pro- ducing the spectral estimates are outlined in the appendix. Also for each sub surface- surface pair, the frequency response and magnification of the subsurface sensor were normalized to those of the surface sensor. All adjustments were made from field- recorded calibrations, and we be- lieve that the precision of the normalizations is within 1 dB. Plots showing the differences between the spectra, estimated from the simultaneous recordings, are given in figures 3.2 to 3.9. The plots depict the surface power spectral density estimate minus the sub-surface power spectral density estimate, expressed in decibels. Mind velocities, recorded during the time frame used for each spectral estimate, are also indicated. In general, the illustrations demonstrate that there is only a small difference between the surface and subsurface recordings, even during times of high wind velocity. These data, summarized in histograms and graphs of cumulative probability distribution in figures 3.10 and 3.11, show that 97 percent of the 71 estimates produce differences of less than 3 dB. For recordings made during high wind velocities, 88 percent show 12 o o 00 o 1^ s o 2 8 < < I J o o M X Ql" < 2 lit >■ Si O s 5 o CO 8 HI D < 2 V) 1X1 t= < 2 K W UJ >- UJ H Q — < Z 5 LU Q - n r J UJ O z UJ DC UJ I O O < co co eo £00000002228"°"*^° OOOO OOOO^^ 010 < • CO 05 O is Si •^ £ co v • £ Si O « to 05 4^ £ 05 s a, w v CO r£ <3 Es ^ is a, 05 H 3 • 05 O ^3 Cl, 05 «K S o is £ 3 Si is to 05 is 3 £ CO 05 CO O is CO £ 05 O £ 05 is 05 v£ O i3i O ^ Ps 05 NOU-ISOd aanindiAiv 13 20 30 40 ~I 1 I I CONTOUR INTERVAL = 1 dB i — i — n WIND AVG. GUST 2 1 10 £,-1.0 -1.0 ■" 25 0) ID 01 •" 50 - 1 < 111 > IL o > 60 < Q 2 _ 2 5 28 70 - 7 3 17 _ 3 5 - 4 10 80 2 2 2 I I I I I I 1.0 2.0 5.0 FREQUENCY (Hz) Figure 3.2. The difference between power spectral density estimates made from simultaneous surface and subsurface recordings at position 2B . The differences are for the subsurface . Band displayed is from 0.3 Hz t.o 5.0 Hz. The discontinuity between days 28 and 30 was probably caused by an adjustment of the EML field amplifier . Ik 20 30 40 WIND -AVG. GUST - 1 — ■■ -3 — » I — m - 1 — • - 4 - 8 - - 12 28 - 13 20 - 9 12 01 10 0> 50 < UJ > u. o > < Q 1 60 70 80 1 1 1 I I CONTOUR INTERVAL = "i — n 2 13 25 15 20 1 4 2 4 <^ - 1.0 A s J L_L 1.0 FREQUENCY (Hz) 2.0 5.0 Figure 3.3. The difference between power spectral density estimates made from simultaneous surface and subsurface recordings at position 03B. The difference are for the surface minus the subsurface. Band displayed is from 0. 3 Hz to 5.0 Hz. 15 50 60 90 100 "i 1 — I — i i i I r CONTOUR INTERVAL-1 dB i — i — n WIND AVG. GUST -»- 6 17 18 33 - 14 32 — S 70 - 5 0) — 2 . - B < 111 > u. O 80 > 2 < a - ■*= y: -i.o 7 15- 12 20-i 10 -1.0* J I I I I I 1.0 FREQUENCY (Hz) 2.0 5.0 Figure 3.4. The difference between power spectral density estimates made from simultaneous surface and subsurface recordings at position 05B. The differences are for the surface minus the subsurface . Band displayed is from . 3 Hz to 5.0 Hz. 16 88 4 90 16 92 01 (D O) «- 94 K < Ul > IL o >■ 90 < Q 98 100 T 1 I I I I | i — r CONTOUR INTERVAL - 1 dB WIND AVG. GUST 17 28 16 33 f J I I I I J L 1.0 2.0 FREQUENCY (Hz) 5.0 Figure 3.5. The difference between power spectral density estimates made from simultaneous surface and subsurface recordings at position 6B . The difference are for the surface minus the subsurface. Band displayed is from . 3 Hz to 5.0 Hz. 17 60 90 100 i — i — r i — r CONTOUR INTERVAL - 1 dB WIND — j HVG. GUST 13 24 _•-, « 22 35 — ■- tf! 1- ' a. < 70 ID 1 — -%*. > U. O 7 — «• > < Q 80 2 - 20 33 Z 13 13 . 18 27 & 2.0 2.0 J I I I I I 1.0 FREQUENCY (Hz) 2.0 5.0 Figure 2.6. The difference between power spectral density estimates made from simultaneous surface and subsurface recordings at position 07B. The differences are for the subsurface. Band displayed is from 0.3 Hz to 5.0 Hz 18 1 1 ■I 1 T I 1 | 1 01A. DAY 38 ~T~ — T" ~T" 3 WIND: AVG. - - 10 mph GUST - - 16 mph CD 2 2 - A ~ UJ o z 1 _ UJ ^^~ 50 — > t- — ■ ^"^ 2 H -i O jl 40 - < £ < I O > 3 K — K O a. UJ U 30 - ca • | 1 | 1 1 o 2 4 6 LL O 20- DIFFERENCE (db) O z 10 — — 1 1- — f— — 1 Figure 3»10. 4 5 6 DIFFERENCE (db) Summary of observed differences between bhe surface and subsurface microseisiaic fields. 7 -| Z O 6 < 5 H > K 4-| LU vt aa 3 H O 2 - 1 — 1 — i uj > > "> < a = 2 2 O o a. 1 i ' i ■ r 2 4 DIFFERENCE (db) 2 3 4 DIFFERENCE (db) Figure 3.11. Summary of observed differences between the surface and subsurface microseismic fields during times of high winds. 21 differences of less than 3 dB. In the dominant teleseisraic signal band (0.8 to 2.0 Hz), 96 percent of the 71 estimates differ by less than 1 dB, and all differ by less than 3 dB. Examples of seismograms recorded during times of high winds, given in the appendix, also show that the sur- face and subsurface outputs are almost identical. 3.2 Seismic Signals Figure 3.12 shows tracings of teleseismic signals recorded by the surface and subsurface sensors of six subarrays. The paired seismograms appear almost identical. 3.3 Other Observations In addition to the data presented in this section, we and our co-workers at EML have examined more than 1,000 seismograms (having a time base of 3 mm/s, showing microseisms, teleseisms, local quarrying blasts, and cultural noise. From studying simultaneous recordings made by surface and subsurface pairs having equal responses and magnifications, we observed, for each pair the following: (1) Teleseismic recordings show almost identical wave forms and amplitudes. (2) MLcroseismic recordings almost always appear identical, even during times of recorded high wind velocities; at times of re- corded high winds, we produced seismograms having a time base of UO mm/s. Exceptions were found on the 06b seismograms, which sometimes showed the surface approximately 6 dB higher 22 10 1 WELL 01A 01B SURFACE 02B 03B 04B 05B 06B 07 B Figure S.12. Tracings of subsurface- and surface-recorded teleseisms . A seismogram from each subarray is displayed. 