Geophys. J.

Jill, (1993) ill 507-516

Coda Q in the eastern Caribbean, West Indies

W. B. Ambeh and L. L. Lynch

Seismic Research Unit, The Uhiversil:y of ih~ West Indies, SI. Augustin«, Trinidad: & Tobag8, West Indies

Accepted 1992 August U. Received 1992 August 4; in original form 1992 February 18

SUMMARY

Coda Q~ Q"" has been estimated from seismograms of local earthquakes recorded 00 short-period seismographs on some of the eastern Caribbean 'islands of the Lesser Antilles arc, West Indies, using the S-S single-scattering model. The region wall subdivided into fOUI distinct subregions in order to investigate lateral variations in c, The Qc values exhibit astrong frequency dependence in the frequency band 1.5 to 16Hz, withrcalculated coda Q at 1 Hz, Q", ranging from a minimum: of 97 .in, the Dominica area toa maximum of 14~5 in the Leewards. The degree of frequency dependence is also largest in the .Dorniniea area en = 1.09) and smallest in the Leewards (n =0:82). Although there appears to besome lateral.and depth variation of both Q", 'and n in the region, this variation does mit seem to be very significant.

Key worm,,: Caribbean coda Q, earthquakes, seismograms.

1 INTRODUCTION

Attenuation of seismic waves, customarily described by the dimensionless parameter Q, caned the speciffc quality factor, expresses the wave-amplitude decay that occurs when a wave pro\Jagates -rfuough I.eal media ,and cannet be attributed to geometrical spreading, Seismic-wave attennation .is ''8 combination of two different loss mechanisms: anelastic absorption, the loss of elastic encr.gy to heat or other forms of energy, and scattering, the defl:eCtionand/or mode conversien of seismic energy due to randomly distributed inhomogeneities in the 'transmitting medium (Dainty 1981). However, accotding to Dainty (_l981) the effuds of intrinsic and seatrering attenuution cannot be directly separated because of similarity in their mathematical form,

Seismic-wave attenuation varies spatially (e.g, Sutton, Mitronovas & "Pomeroy H167; Cheng & Mitchell 1981; Singh & Herrmann 1983; lin & Aki 19&.8) and may also be frequency dependent (Mittl1ell 1980; Rovelli 1984; Kvamme & Havskov 19,89), With. the degree of frequency dependence varying from region to region (Mitchell 1981; Pulli & Aki 1981). Consequently, the degree of spatial variation. of both attenuation and its frequency dependence may be used to infer the nature Of the material conditions in the' earth.

Although seismic-wave artenuation is an important parameter, klIowledge of which is, fprcxample, essential for the prediction, of earthquake ground motion in seismichazard analyses, its nature is not well understood, Furthermore, it is difficult to measure, especially in the frequency band 1 to ::30 Hz; frequencies of interest in short-period seismology and struejural engineering, Hough et al. (1988);iollowing·fue' categories recognized by Cormier

(1982) in teleseismic studies" classified methocisWleil: to study high-frequency attenuation at near 'and regional distances into f9UT -bread classes. These include methods which cancel the seismic source, methods which make .assnmptiens about the source, coda methods and timedomain methods. Many recent determinations of attenuation usIng local earthquakes have involved the use of coila methods because ofthei:r simpliCity and ease of application.

Coda wilYes comprise the Tatter part of seismograms, the part after all the-direct waves STIch as E, S and surface Waves have arrived (Herraiz & Espinosa 1987). Theoretical and observational -studies on coda waves generally seem to suggest that the coda is produced by the interaction between primary wavelJ-and small-scale lithospheric heterogeneities. Scattered body waves can also bege:n'e:ratelL by topography (Spudieli & Bostwick 1987) while; 1m .irtegujar low-velocity surfaee' layer will COnvert ana scatter body waves to surface

waves (Le~ander &. Hill 1985), ~ ,

The most widely applied coda method is that formulated by Aki (1969) and exten{)e'd DY Aki & Chonet (197.)) and Sata (1971). It considers the S~wave coda as consisting of single, backscatrered S~waves from randomly distributed .heterogenefties in the lithosphere, The ienvelopes of the codas of narrow, bandpass-filtered seismograms can then be used to determine the average Q fQr tfie medium fOJ" the particular frequency bands. TlW singJe-scattering model bas been used to d.ete~ealt:enUil.;t1on in: marty Tegions of the world (e.g, Rautian & Khalturin 1978; Reecker et ol, 1982; Pulli 1984; Havskov et. al. 1989; Ambeh & Fairhead 1989) and to investigate temporal changes in coda attenuation as a possible tool for earthquake prediction (e.g. lin & Alci 1986; Lee et at. 1986; Novelo-Caseneva, Berg & Helsley 1990).

Although this simpte model seems 'to-explain much of the

507

508 W. B. Ambeh. and L. L. Lynch.

character of observed coda, there are still significant disagreements especially in the details of tbe modelling process. For example, a point of great contention is how the coda Q'sho'uld be Interpreted. Coda Q as originally formulated Was thought to bearefiection of the scattering effect only (Aki 1969).. Aki .(1980a) and many others, from comparisons of coda Q and shear Wave Q, have suggested that coda Q is more closely related to intrinsic rather than scattering attenuation. This view has been supported by results of analyses by Wu (1985) and Frankel <i Wennerberg (1987) using energy flux models. Langston (1989) has, however, shown in the analysis of long-period Rayleigh waves in western North America that coda. decay with time, which is used to obtain Q, cannot QI;jC -used to dtseriminate between coda mechanisms, i.e, scattering- of anelastic absorption.

The eastern Caribbean is a tectonically complex area Which includes a subduction zone, island arc and ,alltiv€. volcanoes. Earthquakes occurring in the 'ar-ea have focal depths from near surfaee to abol,1t 200'kmaad are monitored by short-period seismographs operated by the Seismic Research Unit of the University of the We-st Indies, Trinidad, and the 'observatoires Volcanologiqnes' in Guadeloupe ami Martinique, No quantitative studies of seismic-wave attenuation in the eastern Ganbbean have been made to date. In this .paR~r> we investigate the .spatial variation of seismic-wave attenuation .and its_ frequency dependence in the eastern Caribbean using the Sato (1977) extension to the S-S single scattering coda model of Aki & Cbouet (1975).

2 MODEL, DATA AND ANALY'SlS 2.1 The model

In the single, backseattering model of Aki .& Chener (1975), coda waves are considered a~ being composed of the superposition of backscattered body waves from numerous randomly distributed heterogeneities in the lithosphere. The model assumes that the earthquake and the station are coincident, i.e. the scatterers are .at tru;ge distances compared to the sonrce-receiver distance. AccordingtoA]d & Chouet (1975) I coda amplitude, A(f II), at frequency. [, and lapse time (traveltime), t, is given by

where C(f) represents the vsoda source factor ,a is a eonsrant that depends on geometrical spreading and takes values of 1.0, 0.5 and 0.15 for body wave, surface wave and diffusion scattering respectively, and Q" is the coda' quality factor.

