DTIC FILE COPYGL-TR-89-0335

NORSAR Basic Seismological Research1 January - 30 September 1989

0WO Editor:

Svein Mykkeltveit

I NTNF/NORSARPost Box 51

SN-2007 Kjeller, Norway

29 November 1989

Scientific Report No. 3

APPROVED FOR PUBLIC RELEASE; DISTRIBUTION UNLIMITED D TICS- ELECTEN. MAR28 1990

GEOPHYSICS LABORATORYAIR FORCE SYSTEMS COMMANDUNITED STATES AIR FORCEHANSCOM AIR FORCE BASE, MASSACHUSETTS 01731-5000

• 03 8 122

SPONSORED BYDefense Advanced Research Projects Agency

Nuclear Monitoring Research OfficeARPA ORDER NO.5307

MONITORED BYGeophysics Laboratory

F49620-89-C-0030

The views and conclusions contained in this document are those ofthe authors and should not be interpreted as representing theofficial policies, either expressed or implied, of the DefenseAdvanced Research Projects Agency or the U.S. Government.

This technical report has been reviewed and is approved forpublication.

ota" t F.nLWaO CI o J AMS F. LEWKOWICZ , /

(CQ~ractmanagerBach ChiefSolid Earth Geophysics Bra Schid Earth.Geophysics Br chEarth Sciences Division Earth Sciences Division

FOR THE COMMANDER

DONALD H. ECKHARDT, DirectorEarth Sciences Division

This report has been reviewed by the ESD Public Affairs Office(PA) and is releasable to the National Technical InformationService (NTIS).

Qualified requestors may obtain additional copies from the DefenseTechnical Information Center. All others should apply to theNational Technical Information Service.

If your address has changed, or if you wish to be removed from themailing list, or if the addressee is no longer employed by yourorganization, please notify GL/IMA, Hanscom AFB, MA 01731-5000.This will assist us in maintaining a current mailing list.

Do not return copies of this report unless contractual obligationsor notices on a specific document requires that it be returned.

1o'*

UncI,-si f iedSECURiTY CLASSifiCATION OF THIS PAGE

Form Ap.rovedREPORT DOCUMENTATION PAGE OMB No 0704-0188

la. REPORT SECURITY CLASSIFICATION lb RESTRICTIVE MARKINGSUnclassified

2a. SECURITY CLASSIFICATION AUTHORITY 3 DISTRIBUTION/AVAILABILITY OF REPORTApproved for public release;

2b. DECLASSIFICATION / DOWNGRADING SCHEDULE Distribution unlimited

4 PERFORMING ORGANIZATION REPORT NUMBER(S) 5 MONITORING ORGANIZATION REPORT NUMBER(S)

GL-TR-89-0335

6a. NAME OF PERFORMING ORGANIZATION 6b OFFICE SYMBOL 7a NAME OF MONITORING ORGANIZATION

NTN F/NORS AR I (f applicable) Geophysics Laboratory

6c. ADDRESS (City, State, and ZIP Code) 7b ADDRESS (City, State, and ZIP Code)

Post Bo: 51 1an.;coin Air Force BaseN-2007 Kieller, Norway assachusetts 01731-5000

Ba. NAME OF FJN[ING"'ONSCI^'C -d -'ICE SYMBOL 9 PROCUREMENT INSTRUMENT IDENTIFICATION NUMBERORGANIZATION Defence Advanced (If applicable) Contract No. F49620-89-C-0038Research Projects Agency NMRO

Bc. ADDRESS (City, State, and ZIP Code) 10 SOURCE OF FUNDING NUMBERS

PROGRAM PROJECT TASK WORK UNIT

1400 Wilson Blvd. ELEMENT NO. NO. NO ACCESSION NO.

Arlington, VA 22209-2308 62714E 9A10 DA BlI11 TITLE (Include Security Classification)

NORSAR Basic Seismological Research- 1 January - 30 September 1989

12 PERSONAL AUTHOR(S)Svein Mvkkeltveit (ed.)

13a TYPE OF REPORT 13b TIME COVERED 14. DATE OF REPORT (Year, Month, Day) 15. PAGE COUNTSCIENTIFIC REP. 3 FROMHLI,1TO_89__0/31 1989 November 29 78

16 SUPPLEMENTARY NOTATION

17. COSATi CODES 18. SUBJECT TERMS (Continue on reverse if necessary and identify by block number)FIELD GROUP =SUB-GROUP -Surface wave modelling; source and structure invor-

I sions; topographic effects; local magnitudes' regionalattenuation, 2etwork detection; Rhase associatior.

19 ABSTRACT (Continue on reverse if necessary and identify by block number) region-spe6ific knowledge.

Sj f - '

This Annual Technical Report describes the work accomplished underContract No. F49620-C-89-0038 during the period 1 January - 30Septembe; 1989. The report comprises five separate investigations,addressing the main topics within the scope of work.

As part of our effort to enhance the performance of the IntelligentMonitoring System, we have embarked on a series of projects to aid inthe understanding of the behavior of the tectonic processes currentlyacting , , thin Norway as well as wave propagation effects for observed

20 D:STPIBI ''"N ,''4 " A9LITY OF ABSIRA.- 21 ABSTRACT SECURITY CLASSIFICATIONEL-UNCLASSIFi-i , I Ir.rTED [ SAME AS RPT l nTir '.SERS UNtLASSIFIED

', N F n '.....-, . , r,,. .. . . .. 22b TELEPHONE (Inhide Area Code) 22( OFFI(E 'YMIi)I.ies i , ',:I- wicz (617) 377-3)028 1 (I../I.1WII

DD Form 1473, JUN 86 v,.vous ediri:ns a'e obsolete I CIJRITY (1tASIiCATON Oi)F Tills PAC,

lnc lass i f i ed

19. (cont.)

seismic phases. In paragraph 2.1 of this report, new information isconveyed about earthquake faulting mechanisms and quantification of thestructural information contained within waveform data observed inNorway.

An analysis of the location capabilities of NORESS and ARGESS, and ofthe 3-component stations within the arrays, leads to the followingconclusions concerning 3-component slowness solutions for P waves: (1)There is a relatively large scatter in the solutions for events fromthe same source region, and (2) there are significant differencesbetween the solutions at the different stations. In the contribution ofparagraph 2.2 of this report, topographic effects have been analyzed bycomparing slowness solutions for P from explosions in three differentsource regions. Numerical modelling demonstrates that surfacetopography can explain about half of the slowness anomalies observed onthree-component stations.

Paragraph 2.3 is a status report on work related to local magnitudesand regional wave attenuation. The work comprises the establishment ofa large data base of regional records, the development of analysistools and strategies, and subsequent data analysis aimed at derivationof regional wave attenuation relationships and the development of a newmagnitude scale for Norway and adjacent areas.

The paper of paragraph 2.4 presents a method for associating phasedetections from a network of stations, which is analogous to theconventional delay-and-sum beamforming commonly applied in arraydetection processing. Examples of application of this method are given,based on data recorded at the three regional arrays (NORESS, ARCESS,FINESA) in Fennoscandia. The problem of continuously monitoring theregional seismic noise field is also addressed, with th- r"pose ofobtaining a quantitative assessment of the upper limit k. i-.,-nitudes ofseiqmic events that would go undetected by a given netwoI.

Two investigations aimed at deriving region-specific knowledge fromanalysis of NORESS data are described in paragraph 2.5. The firstinvestigation is based on a study of the complete NORESS detectionlists for the period 1985-1988, and statistics on phase velocities andarrival azimuth residuals are obtained for events reported in theregional network bulletins. The second investigation is a detailedSttudy of 103 events in the western Norway area. The study shows thatarrival azimuth residuals observed at NORESS for Pn and Sn phases vary

v with source region in the western Norway area.

IIU

, I I

Prefalce

Un rder C'ontract No. F419(20-C(-89-0(t3'1 . NI"! /NORHSAl is (o0 01 i O (

soarch withinn it% wile range of slibjocts relevaint to seismic rironlitonog. 'Ihi omirra.is f tre eserchprogr;iri is On rleveojrpiirg arnd ;1rssssinlg nw'thr'ois for processingo

of data recorded by networks of' sniall-apir I !' irrray a d :I-comrrpimoni st it 'i01i .

for events both ait regional and tolosel.Ilorrc distalnces, III aduditIionr. Irior''gnr

sf'i.smrooj'Icar rele; reir topics ar 'i r

Fail r(rlrt'ivY t('cr'liCaI report under this wmrit c preso''rr)t(, 00or scevmi

mienit of vork. Sruitniarli-s of thIe rostI;ircl ofrt, witin 1 r Jrp'1ifi it.- ;I 0jroJ(

ar:e givoelir m arrirriall technlical rvporl .

'Ilbis -uciert ific Report No. ~3 is t he a fnnual techn iical report for Ire perio 01

Januairy, - :;0 S ept ornb er 19,89. It co)urtt ;r Iris five separat e (-oir t ri b uinrs. ad rossI Trg

!f, dIIi I opi cs wi I t ii Ii the14 scoprc I f wt rto Ii i i coin pt r

NORSAI? Contribuion No. .113

Accession For

NTIS R IDTXJC TABUnannf clO~ lBJu9 Ufication

DistributIOr4

ii ~Ava l a i r

Dist Specil

A.L

Table of Contents

Page1. SUMMARY 12. SUMMARY OF TECHNICAL FINDINGS AND 3

ACCOMPLISHMENTS2.1 Surface wave niodellint for s5urco amd structur, :3

inversion2.2 Topographic effects on. arrays a' v. (d fliree-compo1tt stat (itS 152.3 Local magnitudes and regional wav. attemiation . 272.1 A generalized l)eamforming a0;lq,-ch to real t imo network

d(lection and phase association- 372.5 lIegioii-specific knowledge derived from analvsis of NO H',SS 51

data. V

f .% . - iI ,,r /

'T Vm Ii ll ll m rl mll I ma *1'wl i

1 SUMMARY

This Annual Technical Report describes the work accomplished under ContractNo. F49620-C-89-0038 during the period I January - 30 September 1989. Thereport comprises five separate investigations, addressing tile main topics withinthe scope of work.

As part ol our effort to enhance the performance of the Intelligent MonitoringSystem, we have embarked on a series of projects to aid in the understandingof the behavior of thi( tectonic ,ro(c,('ss currently acting withiii Norway as wellas wayve propagation effects for (,bservd seismic ili~ases. it pirag rai)lh 2.1 o

- this report, new information is conve sd about ,arthquake f;tulting nichianuisand quantification of the structural information contained within waveform dataobserved in Norway.

An analysis of the location capabilities of NORESS and ARCESS, and ofthe 3-component stations within the arrays, leads to the following conclusionsconcerning 3-component slowness solutions for P waves: (1) There is a relativelylarge scatter in the solutions for events from the same source region, and (2)there are significant differences between the solutions at the different stations. Inthe contribution of paragraph 2.2 of this report, topographic effects have beenanalyzed by comparing slowness solutions for P from explosions in three differ-ent source regions. Numerical modelling demonstrates that surface topographycan explain about half of the slowness anomalies observed on three-componentstations.

Paragraph 2.3 is a status report on work related to local magnitudes andregional wave attenuation. The work comprises the establishment of a largedata base of regional records, the development of analysis tools and strategies,and subsequent data analysis aimed at derivation of regional wave attenuationrelationships and the development of a new magnitude scale for Norway andadjacent areas.

The paper of paragraph 2.4 presents a method for associating phase detectionsfrom a network of stations, which is analogous to the conventional delay-and-sum beamforming commonly applied in array detection processing. Examplesof application of this method are given, based on data recorded at the threeregional arrays (NORESS, ARCESS, FINESA) in Fennoscandia. The problem ofcontinuously monitoring the regional seismic noise field is also addressed, with thepurpose of obtaining a quantitative assessment of the upper limit of magnitudesof seismic events that would go undetected by a given network.

Two investigations aimed at deriving iegion-specific knowledge from analysisof NORESS data are described in paragraph 2.5. The first investigation is basedon a study of the complete NORESS detection lists for the period 1985-1988, and

I

statistics on phase velocities and arrival azimuth residuals are obtained for eventsreported in the regional network bulletins. The second investigation is a detailedstudy of 103 events in the western Norway area. The study shows that arrivalazimuth residuals observed at NORESS for Pn and Sn phases vary systematicallywith source region in the western Norway area.

2

2 SUMMARY OF TECHNICAL FINDINGS AND

ACCOMPLISHMENTS

2.1 Surface Wave Modelling for Source and Structure Inversions

Introduction

As a part of our effort to enhance the performance of the Intelligent Monitor-ng System we have embarked on a series of projects to aid in the understanding

of the behavior of the tectonic processes currently acting within Norway as wella-s wave propagation (flects for observed seismic Ahases. .\ pr per under.,t ;ding of dynamic processes within the crust and upper mantl,, requires a detailedknowledge about the occurrence of earthquakes and the nodium through whichthe seismic waves propagate.

Norway and its surrounding areas, in particular the northern continental mar-gin, have until recently been relatively poorly covered in terms of seismic instru-inentation. In the last decade the situation has been much improved, however,through the installation of new regional and local networks and arrays. Thisincreased numbtr of eta tions has resulted in more detailed delineation of seismic-ity patterns as well as improved focal mechanism solutions and local magnitudescales. The methods of matching synthetic waveforms to observed data for the re-trieval of source processes has further increased our ability to discriminate varioussource types from the fundamental ohservations of seismic data. This niethod-olgy is strongly coupled to the determination of crustal structure for which wehave also gained much information.

The purpose of this contribution is to convey significant new informationabout earthquake faulting mechanisms and quantification of the structural infor-mation contained within waveform data observed within Norway and along theNorwegian continental margin areas. l)iscussion of the results of the modellingof specific events is left to a further report. or can be found in Bungun c. al.(1989).

