w.rrn ^-^nr».TP ---^^.--.-^^T-.T™«. "'.fl "WHW WJ.WW'P W"W'!!« T'■.,l■■l>,

AD-773 320

NORSAR RESEARCH AND DEVELOPMENT - 1 JULY 1972-30 JUNE 1973

E. S. Husebye

Royal Norwegian Council for Scientific and Industrial Research

Prepared for:

Advanced Research Projects Agency Electronic Systems Division

1 November 1973

DISTRIBUTED BY:

KHJ National Technical Information Service U. S. DEPARTMENT OF COMMERCE 5285 Port Royal Road, Springfield Va. 22151

J

mm WWBUMiinm i i ^mj-'nJu-Mittri-iini*»TaWi**.'.<-*;-w»

f I III I I I Mil mi i iillili'- ii 11 nt .-iH^MauMm in ■

Hi-i.jjL.J.I ,i»,m MHiv^.J iljivi IIP"" U L mU , ^MHOi^'.Billl^lUU! l|l|RJPjppCTii|ip.l|IWJIJ«*P(l.l lll.ltt-JIIIHIIJUl !,ii!.WW MW litt i.l-,—JUP^iiil.UJIIl.llHJ UlU .1 ^. l I "!<P^ »I'lP^PIil'J m" mi mil UM i i ij, UMIIWJUN up ^n ,.11 ji|N!ll I HIJ I u^ IP^I. .,(» »jp.j lli; |l I ;ll,lll||m

SECURITY CLASSIFICATION OF THIS PAGEfWion Dnlm Enlerad}

20. Cont.

beamforming and optimum signal estimation techfiiques are discussed. The importance of so-cal'ed intrinsic time and amplitude anomalies or wave scattering (Chernov) effects in array data processing are also demonstrated. Finally, work on seismic verification problems is discussed briefly.

,y Si TWRITY ci Assirir ft-rioN or THIS P AGEWKVI nmn Enitttid)

■MMHH I IMJBII

.jiii»mw. ip.ii........ii mmammvrmm in: i \x HI ■ »I—H-J i* ti.wmmmmmrmr*~***<vmim^t^ ■ « ■ " > "

.

NTNF/NORSAR P.O. Box 51 N-2007 Kjeller NORWAY

NORSAR Report No. 6 3 Budget Bureau No. 2 2-R029 3

NORSAR Research and Development

(1 July 1972-30 June 1973)

by

E.S. Husebye Chief Seismologist

1 November 197 3

The NORSAR research project has been sponsored by the United States of America under the overall direction of the Advanced Research Projects Agency and the technical management of the Electronic Systems Division, Air Force Systems Command, through Contract No. F19628-70-C-0283 with the Royal Norwegian Council for Scientific and Industrial Research.

This report has been reviewed and accepted by the European Office of Aerospace Research and Development, London, England.

' n

üum^m~*^*m i

»"■I' l"J II ■ •«•Wl-^l'» M. I IBOT^W^T • i^ni«!««-!«!» << i ,i. <«in,..iiiailw*lim.Liiai.i*iimaHV**m«!Jajliil.l<M " ' ■ I u ■ I

- 11 -

ARPA Order No. 80 0

Name of Contractor

Date of Contract

Amount of Contract

Contract No.

Contract Termination Date

Project Supervisor

Project Manager

Title of Contract

Program Code No. IF10

: Royal Norwegian Council for Scientific and Industrial Research

15 May 1970

$2,051,886

F19628-70-C-0283

30 June 1973

Robert Major, NTNF

Nils Maras

Norwegian Seismic Array (NORSAR) Phase 3

' — •—- -— MM»

UIIUP.MII.I!«.)■■ ui.uw.jni.'miMliii "I 'in in . . i .1 i*i:mm^^W »■iinwMi« •> i inn» in m W^W

- Ill -

CONTENTS

SUMMARY

1. INTRODUCTION

2. RESEARCH EFFORTS

NORSAR Event Detector Threshold Setting and the False Alarm Rate

Event Detector based on Envelope or Incoherent Beamforming

Optimal Beamforming

Intrinsic P-wave Travel Time and Amplitude Anomalies across NORSAR

Travel Time Anomalies across the NORSAR Array

P-wave Amplitude Variation across NORSAR

Seismic Wave Propagation in an Earth which is partly Modeled as a Random or Chernov Media

Seismic Magnitude Investigations

Analysis of the Operational Capabilities for Detection and Location of Seismic Events at NORSAR

Seismic Verification Research

3. MISCELLANEOUS

4. REFERENCES

Page

1

3

3

7

10

14

14

20

25

2 5

28

32

36

39

mutm mamm^mmmmm^ammm nmmm

" ' ' " ■ '" liiouini ill mm ^l i ill I u« I ■■■iiiiinii n ■in ai i.

I '

SUMMAl^Y

The report covers the period 1 Ju.lv 19 7 2 - 30 June 197 3 which is characterized by continuous improve- ments in event detection and location performance during routine operation of the array. The re- search efforts were aimed at improving the event detectability, and the potential exploitation of wave scattering effects for improving NORSAR's event detection and classification capabilities.

Work completed and in progre II. The first section'deals noise level fluctuations pin of the noise. Next, differe forming and optimum signal e are discussed. The importan time and amplitude anomalies (Chernov) effects in array d also demonstrated. Finally, cation problems is discussed

ss is presented in Chapter with seasonal and diurnal

s changes in the character nt types of array bcam- stimation techniques ce of so-called intrinsic or wave scattering

ata processing are work on seismic verifi- bi iefly.

■'■• INTROnUCTION

The report summarizes the NTNF/NORSAK research and

development efforts during the interval 1 July 19 72 to

30 June 1973. In the first part of the period most

attention was given to development work like software

modifications and «implementation of now data processing

routines. For example, a supplementary event detection

processor, based on so-called incoherent beamforming

(Ringdal et al, 1973), was implemented in the on-line

system in September 1972. For the purpose of editing

a daily bulletin of seismic events, the required input

data may be read directly from the detection log tape;.

Thus, a daily list of seismic events is nov; available

every morning and covering the previous 24 hours, even

if the Event Processor (EP) 'is behind its time

schedule. Moreover, a status report of all data channels

is generated daily and is available to users of NORSAR data

■ IIIIWH! W-WW"-." ..I-!.,W«l"^i»'»'W^-"l**l.l><"M"*?,1«'Ji w■ u*'w«r.iiw*rw wwww-w:i^ww.Wp. i n. MIR|1J..,Ji im j,t,.^vf ^v<WI'IAOPI■^■Miif^nHllPFiMIPRi.MMiinilPTlffPniii11--, liii") WI'Mil .MWIWJWIff'rWWWPWWiyWrliWlliifLimWPfliippujW'^W'W.yi»11"1 !.,!■ KDUNM

- 2

The research topics investigated or in progress are mainly

aimed at system improvements and evaluation of the array's

event detection and location performance. Also, some

aspects of the event classification problem have been

considered. Most attention has been given to improving

NORSAR's event detection capability, and the work here

comprised optinum amplitude weighting of subarray beam

signals, predictive decomposition models for explaining

and predicting intrinsic phase shift and amplitude

variations across the array, event detector false alarm

rate fluctuations and signal-noise wavelet classification

The detectability of the so-called incoherent event detec-

tor, part of the NORSAR on-line system, is superior to

that of conventional beamforming in seismic regions

characterized by complex P-signals. An evaluation study

of NORSAR event detection and location capabilities, based

on the array's routine performance in the interval Apr-

Oct 19 72 has been completed. One interesting result here

is that the NOAA-NORSAR m. magnitude discrepancy is a non-

linear function of magnitude, i.e., NORSAR reports rela-

tively too large m values for small events. Moreover,

a bias analysis of NORSAR event magnitude estimates has

also been undertaken. Several kinds of Vespagram analysis

of core precursor phases indicate that these waves are

not explainable in terms of the standard velocity models

for the earth's core, while a realistic alternative is

scattering sources in the lower mantle.

The above research topics and relevant results are described

in the next chapter. In general only the main results are

presented here, as further details are available in NORSAR

Technical Reports or from papers published in professional

journals.

HMMMM MM. ttaalimmHmmmm 'J*^--

PW'M.!IPWflJP^.|PI«li.W W

3 -

2. RESEARCH EFFORT^

■ The research activities in the reporting period 1 July

197 2 - 30 June 19 7 3 have been focused on event detection

and classification problems. Presently, NORSAR reports

in average 20 events per day, but recent research results

indicate that significant improvements of the array's

event detectability is still possible. An important

problem here is the development of objective, criteria

for discriminating between weak P-signals and signal

shaped noise wavelets.

Conc.tioned on the present NORSAR computer configuration,

the most pressing event detection problems are considered

solved. Thus, at the end of the reporting period more

research effort could be spent on seismic event classifica-

tion problems. For example, conventional discriminants

criteria have been adapted to NORSAR data, but also so-

called signal space expansion techniques are under in-

vestigation. In the reporting period a number of visiting

scientists have been doing research at the NORSAR data

center and part of this work is included in this report.

False Alarm_Rate

At NORSAR a significant trend in seasonal noise level

variations occurs, and the same holds on a diurnal basis

as demonstrated in Figs. 1 and 2. For example, extreme

cases with a variation in noise power up to 18 dB in the

frequency band 2.0-3.0 Hz within a few hours have been

observed at a large number of NORSAR short period sensor

sites. This simply means that the array's event detection

capability is lower during winter than summer and

also lower during the day than the night. In the latter

case, there are roughly 20% more events detected during

night time. Relevant data on the phenomena are presented

MMMMMMMH „^-MMMIMBiiM

' ' ■■■■'—-I I »«I^^^^^^HB^^^BMM ■m*»WM» .. PIHIII. LJLMIUIhnilV|IJ,Hd ^ "" '- "WH

- 4 -

10.0 -n 1 1—1 1 1—I 1 1—I 1 1 1—1 1—I—I 1 1—|—I 1 1—1 1—| 1 [—I—T-r

nihmi rrr^irvM/WW^^

0,1 Jj—i—i—i—i—u- _j—i—i—i—i—i—(_ .j—i—i—i—i—i—|—i—1_ -l—i—(._J—j_ 20 40 60 80 100 120 140 160 160 200 22D 240 250 230 300 320

DAY OF YEAR 1972

Fig. 1 Beam average of LTA for the time period G Jan - 23 Nov 1972. (LTA=Long Term Average, which is equivalent to linear power measured in a window of approximately 30 sec.) The sampling rate is 20 s/day and the frequency filtering of the data fron which the LTA is computed .if; 1.2-3.2 Hz.