Aki (1980b) based on reported results ()f good agreement he-tween coda Q and shear wave Q (e.g. Rautian & .Khalturin 1978; Aki 1980a), concluded that coda waves-are made up. -primarily of backseattered S waves. This close agreement between S-wave Q and coda Q has been confirmed by other workers (e.g, Roecker et al. 1982~ -Del Pezzc et at. 19-5'5; Rebollar , Traslosheros &. Alvarez 19.85,; Kvamme & Havskev 1%9)_ Consequently, we assume a .spreading paI-ameter of t.lr = 1. It must, however, be mentioned that Herrmann (1980) also fonndgoQd

agreement at 1 Hz between coda Qand Q from Lg waves,

The assumption -in the model of Aki & Chouet (1975) that the source .and the receiver are located at the same ppin! i:s acceptable for coda waves-recorded long after the passage of the primary waves. In fact, 'Rantian & Khaltunn (1978) state that eq, (1) is valid only for- lapse times greater than twice the S-wave traveltime. However, analysis in this region of the coda is often difficult because of an inadequate signal-to-noise ratio or Short recording times. Sate (i~77) extended the above single-scattering model to incorporate source-receiver offset, thus allowing' one to begin coda analysis immediately after the shear wave arrival. The coda-wave .relationship in the case of non-coincident source and receiver may bestated as

A(r. fit) = C(t):K(r, a) exp (- nlt/ Q,,)

(2)

where a ~ tjt~:I t9 IS the S-wave lapse time. r is the source-receiver distance, K(r, a) =-(1/r){,(l/a) In [(a + l)j(a _l)]}I)-S and all the other symbols have the s-ame .meaning as in eq. (1) .. Taking natural Logarithms of eq. (2) and reanrangingterrns, we obtain

In [A(r, t I t)/k(r, a)] = In C(f) - (;tf to».

(3)

For narrow-bandpass filtered seismograms, C(f) is a constant and hence by performing, a linear regression of the term on the left-hand side of eq. (3) versus t; Q.,can be. determined from the slope which is equal to -1rj..j Q".

2.2 Data

('I)

In order to make a r-egi_onaJ GQmp~Qn of ccda Q in the eastern Caribbean, we searched the data QaSe- of the §"eismic Research. Unit (SRU) of the University of the We$t. Indies, Trinidad, ftiI suitable eYentsLseismograms recorded during the period May 1989 to June 1991. The SRU has operated a telemetred network of- shert-period (1 s) vertical component seismic stations in the English-speaking eastern Caribbean islan-ds since HiS7. Data digitized?t a rate of 100 samples pe.r/~ and 12-li-it resolution has been recorded OR personal CfJmpute:r~. Event's(-seismdgiams were selected on the basis of such factors as fipati'aJ dil!tribution, absence of spikes, good signal-to-noise ratio characteristics, and .epicentral distances Less than about 70 1cm.

The TIna] data set used in this stJIdy consists, of 73 earthquakes with signa! duration magnitudes; MO, nurging from 23"to,3.7 and fecal (;lepths f,vo.m 1 km to 183 km. The earthquake epicentres, and station locarious are plotted in Fig. 1 while detailed information on the events ana stations are listed in Tables 1 and 2 respectively, Ideally, we would have liked the epicentres to bedistriouted unitonnly along the is_land art but, unfortunately, this is not the c.(l,se. Events are concearrared.atthe northern .aad southern ends and this, to a certain, extent, is a .refieetion of the contemporary seismicity pattern in the eastern Caribbean. The lack of data in the Guadeloupe and ~nique areas is: due. to the unayailability of seismograms for stations on these islands while fewer earthquakes €lecU} in the region from Grenada to St Lucia, Usirrgthe available earthquake and station data, we subdivided the eastern Caribbean into four subregions as shown in Eg. land 'I'able 1.

•

16·5 r---.....,...--- ......... --~----,------.

• ... • _:hepB (J)

.. .. .... Q

SKI <:>... .._. II:

""'bA ANG- <

NEV .... ·It", I}-. 3=

BPA"" W

MGHO • W

· CP:~~

o ~ .. 11 DOMINICA DTMT\J

nsc •

'N

~-

•

SLB"~ s~ LUCIA

..

SSV@ & .•

~~ ST· VINCENT

~

"

---- ,p

0 ______

TpFI _ TOBAGO .. ~

BOT

10'5

VENEZUELA

63'0

'W

60·0

Figu~ 1. Map showing the earthquake epicentres and stations-used in this .. study. Dots are shallow earthquakes (0,:;;: Z s 70, km), triangles are intermediate depth earthquakes (70< Z <200 km) and squares are seismic stations. Approximate boundaries of the four subregions are also shown,

2.3 Analysis

Each seismogram time series was first bandpass filtered over eight frequency bands centred at 1.5, 2, 3, 4, 6, 8, 12 and 16 HZ, with bandwidths of 1 2, 2, 4, 4" 8, 8; and 16-Rz respectively, using an eight-pole, phase-free Butterworth filter. Starting at the S arrival, the root-me an-square (rms) amplitudes in each window Were determined for the data in each frequency band, with the window sliding across the time series at 1 s intervals, Time-window lengths for the mIS averaging of 8 s (for centre frequencies fiJf i.s, 2, 3, 4 'and 6lh) and 5 s (for eentre frequencies 8, 12 and 16HZ) were chosen so as to smooth out as much as possible irregularities in the amplitudes of the arrivals. The left-hand sk!e of eq, (3) was then evaluated and plOtted against lapse time. from the pLat, the Iinear portion, wbi.e:h.hai:! tobe at least 30 s long, was visually selected and a straight line fitted to it

S_t~tions

Lan.

(OWl 61. 53 6,2.17 ,ft2.2.5 ;t;0 •. 78 61. 62 61.27 60.55 61.86 62.21 '62.00' 61.89 61.93 6.i. 55 6-1.11. 6::2. Hi' 61. 82 61.77 61.51> 61.28 61.1.3 Eil .• fiO Ei-1.90 6:1.0;4 6:1.80 .6L19 62.11 61. 73 p.lO 6~.17 61.25 '61.68

61. 42 6'0.98 60.95

62 .~O 62'.26 62.11 62.11 62.2'7 62 .. 2J. 6'2.40 62.09 61. 30' 62.'94 62.29 61.94 62.2,4 6,1.99 61.,85 62.27 61.68 61.79 51.76, 6-1. 90 61.71 61.6·7 61.82 61. 18 61. as 61. 92 6,2.04 61. 65 61,.59 62.05 62.26

St. Vinctmt/St:. Lucia (SVL) Region 090'111989 2QO'l 2'3.63 lA'.32 ,60.83 210'9-1989 1656 55.85. 13.44 61. 56 19'07.1990 0624 28.30, l3.37 60.99 25Ji'SI~9'Q 0-509 31. 78 14.0,4 60.5,9 30091990 b6i2 22.90 13.92 60.89 02.1 OJ, 9.90 0358 :4 2.67 13:.79: 60.97 251"21990 2108' 4:4.8'4 13. 366,~. 85 10021991 11134. 39.75 14.13 6,1.10

Depth Md (J<m)

34 ·3.8

76, 3.2

.91 3.3'

5-5 3.0

39 3.1

,44, 2.9

44 3.5 69 2.9 87 3.i

4-8 3.2

15 3.0

.5 3.7

.41 3.2

46 2.6

9: 2, .• 9

II 3.6

2'9 3.1

39 2.5

44 2.4-

10 2.9

61 3.4.

44 3.5

33 2'.4-

26 3.1 49 3.0 76 3.l

8 3.0

4.6 3.0

68 3.'0

32 3.3

33 3.5

9 liS 1.2'8 7.1 3<:1 20 2,7 10

19. 1

lA6 '38,

18 11

3 1'05 13

i.83 14 40 96 19 6'4 78 26 20

128 13

8 71 83 20 25 29 28 3,5 4'6 25 24

Coda Q in Eastern Caribbean 509

Table 1. Parameters of events used.in the. coda-Q analysis, 'l"tin.idad/TQbagq (a'T) Region

Date

Lat.