23 than the subsurface in the band near 6 Hz. (3) Cultural noise (e.g., made by vehicles) generated "within several hundred meters of the site sometimes showed ampli- tudes approximately 6 dB higher on the surface sensors than on the subsurface sensors. Usually, the cultural noise we found, from our audio monitoring procedure, appeared very near the detection threshold. From our data -we judge that cultural noise makes a negligible contribution to the overall noise field. (U) Local quarrying explosions sometimes produce 10 to 1$ Hz P signals having amplitudes noticeably higher on the surface seismograms. This is probably an effect of the free surface. U. PARAMETERS OF THE TIME- VARYING MICROSEISMIC FIELD U.l Estimates by Analog Techniques To reduce the field- recorded data by analog techniques, we: (1) processed the vertical component microseismic recordings with bandpass filters centered at 0.3 Hz and 1.0 Hz, (2) made logs of the root-mean square filter outputs at 7- hour intervals (sample length, approximately 10 to 20 min), (3) applied magnification corrections to the logged meter readings, and (U) plotted the root-mean square amplitudes versus time. The meters used for monitoring the filtered outputs were constructed of 2U a linear law rectifier, designed to detect the absolute average value of a sine wave and convert the absolute average value to the root-mean square value by the meter scale. Mien a Gaussian signal is applied to the input of this meter type, the values read must be adjusted by a fac- tor of 1.13 (Bendat and Piersol, l°6Li, p. 2-17). Because the wide-band seismic noise appears to have a Gaussian probability density function (see app.), we assumed Gaussian input to the root-mean square meters for the data used in estimating statistical parameters of the micro seism field. For other data, showing relative trends of the time varying microseism field, only magnification corrections were applied (i.e., corrections for Gaussian input were applied to figs. Iw5> through U.8 only). Measurements of the root-mean square amplitudes of microseism are presented in figures Lul and I4.2, which show the time- varying trends of the 0.3- Hz and 1.0- Hz microseisms. In both illustrations, the trends at the different recording positions are remarkably similar, demon- strating that amplitudes of the monitored microseismic field were ap- proximately space stationary. A comparison between those charts shows that the time-varying trends of the 0.3-Hz and 1.0-Kz microseismic ampli- tudes are generally similar, although noticeable differences do exist, such as in the 10 to 20 day interval. U.2 Estimates by Digital Techniques Spectral density estimates of microseisms recorded in the LPV's at positions 02B* and 03B are plotted as a function of time and frequency in 25 CO m CO a m in r» CM m (O o o o • o o o o o O O o o o o o o O O O O O O O o o o o o O O o o o o o o o o o o O O O o (0 « CM O 1 1 1 t CM o 1 1 1 (O 9 CM O 1 1 - o o " — o ) r > V o - "' 1 1 1 «» > < - M _ '. • • \- - EC! _ V 2 ' • # ~ o t .> o " CO — O) .; 1 ( , \ o> . <2> s •_ •• \ " o CO : x 'i V i l' ( 1 1 \ o " s S • :' •> ' .i ..■ 1 < $ r ** ) •• o s 5_ * , V '€ • • \l .> 9 •^ - o ~ 10 ; I } <• t ) i i \y ) » i > • 1 o CO S to 0) to o _ in i uj "J - ■ > r so: ID UJ co o \| > s * CO * * 'I • ] > % J SO = Uj Ul .i H> \ oo o o E o to - 1- > \ t ' •• ■ r w * J > s ~ o _ * oo o •■ ,4 *•• . / > ■. 1 to 4^ - > L 1 > / • • t \ s ~ o _ w *■ «• : « r s < ■ i • r, . /' < s •l i o " co~ a i — . * I V \, E -n 2 r i s } •^ r-s En . ~ o _ cm J" V J •: > j.' > /•' 1 o " CM . 01 OB 0) • • - /.' • - . •i • v) — . % oc ^ t^ ~ # i .. < _ .• F • • ui Q) ca o _ r- . • \ # { o *| s to **, U. (35^^ — . O \i - ; • 1*4 — ' \ > _ _ < - - D _ i i o o o o 1 1 o o o o i i i o o o o 1 1 o o o o 1 1 o o o o o o o o O O O O o o o o O O O o o o o o U9 *■ CM O (0 t CM O (Sim a ID 1 IN O •0 * CM O TdlAIV e « n o < D a a ffl «- ^ CM n » o o O o o 26 CO * M O m r- o CO ■* CM _i L m CM o CO ^ CM O J I I l_ CD 9 _l l_ 00 CO o CD 9 (M _l L j! ! s Q -i uj w .! I- > •.' v> o >* CO ' >" 5 f f UJ UJ I- > i \ -i — i — i — r CO 9 CM O ) SO UJ Ul I- > J « £ co CO CO o o CO c • +J • s • Si 0) ?^ CO i 05 0) > H H J <5 5 O 3 K - U o. -• 4 5 6 7 8 9 AMPLITUDE (mju /Hi) I 10 AMPLITUDE (m^ /Hi) Figure 4.7. Density and distribution of mioroseism ampli' tudes in the band near 1.0 Hz. AMPLITUDE ■ 01 X02 - 00X02 - 02X03 I I I I Mill I I I I I I Mill I I 65 0.3 Hz Figure 5.4. Coherence charts and coherence polygon for sub- array 2B . 38 0.1 1.0 £R EQU ENCY (Hz) 2.5 0.1 1.0 2.5 Tf TT cc O 00 Z in 00 z lu 4 UJ m UJ o z < W I I I I Mill I 00X03- 03X04 00X01 02X03 01X02 oi 79, 1449 ——I \\Vpi' i u _ii^> ( Cj)\ 08 ^^) y'o? I I I I Mill I I ARRAY MAP 03B □ 01 O 06 A' Qoo 003 02 2 10 KILOMETERS Figure 5.5. Coherence charts and coherence polygon for sub- array OSB. 39 0.1 1.0 FR EQUENCY (Hz) 2.5 0.1 1.0 2.5 1 I I I I Mil f TTT E W 2 O c/> Z LU v> iii 4 S LU o z < 38, 1415 00X04 02X03 00X03 00X05 04X05 00X02 03X04 .0.1-— ^~~~~ 43, / 0.4 5*< 2300 ~-^£s ^\^J,u ft ii I I I I Mill I I I I I I Mill I I 38 0.3 Hz ARRAY (O)o6 MAP w 04B m oi - j □- H ° 2 10 KILOMETERS Fvgure 5.6. Cohevenoe charts and coherence polygon for sub array 04B. U0 0.1 1.0 FREQUENCY (Hz) 2.5 0.1 1 .0 2.5 I I I f I I I | TTT E je co 2 EC O to z iu CO UJ 03 LU U 2 6 < V) a 65, 1422 I I I I Mill I 65 0.3 Hz 04X05 02X03 01X05 03X04 01 X02 -N- 81, 0450 . 0.4 0.2 ., 02 fl w 63 I I I I Mill I ARRAY MAP 05B O ©1 , o- I □- □ T 04 02 10 ID KILOMETERS Figure 5.7. Coherence charts and coherence polygon for subarray 5B . 1*1 FREQUENCY (Hz) 0.1 1.0 2.5 ] — i i i i iiii| — rp ir O w z tu ai 5 l- a 89 0.4 Hz 00X03 00X04 00X02 01X02 00X01 03X04 02X03 I I I I Mill I I tf? ;t\oN o\* E CT»ON 89 0.3 Hz AP PARENT PRO PAGAT,ON OPTION ARRAY | — | MAP I • |Q» 06B □ 01 Q]o4 Q]00 A 1 03 13 □ 02 10 KILOMETERS Figure 5.8. Coherence chart and coherence polygons for subarray 06B. Shown are polygons for 0.3 Hz and 0.4 Hz; both show the 0.7 coherence contour. 1*2 0.1 1.0 1 — i i i i iiii| — rj FREQUENCY (Hz) 2.5 0.1 I I I I MM 1.0 2.5 H" E Z 2 EC O v> Z ai v> I 4 UJ S H Ui 00 UJ < H (0 72, 1910 \ " i 01 0.! \ °; i i « %i 91, 1432 • 00X04^ JL-01 X05^ 3-00X05^ '0.8 2^5? r? M dP ^-00X01-^- 0.6 . iff? 0-2 0.4 \ ' V 1,12 ^-04X05-^ ' ' ■All I I I I Mill I I I I MM ARRAY s-\ map Ky 07B | B " Q A 01 A" oo □ 03 O 02 10 KILOMETERS Figure 5.9. Coherence charts and coherence polygon for subarray 07B. U3 FREQUENCY (Hz) Figure 5.10. Coherence estimated from two unrelated samples (2) Normally, the coherence displays a strong asimuthal dependence, especially at lower frequencies. This dependence is particu- larly well exemplified by the elongate shapes of the polygons, which indicate that, at the lower frequencies, the highest co- herence normally is in the direction of apparent microseism propagation— as observed previously by Bungum et al. (1969). (3) The coherence trends, shown by the contours on the frequency versus distance charts, are time variant, i.e., at a given sub- array, the charts constructed for different times are somewhat dissimilar (in particular, see fig. 5.5). (h) The distance-frequency plots of coherence often display dis- tinct secondary maxima and minima. Figure 5.11 shows coher- ence, plotted as a function of fiequency only, of selected hh sample pairs used in constructing figure 3>.2. From a compari- son of figure 5.11 with f?.10 we see that the secondary maxima are well above the level of randomness. Secondary maxima also appear in coherence estimates made by Rygg, Bungum, and Bruland (1969). 