IU

3

Ul

ö

3 /, 0 TIME OF WEEK

Fig. 2 1,'I'A in relative units as a function of time of week, where day 1 is Sunday. (LTA as defined in Fig. 1) Average is made over 46 weeks, and a trend-removal is applied to the LTA time series by subtraction of daily averages. The sampling rate is 20 s/day, and the frequency filtering of'the data from which the LTA is computed is 1.2-3.2 Hz.

MM» ^^^^^mjttm MMhifiii WUwi^w.y-.i. .«*■

" '^W^P^m* v^amwrnmowrmm!*** i.iipnmi*llli(Jlip^i,Jf Ji|i,|iwiwWV«Witill'«tWII."P. .■■■■■wp.-i^Mi« ^^mm^^mmvf ^r^^j^w MipiHaiiiiiiMiiiwiiliwlWi

- 5 -

and discussed in a recent report" by Bungum and Ringdal (1973)

Also, a similar study for long period noise is in progress.

An important but mostly ignored aspect of noise level fluc-

tuations is that the statistical properties of the background

noise change too. The same effect is also obtained by using

filters with different passbands. Henceforth, from the

theoretical studies of Rice (194 4) and Cartv/right and Longuet-

Higgins (.1956) we would expect that the noise wave train

maxima would fluctuate between Gaussian and Rayleigh prob-

ability density distributions. In other words, the so-

called false alarm rate would vary, i.e., the number of

times pure noise wavelets trigger the event detector for a

fixed SNR threshold (see Fig. 3) . This phenomenon does not

necessarily mean that relatively more noise wavelets arc

reported as seismic events under adverse noise situations

IMO

«

0-

o

UJ a

au 9 0 too

SIGNAL- TO- NOIGE RATIO MB)

110

Fig. 3 Number of detections as a function of SNR for two time periods-, covering one hour of day time and one hour of night time, respectively. The-dashed line has a slope of -15.0.

as such a decision rests with the analyst. Instead, under

favorable conditions too few events would be reported as

the SNR threshold value in the event, detector would be too

large. The above problem was first considered by Lacoss (1972;

■—■—n—HM—mi ■■—-■ ■,„„, „M, ,,„

imwmwmwv.*."*i in' 'w^iMi'w^if.Mj« •iJii!ni,i.^MJisiviwPP:«^«".ww!"wwiTOi jwwupiJBUuiiinui ipiiiiiiiwiMi'-11 ' ''' ' t^**m^* IIIIIIP mini ■IIIIIIIIIJIU .. in iniffimmmmifii

- 6 -

who forwarded an approximately linear relationship between

the noise stability parameter and the number of false alarms

The stability parameter was defined as the square of the

noise level average relative to its variance. Steinert et

al (1973) have continued this work, and the importance of

the problem and also the potential gain by having a float-

ing detector threshold setting are demonstrated in Fig. 4.

Due to a limited data base the; SNR threshold values range

from 8 to 10 dB in the figure, while the corresponding

values in the operational system are between 10 and 12 dB.

Such an algorithm for the prespecified false alarm rate

is feasible to implement in the NORSAR on-line system, and

the expected gain expressed in equivalent SWR units would

probably amount to around 0.5 dB. In addition, this routine

would permit a more efficient computer capacity utilization.

K)

>

VI in

o 2 IO1

i ' r

J0C|-J-

Hi. tES OLD SYMBOLS c ao da o 8.5 dLi • X so da K £5 d3 t. WO 03

1. 12 ■ 1$ ao

Pig. 4 False alarm rate versus noise stability for different event detector threshold values. Three differenl noise situations vi'.'ro analyzed, each corresponding to one hour of NORSAR on- line processing. For further' variation of the noise structure, three different bandpass filters were also used.

riMMMM ^-(1-^^ laiiahLMt^-..-^ „...■■-......, .,■^■■..^1- k^ ,. m. . ^-iriMfi'ttll

^^^^^^^^^^^^—^^^^^^^^^^^^— ^^^^^^^^^^^^^^^"

5Y£Di-_D2i-9£t2i"_!2ä2S^_LJD_SDYS-':2D£;_0^_Inc9!]£IC!nt: Beamforming

The P-signals recorded by NORSAR are only partially co-

herent across the array. This means that the expected gain

in signal-to-noise ratio (SNR) which is proportional to

the square root of number of sensors used is not Obtained

during array beamforming operations. The corresponding

signal energy loss increases with increasing frequency,

and may severely degrade the array's detnctability of very

short period P-waves. This problem may be partly circum-

vented by replacing or supplementing the array beam traces

with the average of subarray beam traces. The relative

advantages of using the so-called incoherent beams are

modest signal losses, better estimates of the noise

variance and good areal coverage. The noise suppression

is small as compared to array beamforming, but could partly

be compensated for by using high frequency bandpass filter-

ing. As mentioned previously, a supplementary event detec-

tor based on incoherent beams was implemented in the on-

line system in September 1972. Results from the first

two months of parallel operation of the so-called coherent

and incoherent event detectors are presented in Fig. 5

and Table 1. The improvement in the array's event detect-

ability amounts to around 15 per cent. For further

details see the report by Ringdal et al, 1972.

The characteristic feature of a seismic array is real-time

processing of data from a large number of sensors organized

in a certain pattern on the surface of the earth. As is

well known, when sensor separation increases, the signal

similarity, in general, decreases. The consequence here

is that when processing signals from a continental array

or the global seismological network, the signal suppression

is approximately equal to the noise suppression, resulting

in a processing gain close to zero. One possible way to

circumvent this problem might be to replace the individual

signal trace by its envelope as v/e intuitively should

expect this kind of signals to exhibit a large degree of

mtt - -

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^— W«HP^^^^^

- 8

NO, Zone

Name EVENTS Total

COH.BF Only

IKC.BJ Only

' COH.S INC.

COH.BF Total

INC. Totc

BF ■,1

I Jo. No. No. No. No. % No. %

1 Greocc/Turkcy 117 5 45 67 72 62 112 9,; 2 ÜSSR/Centr.Asia 194 28 41 125 153 79 1GG fin J Japan/Kara,/Alcu. 168 40 7 121 161 96 120 76 -1 USA/Cent.America 64 31 1 32 63 99 33 si b Global I

(All events) 1030 242 133 633 905 07 796 77

G Global II

(High Quality 546 24 25 497 521 95 522 35

events) „

TABLE 1

«INCOHERENT (N = i&), oCOHERENT (N.~5 | I ] r x, 1 |«2ofj

• « • e • •

• eo*» ••

• • • o.

• o

L- 1 L_IL_JL 18 22 26

39

3G

i

• INCOHERENT (N»41) oCOHERENT (M-28)

'"1 "i nr f03!l

BO

I« o 0." • e» o o

«:«>

40

30

30°E 40 I L

60 -1 II

60 J2C 100't

Pig. Events reported in the finaJ NORSAR bulletin which were

but not H V^" the COherent or the ^coherent detector, but not by both. The time period covered is 16 Sep . 15 ^v

1972 and typica] SNR detection thresholds were 3.6 (coherent)

in ti; MedinSrent) • Thi'figures siiow dctectio" p-ro™ ' in ltu. Mediterranean and Central Asia s Russia areas.

MMMI MMMBHi ^-^■"—- ^.J *.****.

^^^^^^^^^^^^^^^^^^^^ ^^^^^^^^^^^ ^^^^^^^^ ^^^^^-^^^^^

- 9 -

signal similarity independent of sensor separation, seis-

mometer type, etc. This hypothesis has been tested on

WWSSN station records (two earthquakes and one explosion)

and NORSAR subarray beams from many different events. The

WWSSN beam pattern for an earthquake in Chile is shovm in

Fig, 6. Signal envelope similarity has been calculated

through cross-correlation and coherency analysis. Typical

cross-correlation values were around 0.75 units between

WWSSN envelope signals. Similar results were obtained

by joint analysis of 22 different NORSAR events in the

distance range 3-145 deg. Moreover, using data on the

P-wave amplitude variation in the teleseismic distance

range and the theory for incoherent event detectors

(Ringdal et al, 1972), reliable estimates on multiarray

processing gains are obtainable. It .is interesting to

note that the above method may supplement previously

proposed schemes based on joint detectability analysis

of data from several arrays. The above topic is dis-

cussed in some detail in a recent report by Ilusebye et

al, 19 72. —-^———^

\A/7"^T2?

-2 0 2

REL LONGITUDE (deg)

Fig. G Response pattern for the Greeley nuclear explosion in Nevada 12/20/1960 based on envelope traces for 19 WWSSI

statjons,

Mam •■MMMMaiMB ^ • "-■'■ ~ : ■

. ^^^^^^^^^^^^^^^^^^^—^^^^^^^^^^^^^^^^^^^— T^

.10

9 L? i i IL1 9.1- _ 15:; <} üjf 21; r1 iss

Beamforming or simple delay-and-sum process; ig is exten-

sively used in analysis of P-waves recorded by the large

aperture arrays NORSAR (Norway) and LASA (Montana). When

the underlying assumptions of well-equivalized noise levels

and identical signals between instruments is correct, the

corresponding gain in SNR is optimum. In practice, these

restrictive signal and noise mode]s are not valid, thus

degrading bhe final signal estimate and the event

detectability of seismic P-waves.