Source Mechanism Dctermination

For an area such as the Norwegian continental shelf, with low-to-intermediateseismicity as described above, the earlier (pre 1980) limitations in instrumentalcoverage also severely limited the possibilities for obtaining information aboutcrustal stress conditions through earthquake focal mechanisms. More and betterseismological stations as well as improved methods for source anmalysis throughwaveform modelling have now changed this situation markedly.

A major effort towards obtaining new focal mechanism solutions has resultedin 28 new solutions. Focal mechanism solutions for several of the earthquakes

3

proved to be quite difficult to obtain using the conventional first motion analysis.There either were not enough data to well constrain the nodal planes or errorswere introduced in the take-off angles from the source due to misinterpretation ofthe travel path of the seismic phase because of the complicated geometry of thecrust-mantle interface (as was the case for the event in Figure 2.1.1, Event 5 inTable 2.1.1). In six of these cases, source mechanisms have been retrieved throughwaveform modelling, using the same method as Hansen et. al. (1989) applied fortwo of the largest earthquakes in the area (Events 2 and 5 in Table 2.1.1). Theresults have l)een i'otrnd to he very stable even when data from only one stationhave been used (NOIRSAR long-period for Event 1, NORESS intermediate-periodfor Events 3 and 4).

Since several events were found to have produced fairly simple and wellrecorded long period surface waves, an approach of source mechanism retrievalwas adopted that combines broad-band waveform modelling with the more con-ventional first motion analysis. The method (Hansen, 1989) consists of low passfiltering the broad-band records from the NORESS array (NRS) and Kongs-berg (KONO) to emphasize the low frequency waves up to about 6 seconds pe-riod and then matching the waveforms to synthetic seismograms computed for agiven earth structure and source mechanism in both a forward and inverse sense.Green's functions vere computed using the Locked Mode Approximation methodof Harvey (1981) for a crust-mantle structure derived from Mykkeltveit (1980)for the crust and Stuart (1978) for the upper mantle.

The results of this modelling are illustrated in Figures 2.1.1 (forward mod-elling) and 2.1.2 (inverse modelling) for Events 5 and 2, respectively. Three-component seismograms recorded for each event are shown after filtering androtation to vertical, radial, and transverse components. The corresponding syn-thetic seismograms are shown just below the observed ones. The method wasfirst applied to Event 5 (Figure 2.1.1) where excellent results were obtained. Theinverse procedures were then developed and the data set for the event in Fig-ure 2.1.2 was augmented to include the thret.component Kongsberg (KONO)long period data. To the right of Figures 2.1.1 and 2.1.2 are the focal mecha-nism drawings (lower hemisphere stereographic projections) for the nodal planesobtained from the waveform modelling together with the first motion data. Itis obvious that the first motion data are to some extent discordant, either dueto incorrect readings or false interpretations. Experimenting with different ve-locity models for location, and varying the source depth failed to improve theinterpretation of the first motion data. However, they were useful in helping todiscriminate between say two different waveform solutions and serve to supplyconstraints on further interpretation of velocity models in the area. Several ofthese events were verified with forward waveform modelling even if the inversionprocedure was not deemed warranted by the data. As an example, Figures 2.1.3and 2.1.4 illustrate the fit from forward modelling of two earthquakes (Events 4

4

and I in Table 2.1.1) near the Lofoten Islands at a distance of more than 800kilometers from the recording stations at NORESS (Figure 2.1.3) and NORSAR(Figure 2.1.4). It is very encouraging to find that initial interpretations of thesparse amount of first motion data can be verified to a much higher dgroe ofconfidence.

Due to tlihe complexitN of the crust and upper mantle structure in the vicinrityof the earthquakes, it was necessary to explore the effect of the assunied velocity structure on the low freq;i'ency wavefoni ,nooelling. Since we are ot-sorvingpihases with wavelengths on the order of 20 kilonmeters and longer, miuclh f thecom plexity affecting the high frf', ecy first motion data is Sif,,otlied out. 'llestructures used for (on puti n g svrith. i ic ,; niogranis therefore reflected the a,'er

age, arld more heniogenvous, prop( rti(s s;,npled along tIhe napi(,ity of the travelpath from the earl hquake to the rvceivi:,g stations. lv obs(,rviig the change inthe synthtic waveforms as a fIction of structural model and arthoike soiri edepth it wa.s verified that the solutions 'or the focal mechanisns and source (ldepthsare quite rtolusm ftr the 'ang period data. Variations in the wnm)ression and ten-sion axes of the focal sphere were found to be less than about 5 degrees due tochangc , in velocity models for a single station solution, and were improved wit,the inclusion of a second three-component station (KONO). It was also foundthat source (epthls could be verified from the modelling to a precision of about±5 km.

For Event 2 (Fig. 2.1.2), the waveform fit is excellent. It is seen trom the focalmechaiism plot that the fault plane solution would be difficult to obtain from firstmotion data ale -.. In fact, an earlier interpretation of the first motion data of thisevent indicated a normal fault with a near vertical axis of maximum compression.However, the waveform modelling shown here for this event completely rules outthis type of solution. Instead, the solution for this earthquake clear!y indicatesoblique thrust faulting in response to a regional stre,s field exhibiting NW-SEhorizontal compression.

For most of the 28 new events, a similar amplitude modelling has not beenfeasible because of the very different claracter of the long period waveforms.Small events did not have enough eno_,rgy in the long periods for them to be ob-served above the noise levels, while some larger events produced far too complexwaveforms to be easily interpreted. The complex seismograms are likely a resultof the rapidly changing crustal structure between cvents out to sea and th. seis-rtometers located in southeastern Norway. In these cases, a solution using firstmotions only is given, where a combination of local and teleseismic data helpsin constraining the nodal planes. However, when the events are within local dis-tance ranges, the higher order surface wave trains can be summed to producebody-waves that can be fit to the observations up to 2 or 3 Hertz. Figure 2.1.5shows an example of such an event (Event 6) occurring very near the NORESSand NORSAR arrays with recording stations ranging from about 15 to 85 kilome-

5!! !

ters. Figu re 2.1.5 shows that the velocity model is only crucial to events at largedistances at higher fre2quencies. An approach to improve this result is outlined inthe next section as an improvement to the velocity models.

hImprovcd Veloity Models

Although we have achiev'd good results in inverti:-g for source parameters bymatching synthetic an ' real seismograms, our success has been limited since wehave only been able to do this for a small number of events and at low frecj encies.Since we are using a laterally homogeneous modelling technique, it can be arguedthat such a technique may never be suitable for inversion in situations wherethe structure depth dependence changes significantly a'ong the source-receiverpropagation path, such as earthquakes in the oceanic crust that are recordedwithin the continental crust. Unfortunately the I- ")blem of regional forwardwave propagation modelling through arbitrary inhomogeneous media is effectivelyintractal)le at intermediate and high frequencies, and even at low frequencies itcan only be accomplished by making assumptions on the rature of the mediainhoiogeneities, such as slow and weak lateral variations, that are often violatedin real world <tuations.

The success that we have had demostrates that, at lea-st in certain situations.laterally homogeneous modelling can be used to "image out" both the structureand the source parameters. These situations seem to involve cases where tiesource-receiver propagation paths stay within the boundaries of the continentalcrust, which lends more credence to the laterally homogeneous assumption. Thereare not many crustal events, however, that are large enough to produce significantenergy in the low frequencies. We would like to be able to investigate smallcrustal ev nits, such as local and near-regional explosions, but we can only do th',by looking at higher frequencies where ;to source produces significant energy.When we look at th, events that we can match well at low frequencies, ourmatches deteriorat( rapidly as we inciease the frequency bandwidth, which isnot surprising considering the relatively crude nature of the structural modelsthat we have been using. As we raise the frequency content of our syntheticseisrnograins, we find that they becwne increasingly sensitive to detailed changesin the structural model. Thus it becomes, ,ident that, in order lo match syntheticand real data at higher frequencies, it will be necessary to invert for structuralparameters as well as for source parameters.

We have begun a systematic approach to the comparison-inversion problemwhich will take advantage of some recent work that has been done in this areaand which should l)roide a clear and realizanle path to detailed structural mod-els that will )rC(luce optimal fits between the real and synthetic seismograms.The core of our method involves a modified version of the direct inversion tech-nique develope d by Gomberg and Masters (1988) which uses iterative linear in-version base,' upon differential seismograms that are analytic-Ily computed from

6

a Locked Mode representation of the solution. We will carry out this inversionin stages, starting with very low frequency, band limited comparisons of rea! andsynthetic seismograms which will determine the gross cliaracterijtics of the struc-ture. We will then proceed to open up the bandwidth of comparison :n steps andinvert for increasingly detailed aspects of the structural model while maintainingthe average characteristics that weri. determined in the lower band inversions forboth source and structure. At the same time it may be necessary as we go 1i1)

in frequency to re-estimate the source parameters for each more complex struc-ture iteration. In this manner we hope to obtain the best average veloci ty-depthdependeiicies il specific tectonic provinces.

We have implemiiented the core conputations for the differential seisniogra Mnsat NOIRSAh and (.IRIES based on t lie Locked Mode methods of Harvev (19.8l).At this stage we are able to compute prt irbatoion seislo'p ranIls that (all be usedto determine the effect on the seismoira i of small changes in layer velocities.The current implementation is functionally identical to that of Gomberg atidMasters in that it is a perturbation implenentation which assumies that. onl hythe eigenvalues change, as opposed to a true differential implementation whichaccounts for both changes in the eigenfunctions and eigenvalues. The pertur-bation seismograms become inaccurate as the source-receiver distance decreasesand the frequency increases, so we are in the process of implementing the exactdifferential seisrnogram computations and undertaking a detailed study to showwhere the perturbation approximation breaks down. We have implemented thecore differential computations in a significantly different manner from that of(,omberg and Masters which allows us to do the exact differential computationsin a straightforward manner.

Roger A. HansenDanny J. Harvey, ('Il ES, University of Colorado

Rcference.,

Bungum If., A. Alsaker. L.B. Kvamme and R.A. Hansen (1989): Seismicity andseismot ctonics of Norway and nearby continental shelf areas. Submittedto Journ. of Geophys. Research

Gomberg, J. and T. Masters (1988): Waveform modelling using locked-modesynthetic and differential seismograms: application to determination of thestructure of Mexico Gcophys. J.R. astr. Soc., 94, 193-218.

Hansen, R.A. (1989): Surface wave modelling and source parameter inversionsof Norwegian earthquakes recorded at local and regional distances, EOSTrans. A GU1. iin 1)r- ss.

7

lansen, R.A., I. bunguln and A. Alsaker (1989): Three recent larger earth-quakes offshore Norway, Temra Nova, 1, 284-295.

Ilarvey, D. (1981): Seismogram synthesis using normal mode superposition: thelocked m.ode approximation, Geophys. J.R. asir. Soc., 66, 37-61.

Mykkeltveit, S. (1980): A seismic profile in southern Norway, Pageoph., 118.1310-1325.

Stuart, G.W. (1978): The upper mantle structure of the North Sea region fromRayleigh wave dispersion, Gcophys. J.R. astr. So-., 52, 367-382.

8

No. Reg. Date Lat Lon Depth Mag. T-axis P-axis TypeN E Az Dip Az Dip

I NN 810903 69.62 13.68 12C 4.7 49 38 :30 1 22 IHO2 WN 860205 62.71 4.69 20C 4[9 198 55 303 1P R3 WN 861026 61.83 3.20 1413 4.5 305 78 112 12 1?.1 NN 880131 68.03 9.58 20C 4.3 92 21 329 51 NO

5 WN 890123 61.97 4.42 26A 5.1 239 72 124 8 116 SN 890410 60.61 11.40 22B 1.9 35 42 303 3 RO

Table 2.1.1. Earthquake focal mechanisins coll -ected fr T i stud, with Il,,]

differenit coluiujs indi(cating event Jiumblclr, region (SN, Smollicasteri NorwaWN, Western Norway, NN, Nort hern Norway), (late (year, nnuoiut ih. day), latit ud'.longitude, focal depth (A, precision < 3 kin; B, 4-6 kin; C, 7-9 kin, 1), > 1(1kin), magnitude (NL), T-axis (mininium compression) azimuth and (lip, lI-axis(maximum compression) azimuth and dip, type of mechanism ( N, normal; NO,normal oblique: R, reverse or thrust; HO, reverse oblique).

9

Date: 01 23 89 1406

MRSL

Plan. I -r 0 ~ 6 l- 7.

Plaa 2 2-27 r 44 Sp o~IZ L P+xs.~r 340Pg .

50.0 75. 100.0 125. 150.0 17.20. 225.040 ip7

Flierl g- 0 0.00 8r 0.07P1- .Green Hunreio A6urc d0pth = 30.0.9

Freucignure 2.1.1. =oa mehns3slto0ro.owrdwvfr0mdlig

Event 5, 2:3 January 1989. T'he left framie shows three component,,, (vertical, r;adi~iIalwill traiISVerSe1) of NOR ESS lbroadl band data together with thleoretical svis-inogranis from the double couple moment tensor immediately below each trace.while the right frame showvs the double couple solution from the moment tensor.T'he first motion data not read by the authors are indicated by plus and minussigns for compressions and dilations, respectively.

10

Date: 02 05 86 1753

R3 Max Amp: -2491.3

0 .....PIn.I Sr Z3 . -..... p.16.