TRN

TR~ 'l".CE' TOE

TRN

'['lUI TeE 'I'RN -

BOT

Tc;!E

TeE

T~ TCE TOE

TOE

TRN TC,];; TJ<N

TC.E

TRN TOE' --reB

TRN

~N

'l'lU;

T.RN; T.CE x&t

TRN 'I'FR TlUI

'I'RN

TCE'

TRN

TRN

IJ.'U,

ri&

TRN'

3.4 3.1 3.1 3.4 3.2 2.9' ).0 2.4.

SLB

'SVV SSV SVS SVB

SLB

SLB

SLB

SLB

SLB

DDMM-Y'Y-HRMN S:EC CON)

Sl061939 015. 41.17 10.55 27061989 19.52 52.14 10.56 04'07;],-989 07)6 35.90 10.86 1007.1,98.9 123,0 03.46 10.65 3.L07.Ei'U 08!20 13.03 10'.75 03:i019S9 1903 4.2.4-7 iO.49 20tU989 n073e. TO J.l.l4 291119-89 1020 00.44 11.10 20121989 103.0 04.89 ~0.85 20121'98,9 IBi:5 43.38 .10.83 2'1121-9.89 1516 13.02 11. 21 2·11.2'1989 '03.55 29.5-2 ;10 .16 190.11990 0:14.2 23.2L 10.72 ~10U99(1 22~14 5,o!!.73 Ib.28 08041390 O9r5 35.79 ll.12 "1404i9.90 i232 34.,37 10.99'

~~g:i~:~~ ~~.~~ ~-~ :~I It~~

2201'19'90 0350 :t).!lO LO.S1 05081990 cl1Jl3 54.29 16.'46 '070,91990 0-749' ~6 .07 10.80 11091-9.90 1207 43.97 10.cSl 22091·990 ~OS8o 1'6._85 11.Qa 201ll9'90 1-240 57' .35 10.52 .29111990 1.635' ic2. 801: HI;. 64 j11219ge .Oi~5 29.4.6 H.OS 15011991 '1555 14.62 10.54 0302_1.991 0'311 51.~·7 U. 55 13031991 H46 07.J6 10'.98 30'04.19-91 2224 08.79 11.16 1406L99L 01,55 46,,12 10'.5'8

Dprninioa Region

26'051'989 0'216 17.Il-S 15.39 1l5U1989 23'24 09.88 15 .• 14 12121989 0"256 O'I.39 15.17

2.7 3.1 2.7

DT~T DSC D'l'JIIT OTMT

Iieewards Region,

1205198~ 06e? 25.17 17.00. 2605,n89 012'. 23.59 !7.GS 01061989 0525 4.1 .• i9 16.56 20061989 1638 38.46 17.64. 2106~999 0708 23.66 17.13 01071999 1321 5{l.33 17.03 16071989 0llO,1 15'.59 17.09 17671"999 1·6,34 1·5.~6 17.02 170719'89 2'1G2 56..96 17.32 2107198·9 05;38 09'.36 17.57 0308.1989 1753 15.53 16.99 06'S1989 0148 27.92 17.44 09081989 1906 29.15 17.05 1-4091999 9·213 21.45 16.80' 30081989 10)2 0,6. 9~ 11. 5<L 0,7111989 1305 1'4.79' 17.43 3()1l1989 1435 57.29 17 .. 47 06011990 163"0 47.1516.95 "090,t,19ge ,002,1 ~5 ;06 16."59 06021990 1126 08.91 17.48 066'21990 11033 '17 .. 28' 17.37 17-021990 163'04:i. 70 17. as 25041990 '06-25 28.'0'3 15. 9 a: 03061990 0951 39.59 17 d2 2,20H99.0 084c7 56.65 18 •. 05 260'7199.6 06.4'8' 17.53 17.52 030919,90 002254.44. l.-V .'46 11091990 a8~8 28.68 17.26 17091i~o 0336 A~.76 17.23 24101990 1002. 23.23 f 7. ·36 271019:90 1952 42.51. f7 .10

3.2 3.7 3.4 2.9 3 •. J. 3.1 2.4 3.,7 ~.l 3.2 3.0 3.'2

2.~ 3.1 3;.5 3.2 3.2 2.9 2'.7 3.5 3.0 3'.2. 3.2 2.3 2.9 3.S 3.4- 3.3 2.9 2.'# 3.0

NEV SKl BPAlIlm MGH cBPA NtV' 'BPA NEV

BPA

NBV

KGH BPA BPA

SKI

KGB.

BPA

BPA

a·PA

B.PA SPA BPA

BPA ANG ANG. BPA

'BPA SPA .MGH' BPI. Cl'.B

CPIl"

BPA

B)lA

'Bl'A

'BFA

BU' 'BFA

33·8 5,5 '
67'6 0
.- ~.". 4Ii 3 Hz
Qc'"358
6·9
. ~6.
....
. t'···'·~ ~
5·8
33-8 6Ni 0
6Hz
Qc'867
7'7 FJgIII"e 2. Example of coda-Q analY!liR" as applied to station TRNseiS~Q~~anl fQ[ event ¢f 22tr:HJ9U Q'350 UTe. The plot is fOT the quantity in (ACr,! I ()/k(r, a)] versus lapse time, I. The straigb] lines,sJww the pGrtlOlls,chosento determine Qc'

510 W. B. Ambih and L. L. Lynch

Table z. Listings of stations used in the coda-Q analysis.

Code

Name

'splitaecording tp. depth ranges where- possibJe. This, was also carried au! for eaehof {he subregions. Table 3 is a listing of the average Qc. and their standard errors, where the :s~and'ard -error (SE) is the standard deviation divided by the square 'root of the number of Qc values used in calculating the average,

2.4 VollIIDe sampled by coda ".aves

In order to carry out meaningful comparisons of coda Q from different regions, it is imperative to make estimates of the volumes .sampled by the different groups of data. Aceording to Pulli (1984), th~ volume: sampled by coda waves at lapse time, t, is an ellipsoid whosesmface projection bas serai-maicraxis, al = "V,;t/2, .and semi-minor axis, a2=[(ai)20_(r'2:/4j]P.s where V; is 'the S-wave velocity and r is the source-receiver distaace. Thus Q" villu~ represent an 'average over a Iarge regien.

The average velume sampled ean be assumed to be represented by the average lapse time, Tev, where tev =,tstart + /).-twin{Z'; (start' is the starting lapse time arid 6twin is the window Jength. A reasonable estimate of the maximum depth of the volume, .Zmax, is the average focal depth, Zav, for a .group. of events plus a2. Using S-wave' velpcities of. 4kms-1. and 4.5 kms-1 for shallow (O:sz:s 70 Iari) and intermediate (Z > 70 kro) depth events respectively, together with average source-station distances ana

TRW ~CE BOT 'J1PR 'SVB

~rv

SLB Dse D'llM'l' MGIi OlE:", SKI .Bl'Il AI.';' CPS

ar., "U'lu"t~ne.,. 'l:rl.pid.ad 1D',65

Cl:1aca;cnaca.l:.e, T"r-inida.d .10.7.0

·Baco1etj 'i'o.oago U __ 16-

·E.co" eect , Tobago U ; '18

BelmOnt, se, 'V1.nceo;t- 1j.:17

Wallioou, ~'t. vin<:ent 13.3.