In investigating this problem, we followed the line of

development presented by Christoffersson and Jansen (1973)

where the interest is focused on the relation between sig-

nals at different instruments, i.e., the space spanned by

the recorded signals. In a recent paper Christoffersson

and Husebye (1973) Introduced rrore generalized P-signa]

models during array beamforming and the corresponding

least squares signal estimation techniques were described.

The usefulness of these data processing schemes was also

demonstrated in analysis of more than one hundred LASA

and NORSAR recorded signals (see Tables 2 and 3 and Fig. 7).

For LASA the average gain in SNR relative to that of con-

ventional beamforming for the teleseismic event was around

3.7 dB. Most of this was obtained by accounting for noise

level variations between subarrays, as signal coherency across

the array is good. For NORSAR the corresponding SNR gain was

approx. 2.5 dB, mostly obtained by accounting for the more

complicated signal structures in this case. For local

events, characterized by partly incoherent array signals,

relative SNR gains amounting to 5-10 dB were usually ob-

tained. A great advantage with the signa] estimation

techniques used is that both positive and negative signal

weights are permitted thus partly avoiding destructive

signal interference during beamforming of complicated or

very weak signals.

mm Ua^ta MtfMMMIHMÜ MMWMHBMM .... ■

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^" ******—'

- 11 -

H

U

o h M M W CJ o c;

m H CQ H .0 H ■

H PQ

H

m hi (0 H rCI

< a. H ^

•H C/)

rH ro vT) rsi CN fl o ^o M* ■r r-j ^ ^3, in o (N rH t-^ m t~- o> oi n cri n (.o ixi v? vr co ro isD (N n vj' rH o oi ro t^- p~ o oo o -^ n LO m r-- H

C^J >.£> fi -.r f i r-i LD rvj (N o tN i.D o oi M CM -y »-r n co M LD ^4' H CN

ro (N m yj '3' H o cr\ ^o ^ o f') o o vn n r-i r^ r-< KD CN! ro in rn "ü" m ID i-- ^r ••3' TT co m vr cr> ("- VJ- OJ o oo cn r- co CN CO O i—t ^ CM O\

r-H (N O rH o o CM o r^ o <N o m o (N o cn ll I I I I II I

O r-H rH I

iH ■^ r-< %r o r-^ o vr r-H CM i~~ o u^ m VD o-> CT\ cr> ^ r^ c? r- tH rH o r~- CN o «sr in m m m co -j i^- ui H OD m vo rH in H in in o cn »a1 H

CM C£) CO CM iH O CO r-H (N CO rH v: O ^< i-( r) in (N r- I O CM VT VJ' rH (N I

r~- CN r~ cn t~- cn «r in m CA N CD cr> H in in m i.o rH ro ro co (Ti o CTI VD iH r^ fN M" r-- ro co CM c> co co CM o r-^ ro o-i ro m co r^ o r* c\ CJ

(T\ vr rH r- o »:• ro ro co rH ro CM vr co ro •y o co CM iTi H m in CTJ r^ CD rH t-( fN rH "ST rH Ol rH CO CM rH <N

■tr>

CO

w

■J] S

< Pi

EH

CO

41 o> rj 0

C) ß

Ü)

< jr.

01 4-i

Q

O

m co ro r-- co f-H rH «-r vj' -T rH ro r^ in r- cri in vr m o^ ro i-^ in m co

m ro ro ro ro vi' ■M1 CO ro ri tj' ro ro ro si< ro sy ro «a« ro ro co ro ro ro

CM ö oi CN a\ rsj CM co r- to v^ M' K co öj ro in co o fs! r^ f-- CN m r* cor-icDcoocDcococo co co co oo ci c^ ai co o o o\ co o co co a\

CD 'V co cn o r i r^ rH rH r-H rH r-

co r-H r-H CN ro ro co ro i-~ err rH ro 'T iH o oo rg co rHrHr-Hr-l rHHrHrH (NM.HrHCN rHrH

cn co co co •sp o CM o m n o .o CD co co «sr '^' '3" c^ m GS co H co m • • •^ CD i-

• cn

• oo CN

■ ro

• • • o> lO CO CO

• • • • • • CO

• CO

• CO

• ro

• • o

• • cn

• ro

rH r-( rH in CN rH r-H rH ro o in H in C-J r-\ ro rH in rH •i- m ro CN ro r-i

CD r- "ST rH in CD m rH ro IT) :N CM CO rH in in |-> cn CO ';' m CO r-i CD CO m o

• • rl • ro

• rH • in • LD • rH IN o

• ■ •

r 1 • r-H • CO ■

vr • o * in • r-H • o • IM ■

ro •

CM • ro • to • r>) -.r

rO rH O rH

H rH ro ro M" i-- CM o co ■y o c> o ro o ro "^ co in r-- rH CM co H O rH r-( rH O r-) CM O O O O r-l O CM O rH r-l O rH O O rH

ro ro ro ro ro ro ro ro ro ri ro ro ro ro ro ro ro ro ro n ro ro ro Fo ro r- r* r- r^ r- r^ r^ r^ i^ r-- r- r- r- r- (^ t-- r-- I-» r- I-- r- r^- r-~ r^- r^

CN ro ro ■r-r T m m CD CD CD i^ ro co ro ro o o CN ^ 'j m in o CD CD O O C> O O O O O C) O O O O CJ O r-H r-l rH rH rH r H rH rH rH H

rH rH r-l rH rH rH r-H rH rH rH rH rH rH rH H rH H r-l rH rH r-H rH rH rH rH ooooooooooooooooooooooooo

rH CN ro ^ in co r-- oo oi o H CN ro rr in co r~ oo oi o r-i CN ro -r in H rH rH rH r-l r-( rH rl iH rH CM (N CN (N CM (N

in

ro

PJ

CO I

ro o

ro

rH rd >

G •rl n t.T>

C) tn rö u C) >

■0 QJ

0) U 41 10 c; u r H a' ci

tt) Cii 3 u r 1 n r-l •r 1

Q) 11 ;■> (0 o. W fil

I I ■H

(/) j> (U

N H K li) rH t!l J ^ H' e rH C (/■) ..'. 0 QJ ■H M 'Ü T) l/l o 01 c I) ' 1 .c: | ■

(0 •-■ 4 1 ai V, , . tP

-C QJ in |ii ■rH 1 -1 r^

CM C •rl >1

M H

14 1 11 H 0 ll Ml (11 »

0 1 ro '0 ^—,

w lt-1 () CJ m •H r ( l ;.■"! 4J i-~

in (;' .; 01 Oi j., 0) II ^ ■ H H

r H -Q < O rH

nt \ i |14 >■

i: rH H 01 OJ id (0

G HI ^^ Vl r-,

.0 i; O r 1 • 0!

•■H •H Cl U) KC CO I i M 4-1 H 3

'O c 0 H •!' 0) ;;; *.] til

in > ■H rH o3

P G • >■. C! Q) 0 U) i-; ■O c

in (i rH (i 0 m II C) C) 111 y rt.' J> l/l U)

0' i ' C) ■. 1 0 V4

■rl H r 1 (1 u Q) r. i: '1-1 .c. r-H fl) (11 m u ( t M to r* 0 o '0 ■H ■H I-' Hi o :•: (1

i:

(1J. 4) 0, ■H

ß *— 1 iii . 1 M rM 0 (■; (!) rC ^ ■,( ' * 0) U U

4) • n' T1

W '0 r. ■ ^ n; hi fi ( C) 1 i

1 j»

m Ti •rl 4-1 ■H w < m

Cn in rH Di c

:i-i rH

L-i (ji Vl (0 n 1 ■H (U

. 4 m 0 10 ■H (Ö Li •H tT 1H 4.1 r: ■r| 1 1 C ■H u til M li ■r 1 tf d

•H (11 o ( i !!■ in in l- (>i : i

l< () (!) QJ '-i rn Vl tu d

ri;

r-l i), i: ■H

■4-)

tn U 10 i: (0 Cl

H i i

4J

o r 1

(IJ ^'

H rn in ii i Tl CM

,-1 u ■r 1

(i) Is (M

ii R

in rH

Ü) U) a) 'M .c ^3 V4 '0 0 4J

11

Oi c

■rH

II m i i

,i

4 1

i: rci

'0 •H M H 0 CO c ;j H ro (0

Q) Hi rH QJ

QJ f!l .',' C Cl •P Vl

cO (0 •H ■0 c* ri; •1 1:

VH Ü M

0 (' 11 .G O 114 4J

H Ml V. U) (ll B

rH (!)

(0 (1)

(Ö 4J r-l

-1

• rl V.