Pln 2 : S t o 100 Dip- 38.1 Sip- 62.7

P Axio S tr- 306.9 Pig- 12.0T A.W Sir- 189.2 Pig- 84.7

12 ...--. ~-...P Hera.: A. 302.0 Si.e--0.92

4T Horax: Ali- 212.0 Sill- 0-161

1 3 rj-j T -1.1d~p .. I 5 r1p rj Sl 13 I I I I Hor Dev Sirens- 0.638 Err- 0 000

50.0 75.0 100.0 125.0 150.0 175.0200.0225.0

Time, (sec)

Filter-in g 0 0.00 2 0.07norsarI lhz25 !eb5kono.grn 9/22/9!319:17:44 IN(QM

Figure 2.1.2. Focal mechanism solution from inversion of waveform data,,Event 2, 5 February 1986. The left frame shows three components (vertical,radial and transverse) of NORESS broad band data and Kongsberg Long Perioddata together with theoretical seismograms from the deviatoric moment tensorimmediately below each trace, while the right frame shows the best double couplesolution from the moment tensor. The percent non-double couple for this eventat these low frequencies is less than 1.

Date.- 01 31 88 1853ii lt I III I I II I I llN

Max Amp. -826.15

chae I-

cha: 2.Iz P__ _

IR P

ch: -Plane 1: Str- 25 0 Dip- 76 0 Olp- -0 0Plane 2 Sr- 319.3 Dip- 147.0 Slp- 161,7

ch:6 P Axis . 51- 329.0 Pig- 60.9ITP1 Axis Str- 92.'2 Pig- 24 2

P Horax: AW- 352.1 Bja-0 311I

Otr 26 200. dip - 75 00 alp - -60 00 Moment factor - 0 33900E +03 T Hors.: Asi- 82.1 Sloe- 0 746

III!I IIIII ;IIIII I I I I I I I I I I I i I liorDev Stress-O .628 Err- 0.000

0.0 100.0 200.0 300.0

Time, (sec)

Filterin 0 0.00 4 0.07Creen Funiction source depth = 20.0nlvzlhz20jan3l .grn 9/22/99 17:34:24 ly( R&A

Figure 2.1.3. Focal mechanism solution from forward waveform modelling fora Lofoten event of 31 January 1988. This figure illustrates the application ofthe modelling technuique as applied to the NORESS Broad-Band recording of allevent in Northern Norway at a distance of more than 800 km in the magnituderange of 4.0 to 4.5. The left frame shows three components (vertical, radialand transverse) of NORESS filtered Broad-Band data together with theoreticalseismograms immediately below each trace, while the right frame shows the focalmechanism solution used in the waveform modelling.

12

Date: 09 03 81 1839

NAON

c ha I-

cha 2 ______________________________

LZ P

NA01cha: 3 -

cha 4 ___0LR P

ch..____________ Plane 1; 31- 178 0 Dip- 80-0 Sip- 46.0P ' ne 2: St- 78.1 Dip- 46 9 Olp- 186,0

c F 4 P Axis Str- 300.8 Pig- 21.7L PT Ax~x Slr= 48.C Pla- 38.0P Ho,.,,: Asi -308.1 Oise,- -0.828

Str -178.00. dip 6 0.00 alp - 45.00 Moment factor - 0.95407E-03 T Ho,.,,: Asi- 38. 1 uSie. 0.688p p ii r iiiiiiI II I iiHr De, Stress- 0.707 Err-0.000

100.0 200.0 300.0 400.0

Tirme. (sec)

Filterin ' 0 0.00 4 0.10Green Function Source depth = 10.0nivzl hzI0sep3.grn 10/17/~1 :34AZ NO&A

Figure 2.1.4. Focal mechanism solution from forward waveform modelling fora Lofoten event of 3 September 1981. This figure illustrates the application ofthe modelling technique as applied to the NORSAR LP recording of an event inNorthern Norway at a distance of more than 800 km in the magnitude range of4.5 to 5. The left frame shows three components (vertical, radial and transverse)of NORSAR Long Period data together with theoretical seismograms immedi-atelv below each trace, while the right frame shows the focal mechanism solutionused in the waveform modelling. This figure illustrates the dominance of the Lovewave component for solutions derived from waveform modelling.

13

Date: 04 10 89 2114

Ma. Amp 4648 6

NAIA{

ci,,. 2-V

sZ L.

Plane 12: Sir. 278.6 Dip- 63.6 Oip. 350Plane 2 Sir. 69A Dip- 69.A Sip. 1487P.P : Sir- 301 9 Plg- 2.7

Si £75.60. d~p 63.60 alp 36.00 M~oment ractor 0.36095E-0l 1 TMie Sir. 3.4.6 P11- 42.7P Hor... A.i- 302.6 fia-0.097T Horx Azi 32.8 Si.- 0.539

Lj Hoc DeSrs. 0.788 E",- 0.000

15.0 17.5 20.0 22.5 25.0 27.5 30.0 32.5 35.0

Time (sec)

Filtering 0 0.00 6 2.00Creen Function source depthna Ia.22.gund grn I I/ I /gEl4: 11:47 XOJRI!'f

Figure 2.1.5. Focal mechanism solution from forward waveform modelling fora magnitude 2.1 event located within the NORSAR array. This figure illustratesthe application of the modelling technique as applied to the NORSAR recordingof an event at a distance of 40 kmn and to a frequency of 3 hiz. The left frameshows the vertical component of NORSAR short period data together with asynthetic seismogram calculated with the fault plane solution from the right frameconstrained by the first motion data.

2.2 Topographic effects on arrays and three-component stations

An analysis of the location capabilities of NORESS and AII'ESS, and of the3-component stations within the arrays, leads to the following conclusions con-cerning 3-component slowness solutions for P waves: (1) There is a relativelylarge scatter in the solutions for events from the same source region, and (2)there are significant differences between the solutions at the different stations. Athird result is common for both array- and 3-component solutions: (3) 'There aresignicant anomalies in the slowness of P from particular source regions. For aproper evaluation of NORESS and similar arrays, it is important to understandthe cause of these observational phenomena. In this work we have analyzedtopographic effects by comparing slowness solutions for P from explosions in 3different source regions: near Leningrad, in E. Kazakh, and near Blasjo in S.W.Norway.

.S!owncss analysis

A unified approach to slowness analysis with arrays and 3-component stationsis possible by expressing solutions in terms of a covariance matrix C (e.g., Es-mersoy et al, 1985). Here we introduce C as a function of slowness s by phaseshifting the signals:

C =(S) 1 ,,(w,, s)Fi(ws)dwl2r (1)

whereF.(w, s) = F,(w) exp(iws.xn)

and F, is the Fourier spectrum at channel n. Using C of eq. (1), the generaliza-tion of conventional beamforming is given by the normalized response

P(s) = gtCg/{I g 12 trC} (2)

where g is the predicted displacement vector for slowness s. The generalization(2) can be interpreted as a matched filter since the response P depends on ourchoice of matching the covariance matrix C:

gtCg=tr(CG) with G=gg t

Thus for a 1-component array: gT = (1,....,I)T, and for a 3-componentsensor: gT = (g9,gy,gz) = displacement vector. The latter is a function ofslowness s, and the surface interaction must be taken into account. On thisaccount we may expect 3-component slowness solutions to be relatively sensitive

*to the choice of near-surface model. It is possible to extend the procedure to ageneralization of optimum heamforming. For example, the normalized responseof the maximum likelihood method is

P'(s) = {gtC-Ig}-' Ig 2 /trC (3)

15

Although eq. (3) leads to solutions with apparently higher resolution, thesolutions are less stable and for location purposes stability is more important.Hence we will proceed the analysis based on eq. (2).

The Leningrad events

Figure 2.2.1 shows a typical 3-component record at NORESS from a miningexplosion near Leningrad. Five such events were recorded with good signal-t(-noise ratio, and Figure 2.2.2 summarizes both the array slowness solutions andthe 3-component solutions for 4 stations within NORESS. Standard error barsindicate the scatter of solutions for the 5 events. Error bars of 3-componentsolutions are given for only one station, but the scatter for the other stations issimilar. The theoretical results included in Figure 2.2.2 will be discussed later.The 3-component slownesses show a relatively large scatter, both of solutions atdifferent stations for the sane event and of solutions for different events at thesame station. These solutions take into account wave interaction with a planesurface. An obvious extension is to include the effect of surface topography.

Interaction with surface topography

The usual correction for surface topography implies a time correction for ele-vation, and possibly a particle motion correction for surface slope. These correc-tions are consistent with geometrical ray theory, which requires that topographicrelief be smooth on the scale of a wavelength. However, the topography of theNORESS area is not smooth on the scale of the relevant wavelengths (- 2 kin),and wave scattering effects may be important. The same is true for the ARCESSarea. To evaluate the scattering, we have applied a perturbation method wherethe solution is obtained recursively in wavenumber space (Doornbos, 1988). Forthe present purpose we need the surface displacement u(x, y). The required per-turbation series is then

u(k,, = ~)(, y (4)n =0

where U(kx,ky) is the Fourier transform of u(x,y), and U(') is a function ofU.., 1 < n < n, of the surface topography f(x, y), and of the surface gradientsOf/Ox, Of/Oy. The zeroth order term U° gives the conventional free surfaceresponse for a plane, tjie first order term U1 includes the Born approximation.and the higher order terms account for multiple scattering.

We have synthesized the free surface response at NORESS in the signal band-width (2-4 H1z), assuming the incident wave direction is given by the array slow-ness solution. We then applied the method of equation (2) to the 3 components ofsurface displacement, to obtain theoretical slowness solutions for the 3-componentstation sites. The results are included in Figure 2.2.2. These results explain abouthalf of the observed anomalies. We speculate that shallow subsurface structuremay enhance the surface topographic effect. The results represent a weighted

16

integral over the bandwidth 2-4 Hz. but single frequency solutions give a signifi-cant variation with frequency. This is consistent with the notion of response by arough surface as an interference pattern. Thus, different signal spctra front thesame source region may lead to different slowness solutions, which might explainthe scatter of solutions at a single station.

E. Kazakh cints

Teleseismic P waves from recent nuclear explosions in E. Kazakh havo beenrecorded both at NOR ESS and ARCESS. The P wave spectra from all events aresimilar at NORESS, but at ARCESS we can distinguish two groups of events. Thespectra within a group are similar, but there are significant differences betweenthe two groups. Representative spectra at ARCESS are shown in Figure 2.2.3.and average slowness solutions for each group are plotted in Figure 2.2.1 (a andb). There is a slight difference between the array slowness solutions, but there isa large difference between the 3-component solutions for the two groups. We havealso plotted the theoretical slowness solutions for the 3-component station sites.The assumed signal bandwidth was 0.9-2.5 lIz for the "high-frequency" groupj ofevents, and 0.9-1.6 lHz for the "low-frequency" group. The results suggest thatthe surface topography at ARCESS explains slightly less than half of the observedanomalies. It is also clear that the slowness results for the two spectra are ratherdifferent in accordance with observations. This supports the explanation thatthe often significant variation of 3-component slowness solutions for signals fromevents in the same source region can be related to differences in the signal spectra.

Bldsj0 events

We have analyzed in some detail the NORESS records of Pn from a suite of6 mining explosions in the Bl5sjo area in S.W. Norway. The mining site is about300 km from NOR ESS in an azimuth direction of about 2400. A record sectionfor one of these events with a plot of slowness as a function of time is shown inFigure 2.2.5. Slowness and azimuth of the first arrival are consistent with Pu in aone-dimensional crust-mantle model (e.g., Menke and Richards, 1980). However,it can be seen that the first arrivPl is relatively weak. In fact, this arrival is easilymissed for small events, whence Pn is often associated with the dominant part ofthe wave train. The dominant part about 0.5-0.6 seconds after the first arrivalhas a consistent slowness and azimuth anomaly for all events. Slowness solutionsas a function of frequency as summarized in the slowness/azimuth spectrum ofFigure 2.2.6 show that the anomaly is related to the frequency range 2-4 Hz;this is also the range where the signal has its maximum energy. The nearlyplane wavefront indicates that the anomaly cannot be generated near the surface,and it is suggested that these "Pn" waves are actually the result of scatteringat depth. From the measured slowness and time delay, the scattering sourceis constrained to be within the depth range of the Moho (30-40 km). Hencethe observational results are consistent with scattering by topographic relief of

17

the Moho. An interesting geological aspect is that the inferred location of theproposed topographic feature coincides with the border of the Oslo Graben.

Scattering by Moho topography

Scattering will of course affect all waves interacting with a rough Moho dis-continuity. However, the scattered waves usually arrive in the coda of a relativelystrong primary wave. In contrast, the first arriving Pn is relatively weak dIli,,to the small coelficient of refraction through the Moho, and scattering due totopography of the boundary may dominate the wave train.

'lb illustrate these concepts, we have calculated generalized transmission coef-ficienits for a rough Moho. In analogy to equation (4) for the displacement vector.we introduce a recursive solution for the vector of scattering coefficients:

B ( k , , "y ) : B ( O k , ) ( )

n=0

The components of B( ' ) are just the plane wave transmission coefficientsfor a plane interface. From the components of B we can calculate the energyflux for any wave type. Figure 2.2.7 shows the energy flux of P transmittedupward through a rough Moho, as a function of incident P below the boundary.The topography here is characterized by a correlation length of 5.6 km and anaverage height of 1 km, and the wave frequency is 3 11z. Two modes of scatteringare shown: (1) The specular flux E' in the direction defined by the plane wave-plane interface concept. The specular flux through a rough interface is reducedwith respect to the flux through a plane. (2) The diffuse flux E "C due to multiplescattering in all upward directions. E ' does not exist for a plane interface.The figure illustrates well the sharp increase of the ratio ESCIE as the slownessapproaches the critical value corresponding to Pn, thus supporting in a qualitativeway the scattering model for propagation of this wave. The results suggest thata careful calibration is needed before using this phase for event location andvelocity determination purposes.

D.J. DoornbosE. Odegaard, Univ. of OsloT. Kvaerna

References

Doornbos, D.J. (1988): Multiple scattering by topographic relief with applica-

tion to the core-mantle bouindary, G(ophys. J., 92, 465-478.