S1lIIUIIi.'t, st. vJ.nc""t i3.3,3

!laHond., St. Lllc'ia 13.82

Scott',. !lead, o;>OlIlinica 15.·il

Tete xo'rne, o:omini"" 15.'2J

SL George'" 'ilill.,' 'MDnse-rra't 16.77:

Ginq)'lI1.anQ" .1leyis 17.14

8a.yifor~&; l>t:.Kiets 11.33,

ap;ggy .Pe.a.~~ Antigu~a 1 '7 ~ O'il

Bri~ Bill, Antiqua 11.1J

c9~h;gt.on I B.a.rbud~ 1'"] . ~.4·

B'l. 40 6.1 •. 15 '60:1£ 60.]'8 61. 2'5 '61.2.1 61.19

~tJ!

61.3:<; &~. 21 62-.5; 62.74 61.86 6:1.83 6.J..B~

by least 'squares. FrOID the slope of the line, Q" was calculated. By usingoaly linear windows which wereat least 30 s long, we hope to mfnimize the contribution of terr:m'oral small-scale variations ill coda 'amplitude- which may be caused by interference effects, secondary arrivals, or site eondi ti ons,

Fig. 2 is an example of the application of this. fmalysis procedure. Only Q" values. obtainedfrom least-squares fits. with correlation coefficients greater than or equal to 0,95 were rcJained. The arbitrary value of 0.-95 seems: tq represent an acceptablecompromise between rejecting too much data and obtaining a high scatter Tn, individual Q values for a particular region. For each station, 'an average coda-Q value was calculated at each frequencyfor the data,

5·9

1·5 Hz Qt:-·H:15

, .. ,' .~

.........

67·6

4 Hz Oe-448

.

........ -

.. :.. :",- .. '

....

. .

. ...

. ....

'. e'

..........

4a~------~------~------~--------~------~

o·

6·5

33'S

8 Hz aC~1126

68-8

16 Hz .

7.6'1 ~.. Oc:2516

.... 0:-. ~ •• -., '" • a.· ... ,

..... • •••• ~' ... toO

.. ,,~

65~------~1--------~'------~--------~~"~"~~!

. o 34·4 6&8

LAPSE

TIME (SEC)

7,2

6"7·6

12 Hz Oc:1658

Coda Q in Eastern. Caribbean 511

750:38 (19)

822~42 (10)

1f05:ti9 (4)

811.3.5 (14)

814' (1)

776128 (30)

805£69 (4) -

'h~~ (34)

,555 (1)

718, (')

'741~eo ,'(2),

754:45 'I~)

7S4±33 (3)

8<19

(1)

rncaa :(4)

(7h26 (e)

1023:1'~40

(3)

'GZ3:tl"1O

\~)

1009 II)

pG4±156 (3)

l00e&:111

(4j

1056±341 (2)

1255 (1)

1225 (1)

12.w~lS (2)

805~38 (16)

'853,,116'

(6)

618=4,'

(22)

983:8& (Z)

54,9

I~) 938;;.67 (3)

12.0 1037~72 1,14),

lOO~:c1:1 ,(14)

h37~146

(3) .

12'17::294 (4)

12!:lG!;tliB 17j, -

126&

(1)

1067~1, (11ij

1247±'.294 (4)

tCl!l9~70' (22)

t022 (,1)

1064 (I)

1367 ~1)

1144±1;2 (8)

10'72:145 (3)

1162

(1)

1095:11)5 (4f 1',6±71 (7)'

13504' (1)

13S4- (1)

164'1 II)

1818~e72 (2),

1759;:507 (3)

,98<1

III

2884 (1),

2884 (1l

1',21,,51 (H)'

1176:119 (Il)

,,35:47 (23)

1214±29<l (2\

1274±.29Cl (2)

16,0 1-484':-122 (\~)

Table~. Avcr:ag,e pc ~alueS"'andthei~ standard errors. Valuesin brackets represent the number of Q used iii averagmg., T.T'-T?D1dad/Tobli~o; Sl V't-.5l Vincent; DOM-Dominica; Lwds-e-Leewards. S: shallow

events; I: .intermediatc depth events; A: all events. '

Flr9lqn T~(S)

TFlN{~

TRI\I(A)

TC~(SJ

TOEtAI,

BOTjSJ

TPR(S)

Ti(S)

T.r(",)

SWII)

ssvw

SV8(1)

SlV't(A)

SlB(S)

SLB(I)

DTMf(S)

OOM(A)

MGH(s)

MGH(I)

MGFI(A~

NEV(S)

SKlIA)

SPAtS1

BPA(I)

BPA(Aj'

ANG(S)

ANG(U

194,,16 (4)

152 {1)

1052 '(1)

,76 {II

'00",13 (5)

152 (II

"64-"'\2 is)

',57 (1)

t5] '(I)

157 (l)

20,7,53 (21

2llh53 (2]

21!h30 (5)

239.3& (21

2211211 (7)

2,0 259~13 (lsj

Z59~13 (Hi)

,211:;:7

'(i!)

'2!11l~I06 (2)

255~SO (4)

163 (1)

1'42 (1.1

2~Jl13 (19)

298:.106 (il)

248,14 (21)

350 (tl

3m (1)

2100c16 (2)

197 (t)

,200~10' (3)

242~37 (4)

io7 (1)

23S (1)

221'",4 !21

006.1

'Iaf

3OO~1 (j)

"269 (1)

3017;69 (4)

298m '(12)

317 ,1)

321~17' (1"6)

35!k21 III

358"39 [a)'

asa",6 (1_1.))

243 (I)

207' (1)

327;:t4 (25)

3511·39

(3)

~~13 (28)

436

(1)

3M (1)

35H74

Gl)

3810::36 (4)

296~20 (4)

344 (i)

30!bra (5)

339±22 (9)

295 (1f

,270 ,(1)

'2112,,13 -,(2)

398 (1)

3n<35 (3)

3BO±24- '(4)

:273",27

(4)

.352~1'9 (11)

351 !:59 (6) .

3~52.23 (1-7)

'45ti±26 '(2) -

4_0

'3,98,,'5' '(2D)

362 (1)

390::14 (21)

44,2~19 (9)

404:5 (3)

43,:hI5 (12)

2M (1)

XlI (1):

4G5,12 (3i)

394z'1 (~),

4(loi1=11 (35)

4ea '(I)

Q3 rl)

450~8a (2)'

449!3o (4) '.

!369:f2!!(~J

467 m

~30 '(5)

'415:<24 :(9)

358 (1)

46!h81 (2\

~_58, (3)

400 (,1'1

4e6±42 ):3J

·465l::t7 (4)

.:l45~3A (4)

615 (1)

-521 (I)

.sea!_4? , (2j

-400.:,2\1

(15)

454,,69 (6)

473~24 (21)

S99_±O (21

sea (1)

593~ {:.IJ

FAE':QUENC,!;

$,0

548>:26 (_2D!

449 (1)

S44>:25 (-2<1"

664f25 (10)

576t45, (4)

839...:24 ((4)

304 (1)

S22 11)

~{~F

55h43 (5)

51~O (37)

631 ,(I)

523 ,i)

5461:68 (2)

56h36 (4) -

038::21 (2)

762 (I'

67~43 (_3)

612£35 m

75U209 {21

757't2Q9 (2)

752-

(1)

730'79 (3)~

735:>56 (4)

423::111 (2)

828

(11

872

(1)

850:22

12)

655±46 -('16)

6:10:84

(6]

64!h4,1

c22i

820=18 (2)

sao (I)

74b+J!~ (3j .