'Ü ,(.' Oi ii

C) fO ■ i l ;■* 01 rl ( 1

(; ■(1 in i.

ii i V\ c: rO it i) 0 ü (1) Ql

QJ u ■H 'O 4J I'l (1) 1 > ;l i ;1

11 VI 4 > •r-l [ ■

. i QJ ■rH r -( rj ;i 1^ rH

in (/l i: U. Cl in QJ « D 1: Vl ■H

li, 11 ii i,i (i3 ■o

■MMI ^1 i •■*M^-""—-— ■

I

12 -

_ m rn oo co yj in Ol r- o r~- in r\ r-i in tn öD o vr r- c.\ cn in o OD CD r- CD

H 'U

rH CO CN «y vr vj' U3 O in rn sr r- ai co c:> o CN CN M' CN CJ rH ri CÜ f- CD

H H vr H rn CNj o rn CN CN rl C\J H CN CN rn rn o CM m H rn ws . H ■=r CM rH CN

O c) c.i os r~ CM CN O rH CO rH O) Cn tf Cn rH 01 O CN o i-- CD (N ,_, CO Öl

< LD VT) co o r-j IX) rH CO M1 CÜ CM ^i in cn fl tj' o C iCl CO rH in CD r- CM r 1

H rn • ■ • i • • ■ • • • • • • •

hi

H 'O fl H CM OJ O 1

cn CM rH rH rH rH f\l (NJ r-i CM C-) CM m rH C^J kD

1 1 C_) c 1 (Nl rH CN

Ü n O un r- ■:;■ •■:• H rM O M rn rg o cn co H CN CO H (i co CN CO o 01 in

H cn CN H l^ H «3 O >X) cn o cn CD U? O C0 O o •«T C) CM li1 r-l vr ys CN rl ci

CO H rn • ■ • • • • • • • • ' • • « (/) 1 ) ■j (N O r-l rH O --I rH O r-l H O O CN rH r I O r-l O r-l O rH o o c ) CJ O

w CJ o n,

1 1

00 in ^o er.' c\ r-l .-1 t-- rH 01 rn rH in ix» ri v) ci CM r^ OD cn m in *r r- r , P; r-- M co r> c\ o in m r-j -j" ai o co in co o r- rn in in r^ c> CN CN CG

i ; tn <N n co in m in m rH O 'T rn o o r- cri in o CN CN m r-H •^r V tr »O

r-t r-H m n «j- rn m in CM H H rH

(.. es' 00 CM r-l cn 01 tn t- ui o ^1 Öl CO IT. H O O r-t oi r-- r-l o H r-l "* 1.1 .n • • . f • • • • • « r • • ■■■

f ; vr n vr T tp cn ■■r «sr vr cri vr -»j1 ^r rj* *>r ^r vr ^ cn c* i --j' sf ■«a' T N'

Di w W W W W u :J W M rn cj w w u tq a w w cj w a w W w W ',; tn ai i - m oo ■.' m CTI r-l r-t O O O O CD C \ ' -1 rH O O c-1 ;:) r i ci Gi l-

n 0) OT (■•) •■j' rn rr •," m -r ■r «." «T VT «a" -T tn -J" •T^ M" T T ■„■• *--," r i c 1 ST

n n 'Ü r-l iH r( r-H rH H rH i—1 rl H .H rH f-H rH r-l rH r-H r-H rH r 1 r-l r-H rH r-l r-H

w E-i

i tn

-D tn f. •s g. '. '. ','. ',-'. r: ,-S *-r ».» r» r-• ',» »-,• r^r r- t ^ »:.. <-j «-. »-- *--* f 4 i • <-i » * '-• f i e ' Z ■■ t * /-.

r,I 0) UT i •*■ u i rn ---r CN rn 'T vr in ^r in -f x) rj LJ r^ cn CM •;• -j' t.'l CÜ tH N' 3 .-1 'Ö m rn cn rn H CN on m rn rn n m cn m cn -st' rvi -^r cn rn 'i' rn CN m N' rH

rj

■^ c^ cr\ o o O M* co in o O CN rH r-H 01 (- TJ« 'f O O) c^ CM • ;• o Öl

QJ ■ • • • • • • • • • • • • • C 1; to CN CTl rn u^ O m us in m <xi n r- co r-i rn H co CN -r io O r i iO H CN H

9H ■H O in O OJ rH i* o m r-i rn n ^r o in «r M' CN 'V in o CN o •-!' c-1 rl D ^.' H • • • • t » • ■ • • • • • • cn ftT • t"; 0> H m rn o ■o' m uD in in co r-i i--r-r--fn(--mcocom m 1-- rn rl

h> U CN O CN m r-l in rH rH , 1 CN r 1 rr O CM V m "3" m r-H rH "V o o •r rH ü w U • • • • • • • • • • • ' • " Cn

< Ä oo cn o "3" n in ■n' 'X) o o '0 co r- co iri rn to co o m co CD I- r 1 CN a « H H CN o o o r-H o r-H H r-H rH r-t O H ^4 O O r-H O .H O CD H CM U

0)

K O

ro ro m rn m rn m m rn m cn cn ro rn rn cn m m m rn n m n rn m 01 r- t- r- i- r- f" r-< r- r- r» t- r» r* r» r» r» t- r- t^ r> t- r» r- r- r- 41 N. X W\ N W wwxwwww w w

^* m IJT kO ix) r- co 03 CTi O c^ o O O O rH rH rl CM CN CN n ri •." in ii i Cu

D i-H H rH r-l rH H rl c^; CM CN IM CN CN M C^l CN IN CN CN C'i CM CN CN nj CM

-s •N XW •\ W w W W, \ V v vw^ww H rl rH r-H .-H H r-H r-l r 1 r-l r-H rH r-l rH r 1 rH rH iH r-H rH r 1 f i r-H r 1 r-l

O O o o o o O O O O O O O O O CD O O O O CJ o Ci O o

• n H CN cn vj' in vo r- co 01 CD rH CN rn -^ m in r^ co oi o rH CvJ m 'T in

!Z5 rH rH rH rH f-H rH rH rH rH rH CN CN (N CM CN IN

>i 01 .1.

ii 4-) • ci t?

'Cl rt (1) r. I-H 4J << H t/1 u •rl

ui tu CN

l|-l rj C)

S-l r:

! tu I 1

r-l Ii'. c

CN 1

111

n a' 0)

r 1

• t; ,o 0

H ■a tu tu r.

■r 1 •r 1 H-l rH'

t/1 i '■ C) tn 03 r 1 J?

Q) Ci •H M r: u 4J o1

0 C) •rH ■H

H 1 ?': 0) Dl .!-- iU tu Ul

(1 (1 fli CI a' c. X H-l

Pi •H

CD r; in fli

CJ tu n 'O

i > r-t n< rn Ul rd o tT* ■sl1

iJ ö rQ CQ -.1 li 01 11 < Ji .1'. c C-i i/j Ul 4-1 1 ■

G) Ul 1-

rH • i-l 1^ rj 1 fli i: (1 'tl ■

tn ■rl n IM

•rl t)) 'CI in tu

M v-i .C

n fli fc^

4-)

«; fti 1 1 f. o< ;i , i

1 1 (0 N 1 1 in Qi ai II t;

0) 4-) ■n H .c i;

1) 4-1 )i 10 r0 u 0 11

■rt fli

tu u n

11 in rH fO ß (ii •rH i-i (Ü ,v ,11 tn

rd •rl (Tl ;i 11 öS ;>' -. RC .c; 11 >-l

4-1 fli o 1-1 H 4-1 4-) 11 ai Ul r:

fl) •r| tu r 1 r-l u dl i>. -. i '0 'f. Cil n, ü \'\ Vl

ai 0)

^ < n) IH in m C) dl ai

'ii ;1 Vi in i',) r 1 111

J i 11 ai ^ H CI |.. ai

■ l fli 1/1 . t II< fl) fl! fli 'f 'A i; ;., Ul 10 1 -

MMMMMk , i- _iA .v..»

"—^""^^^ - T 1 I 1

L3

Dlfl 328

[J1Ö 251

020 202

D3B 37J

O'iti 377

DW 212

D5B ]9rj

070 25^ •

OIN 221)

02C 270

03C 178

D^C 201

D?C 2S0

OBC 217

07r; 20^

die 297

D9C 246

IOC 270

11C 235

12C 19Cr

13C: 252

l^C 247

3B 530

30 53*

20 2167

2P 1509

0

-3

20

23

19

D

~2\

?0

-IS

-1

30

%

51

3J

36

iq

-13

-30

-33

-'!C

-53

-31

0

0

0

0

—x/v 'AjV^Aj\^/V^Af\J\J^

-^A/V^/\^AA^/\/lAA/vA^^^^^-vvfV^vv'\

W\AW\AAV\AN^WVVAAA^

—A/v

/v

.'vV-u\/VV--vAAy-V A

0. 22t.!4a

0,29253

-0. Ü3071

-0. 2306&

-0.0B603

0. 17^197

-0,07569

0.20521

-0,3^325

0.05533

0. 24?^

Drills

D.^9711

-0.05705

-0. If350

-0.35^5

C. :J2^C2

-0.13575

-0,3539t

D.057C2

-0.027J3

-0,17255

ß.53 DB

7.ß3 03

0.39 03

2.07

Fig. 7 Very weak Western Russia event recorded by NORSAR, Alternatively the signal:; may reprcsiMit ,i rndc .lobe; df Led ion olf a local explosion in Lho Baltic Sea area. The columns to the left give subarray and array beam codes (the latter equivalent to II A&B, III A&B in Table 2), plot scaling factors and relative time shifts in dsec. The rlghthand column gj vo;; Model IIB subarray weights, SNR and relative gain in dB (m the array beams. The signal portion marked with a line is d.S sec.

am^m __-.-*—MBMH MitiMBtJ^ J—— . - ■ . . .^. .,,■.... ..■.. -ttitiMM^ÜL^tiUm HiMVnähitii^ - -^Ltu*. .. -.

T-™ .1 . iiwJini m||Jip.UPJ&.IJH lil.I..L. HIJIJ|»»IWWPp. - J ij .imiiii w y ^u,^ #.IIHJ,LI^L wp»p«nipuij ii pim ii * ^i iii /. ^«■mpw?*!* I-WJ^

- .14 -

Intrinsic P-wave Travel Time and Amplitude Anomalies

As is well known, P-wave travel times and amplitudes as

observed across an array like NORSAR deviate significantly

from that expected from ray theory and standard

earth models. Except for special subarrav

travel time correction files used for minimizing signal

energy losses during beamforming, effects of the above types

have been mostly ignored in at ray data processing. In

recent months, considerable efforts have been spent on

analyzing the above phenomena, i.e., whether there is a

non-random pattern in bhe P-wave time and amplitude

anomalies, the potential improvements in the array event

detectability and classification performance by taking

such effects info account, and finally to find more

realistic earth models to explain the anomalous P-wave

propagation effects. The results obtained so far will

be briefly presented in the next subsections starting

with time anomalies.