1.0 - 0.8 ~i 1 1 — i i i 01A, DAY 26 01x02 1.0 1.0 2.0 FREQUENCY (Hi) 0.0 T 1 1 1 — I - T 0.1 01A, DAY 26 01x05 _i i i ■ ■ ■ FREQUENCY (Hz) 1.0 2.0 1.0 -i 1 1 — i — i i i i 01 A, DAY 26 03x04 1.0 1.0 2.0 FREQUENCY (Hz) tf 0.8 - 0.6 - -i 1 1 — i — i — i i i 01A, DAY 26 04x05 1.0 0.2 FREQUENCY (Ha* Figure 5.11. Estimated coherence between sensors of sub- array 01A. The sensor pairs are indicated on each graph. Map of 01A is shown in figure 5.2. US Our studies suggest to us that an interference process produces the azimuthal dependence and secondary maxima observed. The following section presents a model giving a physical interpretation of these observations. 5.3 Model of the Coherence Variations An interference pattern of two uncorrelated seismic wave trains cross- ing a pair of sensors is depicted in figure 5.12. Respective wave fronts of the two trains reach sensor 2 t-, and t_ seconds after having crossed sensor 1. The sensors are separated by distance x, and the waves propa- gate at a constant velocity. The wave trains are assumed to propagate un- dispersed and unattenuated between the sensors and ground displacements along wave fronts remain constant with respect to lateral distance from the ray trace. Following the reasoning of Jenkins and Watts (1968 p. 329), one can think of wave trains as passing through a network of linear systems as shown in figure 5-135 two propagating wave trains enter the network on the left, are combined as indicated, and emerge as the signals X-,(t) and Xp (t) detected by two sensors. The outputs are then expressed as: Xj(t) = /J h^OO'Z^t-iOdu + /" h 12 (u)-Z 2 (t-u)du, (5.2) and X 2 (t) = f 1 h^OO-Z^t-iOdu + /p h 22 (u).Z 2 (t-u)du. (5.2) For the model, the impulse response functions h(u) are the dirac delta functions shown in the network diagram. The above equations then become Xj(tO - Zj(t) + Z 2 (t), (5.3) U6 z 1] is the expectation operator. By taking E[X.(t)«X (t)], we obtain the covariance functions Y n (u) = I o ± 2./~ h li (v)-h li (v+u)dv, (5.10) Y 12 (u) = Z a. 2 -/p h li (v)-h 2i (v+u)dv, (5.10) k9 Y 21 (u) = I o^'T h 2i (y)-h li (v+u)dv, (5.10) and Y 99 (u) = 1 a 2 -r h (v)-h fv+u)dv. (5.10) 22 1 21 2i The covariance functions Y 12 ( u ) andY 2 i( u ) are zero for negative values of the time parameter v, thus allowing us to write them in the form Y 12 (") = Zo i 2 /:h u (v)-h 21 (v+u)dv, (5.11) and Y 21 (u) = Io i 2 /:h 2i (v)'h u (v+u)dv, (5.11) Upon taking the transforms of (5.10) and (5.11) we obtain expressions for the power spectra and cross-spectra of the wave trains X^(t) and X«(t) in terms of the transfer functions EL. (f) and EL. (f). Hence, P u (f) - Io ± 2.\n i±(f) \2 t (5>l2) r 22 (f) = I a i 2 -|H 2i (f)| 2 , (5.12) T 12 (f) = I o i 2 'H 11 *(f)-H 2i (f), (5.12) and r 21 (f) - Z o i 2 -H 2 .*(f)-H li (f). (5.12) Then, substituting (5.9) into (5.12) we obtain the power spectra and cross-spectra as functions of the variances a 2 and the travel times t. of i i the respective wave trains Z. (t) crossing the two sensors: 50 r u ('f) = I^ 2 , (5.13) r ,(f) = lo. 2 , (5.13) 22 l ' r i0 (f) - I a 2.e-J2irft t , (5.13) 12 i and ^ 2 j2irft. (5.13) T 21 (f) = I o ± -e i . With the above expressions "we now determine the coherence function for the output -waves X-,(t), X„(t) to be r (f)-r (f) < 12 2 (f) =— -^— , (5.14) r n (OT 22 (f) or since * r 2 ,(f) = r *(f), (5.i5) 21 vw *12 |r 1? (f)| 2 < 2(f) = 1 2 - 12 r u (f)-r 22 (f) (5.16) (5.16) which, when substituting from above, becomes II a i 2 -o k 2 -cos[2TTf(t i -t k )] Ti^TV * Backus et al. (196!*) found that the absolute value of a zero order Bessel function describes the coherence of isotropic noise; for this special case, equation $.]£ approximates the square of a zero order Bessel func- tion with argument 2irfx/V. 51 The spectral phase difference between X-^t) and X2(t) is -Im(r 10 (f)) 6 (f) = tan" 12 1 " }' ' Re(r 12 (f)) ' 2 /I a i •sin(2Trft 1 )v tan 1 4 \, (5.17) l T a z «cos(2TTft ) ; i i and © n (f) - - 12 ( f >. (5.18) $,h Comparison of Models With Observations In this section, we compute the model coherence, described by the positive square root of (£.16) for an array of six sensors positioned as shown in figure 5«lUj all computations are for the center sensor, plus one of the others. The model velocity is 3.5 km/s. As a first example, we take a noise field produced by two equal out- put white sources arranged as shown in figure 5.Hia. The model coherence for this example, figure 5«l5j shows a strong azimuthal dependency, as 1 well as secondary maxima occurring at integral multiples of The amplitude of the secondary maxima, however, are larger than those ob- served from the real data (compare fig. 5. 15 with fig. 5.H). For a more realistic example, we take equally spaced, equal output white- noise sources (fig. 5»lUb) modeling a broad source area. The model coherence for this example (fig. 5»l6) shows features, including azimuthal depen- dency and secondary maxima, similar to those of the real coherence esti- mates (compare fig. £.16 with fig. 5.11). Furthermore, the coherence 52 2 SOURCES POSITION 00 OF ARRAY MODEL N I ARRAY MODEL • 01 1 • 05 -N- 1 •oo • 02 • 04 • 03 5 KILOMETERS . . I 10 25 SOURCES POSITION 00 OF ARRAY MODEL B Figure 5.14. Model of two and 25 sources 53 1.0 lfgNO.8 0.6 w 0.4 LU z o O 0.2 0.0 1 = f» i* -J A L_ A h V i» ♦ l 1 l 2 \ ' / / I I J I f J 00x02 -J L " r 1.0 05 FREQUENCY (Hz) 1 .0 2.0 c^ 0.8 £ ^ 0.6 LU O z LU 0.4 LU I O u 0.2 0.0 1 ^ I /\ I V I <« — 1 2 l = 0.s«: I ' / \ I i T \ / j 00x04 00x05 \ I i * L I I 0.05 1.0 2.0 FREQUENCY (Hz) 1.0 iimN 0.8 * 0.6 U | 0.4 LU z o O 0.2 0.0 0.05 1 r .,1 ' A It t AR / \ lt 1 t 2 , . / \ I I I I I I I II \ J \ 00x01 00x03 1 i [ \i \J 1.0 2.0 FREQUENCY (Hz) ARRAV ' MODE". • 01 I • 05 • -N- I • 04 •oo • 03 • 02 i I 5 KILOMETERS 10 Figure 5.15. Model coherence for the two sources displayed in figure 5.14. The model pairs are indicated on each graph. 