1464~122

(13)

1524±.114 "(3)'

1~:<534

(3)'

15412±2s3 (6)

167,5 (11

1502:;97 (17)

1~'s34 (3)

1,~107

1381 (il

1500 Ii)

2150 (1)

r679o:2il8 (3J

1'164,,1 EI2

(4) -

1964

{II

1;564'=179 (51

1607±134 (8)

1S70 11)

1870 (I]

232S",-1·()t9 (2r

2325;:10~9 (2)

1236

II)

34'55 (I!

3~ ti)

1482!oe:8 (15) .

1624;b133 lSi'

1523±73 (2_11

H36±39!>

(2)

3;1 VJtriation or Qc willi winilow leng1Il,8verage lapse time. epicenti'al diStance, focal depth ~d magnitude

It has been observed in, many Goda-Q studies that Q" increases wi tb increased lapse time and consequently, increased window length (e.g. Rautian & Khaltunn, 1978; Rovelli, 1984; Lee et al. i986; Kvamme & Havskov 1989). Taking into consideration the volumes sampled by coda waves as postarared, by Pulli (1984), Increasing coda Q with increasing- lapse tim"eli has beenrnterpreted - 0 be due ttl increasing Q with d!'lFth (Roecker el·ol. 1982; Rovelli 1984). But Gao ei at. (1983a) have suggested that i;llultiple scattering: might be responsible for the observed increase in Q., while others ~ve attributed it to incorrect coda-model parameterization [e.g. the use of an inappropriate geometrical spreading factor, cr, .in eg. (:1)]. We Rote, however, that Havskov et of. (1989) _a~d Canas et al. (1991) did not observe any .sigrrificant variation in Q", with, lapse time and window length for studies in WashiT)g!on Stat;e, USA, andsouthern Spain respectively.

Q,. values for stations 'IRN and BF-A were plotted, for each frequency, against each of the-following variables:

512 W. B. Ambeh and L. L. Lynch

Table- 3. (Coutinued.) .s~bOO

Ragl"" 1.5 2.0
CPEI\S)
Lwards(S] 216~30 3\l1oU
(5) ,(10)
l'Wards'(I) -2Zl:02S ~S7
(1) (7)
L 'l¥aids(Ji} 2111±19 00i2~21
(9) (17) PR"OUENGY
3.0 4.0 6_0 ao 12_0 16.0
238 302- 387~33 588±2S St,3"lS8 1294
(I) (1) (2) (2) (2) (11
342.·18. "63z22 ~±4'1 ~1:c44 , 12!h52 14OOiso
('_~j (24) (24) (24) (23) ji9)
359;<40 48h39 674;054 ·955,84 1506:<255 1 98,3,;283
IS) ('1)- (11) (111 (9) (9)
348;'17
f28) lapse times, the parameters, which are a reflection of thevolume sampled by cod_a waves .in each subregion, were calculafed. For example, for the shallow earthquakes -in the Trinidad and Tobago region, Qa estimates at an average lapse time of 52s correspond to sampling volumes of about 208.km in lateral-extent, 137J(in in depth and 16 50n km2 in surface area, Sampling volumes for shallow earthquakes in the other subregions are similar to those for the Trinidad and Tobago area. We deduce that the volumes .sampled by the coda waves from shallow events in each -stjbregion appear to be- distinct, with ·linle or no overlap. This may however, notbe the case for the coda waves of intermediate depth events which ample larger volumes.

3 RESULTS

OUI r-esults show that average Qc values in th~ eastern Caribbean at a frequency of 1.5 Hz, range from 152 fOI intermediate depth earthquakes recorded at'TeE, to about 239 for intermediate idepth events recorded at BP A. At 16 Hz, the variation in Qa is from about 1236 (NEV)to 3455 (SKI). Looking at the Qc val ues listed in Table J, there seems to be some variation within particular subregions .i.e, among stations, and between the different subregions. However, before examining the coda Q_ in detail for station-to-station or regional variations, we checked for other possible causes of fluctuations in Qc values; such as window length, average 11lp~e time, epieentral distance, focal depth .and magnitude.

window length, average lapse time, epicentral distance, foo,al depth -ami magnitude. TRN and EPA were- chesen because they had relatively more data compared to the other stations. We found .rro obvieus dependence of Q" values at TRN or BPA en any of the .above variables. Selected examples of these plots are- shown in Fig. 3.

3.2 Qc variation _and frequeney liependenee

In most .regions. where coda-Q measurements have been made, a positive correlation between Q" and frequency has .been noted (e.g. .Aki & Oholiet 1975; - Rantian & Khal turi n 1978.; Roecker et al. 1982; PlIlli 1984' Havskov et al. 19&9; Ambeh & Fairhead 1989). Frequency-dependent Q has also been observed in studies of shear Wave Q using the single-statron method (Aki 1$80a) in an .inversion of P~wave daia fo-r Q (e.g. Thouvenot 1983), aad in shear wave spectral analysis '(e.g, Singh et al. 1982.; Rebollarer al. 1985; .Kvamme & Havskov -1989}. The form of this frequency dependence is generally

Q",=Q"r

(4)

where n has been found to lie mainly in the range 0.5. to 1.1, Stations 'and regions with ~. values from at least, three ev(;f1Js were fitted to eq. (4) and the results are listed in Table 4.

In the Trinidad and Tobago region, there -seems to be no significant difference between average Qc at all frequencies for stations TRN- and TeE within error limits for both sballowand intermediate depth earthquakes. This is confmmed bycalculated Q~ values which lie in the naH9W range 12,8 to 13.~. The variation in the index, n, is also very small: 0.83 for TRN shallow events :and 0,88 for TeE sballowand intermediate depth earthquakes. 'Ihisris inot totally surprising since both TeE and t:RN sample almost the same volurne. It may also suggest that the influence of .recerver site effects are minimal. A sfmilar comparison could not be extended .to .the two Tobago .statioI;IS, TPR and BOT, because their Qc values arefrom a single earthquake only,

Q" at St Vincent stations (~W, SSV, SVB) were estimated from tW0 intermedlate-depthearthqnakes and are generally consistent with each other. No Qc values from shallow' earthquakes- are available for 'these .statlons.

For $LB (St Lucia) , average Q" from lour shallow earthquakes lit 3 to 19 Hz appear to be slightly lower (by ~100) than those for a _single .interrnediate deptb.event. Qt) = 105 and n = 0.94 were obtained for there shallow eve-nts. Although 'none of the earthquakes used for SLB was recorded at any of the 'St Vincent stations and vice versa • we decided, because-of the paucity of:suitable earthquake data in the St Vincent and St Lucia area and the'overlapping coda volumes involved, to treat this area as a single