Esmersoy, 'V.F. Cormnier and M.N. Toks6z (1985): Three component array,

process~ng. In: The VELA 1rogram. ed. A.U. lKerr, DARPA.

Mlenke, W.ll. a:id P.C. Richards (1980): Crust-miantle whispering gallery phases:

A deterministic model of teIosvismic Pik wave propagation, J, Gcophys. Ries.,85, 5-116-5-122.

19

J+>i8|U ! 2 2100

-- F- -- -t -T - 7 - Tr-- r--- -- - -r- --r - -T----T -

C'Z 2I'0

0 too 200

Time (seconds)

Fig. 2.2.1. Typical three- onipouient NORESS records fron event in Leningradre*gion. Sc'aling factoIrs of difrerci coinl , lh t s are shown ti li,' 'f4.

2f

e - ALLV- A03CI~ -C23C

- C43CA-C73C

EA

cj 02

Trb

0.01

i. 000.07 0.08 0.09 0.10 0.11

Sx (s/kin)

Fig. 2.2.2. Slowness -olIntions at NORESS. Average over P from .5 events nearLeningrad. Array solution (ALLV) with standard error bars. Three-componentsolutions at the indicated sites, with standard error bars for site C2. Theoreticalsolutions including response to surface topography are framed.

21

70-

'60-

o-

0

0 2 3 4 5

Frequency (Hz)

Fig. 2.2.3. P wave signAl spectra at ARC SS from nuclear explosionsin Kazakh.

--:from group A (high-frequency signals).-- "from group) 11 ()ow-frequency ,signals.)

2;2

or I I I I I I II I I I

0.02 -- -

V - A03C

* - C23C- C41C- C"7 3C

0.01

E

0.00

W V

-0.01 i.L,

-0. 02

0.07 0.08 0.0) 0.10 0.11

Sx (s/km)

0.02

* - C73C

0.01

EA

0 00 1 #eEl

-0.02 I0.07 0.08 0.09 0.10 o.I1

Sx (s/km)

Fig. 2.2.4. Slowness solutions at ARCESS. Details as in Fig. 2.2.2.a. Average over 4 events from group A in Kazakh.b. Average over 5 events from group B in Kazakh.

23

2829.41-01 2.3 45 68

S 260-0J

S 250- Power.C 240-

:3 230-

N 220

210 .. .. . *'0.5-0

E0 1 2 3 4 5 6 7 a0.080

C) 0.100

0

0.120TO

3: + 91o 2 .. 5 8 7 8

Seconds

Fig. 2.2.5. NORESS record of Pn fromn an explosion in the Bl5sjo area. The toptrace is a single channel seismogram. Array solutions of slowness and azimuthas a function of time are shown below. The symbol size in the plot indicates therelative signal power.

2-4

- 260

C.) 250 0 0 Power*0 0

f 8: ) '0

N 220 o.8- 0< 210'

0.0 1.5 3.0 4.5 6.0 7.5 9.0 10.5

E o.oso - 0.6-0

U0.100

d) 0.120 05

o.,i @ 9 @ 0 a 8 0 o 0 e0 0,.,- 0

010

C 0.:60 0.2-

o 0.|80I I I I I.0 15 30 45 6 7.5 9.0 10.5

Frequency (Hz)

Fig. 2.2.6. Slowness and azimuth solutions of Pn as a function of frequency,from 6 explosions in the B1Isjo area. The symbol size indicates the relative signalpower.

25

q--10-

-20-

0.08 0.10 0.12

Slowness (s/kin)

Fig. 2.2.7. Energy flux at 3 IN through rough solid-solid interfpace with thev~lOitydefsitystrctue /- = 6.8/8.1 km/s, vt"/- =3.7/4.5 km/s, p+/ =

2.9/3.4 g/C1m3 . The boundary roughness is characterized b 'y an average height ofI km and a correlation length of 5.6 km. The scattered flux is upward throughthe boundary.

specular direction; - - -: integrated diffuse flux.

26

2.3 Local magnitudes and regional wave attenuation

Introduction

The research conducted during the last year within the field of regional waveattenuation and local magnitudes has been organized as part of a inor long termplan, with the following done so far:

" Establishment of a large data base of regional records, stored in a commonformat

" Development of analysis tools and strategies, and preliminary analysis ofdata

In the following, we will describe each of these subjects in more details.

Data base of records

For a long time, the network of seismic stations in Norway was limited toa few (2-4) conventional analog stations, and from 1970, the NORSAR array.Then, between 1980 and 1983, the regional Southern Norway Seismic Network(SNSN) was in operation, the Western Norway Seismic Network (WNSN) startedrecording in late 1984, and the Northern Norway Seismic Network (SEISNOR)in early 1987. With NORESS (1985) and ARCESS (1987) in addition, the wholecountry and surrounding areas are therefore now quite well covered.

It is being recognized now, more and more, that well-calibrated magnitudescales are quite important in very many seismological problems and applications.This is obviously true for problems related to detectability and yield estimation,but also to wave attenuation where magnitude in some cases is an importantscaling parameter. It is also being recognized that the development of a newregional magnitude scale requires a very thorough development of its wave at-tenuation terms, and that this requires large amounts of observed data in orderto be sufficiently reliable. The data situation described above for Norway makesthis the right time for pursuing these goals.

To this end, we have undertaken a major effort in terms of establishing a largedata base of regional earthquake records, later to be used in the investigation ofa variety of seismological problems. The data collected so far consists of morethan 200 earthquakes and more than 1000 individual records, with magnitudesin the range ML 1.7-5.3 and epicentral distances in the range 10-2000 km. All ofthese records have been converted to a common format, convenient for interactiveanalysis by program packages which already include all necessary system responseinformation.

27

For the records selected, all epicenters, recording stations and ray paths areshown in Fig. 2.3.1, while the magnitude-distance distribution is shown in Fig.2.3.2. It should be noted here that some of these records (near-field, large mag-nitudes) are clipped, while others (far field, low magnitudes) may be too muchaffected by noise.

Analysis strategies

In following Richter (1935; 1958), the local magnitude ML is defined as (seealso Boore, 1989):

AL = logA - logAo+S (1)

where A is measured amplitude (0-p) in mm on a Wood-Anderson seismometerrecording, S is a station correction, and

- logAo = a log(A/100) + b . (A - 100) + 3.0 (2)

where a and b are coefficients and A is epicentral distance.

With no station correction, this definition gives AL = 3.0 for A = 1 mmat a distance of 100 km. Richter's original A0 (attenuation) values, developedfrom a very small data base of Wood-Anderson seismometer recordings fromsouthern California, are surprisingly correct, as demonstrated recently by Bakunand Joyner (1984) for central California and by Hutton and Boore (1987) forsouthern California.

In developing a new ML scale for a different region, it is first of all necessaryto develop a new attenuation relationship for the area, in terms of the coefficientsa and b in equation (2). A problem in this respect is that regional differencesbetween attenuation at 100 krn makes it very questionable to tie the two scalestogether at this distance. In using a shorter distance, such as 10-20 kin, we thenon the other side run into a problem caused by the fact that Richter's A0 valuesare more poorly defined there.

One way to solve this problem has been suggested by Hutton and Boore(1987), who developed new attenuation relations for southern California froma very large data base. Using their A0 values, which are identical to Richter'sat a distance if 100 km, a magnitude 3.0 earthquake should be recorded withan amplitude of 10 mm at a reference distance (Ares) of 17 km. This distanceis more satisfactory for anchoring purposes, under the condition that our localattenuation function can be evaluated with a sufficient precision down to thatdistance. If that is not the case, a larger Ares must be chosen.

This leads to the following expression for Ao:

- logAo = a. log(A/A,,f) + b. (A - A,*f) + K(Arei) (3)

where K(Ari) is determined from the A0 values of Hutton and Boore (1987);

28

e.g. A,., = 17 kin gives K = 2.0. A regression analysis should then be aimed atsolving for the parameters a, b, S and ML in equations (1) and (3).

The collected data will be read from the data base and checked manually fornoise level, clipping, etc. Fig. 2.3.3 shows in this respect a panel of records froman ML 3.1 earthquake on 17 September, 1988, with epicentral distances from 176to 1308 km. The stations include three from SEISNOR, NORESS and two fromWNSN, and serves as a good illustration of the usefulnes in mixing records fromdifferent recording systems.

The usable records will then be corrected for instrument response (in fre-quency domain) in order to get true ground motion spectra as shown in Fig.2.3.4, or ground motion time series, in the cases when those are needed. Simu-lated Wood-Anderson recordings are then established by applying an appropriatehigh pass filter (2-pole hp at 1.25 lIz, most conveniently applied also in frequencydomain), and the resulting time series is then plotted (on the screen) for manualpicking of maximum amplitude. These amplitudes are then used regressively bycombining equations (1) and (3) to give

N, N,

logAi -a.log(AA,, /A )-b.( A L ) S_ 6Ak+ -1L 145-I"(LAref) (4)k=1 1=1

where

Ai.) simulated Wood-Anderson amplitude of earthquake i at station j

bi= Kronecker's symbol (=1 if i=j, otherwise 0)

N,= number of stations

N, =number of events

The parameters to be determined regressively are a, b, Sk and MLI represent-ing the geometrical spreading, attenuation, station correction and magnitude,respectively.

The true ground motion time series and associated Fourier spectra obtainedfrom the data base serve as important subsets of data suited for further waveattenuation research. Such efforts should emphasize the interrelation of geomet-rical spreading and anelastic attenuation in the computation of wave attenuation,possible local and regional differences in wave attenuation, and also possible az-imuthal effects. In contrast to the attenuation terms in the ML inversion, theterms here can be evaluated as functions of frequency (Dahle et al, 1989).

29

Preliminary results

In the section above, we have outlined a procedure for simultaneous inversionof Wood-Anderson amplitudes vs. epicentral distance in order to estimate M1.magnitudes, station corrections, geometrical spreading and anelastic attenuation.This can be done at the same time or through a multi-step regression analysis.In order to test the stability of our planned procedure we selected a few repre-sentative recordings as shown in Table 2.3.1, including 8 events and 35 records.The reference distance in this case has been 100 km, the estimated magnitudesare given in Table 2.3.1, while the station corrections are shown in Table 2.3.2.

The coefficients a and b in equation (4) are as follows:

a (geometrical spreading) : 0.70

b (anelastic attenuation) : 0.0010

These results are very encouraging in that we for all of the 21 estimatedcoefficients get reasonable values using only 35 independent observations. In gen-eral, however, we must expect this kind of analysis to be more unstable thanshown here, and that therefore the various restrictions for the different parame-ters should be considered closely, in addition to the possibilities for a multi-stepregression analysis. The resulting magnitude formula would read as follows (ig-noring the station corrections):

ML = logA + 0.70. logA + 0.001 - A + 1.5 (5)

We emphasize that these results are preliminary and intended only for testpurposes.

A. AlsakerL.B. KvaimneA. DahleH. Bungum

30

References

Boore, D.M. (1989): The Richter scale: its development and use for determining

earthquake source parameters, Tectonophysics, 166, 1-14.

Bakun, W.L. and W.B. Joyner (1984): The ML scale in central California, Bull.

Seism. Soc. Am., 74, 1827-1843.

Dahle, A., I. liungum and L.B. lvanime (1989): Attenuation models inferred

from intraplate earthquake record ings. Submitted for puhlicatiori.

lIitton, L.K. and I).M. Boore (1987): 1e AIL scale in soithiern Califoria,Bull. Scism. Soc. Am., 77, 207.1-2094.

Richter, C.F. (1935): An instrumental earthquake magnitude scale, Bull. Seism.

Soc. Am., 25, 1-32.

Richter, C.F. (1958): Elemcntary Seismology, W.1I. Freeman and Co., San Fran-cisco, California, 758 pp.

31

Earthquake No. of Old Newobs. ML ML

1986 Feb 5 1 4.8 4.92

1986 Oct 26 1 4.4 4.441987 Oct 31 3 3.6 3.821987 Nov 1 6 3.1 3.181987 Nov 3 5 2.3 2.391988 Jan 23 10 3.4 3.351988 Jan 31 6 3.8 3.651988 Aug 8 3 5.2 5.20

Table 2.3.1. Earthquakes used in testing inversion procedure for ML determi-nations, with number of records for each event, old ML (from current formula).and new ML (from inversion).

Station No. of Correctionobs.

NRS 8 - 0.02MORI 1 0.12FRS 5 - 0.02MOL 4 - 0.09LOF 2 0.17TRO 4 0.20KTK1 3 - 0.18SUE 2 0.05

IfYA 2 - 0.12ODDI 2 - 0.11KMY 2 - 0.02

Table 2.3.2. Station corrections from testing inversion procedure for ML de-terminations, with station name in column 1 and number of records in column2.

32

NIs75.0 /0

S65.0

.- 60.0

55.0

0.0 10.0 20.0LONGITUDE (DEG E)

Fig. 2.3.1. Locations of all events in the data base, connected with rmy paths,to each of the stations which have contributed with recordings.

:33

(X 000 o

S0 0 ,, 'Y

0 )*00 00

co o 0 0 0~X Z%) 0 (0

4.0 0.

.00 0 *N 0 <00* 0

3. 0 0 0~ 00 0Qimu0 O(N 3D3 (<O4GOt>M M E,

*0 cDO ccMD 0c o

(N0 0000

2.0 0 <KCk - cO

Fig. 2.3.2. Magnitude-distance distribution for all of the data in the data Lats((203 events, 1028 records).

KT

25 120 L Ir'J RS A 0

I O0

1988-26 I/O9:SO:OO.O0cl

__sz 10/0't/89 1't:30:S3S NORSAR

Fig. 2.3.3. Recordings at six stations from an ML 3.1 earthquake on 17 Septem-ber 1988, located at 61.420N, 1.570E. The two closest stations are from WNSN,while the others are from NORESS (NRS), and from SEISN OR. The epicentra!distances are SUE 1 76, JIYA 249, MIOL 338, NRS 543, LOF 937 and KTI( 1308kin, respectively.