5k 0.0 0.1 I I I I I I 1.0 FREQUENCY (Hz) 1.0 2.0 MODEL 00x02 1.0 2.0 FREQUENCY (Hz) 0.0 0.1 J i i i '''' 1.0 2.0 FREQUENCY (Hz) ARRAY MODEL • o^ 1 • 05 -N- 1 — • 04 •oo • 03 —i ©02 5 KILOMETERS 10 Figure 5.16. Model coherence for the 25 sources displayed in figure 5.14. The model pairs are indicated on each graph. 55 polygons of the model (fig. 5.17) have shapes similar to those constructed for the real data. Dissimilarities, such as amplitude differences between the modeled and the real data, probably result from the simplifying assumptions made for the model — in particular, the model contains no opera- tion decreasing the coherence as a function of distance between sensors. In the real earth we expect a distance-dependent function to exist. An interference process, modeling the noise field, permits time- varying coherence as measured by the simple cross- spectral method. The change in computed coherence can be caused by (1) change in position of the sources, i.e., t. of (5.6) changes, and (2) differential change in the output of the sources. We, in section 5.2, and Bungum, Bruland, and Rygg (1969) observed time varying coherence at NORSAR, as well as time varying orientations of the polygons — observations suggesting that the source is both broad and mo- bile. Bungum and his co-workers make similar conclusions, enhancing our confidence that the model is realistic. 5.5 Time Parameters of the Ifodel In the development of the coherence model, no attempt was made to fit the resulting model phase spectra to the NORSAR data. A valid compari- son requires both phase estimates with minimum bias (especially at the lower frequencies) and surface-wave dispersion functions for the NORSAR area; neither are available. Even though the model phase spectra may be a poor fit to the real data, some features of the model are noteworthy. 56 0.5 Hz 0.4 Hz Figure 5.17. Coherence polygons made from the model coherence for 25 sources. The propagation direction of the central ray is shown. 57 The phase spectra for the 25- source model (fig. 5.110 is shown in figure £.18. The sensor pair (00x02) alig2ied perpendicular to the center ray produces approximately zero time delay at the lower frequencies, an observation we intuitively expect from the symmetry of the source. At frequencies higher than that corresponding to the first minimum in the coherence (see fig. 5>«l6), the 00x02 phase angle fluctuates between and 2.0 - 1.0 0.0 A VG. OF INPUTS - 1.00 t I J L AVG. OF INPUTS - 0.00 i ' ' ' ' 0.05 0.1 0.5 1.0 2.0 FREQUENCY (Hz) Figure 5.18. Interpreted tau of the 25-sourae model shown in figure 5.14. it , making the time delay difficult to interpret — we assumed increasing phase to plot the time delay. For the other sensor pairs, the conversion from phase to time delay was more straightforward. At the lower frequen- cies, the pairs also produce time delays that are approximately those of the averages of the respective suites of rays (see fig. 5»l8). At the higher frequencies, however, fluctuations in the time delays are less pro- nounced than are those of 00x02. 58 The model velocity was a constant 3«5> km/s, but the time delays (fig. 5.18) vary as a function of frequency.. The set of time delays pro- duce the illusions of a dispersive medium and of a frequency dependent azimuth of approach. 6. MICROSEISM AND METEOROLOGICAL CONDITIONS 6.1 Correlation of Meteorological Activity With Time-Varying Microseismic Trends To search for correlating trends at NORSAR, we compared gross fea- tures shown on surface meteorological maps, obtained from the Norwegian Weather Bureau in Bergen, with microseism amplitudes measured at times corresponding to those of the maps (as described in sec. iul, microseisms were sampled at 7-hour intervals; each map comparison was made to the amplitude of the nearest sample). Selected maps are compared with time- correlating microseism amplitudes, at 0.3 Hz and 1.0 Hz, in figures 6.1 and 6.2. To facilitate referencing specific examples, each map is given an alphabetic designation. The map designations are chronologically ordered in figure 6.3, which also shows chronologically ordered microseism trends . Because of popular concepts of microseism generation by atmospheric- induced marine conditions (Longuet-Higgins, 1950, 1952', Hasselmann, 1963), and because of previous correlations we (Murdock et al., 1968) found from studying a microseism field in Australia (fig. 6.I4), we searched for correlations of variations in microseismic amplitudes with variations of meteorological conditions over the oceanic and Scandinavian coastal areas. W3 found that a large band of amplitudes — representative of the entire range — existed for a given meteorological condition; in particular, 59 20 40 50 60 70 DAY OF YEAR, 1969 too Figure 6.3. Microseisms trends and times described by meteorological maps. The alphabetic designation corres ponds to figures 6.1 and 6.2. representative values of the entire microseism amplitude range exist when (1) low pressure zones are over the east Atlantic Ocean (see map set B, C, F, G, K, N, Q, T, X), (2) high pressure zones are over the east Atlantic Ocean (see map set D, H, I, J, M, V), (3) steep barometric gradients are at Scandinavian coast lines (see map set A, E, P, R, S, T, D, W, X). Although variations in meteorological conditions in the coastal and oceanic areas appear to show no correlation with variations in the 0.3- Hz and 1*0-Hz microseisms recorded at N0RSAR, variations in the barometric gradient on the peninsula near N0RSAR do seem to correlate with micro- seismic variations — low gradients occurring at times when the microseismic amplitudes are small and high gradients occurring at times when the ampli- tudes are large (see map set B, E, F, H, I, L, R, T, U, V, W, X). Similar correlations (fig. 6.5) were found by us (Murdock et al., 1968) from data recorded in Australia— in addition to the correlation shown in fig. 6.U. 60 (Cv ' SKI ™ u/ j "S* 1 n ^^^ 0r~ • fQJb» oiiuw a iooi JS- (H h (-1 h 1_T-I KUONETEM s--P ( r-^V^ J rCLv ( %%\ q w i u-^-^ 3 O CElffEB ^S //^ • rvJ^° l "^* UMW 1000 ^_ KILOKTEn e^-> ( Y" Mt - 2l2 ^" l °'' 2.6 mfi 2.6 mfi 2.7 m/i 2.8 m/i 2.8 mfi 2.8 m^i Figure 6.2, (continued) 66 JJ > ' / / //, >^n~ \0nsdag29/l Kl 04. - '/ icnonb _^S / y*/>v^YTHTTttl±UUu 'TTZ^ /y^- Jlu ^^—^ / ///$ffl^\\ ) w w \% afe ^^T*Vt L pli^s I \\\v a v %$A / ) \#/ U v«M_3^5r^ r M 2.9 m/a 3.1 mil Q 3.1 m/u. 3.2 mil 3.5 mil 3.8 rn/x Figure 6.2. (continued) 67 4.2 mil 4.6 mil 4.5 n\fi Vl.HOV LS. p 'I 4.7 m/i 5.3 mit w 5.4 mi* Figure 6.2. (continued) . 