Coda Q in Eastern Caribbean 513
geQ TRN 6Hz 1800 TRN 12Hz BPA 6Hz BPA 12Hz
1200 1800'
It •
• I
• •
• •
I I I,
Qc I • • I ,
I I, I I
• I I. .... .' , ....
• • • • I I • :I. ,
I •• , • •
• , • I" I I,
I " •
I " • • I
400 500 300 600 I
a 100 a 100 a 120 0 120
TIME WINDOW (SEC)
900 • 1800 1200 1800
• , ,
• ,
,
• I
• • .'
• • , • •
Q,c • •
. ' . • •
• .,. .
.JI , ·t " · ~ .
\ • , • , •
'. " • •
.. '. ." •
•
400 500 300 600 •
0 100 a 100 0 120 0 120
AVERAGE LAPSE TIME (SEC)
900 '. 1800 1200, 18'00
• •
•
, •
• •
• • . "
Qc I • • • •
•
• , • • I. ,,' I
1 • ~ • I , ' . , • .e
4 • • • ~"'" •
• I.' I • I" •
" • • •
• • • •
•• , •
4G0 500 30G 600, •
Q 100 0 10Q 0 100 0 lOG
EPICENTRAL DISTANCE (Kml
900 • 1800 1200 1800
•
• ,
• I,
• "
• ~ ..
Qc • • • I • •
- • .' •
• • . '1' ' .
• • .. ' -, ••
• • • , ..
, • It'
:/' '\ • I
, .' ,
I , • • ,
400 500' 300 1300 •
0 lOG 0 10'0 0 150 0 150
FOCAL DEPTH (Km)
900 • 1800 120:0 1800
II • I
• .'
I • •
, • , •
•
Qc , '. I • •
, I • • It' , •
,
• I' • • •• 1. : · ....,
• • •
• •• " • •
, • • • • • • •
I ,- I
• fl' •
400 5'00 300 600 •
2 4 2' 4 2. 4 2 4
MAGNITUDE ( Mo ) Figure 3. Selected examples of plots of the variation of Qc at TRN and BPA as a funetion ofwindow 1en~h.,,<lVerl!g~ lapse time, epicentral distance; focal depth and magnitude,

subregion. A Q" of U9 was obtained for the St Vincent/Sf Lucia intermediate depth events compared te the lQ§j'o.r the shallow events. Again, cons-idering (he. uneertainties Involved in determining Q and the different depths of' penetration of the coda waves, this difference does not

appear very significant. the values of n, 0.89 for the intermediate depth earthquakes and 0.94 for the shallow eyent$,areruso I'1Ot milch different wirhin error limits.

'Ihe .three earrhquakes 'Used in tpe Dominica regiqp .are shallow and average: {2" at 2 and 16 Hz are 221 ± 14 and

514 W. B. Ambeh and L. L. Lyhch

Table 4. Calculared Qil and II values

Su..tl.,Qnl Sh4:.lll::l .... E'vlllllt::::3 Intet'.lO'le'd.i.ate· EVonts Al~ ~V9_ritS
:a.ef:i~On 0. Q. n Q.' D
~'F.N 131 Q.8'hO.'1 13. O • .831..O.·0J
Tee !n ().8~;O.O4 la (l.8B:!O,.1)8 l.Jl [}..8'.8::D'_Q4
SI:;..Fl 105 0.g·'M.05 io. CL~i5::ti','P"
""t<r ,.5 1.~$'"'1'l. ios 1.06fQ-.ll..
.. N-:EV LlH V.s"'O.O< la. CI :,g..~~.o:~ 14-
!KIa iil' e. ~1~.a..t"'~.a !..3J o.~ !)a:tlJ ~ 0,1,
il?A 10' C!~iI~,:-g,'Q.J 13) 1J·:aSI{I".O' .. , ILUJ:U.Ol
MfG- 2~_9 10 • .7.110.01.
~.: ..:r:e:g:L::il'! u. Q ."6~O. (>2 li~ ll .... t6:!:Q.().S 121j 0.:91 (]_O2
SVJ; .r~"'QD lOS t:;~~:~ 1:l$1 0 .... 8:9 .. 0'.03 ''9 O.!'l_I±.a.O'~
'CIoldini-C!4 91 9·7 cl.09:f:'O--.Q,
Leei>1lLO.o H" 1l.-S2.!O',03 li4 Q'.9_2~O.a~ l .. ·a {I.S-S''io_(]'J 1870 respectively, Two .stations (DTMTand DSC) were used although l)SC c_g~tributed only 3 Qc values. Estimates of Qo and rl are 97 and 1.09 respectively, The shortage of suitable data fur stanons in St Vincent and Dominica may be partialfyattributable to the noisiness of the station sites.

2500 r-.-:T=R:-::':-:-N=-='O:-:A-=D/-;:T=-o::-:a:-:A:-:G:-o---------,

of 51 VINCE~T/ST LUCIA (a)

.OOMINICA

,. LEEWAR,DS

2000

1500

1000

2500

2000

1500

1000

Qc

•

•

2500 • TRINIDAD/TOBAGO
+ ST VINCENT/ST LUCIA (c)
• DOMIN'ICA
X LEEWARDS
2000
•
1500 fd
Qc {!
1000 if
~fl
II
500
'101'
•
....
0
0 4 8 12 16
FREQUENCY ( Hz) 500

Ii" to

a

For the Leewards area, average Q" for stations ern and NEV, whicfi are from shallow events 'only, generally appear to be smaller than those for the other stations in the region for the same depth range, This seems to be valia at most of the frequencies. Very high Q .. values of 2,SS4, 'and 345,5 at 12 and 16 Hz respectively l Were obtained at SKI for: the deepest event (183 km) used in tID:; study. These compare with. intermediate-depth average Q", for this Tegion €)'f 15013 ± 255 at 12Hz ana 1983 ±283 at 16 Hz. Unfortunately, a Q", value fOIl thi:s 185' kin depth event at-another station in the region could hot be determined. However, this was not a unique situation because a 10'§km deep event recorded. at MGH also resulted in Qc estimatesof 2'690 and 3344 at 1.2 and 16B.z respectively. The' Low Qc values for NEV compared to other Leewards stations is reflected ina Qt) of 101, eompared to, for example, 152 fOF shallow .events at BP A. This may beuce to site effects or differences. in sampled coda volumes .. Q" for intermediate depth events at BPA is i18. which is slightly lower than the 152. for shallow events .. Althougb B-rA and ANG are situated only about 13.km ap9Jt On the same island, Antigua, their Q" and n values considering everits at. all depths are 209 and 0.13 (ANG) and 147 and 0.82 (BPA). This may be due to sife effects at ANG since sejsmegrams reeorded at this station seem to display anomalously long codas. Average Qc in the Leewards fur both shallow and intenmediate-deptt; events <1.0 not differ signifiGa:ntly within uncertainty bounds; except perhaps at 12 and 16 HZ. Q" for shallowand Intermediate depth earthquakes are ]45 and 134 respectively, while n is 0.82 for shallow and 0.92 forlnteIillecftate events,

Average Qc fer.shallow, intermediate and all earthquakes for the· 'four subregions+atevplotted as .a, function of' frequem'Y ill Fig. 4. FO"r the shallow events (Fig. 4a), the main fearure js: the fact 'that Qc for the Dorniniea T'egion seems generally to be higher than those for the other zegiens -at frequencies great~r than or equal 1'"0 .6 Hz. This is in spite Of the fact that events in the Derninica region sample a

o

12

16

4

8

Qc

,

•

500

• TRINIDAD/TOBAGO

t Sf VINCENT/ST LUCIA • DOMINICA

,. LEEWARDS

o

4

8

12

16

Figure 4. Average Qc versus rrequeno.y for the different subregions (A: shallow events; 13; intermediate depth eYl::nt,~: C: ail events). Error bars represent ±l standard error.