Array beamforming is a two-step processJ first the in-

dividual subarray beams are formed, and finally the array

beam. In the first case, the necessary time delays to

ensure proper line-up of the sensor signals arc based on

least squares P-wave front solutions. In array beam-

forming the P-wave front solution usud is a first order

approximation/ while the second order terms are the de-

viations between observed and predicted time delays (At.)

using so-called master events. In the latter case,

anomalies amounting to +0.6 sec. have been observed.

This justifies an analysis of the effect of

ignoring second order terms, in subarray beamforming.

▲ ■MMM MftMlH MMMMMM

The results obtained give that the subarray travel time

anomalies exhibit a distinct regional pattern (see Fig.8)

and that 9 5 per cent of the At, observations have values 1 i less than 0.1 sec. The corresponding subarray beamforming

losses are around 0.5-1.0 dB, with the exception of

subarrays 0513 and 07C (located on typical Oslo graben

structures) where signal energy losses may amount to

2.0-3.0 dB.

Since the subarray time anomalies are non-random quantities,

they should be predictable, so the following experiment was

undertaken (Dahle et al, 1973). The travel time variation, T., i

across NORSAR is modeled as a function of two factors, namely,

a trend effect, equivalent to the plane wavefront solution

and a signal or wave scattering effect. The basic idea here

is physically shown in Fig. 9, and the corresponding mathe-

matical formulation is given in eq. (1) .

T. = T +U r. +U r. +S .-In, i o x ix y xy i i

(1)

where r, ,r, arc position coordinates, U ,U are trend ix ly ^ x y

components (slowness), S, is scattering or Chernov (1960)

effect, and n, is the noise at the i-th site. The critical factor ' J.

in this kind of analysis is the autocovariance functions

typical for random media wave scattering models (Chernov

I960), and alternative1'/ observed functional values.

To demonstrate the usefulness of the above approach, the

arrival times at roughly half of the NORSAR sensors

were predicted using observed travel times at the remain-

ing SP instruments. Using actually observed travel time

data as a reference base, most of the intrinsic travel

time anomalies within a subarray are accounted for by

inclusion of the signal effect term in eq. (1) as

demonstrated in Fig. 10. Important, the minimum of the

^aaa^MMMiriMM *

■^H =^=-Zl-=l^.-—^~ ".T-^pyy^yT- ?yv „..■■■m* M ,mmim,m .o»^rwT^ir^, »y-w--w-w—nw—<yr.

- 16

400

320

2^0

LU _J 0= 360

cr

CD

CD 60

0L

-i 100

80

60

10

20

I—I—i—I—I—I i i I I—i i I i i i I J 1 L J I J -300 -200 -100 0 100

FUW*RESIDUAL [MSEC]

200

Fig. 8a Observed distribution of timing errors inside subarrays, Exact arrivals computed by iterative cross-correlation techniques on high quality, signals.

tmmm — UM ■ni n ■ 1 'i i tru-i f r ' fwaiMIinilir J-LMU

" ^^^^WflW^WHSwh* ( -■■■.,.......,. ~rf'li","'r',TT"'rn--rnnr 111 m v,gi^>^>i»Mii#^rTTffWIWMMlW"'WWnn'llff''a"'* *|fif*l'*t*W*',lWWP*'(IW**

17

N I.IJ > ( ) UJ 11.) y* Ifl (K 0.^ ~"\ O !-•

UJ < 0 O c^ CK CE: III ^

0

< z —1 < <r ..J f^t?. z r n tu '-"T

•4 ü

<

^ <

D 0

z X

<

< s «3 0

,}—4

10

m

CN

,^«-0

r^o

(oesiuj nvnaiS^y 3AVM-3NV1d

Lf)

~1 V.T UJ

■^■« — nJwi— iiwwM»wiw«MM«lwiwM«««py»»»wwaii»w>Bw>^MMWi

6 o 5 o LH O LQ Q 1 .— »7- CN 'ill

•<r

CO

— in

— vj

a: o

CM co :z CO

CO

CN

.-i

V( 0

in m >i r: ra a;

>(

11 o u

4) n (Ö a

M

41 in

OJ 1 r-l 0) 0 c

w a)

4-1 ;:: (.; CI 0) U r.

M-l r I

•H T.} :>

M O W

I' 0:

:> (d .0 0

n 0 r. C 1-1 .^

j-i 4' (0

o i-t

0, fli • N --

;>, M H Q) r< > |J f: U c: O '-I E r. o 0 (ü f.:

rc tn 0) r. -.; i_ fd GI r^ in 4J U

CM U

a) +J

to ffi a

I , o -

,0 CO

■.H I»

MMM^ MMM^MMMHI .—.-^«■MIIA, ri^M* __. . - — . * .

1 —i^^^^^M wmmmw*mmm*mim

- 18

LIJ Q Z)

Q.

<

o _J

O

LU 2

ULI > < et:

SIGNAL (OR CHERNOV EFFECT)

.^

^^■i

^^*m0*&0*

ARRAY DIAMETER

TREND EFFECT

-J

DISTANCE ALONG A CROSS-SECTION i

Fig. 9 Decomposition model of compressional seismic wavefield accounting for the intrinsic amplitude and travel, time fluctuations observed across NORSAR seismometer sites.

-—■"■■'-—•'--- - —— - BJifillr-i i i

,, ,, ..,_„.,....,. iimwtmmmmmmm™mm^'mm^~**^***mmmmmmmmmmmarmi^**m^^mm\\imiwmrmm^mim>''*''vi'

SAZ. ALASKA /f

15 ?.'3 URVE PRRflMETER

HOKKAIDO

15 25 MOVE PfiRFMEVER

Fig. .10 Contour plot of sum of squared differences between observed and predicted travel bime, including a scattering effect, measured in per cent relative to the same quantity neglecting scattering. Correlation distance and wave parameter as defined in Chernov (1%0)

1

■MMMMM feHÜHiMl Mif—IIBI« _

^--TW .1 i.tii.i ni wKi'-*mnwr;w,-*

20

sum of squared differences between observed and predicted

travel times was observed for a random (Chernov) medium v/ith

correlation distance of approx. 7 km which is in fair

agreement with similar results obtained for LASA (Aki 1973,

Capon 19 7 3) .

P-wave Amplitude_yariation_acrgss_NgRSAR

The first step in analysis of the signal amplitude

variations across NORSAR was the probability density dis-

tribution of this parameter. It was found to be approxi-

mately lognormal (see Fig. 11) when dominant signal

frequency was larger than, say, 0.9 Hz. This result

was explained by Ringdal et al, 1972, in terms of multi-

plicative response effects of multilayered earth structures

and is in quantitative agreement with the wave scattering

theories of Chernov (1960) and Tartarski(1961).

>-

i u-

08- »— z a 06-

> h~

b m (U- < m o CK Q- a/-

/ s-

,A (-01C,OE7)

-N ( 1.0,0.56)

/OBSERVED

-10 1.0 2 0 RELATIVE AMPLITUDE

Fig.ll ObservGd single sensor amplitude distribution for a Kamchatka earthquake occurring Jan 03 at-06.36.44 GMT (NORSAR bulletin). The amplitude values v/ere measured after applying a 1.0-3.4 Hz bandpass filter. The normal and lognormal distribution func- tions estimated from the observed sample mean and variance are also shown.

unMMjMi ^^.^^MMMM«-

IJIIHI)IU.III ..I/ "■-'" ■■■.,.. uuimipui«..»; ■IH.I.IU.J.III|!U.!>|P»|B"».««"»«> i^, n JUHPUHJ »ipi^ti iijmiHWtJ! u.." ,.i., imwHm\iwm-«™«*miAuwmmmmm!i'mmmm wiWMWiimPIWJiFwmuij." "»'JWI

- 22 -

The prediction experiment of NORSAR observed travel

time anomalies mentioned previously was repeated on

amplitude data and some results are shown in Fig 12.

It should be mentioned that the Chernov

media parameters used in modeling the autocovariance

functions and giving the best fit to the observational

data are similar to the corresponding values ob-

tained in the time anomaly experiment.

The optimal beamforming procedure (Christoffersson

and Ilusebye, 197 3) discussed in a previous section

actually takes advantage of the skewness in the sub-

array amplitude distribution. However, this method

is too complex for on-line data processing, but a viable

alternative is using the simplified scheme of masking

the weakest subarrays as demonstrated in Fig. 13.

Presently, work is in progress to map the NORSAR sub array

amplitude pattern in all seismic regions expressiv for the

purpose of 'zero-one' amplitude weighting during on-line array

beamforming. The subarray amplitude pattern may also

be instrumental in discriminating between very weak

seismic signals and signal-shaped noise wavelets.

Assuming that the above optimal beamforming procedure

is used, the calculated weights should be matched against

that expected for^a given region. Preliminary results

give that this method would be an important diagnostic

tool in classifying weal; signals - noise wavelets.

MUHtfta^HH ättäUhÜBiii/iki^m'ii-iäktrul.i.iuM imfnn m tir iito-i^MÜi ■fttiU'fli'i-iti- vi ' ' .- .