68 •3r- O > IT) O < -.002 Figure 6.4. Comparison between barometric pressure gradient ooauring over the Gulf of Carpentaria and root-mean-square amplitudes at 1.0 Hz recorded at Daly Waters, Northern Territory . •3r- V) o > 2 .05 RMS VOLTS GRADIENTWITHIN 500 KM OF DW3 01 % > E Q < cr .002 31 10 15 20 25 30 O < 10 Figure 6.5. Comparison between barometric pressure gradient in the Daly Waters area and root-mean-square amplitudes at 1.0 Hz recorded at Daly Waters. 69 For further correlations between barometric gradients and microseismic activity, we obtained detailed meteorological maps, compiled at 12-hour intervals, from the Norwegian Meteorological Institute in Oslo. Pilot studies performed (figs. 6.6 to 6.8) included comparisons of the micro- seism trends with the following estimates of barometric trends : Maximum gradient : Px - Pc , R where Px minus Pc is the maximum difference between the pressure at the center of NORSAR (Pc) and that (Px) on a circle of radius R from the center of NORSAR. Maximum-opposite gradient ; Px - Pc + Py - Pc 2R where Py is measured on the circle 180 from Px. Average gradient ; Pn-Pc+Pe- P c + Ps-Pc+Pw-Pc UR where Pn, Pe, Ps, and Pw are measured on a circle of radius R at the four compass directions from the center of NORSAR. N-S gradient : Pn - Pc + Ps - Pc 2R The maximum gradient, determined from measurements made 200 km from the center of NORSAR (fig. 6.8) seems to show one of the better correlations with the microseism amplitudes displayed; the weighted average of the microseisms displayed is (100 X Aj # q + A~ o)/2 sampled at 12-hour inter- vals. Data for days 30 to 100 seem to correlate better than data for days 10 to 30. In some instances, such as between days 35 and UO, the gradient trend seems to lead the microseism trend by 12 to 2k hours. 70 oo 6 - 4 - 2 - 6 CD X CO = E 4 - 2 - 6 " 4 - 2 600 " 400 <" 200 " i :£ 6 1 4 - 2 - 100 km 200 km 300 km —j- 25 30 — j— 35 - 1 - 40 - 1 - 45 ~ r~ 50 55 DAY OF YEAR, 1969 Figure 6.6. Pilot studies performed to find a correlation between barometric gradients and root-mean-square trends The maximum gradients observed for 100 km a 200 km, and 300 km (R) from the center of NORSAR are shown. 71 CO AVERAGE GRADIENT 200 km DAY OF YEAR, 1969 Figure 6.7. Pilot studies performed to find a correlation between barometric gradients and root-mean square trends of microseisms . The radi (R) from the center of NORSAR are indicated on the trends. 72 01 ID 0) K < UJ > O a s s s •^ s H o « 5s £ *K 0) s r« ^ +S Ci ^ <^> SC oa « «K CO O '-y K — ^ » CN SHH'rfm uivqm 01 SW8' rf iu 3IW 39VU3AV S1N3IQV83 SWSI3S0H0IW 73 6.2 Discussion If the correlation we show is meaningful, then it seems to be in conflict with the popular concept of microseism generation by atmospheric- induced marine conditions. Although his conclusions support the popular concept, Santo (1962), by studying microseisms recorded in Sweden, report- ed the following that we believe are consistent with local sources of microseisms : An example showing the poor transmission of micro- seismic waves, especially of short period, is the following. On march 31st, 1962, a cyclone (p c = 980 mb) arrived just at the southernmost edge of Sweden. This produced a remarkable short-period (around 2.5 seconds) microseism storm at Karlskrona, quite near the cyclonic center. But there was almost nothing at Uppsala which is only about 3^0 km apart. The same results are clearly observed in Japan. When a cyclone runs over the Pacific Ocean from south to north, the microseismic storm shifts grad- ually from southwestern stations to northeastern, as if it runs after the cyclone. In other words, micro- seism storms in Japan occur quite independently at many places located near each other. Passages of energy sources across the Baltic Sea hardly increase microseismic amplitudes even at Uppsala, as far as those with periods of more than U or 5 seconds are concerned. However, a small amplitude increase is observed at Uppsala for microseisms of shorter periods . . . (p . 371 ) As to the remarkable storm (microseismic storm) which occurred only at KLruna about 26 d 12 h, the general explanation from the intensity or the move- ment of a cyclone only is impossible. The only indi- cation the writer could find from the meteorological data was a small cyclone (100£ mb) very near KLruna at 26 d 06 h, and that the greatest wind velocity was measured at the northernmost coast of Norway just at that time . . • (p . 361 ) His observations above also suggest to us the possibility that microseisms 74 may be generated over land, as does the correlation we show (fig. 6.8). As an alternate interpretation of the data in figure 6.8, there may be a large time lag between the atmospheric driving force and the marine generation of microseisms (essentially Santo 's hypothesis). The baro- metric gradient near NORSAR, therefore, may be indicative only of past or future marine atmospheric conditions that actually cause the micro- seismic activity — an interpretation supported by the observed westerly approach of the 0.3- Hz microseisms (sec. 5), suggestive of their marine origin. Regardless of the microseisms 1 source, however, the observed correlation may prove a valuable tool in the short-term prediction of impending microseismic levels. 7. OBSERVED EARTHQUAKE SIGNALS 7.1 Data Reduction Techniques For the 20 to 88 day interval, 1969, one seismic channel of each magnetic tape was monitored by an interpreter who listened to the play- back output by using an audio band speaker (playback speed was 80 times real time). Paper seismograms, with magnifications in the range of 2 x 10-^ to k x Kr were produced to show recordings of all sounds inter- preted to be seismic signals. Arrival times of signals seen on the paper seismograms were compared with expected earthquake arrival times, esti- mated from information provided by the National Center for Earthquake Information in their Preliminary Determination of Epicenters bulletins (PDE cards). A list was made containing the following information: (1) earthquakes that produced time-correlating signals at NORSAR, (2) earthquakes that did not produce time- correlating signals at 75 NORSAR, (3) magnitude (nO of each earthquake, obtained from the PDE cards, (U) epicentral distances from the center of NORSAR for each earth- quake location given on the PDE cards, and (5) the focal depth of each earthquake given on the PDE cards. From each paper seismogram showing a clearly recorded earthquake signal, we measured the dominant signal frequency occurring within ap- proximately 2 s after signal onset. An average signal frequency was then computed for each earthquake. 