Coda Q Ul Eastem Caribbean. 515

shallower coda volume eomparedto those in tile 'other regions. Attenuation of waves with frequencies greater than about 6Hz may, therefore, be greater in the Dominica region compared to the other areas in the eastern Caribbean.Tt must be stated once more that the number of events used in the Dominica region is small and consequently some caution has 10 be exercised in the interpretation and conclusions, POl' the Intermediate depth earthquakes (Fig. 4b). average Q" for the Leewards are larger than those 'for the Trinidad and Topago .region at 31j frequencies although there is a significant overlap of enror bars at higher frequencies. Fig. 4(c)" which shows data for all events, seems to reflect on average trends, already described for the shallow- and intermediate depth events.

such as folding, faulting or subduction are likely to introduce strong heterogeneity, show significant frequency dependence- of Q". We note that reservations as to the validity of these correlations exist. For ex amgle , coda Q measurements in New England, USA (Ptilli 1984). which is a: non-tectonic region, ·show a very strong frequency dependencear lapse times less than 100 s,

The- Qc measurements in the eastern Caribbean, exhibit a strong frequency dependence(n = O. g~ 1.1), which if we assume the validity of the above eorretation, would not besurprising because of the presence 0f: a subduction zone. However, within the-eastern Caribbean, there seems to be no obvious cOl'!elation between the level of seismic activity andQ,)n since the most seismically active subregions (Leewasdsend the Trinidad/Tobago areas) have about the same 1I and slightly higber Q-.,. than rhe St Vincent/fit Lucia area where ~eismic activity is significanfly less, The Dominica region which may be slightly lei's, seismically active than tile Leewards and Trinidad/Tobago areas has a lower {2" and higher n. The frequency dependence of. Q~ in three subduction-zone environments in other parts of the world: Adak Islanas, Aleutians (Scherbanm & Kisslinger 1985, 1987), Washington State, US4 (Havskg:y etal. 1989) and Hindu Kush (Roecker et al. 1982), average abour 60 for ~ and.Lfor n. The Q" values in these areas are.lower than those obtained in the eastern Caribbean.

Coda Q of'loeal earthquakes recorded by short-period seismic stations on some eastern Caribbean islands was determined using the' single-scanenng: model. The lateral and depth variation of Q, in the region~eeID5 to be relatively small. Q" is a minimum in the Dominica region (~=97) and a maxirrium un the Leewards. (Q" = 14S) while the degree of frequen,cy dependence IS largest in the Dominica area (n = 1.09) and smallest in the Leewards (n = 0.82), The possibility of the existenee €5f site effects at stations stich as ANG and NEV cannot be totally discounted. The use Of shorter lapse. times With the consequent $aJTipIingof shallower structures may have yielded larger I_egional variations. However, thesimilaritj in. average Q",., may be an .indieatron that average attenuation properties in the crust and uppermost mantle in t1ie eastern Caribbean 'are similar.

,4 DISCUSSION ANI) CONCLUSIONS

Acompreheruive review €If 0000 studies and the problems associated with the interpretation of Q" was made by Herraiz & Espinosa- (1987). While there seems to be 00 doubt that coda originat-es frotn waves scattered from Lithospheric heterogeneities, there is still no CQIIlIDQn1y accepted view as to. what Qc represents. 10 this study, we used the extension of 5_ato (1977) too the single S-S backscattering model of AId &. Chouet (1975) and consequently. our interpretation of the results are ultimately constrained by the assumptions and limitations of '-this model.

Several studies have questioned the validity of the single-scattering approach in coda studies, Accordiug to. Gao et ul. (1983a, b), the effects of multiple seatrering are dominant at lapse times greater thaa -abour WO s, but Spndich & Bostwick (1987), in an analysis of Mosgan Hill. California, Jocal earthquakes, fQtlnd that the early pcrtions. of their coda were .dominated by multiple-scattering from near-site heterogeneities. Qc as onginally.fonnulatedby Aki (1969) was thought to reflect a predominantly scattering contribution. However, Frankel & Wennerberg (1987), using an energy-flux model. sug_gest that Q" is a measureof intrinsic rather than scattering attenuation. The physical significance of the frequency dependence of Q" if; also a point pf contention. Da:inty (19.81) suggested that the frequency dependence of Q" between 1 and 20_Hz is mainly due to scattering, with anelastic attenuarion. l:!eing relatively frequency independent. Richards & .Menke (19~3) and Erankel & Clayton (1986), however, conclude that frequency dependence observed in £00.a analysis using the single searrering approach may be due to multiple scattering.

Empinieal oonelations between the coda Q atl Hz. QUI in an area, and the level of.current tectoni-c activity.in the area, have been made. Seismically' active regions, or areas' with a high intensity of tectonic activity, 'seem to be associated With low Q., values (e.g, Aki 1980b; Roecker et uL19~2; Singh &. Herrmann 1983; Van Eck 1988'; Jin &.Aki 198B; Cauas et at. 1991). Comelatians between, the degree of frequency dependence of Qc and the lev-el of tectonicactivity in the area of measurement have also, been .made in compilations of Qc meaSUIeIDc~ts by several workers (e.g, Aki 1980b; Prilli & Aki 1981; Jin, Cao, & Aki 198$) .. In general, tectonically stable r-egiofl.s. were found to exhibit almost no frequency dep-endence wJ1ile: active areas in which processes

A CKN'OWLEUGMENTS

We are grateful to the staff of the Seismic Research Unit Eli the University qf toe West 'Indies, trinidad, fCir help in the data acquisition. Thanks to M[~ Yvonne Joseph fat typing the manuscript .and Mr Godfrey Almorales tor help in d(~fting the diagrams .

RE,F_EQE_NCES

Aki, K., 1969. Analysis of the seismic coda of.Ioea! earthquakes as scattered w}!.ves, 1. geophys. Rf!S .• 7" 615-6J1.

Aki, K., 19S0a. Attenuation ~of shear: wav.es in. the lithosphere for frequencies from 0:05 to 25Hz, Ph.ys: Eizrth planet. Jnieriors n, .5Q-;60.

Aki, K .• 1980b. Scattering- and attenuation of shear wave" in the lithosphere, J. geophys, Res:, 8-5, 6'49~6504.

Aki, K. & Chouet, B.; 1975_ Origin of coda waves: _SOUTCe, attenuation and scattering- 'effects, 'J' geophys. Res., l!}O, 332.2:_3342_

516 W. B. A-mbeh and L. L. Lynch

Ambeh, W. B .. <ll Fairhead, J. D., 1989. Coda Q estimates in the Mount Cameroon volcanic region, West Africa, Bull, seism. Soc, Am., 79,.1589-16"00.

Canas, J. A" Pujades, L, Badal, .J., Payo, G_, Fernando de Miguel. Vidal, P., Alguaeil, G., Iban_e,ztJ.~ Morales, 'L, 1991. Lateral variation-and frequency dependence ef cgda·Q ill the southern. pru:t of Thtiria, Geophys. J. Int., 107, 57-&:..

Cheng; C. C. & Mi;!-cheU, B. J.; 1981. Crustal Q structure In the United States from rnulti-mode.surtace waves, BuTl. seism. .. Soc. Am., 71,161-L&1.

Cormier, V. F." 1982. The effect of attenuationon seismic body waves, Bull. seism. Soc. Am., 72, 8169-20(;:

Dainry, A. M. 198L 'A scattering model to. explain seismic Q observations in me lithosphere ~e-tyveeD 1 and 30 Bz, Gegphys .. Res. Len .. 8, 112fi-1128.

Del "Pezz{),,;E., De. Natale, G., Scarcella, G~ & Zollo, /\_., 1985. Qc . of three-component seismograms af: volcanic microearthquakes at Campi Flegrei voleanic area, southern Italy, Pageoph, 12-3, 6&3-696.

Frankel, A. & Clayton, R. W.o 1986. Finire-dllference simulations of seismic seattering: Implications for tile, propagation of short .. period.vsesirnlc waves and models of crustal heterogeDoily, J. geophys_ He.1'., !lJ,. 64(i5-6489.

