seismic tomography constraints on the deep structure of...

647

Izvestiya, Physics of the Solid Earth, Vol. 40, No. 8, 2004, pp. 647–659. Translated from Fizika Zemli, No. 8, 2004, pp. 41–55.Original Russian Text Copyright © 2004 by Dorbath, Arefiev, Rogozhin.English Translation Copyright © 2004 by

MAIK “Nauka

/Interperiodica” (Russia).

INTRODUCTION

The Spitak earthquake of 1988 is one of the mostcomprehensively studied earthquakes on the territory ofthe former USSR. Epicentral in situ observations wereconducted by seismological groups from several repub-lics and countries. In cooperation with the EpicentralExpedition of the Institute of Physics of the Earth, theseismologically best equipped research group fromStrasbourg University installed about 40 temporaryseismic stations in the epicentral zone. These stationshave yielded thousands of aftershock records, whoseprocessing provided a basis for an earthquake catalog.In the early 1990s, these data were used for the study ofthe source zone by seismic tomography methods exist-ing at that time [Dorbath

et al.

, 1994; Treusov

et al.

,1993]. Solving the tomography problem by a modifiedSIRT method [Sluis and Vorst, 1987; Trampert andLeveque, 1990], Dorbath

et al.

[1994] gained con-straints on the 3-D velocity structure of the Spitaksource zone, which showed that (a) low velocity zonescorrelate with sedimentary basins; (b) the majority ofaftershocks are localized in high velocity zones, whichcan be attributed to outcrops of more rigid deep rocks;and (c) the

b

value is larger in higher velocity zones. Weshould also note that the surface rupture of the obliquefault is not observable in maps of velocity anomalies.Applying a different approach of tomography research,Treusov

et al.

[1993] selected 85 of all available earth-quakes (about 3000), and only these events, recordedby at least 12 network stations, were processed. Theresulting velocity distribution showed that the depthinterval of 0–8 km is best constrained by experimentaldata. Two upper slices have the following common fea-ture: velocities first decrease and then rise in the direc-tion from northwest to southeast, whereas, in the direc-

tion from southwest to northeast, they first rise and thendecrease.

Thus, the previous results of seismic tomographyresearch have only revealed that the general pattern iscomplex (mosaic); moreover, the inferred velocity fieldhad no signatures of fault zones and, more importantly,of the surface rupture of the Spitak earthquake. Furtherdevelopment of tomographic methods, as well as theaccumulation and analysis of additional geological andgeophysical information, enabled a more adequateapproach to the description and interpretation of thedeep structure of the Spitak earthquake source zone.

GEOLOGICAL AND GEOPHYSICAL DATA

The Spitak earthquake of December 7, 1988, had themagnitude

M

s

= 6.9, and its MSK-64 intensity at theepicenter reached 10. The highest intensity (9 and 10)contours are ellipsoidal and trend in the WNW direc-tion consistent with the general structural strike in thispart of the Lesser Caucasus [Shebalin, 1991]. The faultof the main shock extended to a depth of 13–14 km andreached the Earth’s surface in the form of an elongatedzone of coseismic ruptures (Fig. 1). The earthquake waspreceded by a weak foreshock (

M

~ 3

) and was fol-lowed by numerous aftershocks [Arefiev

et al.

, 1991;Kondorskaya

et al.

, 1991].

The whole epicentral area of the Spitak earthquakeis located on the western segment of the Sevan–Akeraand Miskhan–Zangezurskii structural–lithologic zonesof the Lesser Caucasus. Structurally, the first zone is apart of the large Sevan synclinorium and Bazumskiianticlinorium bounding the synclinorium to the north.To the west, the Sevan synclinorium extends into a vastarea of volcanic plateaus, separating the Akhalkalaki(Dzhavakhet) and Armenian plateaus, and is overlain

Seismic Tomography Constraints on the Deep Structureof the Spitak Earthquake Source Zone

K. Dorbath*, S. S. Arefiev**, and E. A. Rogozhin**

* Institut

de Recherche

pour le Développement, École et Observatoire des Sciences de la Terre, Strasbourg, France** Schmidt United Institute of Physics of the Earth, Russian Academy of Sciences, Bol’shaya Gruzinskaya ul. 10, Moscow,

123995 Russia

Received March 4, 2003

Abstract

—The Spitak earthquake of 1988, one of the most comprehensively studied earthquakes on the terri-tory of the former USSR, was examined by an epicentral expedition that gathered unique data on both the sur-face rupture produced by the earthquake and its aftershocks. Based on these data, the local seismic tomographymethod was used for constructing a 3-D velocity model and refining the coordinates of aftershock hypocenters.Analysis of these results revealed the presence of lower velocity bands confined to active faults in the Spitaksource zone. On the whole, the study region exhibits a high degree of velocity inhomogeneity.

648

IZVESTIYA, PHYSICS OF THE SOLID EARTH

Vol. 40

No. 8

2004

DORBATH

et al

.

Fig

. 1.