7.2 NORSAR Detection Threshold for Tele seisms From the earthquake list described above, we constructed figure 7*1* which shows the detection threshold of a single NORSAR seismometer for shallow (h£50 km) tele seisms, and figure 7.2, which shows the detection threshold for deep (h>50 km) teleseisms. In these illustrations, trend lines describe both the upper magnitude limits of teleseisms not record- ed and the lower magnitude limits of teleseisms recorded. In figure 7.1* the trend lines show the following three approximate linear trends: (1) the threshold increases from U.6 at 20° to U*8 at 68°, (2) the threshold in- creases from U.9 at 80 to 5.7 at 100°, and (3) though poorly defined, the threshold decreases from approximately 5. 8 at 137° to U.8 at 15>0°. In the interval between 68° to 80°, the linear trends are interrupted by a de- crease of approximately 0.5 unit, followed by an increase of approximately 0.5 unit. (These features are also shown in fig. 7.2, detection thresh- old of deep earthquakes). No shallow events were recorded in the range 110 to 136°. The detection threshold graph for deep events (fig. 7.2) 76 PKP PHASES UPPER LIMIT OF SIGNALS NOT RECEIVED LOWER LIMIT, RECEIVED • • • • • • • • • • • • • — r - 110 T" 120 "I - 130 T 140 T 150 a single. NORSAR seismometer for shallow trends are circled. 11 NORSAR, (3) magnitude (nO of each earthquake, obtained from the PDE cards, (U) epicentral distances from the center of NORSAR for each earth- quake location given on the PDE cards, and (5) the focal depth of each earthquake given on the PDE cards. From each paper seismogram showing a clearly recorded earthquake signal, we measured the dominant signal frequency occurring within ap- proximately 2 s after signal onset. An average signal frequency was then computed for each earthquake. 7.2 NORSAR Detection Threshold for Teleseisms From the earthquake list described above, we constructed figure 7.1* which shows the detection threshold of a single NORSAR seismometer for shallow (hl^O km) teleseisms, and figure 7.2, which shows the detection threshold for deep (h>]?0 km) teleseisms. In these illustrations, trend lines describe both the upper magnitude limits of teleseisms not record- ed and the lower magnitude limits of teleseisms recorded. In figure 7.1* the trend lines show the following three approximate linear trends: (1) the threshold increases from U.6 at 20° to U.8 at 68°, (2) the threshold in- creases from U.9 at 80 to £.7 at 100°, and (3) though poorly defined, the threshold decreases from approximately 5.8 at 137° to ii.8 at l£0°. in the interval between 68° to 80°, the Linear trends are interrupted by a de- crease of approximately 0.£ unit, followed by an increase of approximately 0.^ unit. (These features are also shown in fig. 7.2, detection thresh- old of deep earthquakes). No shallow events were recorded in the range 110° to 136°. The detection threshold graph for deep events (fig. 7.2) 76 PKP PHASES □ SIGNALS RECEIVED • SIGNALS NOT RECEIVED DISTANCE Idegl Figure 7.1. Detection threshold of a single NORSAR seismometer for shallow teleseisms . Deviations from the trends are circled. 11 PKP PHASES LOWER LIMIT OF SIGNALS RECEIVED UPPER LIMIT, NOT RECEIVED •• • • •• • xy,/ • • • a 110 T" 120 130 M T 140 150 shows trend lines somewhat similar to those in figure 7 «l5 in general, however, the trend lines are approximately 0.3 unit lower than those drawn for shallow events. 7.3 Average Signal Frequencies of Teleseisms The density and distribution of the observed shallow teleseismic signal frequencies are displayed in figure 7.3. As depicted there, 90 percent of the average frequencies are in the band above 0.8 Hz; 85 per- cent are in the 0.8- to 2.0- Hz band; none are in the band above 2.U Hz. Displays for signals received from deep earthquakes are given in figure 7»U. Here, 95> percent of the average frequencies are in the band above 0.8 Hz, and none are observed in the band above 3.2 Hz. Figure 7«5 shows the average frequencies of shallow earthquake sig- nals plotted as a function of their corresponding PDE- listed magnitudes for earthquakes having epicentral distances less than 90 from N0RSAR. For earthquakes having magnitudes less than £.0, approximately 80 percent show average frequencies' in the 1- to 2-Hz band, with a range of 0.8 to 2.I4 Hz. A similar display for deep earthquakes is given in figure 7.6, showing that they have signals mainly in the 1- to 3- Hz band, with a range of 0.9 to 3.2 Hz. 7.U Discussion By studying seismograms produced by a single sensor positioned in a 60-m well, Kelly (1966) investigated the detection threshold at the Large Aperture Seismic Array (LASA) site located in eastern Montana. From his study, made for the distance range I4O to 90° from LASA, he 79 PKP PHASES DISTANCE (dog) Figure 7.2. Detection threshold of a single NORSAE seismometer for deep teleseisms . Deviations from the trends are circled. 78 shows trend lines somewhat similar to those in figure 7.1; in general, however, the trend lines are approximately 0.3 unit lower than those drawn for shallow events. 7.3 Average Signal Frequencies of Teleseisms The density and distribution of the observed shallow teleseismic signal frequencies are displayed in figure 7.3. As depicted there, 90 percent of the average frequencies are in the band above 0.8 Hz; 85 per- cent are in the 0.8- to 2.0- Hz band; none are in the band above 2.U Hz. Displays for signals received from deep earthquakes are given in figure 7.U. Here, 95 percent of the average frequencies are in the band above 0.8 Hz, and none are observed in the band above 3.2 Hz. Figure 7.5 shows the average frequencies of shallow earthquake sig- nals plotted as a function of their corresponding PDE- listed magnitudes for earthquakes having epicentral distances less than 90 from N0RSAR. For earthquakes having magnitudes less than 5.0, approximately 80 percent show average frequencies' in the 1= to 2-Hz band, with a range of 0.8 to 2.U Hz. A similar display for deep earthquakes is given in figure 7.6, showing that they have signals mainly in the 1- to 3- Hz band, with a range of 0.9 to 3.2 Hz. 7.U Discussion By studying seismograms produced by a single sensor positioned in a 60-m well, Kelly (1966) investigated the detection threshold at the Large Aperture Seismic Array (LASA) site located in eastern Montana. From his study, made for the distance range U0 to 90° from LASA, he 79 40 -1 30- I- Z in > V>- 10 - 1 2 FR EQUENCY (Hz) FREQUENCY (Hz) Figure 7.3. Density and distribution of shallow teleseimio signal frequencies . 