^Ll1 ■ ' imsmmEn?.rr.~r-:. —i 'jx-.ra

S.E. ALASKA

15 25 WAVE PiWIETER

HOKKAIDO

15 25 UFIVE PflRRMETER

45

Fig. 12 Contour plot of sum of squared difforences between observed find predicted log-amplitudes including a Chernov (scattering) effect, measured in per cent relative to sum of squared differences between observed and mean log-amplitude. Correlation distance and wave parameter as defined in Chernov (1960).

fMWMillililMllli in in inniliMiM

—^- .... .... ...... ,,.. |J'«I > ' J W IfKWiPMPIIllPWnpWlSIB.H.I IUll|»J,«nmwpwWWIB»(>™wiWW!IW«IJBPI 11,1 ■'■"H I l IMWI» lW»^^WW^IWW«IWiWW«PW»*

24 -

(WESTERN RUSSIA)

U* ! ' ' ' ' ' ' ' ' ' ' ' M I I I I I I I 15 10 15 20

u

12

-i' i' i * 1111''' n'1''' i Li

I©

I I I I I I I I I I I I M I I I l^l. 15 10 IG 20

16 Ii i i1iii i 1i iii 1 i rrrp

© 15 f—s^^- .^

E •D

-#■" ^

S '3

u UJ

g 12 i . 11

10

—®. —®i ® ~ -1 1 1 1 1 1 1 1 1 1 1_L L 1 1 1 1 1 1 1 1

5 '10 15 20 NO OF SUEWRRAYS

K J l 1 1 | l

© M i | m ■ 111 '■'IT

12

4 -

10 41 -

8 \ i^

-

6

\^\

-

/ \~ ">.

?

—®. 1 1 M 1 1 i i i i i i. i

^■

1111 Mil 5 10 IS 20

NO OF SUBARRAYS

(GREECE-TURKEY)

Fig. 13 Relative gain in event dctoctability using the definition of Rinqdal et a] (1972) as a function of no. of NORSAR subarrays for different bandpass filters. The (a) and (c) figures correspond to envelope beamforraing using (1) 1.6- 3.6 Hz, (2) 1.8-3.8 Hz and (3) 2.0-4.0 Hz bandpass filters. The (b) and (d) figures correspond to conventional beam- forming using (4) 1.2-3.;; Hz, (5) 1.4-3,4 Hz, and (6) 1.6-3.6 bandpass filters. The results were averaged over 12 and 11 events respectively for the Western Russia and Greece-Turkey regions. All the events analyzed were very weak, i.e., having SNR values between 2 and 4 and thus mostly reported by the envelope detector.

H z

i^Mta ■"''-■"-' "- '-■l

-*T'l'*~r?tr^7rr. ■ !■ ^- ■.■ i - -, HI ZPIffV -=^-7 '•ff^SM .Tn^rxwfup^iuf^uUlll.llUI

25 -

Seismic WQVG_Propa2ation_in_an_Earth_which__is_partlY M2^1§d a§_ä_B§Dd2n)_2r_9l}9£L12Y_^9di2

The typical features of the Chernov media are small perturbations amounting to a few percent of P and S velocities, density and the elastic parameters

u and X. The corresponding wave scattering effect 1 x could be significant as shown by Haddon (1973) in a

theoretical study of this problem. In case of seismic

arrays this kind of data represents an excellent tool

for observational evidence on seismic wave scatteiing

hypothesis due to the large number of seismometers

within a small area. For example, based on a thorough

analysis of NORSAR recorded core precursor waves,

Doornbos and llusebye (1972) concluded that the standard

P-velocity model for the Earth's core probably was not

quite correct. In more recent studies both Doornbos

and Vlaar (19 7 3) and Haddon (19 7 3) attributed the

above precursor waves to scattering effects in the

deep mantle. Moreover, Aki (1972) and Capon (1973)

have used array data for mapping the extent for which

the crust and upper mantle beneath LASA could be con-

sidered a Chernov or random medium. In short, based

on the evidence briefly discussed above, we feel con-

fident that wave scattering effects could be used as

a diagnostic tool, in detecting and classifying weak

seismic signals. We are planning to investigate most

aspects of wave scattering effects, partly in coopera-

tion with Dr. A. Christoffersson, Uppsala University,

and Dr. R.A.W. Haddon, Sydney University.

Seisniic_Maqnitudc_ Investigations

The m, magnitude parameter, measured on records of short

period P-waves, is a convenient and widely used tool

for ranking of earthquakes. More recently this parameter

■ i fMiilMtMii iiiimiliMi *""-""-"■' tti :.:■■

P»m\W^i'.-' iii . iij/-r.-wr f. >l V- .' .1-HmMftl^WW^n.ÄiW"».'« 11 "i 1 ^"«P^'1-,11 *>'" - J i,l»JV!4..1i«J '■■i-^W\ J.IPMFP ^11. »JU ^.ifi^Tr». <WT I ,11, .(..Jl II HI |.|l|lil|||imii!.||i

- 26 -

has become of critical importance in evaluating event

detection and discrimination capabilities of various

kinds of seismological stations and networks. The

problem of a possible bias in the NORSAR estimation

procedure of m, magnitudes and also that used by the

International Seismological Centre (ISC) in Edinburgh

have been investigated by Husebye et al, 197 3. The

main results obtained are as follows (see also Table 5

and Fig. 14 ) .

No. of Subarrays

3

6

9

12

15

18

Operational 19<No<22

dm(loss) (m.-units)

0.28 i 0.06

0.23 ± 0.0 5

0.20 i 0.04

0.16 i 0.04

0.14 4 0.04

0.11 i 0.03

0.08 ± 0.03

Estimated skewness of subarray max. power distribution

Estimated skewness for log- transformation of max. power

Correlation between signal loss and skewness effects

Sample size

dm(skew) (m, -units) b

0.0 '■0.01

-0.01 t 0.01

-0.0 2 > 0.01

-0.03 i 0.02

-0.04 ' 0.02

-0.05 ± 0.03

-0.07 i 0.03

1.26 i 0.6 3

-0.101 0.42

-0.40 corr. units

222 events

TABLE 5

Estimated magnitude biases due to subarray power loss and skewed maximum power distribution, conditioned on the number of subarrays. The latter parameter represents a decreasing ordering of subarrays

based on maximum power ranking.

M^y ^ia,^*ttmmm^m.^mm*—

BiWfpii,- i, iwi^n,, ly.r »1^lu..*1,Hr,1..ww-;,pw.„ iMiinvvm^^^nngp^piR ii. uii i i i ..fwmmmm'Wf u. ■•MR«>IU „ ll.ll l.miiu^«!^)! ■ WWJ .Wll .l.lllllllijllllMHimii.i ■ if. iimiwiMHi.u» Pi-«.--i n. wm I..IIIII,II. IM m^ß»m^m'inpmwm^m^

2 7

-1.& -as 00 0.8. 1.G0 PREDICTED MAGNITUD: [reLscale!

2i

Fig.14 A comparison between event magnitudes as predicted from a multivariate analysis of ISC data for Japan, and that of the individual stations used in the analysis. The relation- ship betwe ir ISC reported magnitudes and predicted magnitude is also given. Dots are observed points for this line. This figure .is based on 40 events occurring in the Japan region in 1968 and reported jointly by the 9 stations listed on the figure.

The signal energy losses observed during NORSAR P-wave

bearnforming do not'in average affect its event magnitude estimates due to a skew, approximately lognormal, P-

amplitude distribution across the array. A comparison

between NORSAR-NOAA magnitude gave that the difference

is largest at m, ~4.7 and then tapers off towards both

small and large event magnitudes. A multivariate analysis

of ISC data for Japan and the /Aleutian Islands gave a

consistent and linear relationship between the ISC event

magnitude and that predicted from subsets of 5-9 sta-

tions in the m, 4.0~6.0 magnitude range investigated.

In this respect the ISC reported magnitudes are con- side rod unbiased. We also found that the magnitude

observations may be approximated by a normal distribu-

tion. In many cases the magnitude station

m ————— "—■'■' ...

!.,.,..""■-."■■.■■-■. ^u «'"I1-1 I..,II.I.I.K..«,BI.I™..!.IIJII iMni'PnmiiipniMiwnui'.'i.iiiviwniiPL.inuiii 11 gin>uai<wiwjiiwijii vtmum'mmimi'ii^^^mi^Hiiimmmmmmmmimmmmii'mm'i

- 28

correction term v/as not a constant but a function of

event magnitude. This phenomenon is quantitatively

explained as the combined effect of the seismic spectra

scaling law (Aki, 1967, 1972) and the crust-upper mantle

transfer function.

äD^_I.,2£ätion_of_Seisnüc_Events_at_N0RSAR

The evaluation of the NORSAR event detection and event

location capabilities seems to be a popular topic as

a number of scientists recently have worked on this

problem, namely, Shlien and Toksoz (1973), Rlngdal and

Whitelaw (19 73) and Bungum and Ilusebye (19 73) . The

differences in the results presented by the various

authors are mainly due to data bases covering different

time intervals. However, as Bungum and Ilusebye (]9 73)

used both the most extensive and recent data, i.e.,

after improved time correction files, better bandpass

filters, incoherent beamforming, etc., had been incor-

porated in the array's on-line system, we prefer to

give a brief summary of their evaluation of NORSAR's

event detection and location capabilities (see also

Table 6 and Fig. 15 and 16).