35

MOL SO s S-record (338 k1988 261/ 9:SI: i.000

1.E-'t

E

w

_j I.E-0s

w

0

1 '-.0E-06

0.1 1.0 10.0 100.0

FREQUENCY

KTK 8S s S-record 1308 k1988 261/ 951:21.680

I .OE-0S

E

j i.O E -06 ... ...........I..... ......

< IlpVII1~ 1

0U- i.0E-07.......... ..... ....... ...... ....... ...................

.E-80.1 1.0 10).0 FR0N 0 0

Fig. 2.3.4. Fourier acceleration ground mnotion L9 spectra for the recordings atthe stations MOT, and KTI( in Fig. 2.3.3, plotted together with similar estimatesfrom the precedjing noise. It is seen that KTI( (recording an AIL 3.1 earthquakeat a distance of 1:308 ki) has a positive signal-to-noise ratio oniy around 2-6 lIz.

316

2.4 A generalized beamforming approach to real time networkdetection and phase association

Introduction

The objective of this study is to investigate methods for multi-array detection,phase association and location, using as primary data sources the three regionalarrays (NORESS, ARCESS, FINESA) in Fennoscandia. Th emphasis is onapproaches which are suitable for incorporation in an expert system environnnt.As part of this effort, the problem o(continuously monitorinzg t le regional seismicnoise field is also addressed, with th,- purpose to obtain a qmauaItnitaitive assssuneuul,of the upper limit of inagnitudes of seismuiic events that would go undetected bysuch a network.

A more exhaustive treatment of the subject of this contribution is offered inour Scientific Report No. 1 under this contract (GL Report Number: GL-TR-89-0171 ), which will also appear in the December 1989 volume of the Bulletin of the.ciswologica' Soriety of America.

General approach

In the processing of seismic network data, individual phase detections corre-sponding to the same seismic event must be properly associated and grouped to-gether. This is today usually done starting with an initial trial epicenter and thenapplying various search strategies supplemented by combinational techniques.

This paper presents, and gives examples of application of, a method for as-sociating phase detections from a network of stations, which is analogous to theconventional delay-and-sum beamforming commonly applied in array detectionprocessing. A number of beams are steered to a predefined grid of aiming pointsin a geographical space. Each beam has an associated set of time delays, whereeach delay corresponds to the predicted travel time for a given phase at a givenstation.

We assume that the data of each network station is initially subjected to adetection processing procedure, whereby a list of phase detections and attributesis generated. The beamforming process, for a given beam, at time '', can bedescribed as looking for a pattern of detections/non-detections that matches thepredicted pattern for a hypothetical event with origin time T and location withinthe beam region. The actual beam value is derived from probabilistic consid-erations, and in essence describes how well the observed pattern matches theprediction. By moving along the time axis, we thus obtain a beam trace that canbe subjected to standard threshold algorithms for detection. The process can besupplemented by various individual "quality of fit" measures calculated at eachtime point.

37

After processing individuai beams as described above, a grouping/reductionprocess is applied to the set of all beam detections to eliminate side lobe detec-tions. This then results in an event list, comprising origin times, locations anda list of associated phases. Further refinement can then be achieved by standardtechniques for accurate hypocenter determination, magnitude computation, finalconsistency check, etc.

The generalized beamforming approach also provides a convenient tool tomonitor continuously the seismic noise field associated with a given beam. Anapplication of particular interest in a monitoring situation would be to calculate.at each step in time, upper confidence limits for the magnitude of non-detectedevents for each beam. This would be useful to obtain a realistic assessment ofactual retwork detection capabilities, at any given point in time. The paperpresents an example of practical application of this approach.

Regional phase association

The method has been applied to a data base comprising 24 hours of record-ings from the regional arrays NORESS, ARCESS and FINESA, with a beamdeployment covering Fennoscandia and adjacent areas.

A RONAPP-type detector was first applied to each array individually, usingthe broad-band F-K method to obtain phase velocity and azimuth for each de-tected phase. The resulting detectin lists then provided the input to the networkprocessor.

The beam grid used for network processing is shown in Figure 2.4.1, andcomprises altogether 121 aiming points, approximately equally spaced. Typicaldistance between aiming points is 150 km.

In the network beamforming process, a simple model of assigning 0/1 proba-bilities to individual phases at each station was used. We required that estimatedphase velocities, azimuth, dominant frequency and arrival times fall within pre-defined ranges for a phase detection to be accepted for a given beam. Thesetolerance ranges are specified in Table 2.4.1. Note in particular that only verygeneral criteria are applied, and we have made no attempt to optimize perfor-mance by regionalization.

With this simplified model, the network beamforming process in practice wasreduced to, for each beam and each time T, counting the number of phase matchesfor a hypothetical event located in the beam region and having origin time T.The detection threshold was set equal to 2. Thus, all occurrences of two or morematching phase detections were flagged as potential events. A typical beam traceis shown in Figure 2.4.2.

A grouping procedure was then applied to the overall beam detection list.This was done by successively linking together entries in the beam detection list

38

in such a way that a new entry would be linked if it had at least one individualphase detection in common with a previous entry in the group. The maximumallowable duration of a group was set to 10 minutes (in practico, the longestduration was 7 minutes for this data set). In order to resolve obvious multipleevents, groups were split up if two P-detections from the same array occurredwith more than 30 seconds arrival time difference.

The results are summarized in Table 2.4.2. It is important to note that thetotal of 91 groups comprise all possible events that could be associated, giventhe station detection lists. Also, a scrutiny of tihe data shows that only 3 ofthese groups contain multiple events, all of these being small presumnid nining

ex plosoiis 5sii ( y one array only.

Some of the entries in Table 2-1.2. e.g., those generated from two secondaryphases, are probably questionable seismic events, and even if real, may be im-possible to locate accurately. An upper magnitude limit could be estimated forsuch events, in order to determine whether further detailed analysis is desirable.However, the large majority of the entries appear to correspond to real seismicevents, and tie groul)ing proce(dure facilitates the subsequenmt detailed analysis ofthe associated phases.

The network beamforming procedure gives an initial estimnateof event locationby selecting the "best beam" in each group. This is defined as the beam withthe greatest number of associated phase detections, and if equality, the smallestaverage time residual of the detected phases. Since the initial beam grid is verycoarse, we applied a beampacking algorithm for each detection group, using agrid spacing of 20 km in order to improve the location estimate.

The results of this automatic procedure are shown in Table 2.4.3 for thoseevents for which independent location estimates were available. We note that theestimates are very consistent (median difference 40 km), and thus the beam resultscan be used as a reasonable first estimate of event location. For more accurateresults, available techniques for accurate hypocenter location (e.g., TTAZLOC)should be used.

Continuous monitoring of uppcr vent magnitude limits

As a second application of the generalized beamforming procedure, we nowaddress the problem of monitoring the noise levels on each beam, and use thisinformation to assess the size of events that might go undetected.

In formulating the approach, we consider a given geographical location, anda given "origin time" of a hypothetical event. Assume that N seismic phases areconsidered (there might be several stations and several phases per station).

For each phase, we have an estimate Si of the signal (or noise) level at thepredicted arrival time. For P-phases, Si might be the maximum STA value (1 sec-

39

ond integration window) within ± 5 seconds of the predicted time. For Lg, S,might be the average STA value over a 10-20 seconds window.

We assume that the network has been calibrated (or alternatively that stan-dard attenuation values are available), so that magnitude correction factors (b,)are available for all phases. Thus, if a detectable signal is present:

mi = log(Si) + bi (i= 1,2, ... N)

Here, ini are estimates of the event magnitude m. Statistically, we can considereach ut1 as sampled from a normal distribution (m,r). (A standard value of r0.2 seems reasonable for a small epicentral area.)

Let us now assume a "noise situation", i.e., that there are no phase detectionscorresponding to events at the given location for the given origin time.

We then have a set of "noise" observations a,, where

ai= log(S - i) + bi (i = 1,2,...N)

If a hypothetical event of magnitude 7n were present, it would have phasemagnitudes mi normally distributed around m. We know that for each phase,

1rti C: (i = 1,2,... N)

Let us look at the function

f(m) = Prob(all 7i < a,/event magnitude m)

For each phase

fi(m) = Prob(ni ai/m) =1 - ( n - al

where 4 is the standard (0,I) normal distribution.

Thus, assuming independence,

N

f(-n") 171 fi(,n)

'1he 90 per cent upper limit is then defined as the solution of the equation f(7)S1- 0.90

It is important to interpret the 90 per cent limit defined above in the properway. Thus, it should not be considered as a 90 per cent network detection thresh-old since we have made no allowance for a signal-to-noise ratio which would be

-10

required in order to detect an event, given the noise levels. Rather, the computedlevel is tied to the actually observed noise values, and to the fact that any hypo-thetical signal must lie below these values. Our 90 per cent limit represents thelargest magnitude of a possible hidden event, in the sense that above this limit,there is at least a 90 per cent probability that one or more of the observed noisevalues would be exceeded by the signals of such an event.

As an application of the method, we selected an area as shown in Figure 2.1.3situated at 'imilar distance from the three arrays. For each of the three arrays,one Pn beam and one Lg beam were steored to this location. The beam traceswere filtered using, the frequency bands 3-5 liz ( Pn) an(1 2-1I llz (l,g). Magniti,,decalibration vallis (Ibi) were obtained by processing l)reviollsly r'corded evnt Soknown magnitude (ML) and at similar distance ranges, and then determining bivalues independently for Pn and Lg.

Based on these input traces from the three arrays, a network beam was thenformed, using time delays for each phase that corresponded to the given location.Arrival time tolerances were set to ± 5 seconds for P-phases and ± 10 secondsfor secondary )hases. This is roughly consistent with a beam radius of 50 km asshown on the figure.

We chose to analyze a 3 1/2 hour interval during which four regional seismicevents of ML > 2.0 were reported in the Helsinki bulletin. No events were re-ported near the beam area in this period. Figure 2.4.4 shows 90 per cent uppermagnitude limits as previously defined, plotted as a function of time. In this fig-ure, only the Pn phase has been used, and the three arrays are shown individuallyand in combination (bottom trace).

It is clear from Figure 2.4.4 that when considering individual arrays only,there are several possible time intervals when relatively large events (Al 2.03.0) might go undetected because of signals from interfering events. However,when the Pn phases are combined, these instances occur much more seldom.

Figure 2.4.5 shows a similar plot, but this time including both the Pn and theLg phase for each array. Even on an individual array basis, this causes substantialreduction in the upper magnitude limits. For the combined plot (bottom traceof Figure 2.4.5), which takes into account all 6 Pn and Lg phases from the threearrays, we see that the upper limit is well below ML = 2.0 for the entire timeinterval. Thus, we may conclude that, at the specified level of confidence, noevent of ML = 2.0 or higher occurred in the beam region during the time periodconsidered.

Conclusions

With regard to phase association, the generalized beamforming techniqueprovides an effective method to group all combinations of individual phase de-

41

tections that could possibly correspond to the same seismic event. At the same

time, preliminary estimates of epicenter and origin time are obtained.

The primary importance of this would be to obtain a starting point for subsequent detailed interactive anaiysis aimed at precise determination of sourceparameters. In particular, expert system approaches (either script-based or rule-based) could be invoked at this stage. The advantage of applying the generalizedbeamforming as the first step is to reduce the amount of combinational process-

ing that would be necessary otherwise. It is here noteworthy that the processingload when applying generalized beamforming increases in a linear fashion whenthe number of individual phase detections increase, whereas combinational pos-sibilities tend to increase exponentially. While we have in this paper used only athree-array network, the extension to larger networks is clearly straightforward.

The application of the method to provide continuous monitoring of uppermagnitude limits at specified beam locations provides a useful supplement tostandard statistical network capability studies. In particular. this applicationwould give a way to assess the possible magnitude of non-detected events duringthe coda of large earthquakes. In such situations, it would be appropriate to useglobal network data and include as many relevant phases as possible for eachnetwork station. For example, while an expected P phase at a given station maybe obscured by the earthquake coda, later phases such as PcP or PP may be lessinfluenced, and the noise level at their respective expected arrival times wouldtherefore provide important information as to the size of possible undetectedevents.

As a final comment, we note that the approach presented here to upper limitmagnitude calculation could be applied to extend the utililty of various discrim-inants, such as A, : _1b. For small explosions, surface waves frequently are tooweak to be observed at any station of the recording network. Obtaining reliableupper bound on M, in such cases would expand the range of usefulness of thisdiscriminant. In practice, an "upper bound" for single-station measurements hasoften been given as the "noise magnitude" at that station, i.e.. the M, value thatcorresponds to the actually observed noise level at the expected time of Rayleighwave arrival. The proposed procedure will include this as a special case of a moregeneral network formulation.

In future studies, we plan to investigate the applica'iion of more sophisticatedprobabilistic models in the generation of beam traces and the continuous ex-traction of features associated with the individual beams. Application to largernetworks, including teleseismic monitoring using global network data, will alsobe considered.

Frode RingdalTormod Kvaerna

.12

Phase TypePn Pg Sn Lg Rg

Distance interval') 160-3000 0-600 160-3000 0-2000 0-400(kin) for which a pliaseis acceptedMaximum allowable 15 20 30 35 40deviation from p~redicte(darrival time (s)Maxi mun, allowabhle 20 20 20 20 2)azinrut h deviation

(degrees)Acceptance limits for 5.8-14 5.8-10 3.2-5.8 3.0-5.0 2.5-3.7apparent phasevelocity (km/s)Acceptance limits2) 0.5-20 0.5-20 0.5-20 0.5-20 0.5-20dominant frequency(liz)

1) For NORESS, the Rg phase is not included in the phase table

2) For FINESA, a lower frequency limit of 0.9 Hz is used for all phases.