1.0 -i FREQUENCY (Hz) Figure 7.4. Density and distribution of deep teleseismio signal frequencies . 80 reported (p. 7): Our conclusion is that the 7$% detection threshold on a single sensor is at a magnitude of 1*.3 at the C. and G. S. level. For the corresponding distance range, data portrayed in figure 7.1 (shallow teleseisms) suggest a detection threshold of U . 7 to $.k for NORSAR; we detected only 2 shallow events having a reported magnitude equal to, or less than, U.3. 8 -t 7 - LU o p H Z C3 5 - |e I 0.3 m b - 5 19 |«-18-^-10^.2^| fENTS >|«- 31 ^|^21-^4^| ALL EVENTS • • • • •• • ••• •• •• • m • • • • • • m •• • • ••••• • • • • • • • 1 1 I I I I 1.0 FREQUENCY (Hz) I 3.0 Figures 7.5. Signal frequencies of shallow teleseisms plot- ted as a function of their PDE-listed magnitudes . Relative to the shallow events, deep events seem to show a lower detection threshold (figs. 7.1 and 7.2). The lower threshold is probably due, at least in part, to "one relatively higher signal frequencies of the deep events. In general, the microseism amplitudes decrease propor- tionally to an increase in frequency, producing a better signal-to-noise ratio for deep events than for shallow events. "We must emphasize that the frequencies described in this section are average values. Some shallow teleseismic recordings showed frequen- cies as high as 3» $ Hz. 7 — | E in Q Z < S 4 — |< 1 »|<- 6 ->|«-5-»|<-3-»|«4*|«n| ALL EVENTS \i " 3 »|<— 14 -»|<-11 ->|*-4 -*|<-4*|<2*| » « • • • • • • • • t — i — r 0.4 I I I I 1.0 FREQUENCY (Hz) I 4.0 Figure 7.6, Signal frequencies of deep teleseisms plotted as a function of their PDE-listed magnitudes . 82 8. SUMMARY The following is a summary of observations and interpretations made from a study of seismic data recorded by EML at NORSAR: (1) Quite small differences exist between the surface and subsur- face (-60 m) microseismic field, even during times of high wind velocity. For the dominant P signal band (0.8 to 2.0 Hz), the surface and subsurface field normally differs by less than 1 dB (see figs. 3.2 to 3«H* and seismograms shown in app.). Environment of a typical recording site is illustrated in fig- ure 2.2; the EML instrument package, in figure A.lj the responses, in figure A. 2. (2) Teleseismic signals produce almost identical recordings from the surface and subsurface seismometers (see fig. 3.12). (3) Between adjacent subarray sensors, the coherence of microseisms, in the P signal band, approaches values we estimate for unre- lated samples (see figs. 5.10, £.2 to 5*9). As indicated by the coherence polygons (defined in section 5.1), the coherence has a strong azimuthal dependency. (h) Iftfe interpret the NORSAR noise field to be an interference pro- cess (fig. 3>.lU)j composed of waves approaching from a broad band of azimuths. A mathematical model of the noise field (sec. 5.3) produces coherence graphs similar to those estimated from the real data (compare figs. 5.11 and £.16). For a con- stant model velocity the computed time delays are frequency dependent (fig. 5.18). (5) The NORSAR micro seismic field is approximately space stationary (see figs. U.l through U.li). (6) The average microseisraic amplitude at 1.0 Hz is approximately 6 ray/Hz RMS (sec. U.3). (7) There is a weak correlation between microseismic trends and regional barometric gradient trends (figs. 6.6, 6.7, and 6.8). Similar correlations were found by us and others by studying data EML recorded in Australia (see figs. 6.U and 6.5). (8) Approximately 90 percent of the shallow teleseisms recorded have average signal frequencies (estimated from paper seismo- grams) greater than 0.8 Hz; 85 percent of the average frequen- cies are in the 0.8- to 2.0- Hz band (see fig. 7.3). Signal frequency may increase as a function of decreasing source magnitude (see figs. 7.5 and 7.6). (9) For the distance range 20° to 90°, the single seismometer shallow-teleseism P detection threshold ranges from U.6 to $.h mfc,; the threshold for deep earthquakes is somewhat lower (see figs. 7.1 and 7.2). 9. ACKNOWLEDGMENTS This work was supported by the Advanced Research Projects Agency (ARPA) and was monitored by the Electronics Systems Division of the United States Air Force. Mr. Rudolph Black of ARPA, and Lt. Col. W. R. Lauterbach, ESD Program Manager, defined the broad project objectives and participated in program planning. During the interpretation phase of the 84 project, Lt. Col. J. X. Brennan was the ESD Program Manager. The Norway field program was coordinated by Major j6» Brandt zaeg of the Norwegian Defense Research Establishment, and Lt. Col. N. Orsini, assist- ed by Captain R. Jedlicka (both of ESD). The EML efforts were guided by Dr. Don Tocher, Director. Professor Markvard A. Sellevoll, Director of the Seismological Observatory at the University of Bergen, contributed support including the use of his playback facility. Our discussions with him and his staff were very informative; particularly stimulating were those with Professor Sellevoll and Messrs. Hilman Bungum, Leif Bruland, Eivind Rygg, and Reidar Kanestr*$m. The cooperation of Professor Sellevoll allowed data presentations to the NORSAR managers during March 1969, and permit- ted quality control tests of the EML gear during the field program. Mr. Lars Dalland of Tele-plan worked with the EML field crew, pro- viding engineering consultation and acting as liaison with the construc- tion crews. The Tele-plan tests, calibrations, and other contributions were directed by Mr. Thorbj/rfrn Johansen, assisted by Mr. Bjarne Goplen and Mr. Gunnar Hansen. The cooperation of Mr. Johansen and Mr. Dalland permitted our success in data aquisition. EML personnel who made contributions included Lt. Thomas Kalil, who performed almost all of the EML data reduction and quality control at the University of Bergen facility, and who advised and assisted in prepara- tion, installation, and maintenance of the EML instruments; Mr. Ralph Humphrey, who, to make our equipment NORSAR compatible, designed new instruments, modified existing ones, and also supervised instrument con- 85 struction, preparation, and installation; Mr. Erick Young and Ensign Raymond Reilly, who wrote computer programs, reduced a large portion of the data at EML, and made valuable suggestions that were incorporated into the data reduction and analysis programs; Mr. Michael Blackford, who modi- field the spectral analyses program; and Mr, "Wesley Hall, who assisted Mr, Humphrey in construction and preparation of EML instruments; Miss Nancy Crossley and Miss Susan Larson, who prepared the manuscript. Almost all of the artwork and drafting was by Mrs. Patricia Sposato, who also contributed technical assistance. 