Based on one year of data, Apr 1972-Mar 1973, the routine

event detectability of the NORSAR array in Norway has

been investigated in terms of 50?; and 90?, cumulative

detectability thresholds which were derived from

frequency-magnitude distributions. The best performance

was observed for events in Central Asia and adjacent

regions where the 90% cumulative detectability values

are in the rnngo 3.6-3.8 NORSAR m. values. For tele- J b seismic events the value is 3.8. For events with

\

MaMMaHllaa — aa^riMMHM^teial

■' "■ '■ <''"'"• "■"I"1'!'11 '"■" * wmm

- 29 -

rl n

4J 3 E

M

M-l

•H Q

CJ O

•r| 4J ft) U Q

tn

c (i) >

o

"* %- n in ^ TM in co r ■ •j' n r~l lO n ro o .H CN rH r~\ r-l CD o C5 O CN CM CNJ o .H

o Ü o o O o O o o o o O o o O +1 + 1 +1 +1 •II +1 +1 +1 +1 ■ii HI HI •II ■II +1 01 10 Cl r^ «* o (N o O .-,)• O^ CM r-l H O l- (N in rN n uo t*- r^ m 0> Ü o O LO r- o in o o ro ^ r-i ro

o o (M

CN CO

* CN|

en o

o O

CO o in n o

o HI

o +l

U3

O +1 ra CNJ

o +1 CO en

o HI

O

o HI Cl r\i

O +1

c^

o +i CM

o ■II

o HI CO

O HI

o HI LO

HI o HI CO

O I

CN

I

O

I o o o o

I o

I O .-(

I o

I

CN rH

HI

o CN

+1 in

in CM

HI CO I-- CN

.-I

.-( HI r-l •-I

O CM

HI ID

in CN

11 CO

CD CO

rH r-H

HI

r-l

O CN

+1

co

en

HI en «a1

en

IS) o un

HI +1

CN in

(N (N in

rH

HI

CN

in r-l

HI

en

rl

HI i-O CN CN

I

rH +1 m CO

en CN

HI

n HI Cx! CN

•fl

O I

•X-

ro CM

HI

en r-H

CN

HI .H yo

i

* CO

HI

n i

o r-«

HI CN in r-H

I

rH O n co

+i +i

rH r-H CN

I

■*

co

+1 o

o

n o o rH i'l m CO

o CA

CD in UD

o CO

in

o CN

O

>^ rn

CD r I

O

CN

o CO

o H 0D

CD CO UD

10 10 o ro CO m r-H H sj<

O CO (O

o CN CN

CD in

in o

o in m en

CD

CO

CO CO

o c>

in en ro ^o

o CN rH

10 o CN

CN CD in L0 CN

CD

y. to IÖ

. H <

I Ifl e at •r|

41 :i t> r(

It

.a ö U o y.

.i-i HI

o u

0

'0

4-1

< I

4) Ifl re)

W ü)

r-4

T) •.I

-I CN CO

« rl (/I (0 :i « 4) ■Jl 11)

M

1.0

4) i: o U

tn

4J in nl 01

.c: 4-)

n o w

co

a, •H rH ■rl x: Pi

I C) 3

■;i

■ Cl

;>■;

i r; id {'. ill

!■

U (])

W l

3 O

.1)

u 0)

f.: fi a y, i

CO

CD HI H

O

ID CD

+1 HI +1 41 CN O OJ "^ CN CO 01 CD <* ij« (v) m

in CM

■H

rH rH

I

o o OD (N >^ o

in n •vT CM

H ro

ro ro rH en CO CO 01 o rH rH ^<

rH

I ,c: 4-J

0 to

H CN

a

o OD

I c o ro

<1) t)l c fU

■ II 1/1

■11 Cl

.H

CD ro rH

CD

01

4) «)

■H Q

in rH

14 0) 0 V( VI

0) 0) 1-.' X .c 0 4-) .i

■H C) 4J Rl in m 0 r. •H 4J o . > •H rH OJ V &i 0) -s 10 0) >

c U 01 Tl 0 rH u ■H 0) (TI 4J .n OJ

■■o 3 4J u ß rH p S 0 W 01

4J 1/1 ry '0 in H ■H

Vi Iß MH T) 0) S G C 4-1 0 (0 c y. o

0) in 01 u •H o» tJl •H >H m (0 u, tti en (H Cl) 4-) tt) to ra > n; (Ö rj a a

") C 0 VI (ii < ß o 4J (0 2 i:

k. 'Cl in o Q) c 0) -H U n; 4J Ml n 113 -H

CO 0) yi ^ •H &

w c> ü •l-' H ►J MH S3 in tn m "H (D ^ -rH G 0 EH 'O 0) M M

0) 0) (C e j 4-1 Cl 4J 4-) tn

■H 0) Ifl (1) 4-) fl r-l U rj a U 0) 01 OJ 0 u £ u

rH c •P 0) 01 MH

01' M n MH o 0) ■H -.H GD HH 'O

MH tn fO •rl u r, c TJ 0 4J ro M 3

Ä V4 s

s 4-1 0) -H 3 N

•H r '0 fl rü ■H H Q) N rj U B (0 1) 0

U-l 'd (C 0) O d 4J U

(0 to a W (Ü 01 01 0) 4J 4J o si m rö c 4-1 -H 0 id v

■H 4) .r. 4» tn 4J Ql t/1 ■H •H rc; w '0 S 4J

T--* «"MT"-f^rr-! -——-'-^^-■-■u>r^-w--wi jj^ipuL.« ^j ,,Mr I^.I^JJ^..;,^ ^L« ^.^ijip.Ri.uj^iKf1 IIPIIWU»4W^<M,n^fK"^ ■^il,*^l^^<"-^^^B.<*t"^ ■ I'-I'IJ'-I"—'" TT??

- 30 -

tn C

IU 0 r.. D Oi — a) a; o M r- •t 4J

Lfl H Ä

ai P X o JJ u

Ä V< +J 0

+J

c "w 0

•,H W 4J OJ 3 H

43 lo ■M O

4J W W

-H +J

(Ö 0) 0)

-O H 0

4-J 0) n

r; nj & rj U) B QJ I C ;>. -H U H a a -P 3 Ä

a' -H m ^ n u

M-l M (Ö

H to (0 rH 4J 0) (0 c x: o 0) H

111 M rj o 3 C y. •H o < •0 ^ C

ai

4-) >4

• (1) ^.0 >

0) QJ

a) r: a) > -M s •rl 4J 4-i TI a) ri c - n MI 4-) 'J Q) IX) U '0 V

en •r4 IK

S1N3A3 dO UBQUnN S1N3A3 JO UBeWON UM3A3 JO aaawnN

A ■MM aaMaa^au — -■ ■■1 ■

I . . »WP ■.UT!^l1!'UifU^WPI

■—■- - ...-—-» i tummmmt.mmmm i»,id|l|.-|||UII«|plW|B«

31 -

III O

5 o 3

1.0

OS -

-05

<

1 ■ i i

/ 262 101

•

ß- ^>—Q^ ^L —C> 0

— ^iT —^

—i—,— -»— —i- 1— -1 1

■

is W 45 BO NOAA MAGNITUDE

55 6.0

r.) a »-

o •A

-0.5

-1.0 35 ^Q 45 50

NOAA MAGNITUDE 6.Q

Fig. 16 Average difference between NOAA and NORSAR body wave magnitudes as a function of NOAA magnitude for all events with epicentral range 30O-90 and 110O-180O

from NORSAR {regions 14 and 15 in Table 6). The averaging is done over bands of 0.3 magnitude units, the number at each data point gives the number of events, and the upper and lower bounds are the standard deviations.

Ml .i^.

UPjwiiwflSBWiwPMms'-puMW«^ iWB^gmm&i i .uuijBmnimmpi "•M' »"»"'"•■■ wwwww nn Wi ii,iL.g<iin ■ i LIII » mm

- 32 -

above 4.0 NOAA reports the larger n^ value, while

NORSAR reports the larger for events below n^ 4.0.

The accuracy of NORSAR-estimated epicenter solutions

as compared to those of NOAA were also investigated.

The best results were found for Japan and Central Asia,

where the median location difference is 9 5 and 105 km,

respectively. For teleseismic events, the value is

145 km. The biased errors in the location estimates

are demonstrated to have been eliminated for most of

the regions considered. Finally, improvements of

the present NORSAR event detectability performance are

discussed in view of recently developed array data

processing techniques.

Seismic Verification Research

So far our main research efforts have been aimed at im-

proving the event detectability and location capability

of the NORSAR array. Some investigations on the array's

capability to discriminate between earthquakes and ex-

plosions have already been undertaken, although at the

present stage only conventional classification criteria

have been used in the analysis of relevant NORSAR data.

The main problem with the application of m^ : Ms criterion

is to be able to detect the surface waves from small ex-

plosions and earthquakes. Three major signal enhancement

techniques have been used in analysis of NORSAR surface

wave data with good results, namely:

Bandpass filtering centered at around 20 seconds,

reducing the G second microseismic energy, which

sometimes can be very s-trong in the winter.

mtm ■MtfUMMHI -^-

Kvm^'^m^mmm^mwi' i^ammumi mi i ■. mi i ■■-"—■• | || i ■■ ■ "

- 33

Boamforming, which works well if the noise is

separated in azimuth from the signals. The signal

similarity is always high.

Matched filtering, which is a master event technique

that takes advantage of the time invariance of

recorded Rayleigh waves.

Fig. 17 shows the results from an m. : M study where b s J

those techniques have been applied for signal enhancement.

The lowest M reported is 2.51 however, at other times

M^ 3.5 may not bo detected due to variations in the

background noise. Preliminary results from an m. : M

study at NORSAR have been published by Pilsen and Bungum

(1972), and this work is continuing.

V >

5-

<s

• • > I M * » 0 0

' " o

SF%

4 5 6 7 rru

Fig. 17 Body wave magnitude inh versus surface wave magnitude Ms

for events located by NORSAR i.n Centra] Asia during 1971 and 1972. Open circles indicate presumed nuclear explosions

aMM-MH. MMMMMH

i ■ üuiwniiiu p» i i in i i iiu iiiiiiiijiiiua •■! < i .in iidnil'PaWW'i1»"*1 i n m mi ■■n n iiim^mm^^*^ n ■ i i i imwtm m m 'jmmmm.

- 34

The n^ : Ms discrimination criterion works well for

larger seismic events. However, the difficulty of

detecting surface waves in the low magnitude range

necessitates investigations of event classification

capabilities based solely on P-waves. Modified versions

of the complexity and t1 ird moment of frequency

criteria have been tested on NORSAR recorded events -

earthquakes located in Eurasia and North America. In

the latter region event discrimination using P-waves

only is relatively poor, and also the array's event

detection capability is not specially good. On the

other hand, preliminary results indicate that fairly

good discrimination between earthquakes and presumed

underground explosions is achievable for Eurasia. This

statement is restricted to event distances larger than

around 3,000 km from NORSAR, as the modified complexity

criterion does not give satisfactory results for shorter

distances. The principal investigators here are I.