Table 2.4.1. Acceptance limits for parameters used in the network beamformingprocess.

43

Number of phases forbest beam in each group

No. of phase groups: 2 3 4 5 6 7 8NORESSonly 18 13 4 1 0 0 0 0ARCESS only 34 19 10 4 1 0 0 0FINESA only 14 13 1 0 0 0 0 0Two arrays 17 9 4 3 0 1 0 0Three arrays 8 0 0 2 0 1 3 2Totals 91 5'1 19 10 1 2 3 2

Table 2.4.2 Phase groups associated by the network beaniforning proceduresfor a 24-hour interval.

,1.1

Event Date Time Network Mag. No. of No. of Beamforning 'Err,"No. Lat. Lon. AL phases arrays Lat. Lon. (kin)

1 88/03/17 08.40.25.0 57.73 11.03 2.5 7 3 57.9 10,1 3112 08.46.18.7 58.07 11.36 2.6 6 2 57.9 10.83 09.07.10.3 58.08 11.43 2.7 8 3 57.8 10.8 174 10.21.23.0 69.60 29.90 2.9 8 3 69.6 30.5 2:;5 10.27.20.0 59.20 27.60 2.3 4 2 59.5 27.5 316 10.46.21.0 59.20 27.60 <2.0 2 1 59.8 28.7 9(17 11.18.48.0 59 30 27.20 2.3 5 3 58.9 26.7 58 11.54.41.0 65.80 24.70 <2.0 5 1 66.6 24.,1 , }9 11.57.57.!) 60.57 8.36 1.8 2 1 60.6 8.1 H,

1( 12.02.36.0 59.10 28.50 2.1 3 2 59. 5 2S.2 '21111 12.42.22.!9 59.78 10.76 23 3 1 59.5 10.0 .-'212 14.13.14.0 58.33 6.28 2.4 4 1 58.0 6.1 3",13 14.21.08.0 60.90 29.40 2.3 3 2 61.3 29.1 ,1714 14.33.48.3 59.06 5.88 2.2 2 1 58.9 3.3 1.1115 18.58.08.1 59.68 5.57 3.2 7 3 C0.0 5.7

Table 2.4.3. Location estimates obtained automatically from the ea)mpackiiig

procedure compared with independent network locations from the lelsinki and

Bergen bulletins. Note the good consistency, especially for events with more than

one detecting array.

45

INITIAL BEAM GRID

-70.0

0

00

w

55.0

10.0 20.0 350.0

LONGITUDE (DEG E)

Fig. 2.4.1. Beam grid used in the generalized beamnforining procedure for thepurpose of associating regional phases from NORESS, ARCESS and FINESA.Thle location of the three arrays is shown on the map.

46

BEAM LOCATION: 57.ON 1O.OE

WJ 8-

U,,

ILL0 4

to 2 ------- -

08.00 09.00 10.00 11.00

ORIGIN TIME

Fig. 2.4.2. Example of typical output trace for one network beam (steered to

57vN, 10°E). In the 3 1/2 hour interval shown, there were 4 confirmed seismic

events located in the beam region. These were all correctly detected (arrows),

and no false phase associations occurred during the interval.

47

CONTINUOUS THRESHOLD MONITORING

(Dw0

w0

10.O2.0 30.0

LONGiT-UDE IDEG E)

Fig. 2.4.3. Location of the ',earn area used in the example of continuous moni-

toring of upper magnitude limits on non-detected events. The area covers a circlo

of ap~proximately 50 kin radius, and is situated at similar distance',s om the three

arrays.

CONTINUOUS THRESHOLD MONITORING - PN PHASE

90% PROBABILITY

40 NORESS

3. L) --- - - - - --0-- - -- - - - - - - -

20

0.0 1_____ 1 _ _ . .

080.C 09.00 10.00 11.00 HOUR

C0 ARCESS

2.0[

1.0 L _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

08.00 0900 10.00 11.00HOUR

40 FINESA

O83. O __ __ _.__ __ __ __I0.__ __ __0 _ __ __ __ _ tO0 __

2.0------------------------

0800 09 003 10.00 1100

4.0 THREE ARRAYS

1.0

08.00 0900 10.00 11.00HOUR

Fig. 2.4.4. Results from the continuous threshold monitoring of the area shownin Fig. 2.4.3 for a 3 1/2 hour period, using Pn phases only. The top three tracesshow, for each array, the largest magnitude of a possible non-detected event(confidence 90 per cent) as a function of time. The bottom trace shows the resultof combining the observations from all three arrays (Pn phase only) as described

in the text.

49

CONTINUOUS THRESHOLD MONITORING -PN AND LG PHASES

90% PROBABILITY4.0 NORESS

S 3.0 -- - - -_-- - -- -- - - - -- - -- -- _ _- -

2.0 - - - - - -

.0 -- -_-----------

0.0 1 -- - - -- I I08.00 A300 10-0 11.0

THRE ARRYS

0.00

2.5 Region-specific knowledge derived fr-om analysis of NORESSdatzi

The iiitelligeiit Nliiitoriiig System (I NIS) that is ((lii (lll I"' hoiiig Instalild ait

NOISA , provides funlctionis thaltallow specific regioli-(lcpelil-ilt kilowledgev to

bce taken illt()account ill the atitojiatic piocessin"gOf iiiilt i-dlraY dlata~. To lo~vidlt'

sulch kniowledge mi' eilliaili(lllet (if the perforiitance or I NiS 1,IIl(d wIIvillal

knoi wIlgo IEV tht lii iiefi t of I NIS.

.'tall'.,ics u it it(/1 nal phwase frib ull S froiii NAC) S IL. (t oll /I.Sts

T'he NO() RESS (let ectionl lists l'Or the perioli 19z 5 to I 9S,1 cuiitai mnore t han200,000( ewl is. comlliig teleseisiiiic, regionial anid local phlase arrivals. HIe(

iietwork bulleti us for this four-v ear period p~ublishied bY the Un iversi t of'I ergeml,

Nor%%iv. anld thle University of Hlisinki, Finlaind, conitaini altogetlftr 1(i.Oi( evnts

hat are locall 01 legiollal to ile NOkESS arra.Y. For each (of" thlese lti.UO() evets.

wxe caltliilathi t Ii expectedi arruiVad timesC- at N OVH USS Ior th10 regolial pidits

fit. Pg. -S and Lg andi tllen searched tile detection lists to seewhthecr tile

teortiLical arrival timies matched tihose of actual phlase ariva l, ietei(e onl tlie

NO lS' ;Ia 'av. C ertainl criter-ia were estab~lislhed Ill t his- regard. anld. e.g.. lbr1

IleIt'iilliaso t he ,ewWert that thev arrival timeis slioiiih n1ot d'v iate by vImore t hami

6 secolitI f'rom thle p reiIcted ones, the phi ase velocity ( routi nelY deterin ied by

fre Ielcy-~vve 1111he (-k) analysis for each detected signial ) shomld be wit iillthe itit ervih 6- 15 kiii/s. and tile arrival atzimauitlis (also (le-tlriiiimiil by fL k malysis)

should be wvithini .10" of the lpmedictedl values.

Thiis process of illrgiig of the iictwork bulletins wit h thle NO)ES I? d5(eteit lil

lists prodwoCs a wvealth of informiatio oil attribuites like phlase 'velocitins. aival

meTI resolh;II ialllad arriVal azilulitlI resitilis relative to tilie iitwork locail 11.

Ihief !c.su111 mla also be used as a basis for comipili ng data bases of' iltemest hg

events" froill spe(Tifc soulrce regionis.

As" a, resIllt of' t lis mtergi ng process, approxi matelyv 9 .Ott arrivals d etectel o~ii

the NORESS array could be classified as either Pui, Pg, Sn or Lg origitiating from

these 16,000) events. Figure 2.5.1 (top) shows the number of I'u pliases detectedat NORESS froml these events amnd ]low the correspominig evenits are tlistribuitedonl a gritd of' lox 2" blocks (north-son thi antI east- west, respecti vev ). Thie regioina roundi~ Bergeni Ol thle westernl coast of Norway, the Estoia -Leiiligrail regioli ofthle wvestern UiSSRt. thie iillaiid- P551? border region at1 ;inililld 30'T. GA"N . -Ili

aged III hurt hevll Swedenl. and the Kolai I'eniimsula of' the USSR~ ltaind out. tli'aiYill this Iliap. '['lie luatteril observed partly reflects the Stl loil (1ist ihilt ioil of* t hit

yejporti ng agenicies, )u t mnore b)asically gives anl overview of the wl) m iling areas,inl the Ntrdtc couintries andl( the northwestern tTSS I. Figure 2.7. 1 ( botton ) almoshows the average phase velocity for Pan phiases from events withiun eAch bloc1k. Wosee that these velocities fall within the expected range atrounad S kmi/s for lim itregions of this miap. A ii exception is the Estoitia- Leniingrad area of the west (niUSSR, where I'ti phiase velocities averaging ituore than 10 kin/s are observed.Arrival azimid.N t resiu als for the IPl phases are shown inl Figure 2.5.2. It is seenlhat the high Ili Pnphase velocities observed for the Estonia- Len ingrad region are

accotmipaied liv a rather comuplex patterni of azitiiutli residluals.

Th I zjinit Ii esidutliI for the L~g phlase ate shiowi iii 1igure 2.5.3.11 Is k huh(-

\v(Jlliv1%. t Ilie Fs.-loliau Leinigraid regioli exhibits ratlher iunderate L," aziul iI

It shouild he tinted that phatse velocities all(l arrival aiiilnt hs- for t-is studyvwere derived tisilug, the cotnvenionial or tiarrow-batid f-k analysis miethiod. Sill",1989, the broad-hatnl f-k met hod is ulsed iii the analysis of aill allaY (dat a atNO(RHSAI?. an tulwill also be used b)y the IMS. Several stutdies have test li fed thathe broad- band app~roachl is mote stable than thle narrow-blalid method. so a tnewnvestigation along tile lilies of this study should be untdertaketi. based oti t h,iletctioil iIAts for 1 989.

A1 (ctoilcd oluudy cof 10Y cvcnis in 1/ic western Vorwnyo~ a(a~

Aijiother inivestigaltion aiined at derivintg regioti-speci lic knowledge from antal-

Ysis of NO I? FSS dlata was a detailed stiidY of 103 events itt the wvest era Nor wayarea. Tlhle locations of' thlese events are sliowi inl Figure 2.5.4. T[le large majol--ty of the evenits are inl the distance interval 250-700 kin aind aziu ith interval

21 0"-330" . icla tiv ye t NO RESS . All phiases t hat could be picked for each oJf thle10.1 evetts were st ibjecteol to f-k anialysis. using bo0th the harmow- halil an bit o lIdl-biandi estitiuat iou titet hod. The arrival azimuths derived fromtiliese antalvyses wer,'

cotmpared withI airival azimhitis calculated ftrom the nletwork localtions ptublishuedlby the Lii vcr-itv of Bergen. Norway. and the Brnitishi Geological Suirvey

A rrival azitliitt I residulals ((]definled ats azimuth estitiatei front f-k atnal, .si-ttuis azi lil I accordling to thIe nletwork btillet iii) for thlese evetits for. the lPti. -

lPg. Sti aild Ig phase.'s are plot ted as function of network azitlittlis itt ligate 2.5.5fot I lie lroad-lbal ( st iliatiot mtethiod, atid] iii Figure 2.5.6i fot thIe itarrow-hlaid('stitiatioti ituil lad. It is, clearl , sent that the broad-band method provides themtore stable end ititsatil tlie stanldard deviation of the azitutt residuals forlie lur ad- hantd tiet hol were found( to be of the order of 5-7 olegrees. as, opposedl

to 7-1l degrees for thle tia-row-batid miethiod. The smallest sI i tdard deviat imuiswe-folitld F(au' t11e Arut phlases Pg and~ L~g.

-.\1 iiituI't tu observation is that for. tile Pl and Sil hss the' a/iiuthsu'stitlittdu bY thin broald-banld ittetliod ) deviate svttuftc lhy mltt the trt

values. in tile sense that the residuals are negative for azimut ls in the I ;inge 2 10"-

2700 and positive for azimuths between 2700 andt 320". These resIlls in(ic at I fiatlaterally v arying structures near the crust-mantle boundary rather I haii .t ru(t urwithin th (rust are important in the sense that they influence the azinitulhisestimated at NORESS. More work is needed in order to galin further insight andunderstanding of these effects.

Conclusion.,

''lie merging of the complete NORESS detection lists for the piriod 1985-1988with lle rregional network bulletins results in a largo aumount of s ;iti.,t ical infor-unat ioll ol (liaracteristics of regionall p)hases in 1elulloscaidi ,aif ;adj.ient areas.

Siulilai infornlalon will result from nerging with the AlW( "SS mid II N'.S:\ de-

fectioni fists. The knowledge galied from this exercise will bc ,,f (el,",a.'ue auindilinporltlce I, I' IMS, and it is aiticipated that this knowlelge will be represeited in I.NI S's knowledge base.

A special sI udy of 103 events recorded at NORESS from the westeri Norwaya rea is, aiiot.her example of what kind of studies that must be undertaken in orderto obtain region-specific knowledge needed by the IMS. Among ot her results, thisspeci-il stuidy established that the arrival azimuth residuals observed at NORESSfor the Pn and Sn lhases vary ratlher systematically with source region in thewestern Norway area.