10. REFERENCES Backus, Milo, John Burg, Dick Baldwin, and Ed Bryan (196U), "Wide-band extraction of mantle P waves from ambient noise, Geophysics, XXIX , No. £, 692. Bendat, Julius S., and Allan G. Piersol (196U), Design considerations and use of analog power spectral density analyzers, Minneapolis- Honeywell Regulator Company, Denver, Colorado. Bungum, HLlmar, Leif Bruland, and Eivind Rygg (1969), Seismic noise struc- ture at the Norwegian Seismic Array, Sci. Rept, No, ii, Seismological Observatory, University of Bergen, Bergen, Norway. Hasselmann, K. (I963), A statistical analysis of the generation of micro- seisms, Rev. Geophys., 1, No. 2, 177-210. Jenkins, Gwilym M. , and Donald G. Watts (1968), Spectral Analysis and Its Applications (Holden-Day, San Francisco, California), Kelly, E. J. (1966), LASA on-line detection, location and signal- to- noise enhancement, Tech, note 1966-36, Lincoln Laboratory, Lexington, Massachusetts. Longuet-Higgins, M. S. (1950), A theory on the origin of microseisms, Phil. Trans. Roy. Soc. London, A, 2ii3, 1-35. Longuet-Higgins, M. S. (1952), Can sea waves cause microseisms? Symposium on microseisms, U. S. Natl. Acad. Sci. Publ. 306, 7U-93* 86 Murdock, James N., John H. Pfluke, Roger Kraynick, Michael E. Blackford, and James D. Van Wormer (1968), Microseisms and teleseisms recorded in Australia, ESSA Tech. Rept. ERL 66-ESL U (U.S. Government Printing Office, Washington, D.C.). Rygg, Eivind, HLlmar Bungura, and Leif Bruland (196°), Spectral analysis and statistical properties of microseisms at NORSAR, Sci. Rept. No. 1, Seismological Observatory, University of Bergen, Bergen, Norway. Santo, Tetsuo A. (1962), Energy sources of microseisms in Sweden, Annali de Geophysica, XV, No. h, 335-377. APPENDIX A.l Field Instrumentation, Field Tests, and System Frequency Responses The following describe the NORSAR and EML systems: Seismometers: Hall-Sears HS-10-1/A. Texas Instruments Incorporated Model RA-5 (located in the vaults). EML Seismic Data Constructed by EML. NORSAR Seismic Data Amplifiers : Amplifiers : Primary Time Base Generators : Tape Recorders: Anemometers: Climet Model 011-1. Output frequency proportional to wind velocity. Astrodata Model 6195 time code generator, 100-Hz carrier, directly recorded. Precision Instruments Model 5100, record speed 15/160 ips, FM center frequency 8U.U Hz. The following describe field tests and system frequency responses : rstem Calibrations: Signal generator sine waves (20.0V P-P) normally for frequencies of 0.3* 0.5> 87 NORSAR Seismic Data Amplifier Calibrations ; ML Seismic Data Amplifiers : Seismometer Free Period: System Noise Tests ; Compensation ; Filter Settings for Digitizing ; 0.8, 1.0, 1.2, 1.5, 2.0, U.O, 8.0, 10.0 Kz. Normally made for each NORSAR element every 10 days. Same as above, but -with seismometer- equivalent impedence emplaced at input to the amplifiers. Also, gain tests at 1.0 Hz performed at 10-day inter- vals. Gain tests made at 10-day intervals, 1.0 Kz. 1.0 Hz, from Iissajous pattern and frequency response tests conducted by Tele-plan, a NORSAR contractor. Seismometer-equivalent impedence em- placed at input to NORSAR amplifiers. Normally, conducted at 10-day intervals for each element. For power spectral density estimates and seismogram displays — none For coherence estimates — discriminator center frequency controlled by fre- quency of time code carrier. For power spectral density estimates only — low pass at 5>.£ Hz, 36 dB/octave. For coherence estimates — low pass at 2.2 Hz, 36 dB/octave. Anemometer Channel Electrical analog of the anemometer Calibrations : output made for wind velocities of 1 to 30 mph. One test made for each system. The EML field system is displayed in figure A.l. Typical system frequency responses, computed from the composite field calibrations of EML and Tele-plan, are given in figure A. 2. Shown are frequency responses for the NORSAR-EML system, the NORSAR-EML University of Bergen system, and for the NORSAR-EML- GDC (General Dynamics Corporation, the digitizing contractor) system. Not shown is the NORSAR-EML- GDC system used for digitizing data for coherence estimates^ the 3-dB point for it is at 2.2 Hz, 36 dB/octave. A. 2 Digital Data Reduction Program t , Written at ESSA Research Laboratories, Boulder, Colorado. Modified for EML use by Harold Williams and Roger Kraynick under the direction of John H. Pfluke. Literature referenced: Blackman and Tukey, 1958. Sample length ; For power spectral density estimates — 170 s sampled at 20 sample s/s, 5>11 digital counts for full FM deviation. 89 +s S a R O Sh CO +i +S « K O ■t» +i a +i g 0) E 3 5^ +s CO £ S: O +i o K o o to +s to *«J +s s 90 o o -i HI > _ o < I- Q — V) O Q. > O Z LU 3 a © © Z LU LU o < -I Q. in i i i i — i — r i ii i i i i — i i aamndi/w 3Ai±vn3u 1 1 1 1 i i i — r 6 > o z LU D LU EC LL o 6 Si 4^ 0) fc G t-q £ fel 6~i to iM to S ^ tjl CO &4 to CO I M <■« 4i £ <3 co 0) CO £ • O co CO CO 05 £ Ph O a, £ co CO 0) C31 fc 91 Prewhitening : Spectral Estimates ; Number of Lags : Smoothing Spectral Computations : Magnification Corrections t Frequency Response Corrections: For coherence estimates — 3U0 s sampled at 10 sample s/s, 511 digital counts for full FM deviation. A, = 0.5 A, - 0.5 A, -. A is the amplitude of the kth point of the time series. From auto- correlation and cross- correlation estimates. 100. Parzen lag window. 1 - 6*1 * 6^ V °p, k ± ■% 2 2(l-£ )3, m^ k■■ i - .■■■■■ ■■■■■■■■■■~iaji I w4*« SEISM06RAM la 95 0.4 — 0.2 0.0 -10 0.0 10 PERCENT FM DEVIATION Figure A. 3, Probability density distribution of a typical miaroseism sample. 94 POSITION II WIND I6mph SEISMOGRAM la 95 «-o» i ■ i i i . I i a i > i ■ i ; i i c m i i > l i ■ t-oai m i.,.. » -t , i . i . e i t>u c ; m i , . . ^: i ; « i 1 o i i ci i i i . i i i i ^-- i i 1 1 i a 1 1 ■! I ■ i^-m^4i \/Aa\/v GPO 830- 195 SEISMOGRAM 2b 97 POSITION MAG. WIND 34mph OIB 290 K M- IOS SEISMOGRAM lb 96 POSITION MAG. 02B ■^Iv".^>^X-.«^.\-.^^.-^-^IC--^>^