Noponen and D. Rieber-Mohn.

In addition to the seismic noise there is for long

period waves an important limiting factor for the

detectability in the fact that waves from two events

are very often interfering with each other, maybe

as much as 20 per cent of the time. The long period

coda from a large event may last for hours, and another

complicating factor is that the energy is often scattered

in azimuth through reflections and refractions at con-

tinental margins. A study is now in progress, under-

taken by H. Bungum and J. Capon (M.I.T. Lincoln Lab),

where the energy distribution in the coda for a number

■tiMaa. ■'■-' ■■ ! '■ . Hi- i mil

- 35

of carefully selected events is studied at 20 and 40

second periods. The advantage of working at 40 second

periods is that the multipathing there is much less

severe and that the coda fall off more rapid1y. On

the other hand, some events may have energy only around

20 second periods. The results for NORSAR are. comparable

to those previously obtained for LASA by Capon (1972),

although it seems that NORSAR data gives less variation

in the way the coda around 40 second wave periods fall

off with time.

-—'■•"^'■- ■

- 36 -

3. MISCELLANEOUS

During the reporting period a number of scientists,

whose names are listed below, have visited NORSAR Data

Processing Center, Kjeller, for special research purposes

D. Doornbos Utrecht University The Netherlands

I. Noponen Seismological Institute Helsinki, Finland

R.M. Sheppard M.I.T. Lincoln Lab Cambridge, Mass., U.S.A.

II. Oh lender f Institut für Geophysik Kiel, West Germany

H. Korhonen Oulu University Finland

S. Pirhonen Seismological Institute Helsinki, Finland

Professor Tsujiura, Tokyo, Japan

M.L. MaKi Seismological Institute Helsinki, Finland

A. Christoffersson Statistical Institute Uppsala, Sv/eden

E. Hjortenberg Geodetic Institute Copenhagen, Denmark

J. Capon M.I.T. Lincoln Lab Cambridge, Mass., U.S.A.

3. Vermeulen Utrecht University The Netherlands

1 Jul 8 Dec

12 Jun

1 Jul 16 Feb 10 Jun

11 Sept 19 72 2 2 Doc 19 7 2 2 7 Aug 19 7 3

19 Dec 19 72 2 5 Feb 197 3 20 Jun 19 7 3

18 Sep - 27 Oct 19 72

8 Sep 19 72

16 Nov - 15 Dec 19 72

6 Nov - 20 Dec 197 2

20 Dec 19 7 2

12 Dec - 15 Dec 1972

12 Feb - 23 Fob 19 7 3 13 Jun - 7 Jul .1.9 73

21 Feb - 14 Mar 197 3

16 May - 4 Jun 197 3

18 Jun -■ 14 Sep 19 7 3

MM^M ■■

.-MMMMH-M

- 37 -

NOKSAR scientists participated in the follov/ing seminars,

congresses and meetings in the period 1 July 197 2 --

30 June 197 3.

13th General Assembly of the European Seismological Com-

mission in Brasov, Romania, 30 August - 5 September .19 72.

Participants: K.A. Berteussen, 11. Dungum, H. Gj0ystdal,

and E.S. Husebye. Altogether the NTNF/NORSAR group gave

seven talks, which are listed below:

K.A. Berteussen and E.S. Husebye, Seismicity in

terms of event detection thresholds

H. Bungum, Event detection and location capabilities

at NORSAR

H. Bungum, Array stations as a tool for microseismic

rcseax'ch

H. Gj0ystdal, E.S. Husebye and D. Rieber-Mohn,

One-array and two-array location capabilities

H. Gj0ystdal and E.S. Husebye, Noise suppression

problems

E.S. Husebye and F. Ringdal, Multiarray processing

problems

F. Ringdal and E.S. Husebye, Event detection problems

using a partially coherent array.

Norwegian Geophysical Society in Nesbyen, Norway, 2-5 October.

Participants: II, Bungum and E.S. Husebye. One talk was

given.

American Geophysical Union, 54th Annual meeting, Washington,

D.C., April 1973. Participant; K.A. Berteussen. One talk,

"Bias analysis of NORSAR and ISC reported P-wave magnitudes",

by K.A. Berteussen, A. Dahle and E.S. Husebye was presented.

■HMH—M mmm

- 38

Fourth Nordic Seminar on Detection Seismolc jv, Helsinki,

Finland, 12-14 June. Participants: il. Bungum, A. Dahle,

H. Gj0ystdal, E.S. Husebye, N. Maras, D. Rieber-Mohn,

0. Steinert. The NTNF/NORSAR group gave ten talks, which

are listed below:

E.S. Husebye, A. Dahle and K.A. Berteussen, Analysis

of possible non-random errors in NORSAR event magnitude

estimates

D. Rieber-Mohn and I. Noponen, New short period

discrimination criteria used on NORSAR events

H. Bungum and F. Ringdal, Diurnal variation of

seismic noise and its effect or detectability

0. Steinert, E.S. Husebye and 11. Gj0ystdal, Noise

stability and false alarm rate at NORSAR

E.S. Husebye, F. Ringdal and J. Fyen, On-line event

detection using a global seismological network

H. Gj0ystdal, Array detection and location capabilities

for events in Central Asia

A. Dahle, P-signal variations within NORSAR subarrays

F, Ringdal, E.S. Husebye and A. Dahle, Event detection

problems using a partially coherent seismic array

A. Christoffersson and E.S. Husebye, Amplitude weighting

for optimal gain in SNR during array beamforming

0. Steinert, Stability of array performance

Norwegian Geotravers Meeting, Bergen, Norway, 4-5 May

1973. Participants: K.A. Berteussen, II. Bungum, A. Dahle,

H. Gj0ystdal, E.S. Husebye. Three talks wore cfiven.

mmmmm MHMMiiMMI

- 39

REFERENCES

Aki, K.: Scaling law of seismic spectrum, J. Geophys

Res., 72, 1217-1231, 1967.

Aki, K.: Scaling law of earthquake source time function,

Geophys. J., 31, 3-25, 19 72.

Aki, K.: Scattering of P-waves under the Montana LASA,

J. Geophys. Res., 78, 1334-1346, 1973.

Bungum, H., & E.S. Husebye: Analysis of the operational

capabjlities for detection and location of seismic

events at NORSAR, Bull. Seism. Soc. Am., in press,

1973.

Bungum, II., & F. Ringdal: Diurnal variation of seismic

noise and its effect on detectability, In preparation,

19 73.

Capon, J.: Characterization of crust and upper mantle

structure under LASA as a random medium, Bull. Seism.

Soc. Am., In press, 19 73.

Cartwright, D.E., & M.S. Longuet-Higgins: The statistical

distribution of the maxima of a random function,

Proc. Roy. Soc. Ser. A, 237, 212-232, J95G.

Chernov, L.: Wave propagation in a random medium, Dover

Publications, Inc., New York, 1960.

Christoffersson. A., & E.S. Husebye: Optimum signal

estimation techniques in analysis of NORSAR and

LASA array data. In preparation, 197 3.

Christoffersson, A., & B. Jansson: Estimation of signals

in multiple noise. To be published as a NORSAR

Scientific Report ~ In preparation - 197 3.

■MM . „-anaH

- 40

Dahle, A., K.A, Berteussen & E.S. Husebye: Decomposed

prediction models for travel time and logarithmic

amplitude across seismic arrays;, In preparation,

19 7 3.

Doornbos, D., & E.S. Husebye: Array analysis of PKP

phases and their precursors, Phys. Earth Planet.

Interiors, 5, 387-39(), 1972.

Doornbos, D., & N.J. Vlaar: Regions of seismic wave

scattering in the earth's mantle and precursors

to PKP, Nature, Physical Science, 243, No. 126, 58-61,

1973.

Filson, J., & n. Bungum: Initial discrimination results

from the Norwegian Seismic Array, Geophys. J.R.

Astr. Soc, 31, 315-328, 1972.

Iladdon, R.A.W.: Scattering of seismic body waves by small

random inhomogeneities in the earth. In press, 197 3,

Husebye, U.S., P. Ringdal & J. Fyen: On real time process-

ing of data from a global seismological network,

NORSAR Technical Report No. 43, NTNF/NORSAR, Kjeller,

Norway, 19 72.

Husebye, E.S., A. Dahle & K.A. Berteussen: Bias analysis

of NORSAR and ISC seismic event m magnitudes,

In preparation, 19 73.

Lacoss, R.T.: Variation of false alarm rates at NORSAR,

Seismic Discrimination, Semi-annual Technical Summary,

Lincoln Lab., Maas. Inst, of Tech., Cambridge, Mass., 53-

Juno 19 72.

- 41 -

Rice, S.O.: Mathematical analysis of random noise, Bell

Syst. Tech. 23, 282-332, 1944.

Ringdal, F., E.S. Husebye & J. Fyen; Event detection

probier.,-, using a partially coherent seismic array,

NORSAR Tech. Rep. No. 45, NTNF/NORSAR, Kjeller,

Norway, 19 72.

Ringdal, F., & R.L. Whitelaw: Continued evaluation of

the Norwegian short-period array, Texas Instruments

Special Report No. 9, Alexandria, Va. , 19 73.

Shlien, S., & M.N. Toksöz: Automatic event detection

and location capabilities of large aperture seismic

arrays, Bull. Seism. Soc. Am., 63, 1275-1288, 1973.

Siegel, S.: Nonparametric statistics for the behavioral

sciences, McGraw-Hill, Intern. Student Ed., New York,

1956.

Steinert, 0., E.S. Husebye & H. Gj0ystdal: The NORSAR

array false alarm rate and its noise stability,

In preparation, 1973.

Tatarski, V.l.: Wave propagation in a turbulent medium,

McGraw-Hill Book Co., New York, 1961.

■HMMM MmtltMIMtM