S. MykkeltveitS. Kibsgaard. 1Univ. of Oslo

T. Kvaerna

53

Pn

Numbet ot defections

M3:NOBSI,10 lO - VOBS 100 =100,-NOBS

(0

w0

10.0 20.0 30.0

LONGITUDE (DEG E)

Pn

Phase velocities

O VEL'8 M8.sYEL<9 =9 1VEL<10l 10-l/EL<71

60

ww 0

S0

'0.0 20.0 30.0LONGITUDE (DEG El

Fig. 2.5.1. 10o): Nultllir of I'l !i~~NB~)uv'Id 11 NOIR ISS fromt the(Vaiousl~ Siource areas. lthI"ui: ( nit j lIiilg average hi (('locil I(,,, ( \' L). 'I'hlelocationl of' NORESS ~s indcae. bYIj 1( ;ij SI)'(il sy'Ilt11.

pn

Negativeazimuth

residualsS

60

Azimuthswithin

2.5 degreesof 'true' values

W

00.0 20.0 10.02.5<ARIS.O VR7

Positiveazimuth

residuals

LONGiTuoe (DEG El

Fig. 2.5.2. Azimuth residuals (Alt = estimated azimutb in is 'I ime' azi mmmii timI', degrees) for Pui phiases detected at NORESS.

55r

Lg

=-OO ARCE.7 .7S51ARSO S-0<Aft-25

Negativeazimuth

residuals

Azimuthswithin

2.5 degreesof 'true' va~ues

10.0 20.0 50.0

7S<AR-, . 5,V Aftl., M .S<ARrO 0

Positiveazimuth

residuals W60.0

10.020'0 50.0

LONG ITJC)E (DEG E I

Fig. 2.5.3. Sautej a;v, . .52 hult for telgpitaIsV.

62.0

S7.0

S2.0

0.0 "1.0

Fig. 2.5.4. The Iiiap shows the location of the 103 ewonts ini the western Norwa*yregion used in this study. The location of NORESS is indicated by a specialsymbol.

57

BROAD IST.(Pft) -AZIMUTH MITIVORK SOLUTION.

10.0 ~ i--- _

14.0-~-- _

0.0

30.0 120.0 150.0 180.0 21 0.0 240.0 2 70.0 300.0 330.0

BROAD ZST.(Pj) -AZIMUTH NETFORK SOLUTION.

Z0.9~

0t.0- 8

90.0 920.0 1SO.0 180.0 £10.0 240.0 270.0 300.0 330.0

BROAD fST (Sa) -AZIMUTH NfITWORK SOLUTION.

20.0-

-10.0 --

0.0 10.0 150.0 110.0 210.0 240.0 270.0 300.0 330.0

Fig. ~ ~ BOA 2.5.5.g Th fi AZhw h azIiihr IMT NETWOR St i IIOiLUION.IIv*i il I s

20.0 the br a*al 3slla il meId _____-i lal l, iol fo lT p t

bottom ________ Im_________ the_________ P 11,________ ____________and__________________

5 0 ~ 0j__S

I I %a

_B_ EST.(P) -Ilf _ET_. SOL-UTION_.

-2.0--1 ________

*° I ______________ ____

90.0(200 f0.01800 Zf 0 0 240. 0 270.0 300.9 330.0

MBFE EST.(g) -AZIMUTH MEF. SOLUTION.

210.0- I - __ o__ --

_ _ I _ _ I ,_ _ _ _ _ _ _ _ _

-2 .9 • -i . . a . • i • • i ' . '

10.0 t20.O 150.0 a0. 0 zf0.0 240.0 170.0 300.0 330.0

KBfI EST .(P) AZIMUTH NETY. SOLUTION.

200 I l

4.0-

a' , • • I • • 0 * . Ii '

0.0 fZ0.0 50.0 80. 0 210.0 Z 40. 0 270.0 300,0 $30.0

NBFE' LST'.(L) - AIZIMlUTH I TV'. SO LTION.

o~o. i _ i ___ _ __.. o o

- __- I __ _ ___i

i i TTi ti . . .

10.0 120.0 950 .0 980.0 2 10.0 240.0 270O.0 300,0 310.0

Fig. 2.5.6. Same as Fig. 2.,5.5, but for the narrow- band estin ;tioI metho(t.

59

CONTRACTORS (United States)

Prof. Thomas Ahrens Prof. John FergusonSeismological Lab, 252-21 Center for Lithospheric StudiesDivision of Geological & Planetary Sciences The University of Texas at DallasCalifornia Institute f Technology P.O. Box 830688Pasadena, CA 91125 Richardson, TX 75083-0688

Prof. Charles B. Archambeau Prof. Stanley FlatteCIRES Applied Sciences BuildingUniversity of Colorado University of CaliforniaBoulder, CO 80309 Santa Cruz, CA 95064

Prof. Muawia Barazangi Dr. Alexander FlorenceInstitute for the Study of the Continent SRI InternationalCornell University 333 Ravenswood AvenueIthaca, NY 14853 Menlo Park, CA 94025-3493

Dr. Douglas R. Baumgardt Prof. Henry L. GrayENSCO, Inc Vice Provost and Dean5400 Port Royal Road Department of Statistical SciencesSpringfield, VA 22151-2388 Southern Methodist University

Dallas, TX 75275

Prof. Jonathan Berger Dr. Indra GuptaIGPP, A-025 Teledyne GeotechScripps Institution of Oceanography 314 Montgomery StreetUniversity of California, San Diego Alexandria, VA 22314

La Jolla, CA 92093

Dr. Lawrence J. Burdick Prof. David G. ffarkriderWoodward-Clyde Consultants Seismological Laboratory566 El Dorado Street Division of Geological & Planetary SciencesPasadena, CA 91109-3245 California Institute of Technology

Pasadena, CA 91125

Dr. Karl Coyner Prof. Donald V. tleinbergerNew England Research, Inc. Seismological Laboratory76 Olcott Drive Division of Geological & Planetary ScicncesWhite River Junction, VT 05001 California Institute of Technology

Pasadena, CA 91125

Prof. Vernon F. Cormier Prof. Eugene HerrinDepartment of Geology & Geophysics Institute for the Study of Earth and ManU-45, Room 207 GeophysicalLaboratoryThe University of Connecticut Southern Methodist UniversityStorrs, CT 06268 Dallas, TX 75275

Prof. Steven Day Prof. Robert B. HerrmannDepartment of Geological Sciences Department of Earth & Atmospheric SciencesSan Diego State University St. Louis UniversitySan Diego, CA 92182 St. Louis, MO 63156

Dr. Zoltan A. Der Prof. Bryan IsacksI[NSCO, Inc. Cornell University54(X) Port Royal Road Department of Geological SciencesSpringfield, VA 22151-2388 SNEE Hall

Ithaca, NY 14850-1-

I)r. Rong-Song Jih Dr. Gary McCartorIclcdT'ic Geotech Mission Research Corporation31I Morugomery Street 735 State StreetAlexandria, VA 22314 P.O. Drawer 719

Santa Barbara, CA 93102 (2 copies)

Prof. Lane R. Johnson Prof. Thomas V. McEvillySeismographic Station Seismographic StationUniversity of California University of CaliforniaBerkeley, CA 94720 Berkeley, CA 94720

Prof. Alan Kafka Dr. Keith L. McLaughlinDepartment of Geology & Geophysics S-CUBEDBoston College A Division of Maxwell LaboratoryChestnut Hill, MA 02167 P.O. Box 1620

La Jolla, CA 92038-1620

Prof. Fred K. Lamb Prof. William MenkeUniversity of Illinois at Urbana-Champaign Lamont-Doherty Geological ObservatoryDepartment of Physics of Columbia University1110 West Green Street Palisades, NY 10964Urbana, IL 61801

Prof. Charles A. Langston Stephen MillerGeosciences Department SRI International403 Deike Building 333 Ravenswood AvenueThe Pennsylvania State University Box AF 116University Park, PA 16802 Menlo Park, CA 94025-3493

Prof. Thorne Lay Prof. Bernard MinsterDepartment of Geological Sciences IGPP, A-0251006 C.C. Little Building Scripps Institute of OceanographyUniversity of Michigan University of California, San DiegoAnn Arbor, MI 48109-1063 La Jolla, CA 92093

Prof. Arthur Lerner-Lam Prof. Brian J. MitchellLamont-Doherty Geological Observatory Department of Earth & Atmospheric Sciencesof Columbia University St. Louis UniversityPalisades, NY 10964 St. Louis, MO 63156

Dr. Christopher Lynnes Mr. Jack MurphyTeledyne Geotech S-CUBED, A Division of Maxwell Laboratory314 Montgomery Street 11800 Sunrise Valley DriveAlexandria, VA 22314 Suite 1212

Reston, VA 22091 (2 copies)

Prof. Peter Malin Dr. Bao NguyenUniversity of California at Santa Barbara GL.LWHInstitute for Crustal Studies Hanscom AFB, MA 01731-5000Santa Barbara, CA 93106

Dr. Randolph Martin, II Prof. John A. OrcuttNew England Research, Inc. IGPP, A-02576 Olcott Drive Scripps Institute of OceanographyWhite River Junction, VT 05001 University of California, San Diego

-2- La Jolla, CA 92093

Prof. Keith Priestley Prof. Clifford ThurberUniversity of Nevada University of Wisconsin-MadisonMackay School of Mines Department of Geology & GeophysicsReno, NV 89557 1215 West Dayton Street

Madison, WS 53706

Prof. Paul G. Richards Prof. M. Nafi ToksozLamont-Doherty Geological Observatory Earth Resources Labof Columbia University Massachusetts Institute of TechnologyPalisades, NY 10964 42 Carleton Street

Cambridge, MA 02142

Dr. Wilmer Rivers Prof. John E. VidaleTeledyne Geotech University of California at Santa Cruz314 Montgomery Street Seismological LaboratoryAlexandria, VA 22314 Santa Cruz, CA 95064

Dr. Alan S. Ryall, Jr. Prof. Tcrry C. WallaceCenter for Seismic Studies Department of Geosciences1300 North 17th Street Building #77Suite 1450 University of ArizonaArlington, VA 22209-2308 Tucson, AZ 85721

Prof. Charles G. Sammis Dr. Raymond WillemanCenter for Earth Sciences GL/LWHUniversity of Southern California Hanscom AFB, MA 01731-5000University ParkLos Angeles, CA 90089-0741

Prof. Christopher H. Scholz Dr. Lorraine WolfLamont-Doherty Geological Observatory GlILWHof Columbia University Hanscom AFB, MA 01731-5000Palisades, NY 10964

Prof. David G. Simpson Prof. Francis T. WuLamont-Doherty Geological Observatory Department of Geological Sciencesof Columbia University State University of New YorkPalisades, NY 10964 at Binghamton

Vestal, NY 13901

Dr. Jeffrey StevensS-CUBEDA Division of Maxwell LaboratoryP.O. Box 1620La Jolla, CA 92038-1620

Prof. Brian StumpInstitute for the Study of Earth & ManGeophysical LaboratorySouthern Methodist UniversityDallas, TX 75275

Prof. Jeremiah SullivanUniversity of Illinois at Urbana-ChampaignDepartment of Physics1110 West Green StreetUrbana, IL 61801

"3-

OTHERS (United States)

Dr. lvonem Abdel-Gawad Dr. Stephen BrattRock,,ell International Science Center Science Applications Int'l Corp.1M19 Camino Dos Rios 10210 Campus Point DriveThousand Oaks, CA 91360 San Diego, CA 92121

Prof. Keiiti Aki Michael BrowneCenter for Earth Sciences Teledyne GeotechUniversity of Southern California 3401 Shiloh RoadUniversity Park Garland, TX 75041Los Angeles, CA 90089-0741

Prof. Shelton S. Alexander Mr. Roy BurgerGeosciences Department 1221 Serry Road403 Deike Building Schenectady, NY 12309Tlhe Pennsylvania State UniversityUniversity Park, PA 16802

l)r. Ralph Archuleta Dr. Robert BurridgeDepartment of Geological Sciences Schlumberger-Doll Research CenterUniversity of California at Santa Barbara Old Quarry RoadSanta Barbara, CA 93102 Ridgefield, CT 06877

Dr. Thomas C. Bache, Jr. Dr. Jerry CarterScience Applications Int'l Corp. Rondout Associates10210 Campus Point Drive P.O. Box 224San Diego, CA 92121 (2 copies) Stone Ridge, NY 12484

J. Barker Dr. W. Winston ChanDepartment of Geological Sciences Teledyne GeotechState University of New York 314 Montgomery Streetat Binghaniton Alexandria, VA 22314-1581Vestal, NY 13901

Dr. T.J. Bennett Dr. Theodore CherryS-CUBED Science Horizons, Inc.A Division of Maxwell Laboratory 710 Encinitas Blvd., Suite 20011800 Sunrise Valley Drive, Suite 1212 Encinitas, CA 92024 (2 copies)Reston, VA 22091

Mr. William J. Best Prof. Jon F. Claerbout907 Westwood Drive Department of GeophysicsVienra, VA 22180 Stanford University

Stanford, CA 94305

l)r. N. lB., vas Prof. Robert W. ClaytonGeophysical institnte Seismological LaboratoryUniversity of Alaska Division of Geological & Planetary SciencesFairbanks, AK 9970. California Institute of Technology

Pasadena, CA 7'1 125

Dr. G.A. Bollinger Prof. F. A. DahlenDepartment of Geological Sciences Geological and Geophysical SciencesVirginia Polytechnical Institute Princeton University21044 Derring Hall Princeton, NJ 08544-0636Blacksburg, VA 24061 -4-

Prof. Anton W. Dainty Prof. Stephen GrandEarth Resources Lab University of Texas at AustinMassachusetts Institute of Technology Department of Geological Sciences42 Carleton Street Austin, TX 78713-7909Cambridge, MA 02142

Prof. Adam Dziewonski Prof. Roy GreenfieldHoffman Laboratory Geosciences DepartmentHarvard University 403 Deike Building20 Oxford St The Pennsylvania State UniversityCambridge, MA 02138 University Park, PA 16802

Prof. John Ebel Dan N. HagedornDepartment of Geology & Geophysics BattelleBoston College Pacific Northwest LaboratoriesChestnut Hill, MA 02167 Battelle Boulevard

Richland, WA 99352

Eric Fielding Kevin HutchensonSNEE Hall Department of Earth SciencesINSTOC St. Louis UniversityCornell University 3507 LacledeIthaca, NY 14853 St. Louis, MO 63103

Prof. Donald Forsyth Prof. Thomas H. JordanDepartment of Geological Sciences Department of Earth, AtmosphericBrown University and Planetary SciencesProvidence, RI 02912 Massachusetts Institute of Technology

Cambridge, MA 02139

Prof. Art Frankel Robert C. KemeraitMail Stop 922 ENSCO, Inc.Geological Survey 445 Pineda Court790 National Center Melbourne, FL 32940Reston, VA 22092

Dr. Anthony Gangi William KikendallTexas A&M University Teledyne GeotechDepartment of Geophysics 3401 Shiloh RoadCollege Station, TX 77843 Garland, TX 75041

Dr. Freeman Gilbert Prof. Leon KnopoffInst. of Geophysics & Planetary Physics University of CaliforniaUniversity of California, San Diego Institute of Geophysics & Planetary PhysicsP.O. Box 109 Los Angeles, CA 90024La Jolla, CA 92037

Mr. Edward Giller Prof. L. Timothy LongPacific Sierra Research Corp. School of Geophysical Sciences1401 Wilson Boulevard Georgia Institute of TechnologyArlington, VA 22209 Atlanta, GA 30332

Dr. Jeffrey W. Given Prof. Art McGarrSierra Geophysics Mail Stop 97711255 Kirkland Way Geological SurveyKirkland, WA 98033 345 Middlefield Rd.