Frankel, A. & Wennerberg, L .• 1987, ~n~rg_y-B~ model of seisrme coda: separation ef scattering and in trinsic -attenuation, Bull. seism. Soc. A.rn., 77 1223-125,1.

Gao, L. S., Lee. L. C,' Biswas, N, N. & Aki, K, 1983a.

Comparison of the effects between single and multiple scattering on coda waves for earthquakes, 13f!.n seism. Soc. Am., 73 377-389,

Gao, L.$ -, , Biswas, N. N., Lee, L. C. & Aki.K., 1983b. Effects of multiple scattering oil coda waves in rlin:~e;d.irriensional

medium. p_(/,geoph, m,3-15. .

Havskov, J" Malone, S., McCIUT-g, D. & Crosson, R, 1989. Coda Q (or the state of Wllshington, Bull .\dsm.$oc: Am., 7·9J 1024-.1018.

H¢rr~lz, M. & Espinosa, A. F., 19B7, Coda waves: a review, Pageoph, US, 499-577.

Herrmann, R. B., 1980. Q estimates using the ccda of Iecal earthquakes. Burl seism. §6C. Am., 70, 447-468.

Hough, 8. 8., Anderson, J. G,. Brune, J" Vernon, F., Berger, J., .Flercher, J., 1ia3T, L. Hanks, T. & BakeT~ L" -1988. Auenuation near Anza, California. BIII(. sei,lrif. Soc. Am., 78~ 6n-6~1,

Jin, A. &- A-ki, K, 19&6. Temporal change in coda Q befoJe tb~ Tangshan earthquake ofl976· and lhe-Hl).icheljg earthquake of 1915, J. geophys. Res" 91, _665-613.

TIn. A. &' Aki, K., 11188. Spatial and tempera] correlation between coda Q and seismicity in: China, Bull. seism. Soc. Am" '78, 741--769.

Jin, A .• Cao, T. ~ Aki, K.. L98S. Regional change. incco.da Q ill the oc~iclitbospherc, J. geophys.Re.s., 9.0, 865i-8659.

Kvamme, L. B. & Havskov, J., 1989. Q ill southern. Norway, Bull. seism. Soc. Am.. 79, 157.5-1588.

Langston, C. A., 1999. Scattering of long-Pl:.riol:! Rayleigh waves in Western North, Arn~rica; and the interpretation. of coda Q measurements, Bull. seism. Soc. Am" 79, 774-78"9.

Lee, W. R. K_, Aki, K. Chouer, B., Johnson, P., Marks, S., Newberry 1. T., Ryall,A. S., Sfewart~.S. W. & Tottiagham D. M., 1986. A preliminary study qt oorla-Q in California and Nevada, Bull. Sillsm. Soc. Am., 76, 1133-l_14:l.

Levander, A. R. & Hill, N. R., 1985. P-SV resonances in irregular

low-velocity surfacelayers. Bull . seism. Soc. Am., 15,847-864.

Mitchell, B. J., 1981'), Frequency dependence of she_ar wave internal friction in the continentalcrusr of eastern North America, I. ,C'l.epliy.r" Res., 85, 5212--S218~ "

Milchell, B. L, 1981. Regional variation and frequency dependence oilQ/3 in rhe crust ol.tbe United States, Bull. s~r.rm. Soc. Am., 71,1531-153"8.

Novelo-Casaneva, D. A., Berg, E. & Helsley, C. E., i990. S;wave coda Q from -3 to 20 HZ and P-wave Q Tar foreshocks and aftershocks of tile Petatlan earthquake, 1. geophY.r. Rl's.95, <t787-4795.

Pulli, J _ J. 15184. Attenuation ef coda waves in New England, Bull. seism. Soc. Am., 174, 1I4-9'-1166.

l'ulli, J_ J. lfi,_ Aki, 1(:., 1981. Attennation o{ .seismic wave's ill the litlii:lspbere: eonrparison of active and ~t.uble~reas, In F;_artJiqudldii and. earthquake. ellik~ring: The ElfStern United Slates, pp. 121-141, Deavers, 'J, E., ed. Ann Arbor SCience Publishers Inc., Ann Arbtlr,Micliigan.

Rautian, T. G. & Khalturin, V. t., 1987., Tile use of the coda [OI the determination of th~ earthquake source 'spectrum. Bull, seism. Soc. Am." 68, 923-94;3.

Rebollar, c. J., Transioshetos, C. & Alvarezv Rc, 19&.,'i. Es hnates ofseismic wave .attenuation in northern .Baja California, Bull. seism. See. Am., 75, i37i-i382.

~fhards,~P. G'. & Menke, W., 19:83. The apparent auenuatian of a scartering-medium, Bull. seism. Soc. Am., 73,1005-102.2.

Roecker, S. W., Tucker, B., King, J. & Hatzfeld, D .• 1'9B2_ E~ti.mates of Q in Central, Asia as a function of fre_gu~nc_y and depth ulliQg the ctida of locally recoIAe_d earthquakes, Bull.

seism. So£: Am., 72, 129-149. .

Rovelli, A" 19114. Seismic Q for the-lithosphere of the Montenegro region (Yugoslavia): frequency, depth and ;time windowing effects, Phys~ Earlh planet. ill:teri(Jl's, 34, -,159-172.

Sate, H. 1977. Energy propagation including scattering effects.

Shigledso:tr6"plc, scattering approximatiou, ]. PhY:j. EartJi, 1$, 27-41-

Seherbaum, F~ &. Kisslinger, C., 1985. COdaQ in the Adak seismic - zone, BIUt. seism. Soc. Am., 75. 615;.620.

Scherbaurn, r. & Kisslinger, C.; -1987. Correction to "Cocta Q in the Adak S'eismlC zone", Bull. Seism. S:oi;. Am .• 77, 22.46.

Singh, S. x., Apsel, ~. J., Fried, J. & Brune, J. N·., 1982. Spectral 'nttcnuauon of sa wavesalong the Impenal fault, Bull. seism .

Soc. Am., 75. 200~-2016. .

Sm-gIl, -S . .K .. & Herrmann, R.B., 1983. Regioualizatien of: '-T.jl.$taI coda Q in the eontinental United States, J.geophrys. Res., 88, 527--:538.

SBudichl'. ~Bostwick, T., 1987. Studies of. the seismie coda using eanhq,uak€l clusters as, deeply burned sei~og~apti 'arrays. 1, gli(}-pILY~. Ri!s_ .• 92, 10 526-.lQ 546, -.

Sutton, G. u., Mitrcnovas, W. & l'o1llCrOy~P. W., 1967. Short pc'riod seismic energy radiation patterns from umfergrQund nuclear explosions and small magnitude earthquakes, Brill. seism. Soc, Am., 57, 24§---1.67·.

Thouvenot, F., 1983. Frequency dependence of the quality fader itt the uppercrust: a deep seismic-sounding appEQacli, Geophys. I.

R-. astr .. }roc., 73,-427'-447. -

Van Eck. 'T., 19~{1. Attenuation of coda waveS in the Dead Region,

Bull. seism. Soc. Am., 18, 77Q-179. .

Wu, R. S.) 19~5. Mtiltipie scattering and ,energy transfer of seismic waves-separation -qJ scattering effect nom intrinsic ,*.enlla, non. LThe(;JI:etical modelling. Gi!opiry;s. J. R. astr. Soe, ,82, 57--80.