Menlo Park, CA 94025

Dr. George Mellman Dr. Richard SailorSierra Geophysics TASC Inc.11255 Kirkland Way 55 Walkers Brook DriveKirkland, WA 98033 Reading, MA 01867

Prof. John Nabelek Thomas J. Sereno, Jr.College of Oceanography Science Application Int'l Co:p.Oregon State University 10210 Campus Point DriveCorvallis, OR 97331 San Diego, CA 92121

Prot. Geza Nagy John SherwinUniversity of California, San Diego Teledyne GeotechDepartment of Ames, M.S. B-010 3401 Shiloh RoadLa Jolla, CA 92(Y)3 Garland, TX 75041

Prof. Amos Nur Prof. Robert SmithDepartment of Geophysics Department of GeophysicsStanford University University of UtahStanford, CA 94305 1400 East 2nd South

Salt Lake City, UT 84112

Prof. Jack Oliver Prof. S. W. SmithDepartment of Geology Geophysics ProgramCornell University University of WashingtonIthaca, NY 14850 Seattle, WA 98195

Prof. Robert Phinney Dr. Stewart SmithGeological & Geophysical Sciences IRIS Inc.Princeton University 1616 North Fort Myer DrivePrinceton, NJ 08544-0636 Suite 1440

Arlington, VA 22209

Dr. Paul Pomeroy Dr. George SuttonRondout Associates Rondout AssociatesP.O. Box 224 P.O. Box 224Stone Ridge, NY 12484 Stone Ridge, NY 12484

Dr. Jay Pulli Prof. L. SykesRADIX System, Inc. Lamont-Doherty Geological Observatory2 Taft Court, Suite 203 of Columbia UniversityRockville, MD 20850 Palisades, NY 10964

Dr. Norton Rimer Prof. Pradeep TalwaniS-CUBED Department of Geological SciencesA Division of Maxwell Laboratory University of South CarolinaP.O. Box 1620 Columbia, SC 29208La Jolla, CA 92038-1620

Prot. Larry J. Ruff Prof. Ta-liang TengDepartment of Geological Sciences Center for Earth Sciences1(X)6 C.C. Little Building University of Southern CaliforniaUniversity of Michigan University ParkAnn Arbor, MI 48109-1063 -6- Los Angeles, CA 90089-0741

Dr. R.B. TittmannRockwell International Science Center1049 Camino Dos RiosP.O. Box 1085Thousand Oaks, CA 91360

Dr. Gregory van der VinkIRIS, Inc.1616 North Fort Myer DriveSuite 1440Arlington, VA 22209

William R. WalterSeismological LaboratoryUniversity of NevadaReno, NV 89557

Dr. Gregory WojcikWeidlinger Associates44 10 El Camino RealSuite 110Los Altos, CA 94022

Prof. John H. WoodhouseHoffman LaboratoryHarvard University20 Oxford StreetCambridge, MA 02138

Dr. Gregory B. YoungENSCO, Inc.5400 Port Royal RoadSpringfield, VA 22151-2388

-7-

GOVERNMENT

Dr. Ralph Alewine II Paul JohnsonLDARP NMRO ESS-4, Mail Stop J97914(X) Wilson Boulevard Los Alamos National LaboratoryArlington, VA 01731-5000 Los Alamos, NM 87545

Mr. James C. Battis Janet JohnstonGL/LWH GL/LWHHanscom AFB, MA 22209-2308 Hanscom AFB, MA 01731-5000

Dr. Robert Blandford Dr. Katharine Kadinsky-CadeDARPA/NMRO GILWH14(X) Wilson Boulevard Hanscom AFB, MA 01731-5000Arlington, VA 87185

Eric Chael Ms. Ann KerrDivision 9241 IGPP, A-025Sandia Laboratory Scripps Institute of OceanographyAlbuquerque, NM 01731-5000 University of California, San Diego

La Jolla, CA 92093

Dr. John J. Cipar Dr. Max KoontzGLILWH US Dept of Energy/DP 51lanscom AFB, MA 01731-5000 Forrestal Building

1000 Independence AvenueWashington, DC 20585

Mr. Jeff Duncan Dr. W.H.K. LeeOffice of Congressman Markey Office of Earthquakes, Volcanoes,2133 Rayburn House Bldg. & EngineeringWashington, D.C. 20515 345 Middlefield Road

Menlo Park, CA 94025

Dr. Jack Evernden Dr. William LeithUSGS - Earthquake Studie U.S. Geological Survey345 Middlefield Road Mail Stop 928Menlo Park, CA 94025 Reston, VA 22092

Art Frankel Dr. Richard LewisUSGS Director, Earthquake Engineering & Geophysics922 National Center U.S. Army Corps of EngineersRestonr, VA 22092 Box 631

Vicksburg, MS 39180

Dr. T. I lanks James F. LewkowiczUSGS GL-.LWHNat'l Earthquake Research Center Hanscom AFB, MA 01731-5000345 Middlefield RoadMenlo Park, CA 94025

Dr. James Hannon Mr. Alfred LiebermanLawrence Livemiore Nat'l Laboratory ACDA/VI-OA'State Department BldgP.O. Box 808 Room 5726Livermore, CA 94550 320 - 21st Street, NW

Washington, DC 20451

-8-

Stephen Mangino Dr. Frank F. PilotteGL/LWH HQ AFTACIITHanscom AFB, MA 01731-5000 Patrick AFB, 1. 32925-6001

Dr. Robert Masse Katie PoleyBox 25046, Mail Stop 967 CIA-OSWR/NEDDenver Federal Center Washington, DC 20505Denver, CO 80225

Art McGarr Mr. Jack RachlinU.S. Geological Survey, MS-977 U.S. Geological Survey345 Middlefield Road Geology, Rm 3 C136Menlo Park, CA 94025 Mail Stop 928 National Center

Reston, VA 22092

Richard Morrow Dr. Robert ReinkeACDAIVI, Room 5741 WI_/NTESG320 21st Street N.W Kirtland AFB, NM 87117-6008Washington, DC 20451

Dr. Keith K. Nakanishi Dr. Byron RistvetLawrence Livermore National Laboratory HQ DNA, Nevada Operations OfficeP.O. Box 808, L-205 Atm: NVCGLivermore, CA 94550 P.O. Box 98539

Las Vegas, NV 89193

Dr. Carl Newton Dr. George RotheLos Alamos National Laboratory HQ AFTAC/TGRP.O. Box 1663 Patrick AFB, FL 32925-6001Mail Stop C335, Group ESS-3Los Alamos, NM S i45

Dr. Kenneth H. Olsen Dr. Michael ShoreLos Alamos Scientific Laboratory Defense Nuclear Agency/SPSSP.O. Box 1663 6801 Telegraph RoadMail Stop C335, Group ESS-3 Alexandria, VA 22310Los Alamos, NM 87545

Howard J. Patton Donald L. SpringerLawrence Livermore National Laboratory Lawrence Livermore National LaboratoryP.O. Box 808, L-205 P.O. Box 808, L-205Livermore, CA 94550 Livermore, CA 94550

Mr. Chris Paine Dr. Lawrence TurnbullOffice of Senator Kennedy, SR 315 OSWR/NED

Central Intelligence Agency, Room 5G48United States Senate Washington, DC 20505Was.hington, DC 20510

Colonel Jerry 1. Perrizo Dr. Thomas WeaverAFOSR/NP, Building 410 Los Alamos National LaboratoryBoiling AFB P.O. Box 1663, Mail Stop C335Washington, DC 20332-6448 Los Alamos, NM 87545

-9-

J.J. Zucca Defense Technical Information CenterLawrence Livermore National Laboratory Cameron StationBox 808 Alexandria, VA 22314 (5 copies)Livermore, CA 94550

GL/SULL Defense Intelligence AgencyResearch Library Directorate for Scientific &Hanscom AFB , MA 01731-5000 (2 copies) Technical Intelligence

Washington, DC 20301

Secretary of the Air Force (SAFRD) AFTAC/CA(STINFO)

Washington, DC 20330 Patrick AFB, FL 32925-6001

Office of the Secretary Defense TACTECDDR & E Battelle Memorial InstituteWashington, DC 20330 505 King Avenue

Columbus, OH 43201 (-inal Report On!y)

I IQ DNA Mr. Charles L. TaylorAttn: Technical Library GILWHWashington, DC 20305

Hanscom AFB, MA 01731-5000

DARPAIRMO/RETRIEVAL1400 Wilson BoulevardArlington, VA 22209

DARPA/RMO/Security Office14(X) Wilson BoulevardArlington, VA 22209

Geophysics LaboratoryAttn: XO[lanscom AFB, MA 01731-5000

Geophysics LaboratoryAttn: LWIlanscom AFB, MA 01731-5000

DARPA/PM1400 Wilson BoulevardArlington, VA 22209

-10-

CONTRACTORS (Foreign)

Dr. Ramon Cabre, S.J.Observatorio San CalixtoCasilla 5939La Paz, Bolivia

Prof. Hans-Peter HarjesInstitute for GeophysikRuhr University/BochumP.O, Box 1021484630 Bochum 1, FRG

Prof. Eystein HusebyeNTNFINORSARP.O. Box 51N-2007 Kjeller, NORWAY

Prof. Brian L.N. KennettResearch School of Earth SciencesInstitute of Advanced StudiesG.P.O. Box 4Canberra 2601, AUSTRALIA

Dr. Bernard MassinonSociete Radiomana27 rue Claude Bernard75005 Paris, FRANCE (2 Copies)

Dr. Pierre MechelerSociete Radiomana27 rue Claude Bernard75005 Paris, FRANCE

Dr. Svein MykkeltveitNTNF/NORSARP.O. Box 51N-2007 Kjeller, NORWAY

FOREIGN (Others)

Dr. Peter dasham Dr. Fekadu KebedeEarth Physics Branch Seismological SectionGeological Survey of Canada Box 12019I Observatory Crescent S-750 Uppsala, SWEDENOitrtwa, Ontario, CANADA KiA 0Y3

Dr. Eduard Berg Dr. Tormod KvaernaInstitute of Geophysics NTNF/NORSARUniversity of Hawaii P.O. Box 51Honolulu, HI 96822 N-2007 Kjeller, NORWAY

Dr. Michel Bouchon Dr. Peter MarshalI.R.I.G.M.-B.P. 68 Procurement Executive38402 St. Martin D'Heres Ministry of DefenseCedex, FRANCE Blacknest, Brimpton

Reading FG7-4RS, UNITED KINGDOM

Dr. I lilmar Bungum Prof. Ari B-n-MenahemNTNF/NORSAR Department of Applied MathematicsP.O. Box 51 Weizman Institute of ScienceN-2(X)7 Kjeller, NORWAY Rehovot, ISRAEL 951729

Dr. Michel Canpillo Dr. Robert NorthObservatoire de Grenoble Geophysics DivisionI.R.I.G.M.-B.P. 53 Geological Survey of Canada38041 Grenoble, FRANCE 1 Observatory Crescent

Ottawa, Ontario, CANADA KiA 0Y3

Dr. Kin Yip Chun Dr. Frode RingdalGeophysics Division NTNF/NORSARPhysics Department P.O. Box 51University of Toronto N-2007 Kjeller, NORWAYOntario, CANADA M5S 1A7

Dr. Alan Douglas Dr. Jorg SchlittenhardtMinistry of Defense Federal Institute for Geosciences & Nat'l Res.Blacknest, Brimpton Postfach 510153Reading RG7-4RS, UNITED KINGDOM D-3000 Hannover 51, FEDERAL REPUBLIC OF

GERMANY

)r. Roger I lansenNTNV/NORSAR Prof. Daniel WalkerP.0. Box 51 University of HawaiiN-2(X)7 Kjeller, NORWAY Institute of Geophysics

Honolulu, HI 96822

Dr. Manfred HengerFederal Institute for Geosciences & Nat'l Res.Postfach 510153D-3000 Hanover 51, FRG

Ms. [Eva JohannissonSenior Research OfficerNational Defense Research Inst.P.O. Box 27322S- 102 54 Stockholm, SWEDEN -12-