Str

uctu

ral–

geol

ogic

map

(a) a

nd c

ross

sec

tion

(b) o

f the

sou

rce

zone

[Rog

ozhi

n an

d Fi

lipp,

199

1]: (

1

–

5

) the

Qua

tern

ary:

(

1

,

2

) Hol

ocen

e an

d Pl

eist

ocen

e–H

oloc

ene

allu

vium

,pr

oluv

ium

–del

uviu

m, d

eluv

ium

–col

luvi

um, a

nd tu

ffac

eous

hor

izon

s, (

3

,

4

) L

ower

–Mid

dle

Qua

tern

ary

lava

s an

d tu

ffs,

(

5

) an

omal

ous

terr

aces

upl

ifte

d ab

ove

the

flood

plai

n of

the

Pam

bak

Riv

er; (

6

) st

rike

s an

d di

ps o

f M

esoz

oic

and

Cen

ozoi

c (p

re-Q

uate

rnar

y) b

eds;

(

7

) ax

es o

f an

ticlin

es; (

8

) ax

es o

f sy

nclin

es; (

9

) st

ratig

raph

ic c

onta

cts;

(

10

) ba

salti

c ho

rizo

nat

the

Eoc

ene

base

; (

11

) fa

ults

; (

12

–

18

) va

riou

s ig

neou

s ro

cks:

(

12

) gr

anite

–gra

nodi

orite

s, (

13

) al

kali

rock

s, (

14

) si

lls o

f ga

bbro

porp

hyri

tes

and

gabb

ro–d

iori

tes,

(

15

) Pl

ioce

neba

salts

, (

16

) ser

pent

iniz

ed u

ltrab

asic

rock

s, (

17

) sub

volc

anic

intr

usio

ns a

nd d

ikes

of a

lbito

phyr

es, (

18

) nec

k fa

cies

of v

olca

noes

; (

19

) Pro

tero

zoic

and

Low

er P

aleo

zoic

met

amor

phic

sequ

ence

s; (

20

–

23

) co

seis

mic

def

orm

atio

ns o

f th

e Sp

itak

eart

hqua

ke: (

20

) m

ajor

rig

ht-l

ater

al s

trik

e-sl

ip–r

ever

sed

faul

ts, (

21

) sh

ears

, (

22

) sh

akin

g fr

actu

res,

(

23

) la

ndsl

ides

, (

24

)cr

oss

sect

ion

line.

❙

❙

▲▲

▲▲

▲

▲▲

Gek

hasa

rG

ekha

sar

Gek

hasa

rN

alba

ndN

alba

ndN

alba

ndU

pper

Pam

bak

Upp

er P

amba

kU

pper

Pam

bak

basi

nba

sin

basi

n

▲▲

▲▲

▲

▲▲

▲▲

▲▲

▲▲

▲

Len

inak

anL

enin

akan

Len

inak

an

▲▲

▲▲

▲▲

Spita

kSp

itak

Spita

kK

irov

akan

Kir

ovak

anK

irov

akan

PPPaaammm

bbbaaakkk

RRRiiivvveee

rrr

▲ ▲ ▲

▲

▲▲

▲

▲▲

▲

▲

▲▲

▲

▲

❙❙

❙

❙❙

❙❙

❙

❙

❙

❙

❙❙

❙

❙

❙❙

❙

❙

❙

❙❙

❙

❙

❙

❙

❙

❙

❙

❙

❙

❙

❙

❙❙

❙

❙

❙❙

❙❙

❙

❙❙

❙

❙❙

❙

❙

❙

❙❙❙

❙

❙❙ ❙

❙

❙❙

❙

❙❙

❙❙

❙

❙

❙❙

❙❙

❙

❙

❙

❙

❙

❙

❙

❙

❙

❙

❙

❙❙

❙

▲▲

❙❙

❙

❙

❙

❙

❙

❙❙

❙

❙❙

❙

❙

❙

❙

❙

❙

❙

❙

❙

❙

❙

❙

❙

❙

❙

❙

❙

❙

❙

❙❙

❙

❙

❙

❙

❙

❙

❙❙

❙

❙❙

❙❙

❙❙

❙

❙

❙

❙❙

❙

❙

❙

❙❙

❙❙❙

❙

❙

❙

❙

❙

❙

❙

❙❙

❙

❙

❙

❙

❙

❙❙

❙

❙

❙

❙

❙

❙❙❙

❙

❙

❙

❙

❙

❙

❙

❙

❙

❙

❙

❙

❙

❙ ❙

❙❙

❙

❙

❙

❙

❙❙

❙

❙

❙

❙

❙

❙

❙ ❙

❙

❙❙

❙

❙

❙

❙

❙

❙

❙

❙

❙

❙

❙

❙

❙

❙ ❙

❙

❙ ❙

❙❙

❙

❙

❙

❙

❙❙❙

❙❙

CCChhhiii

ccchhhkkkhhh

aaannnRRR

iiivvveeerrr

PPPaaammm

bbbaaakkk

RRRiiivvv

eeerrr

SSSpppiiitttaaakkkRRRiiivvveeerrr

(a)

(b)

km 2 1 0

N

1 2 3 4 5 6 7 8 9 10 11 12

13 14 15 16 17 18 19 20 21 22 23 24

20

2 km

Q

4

Q

3-4

Q

2

Q

1-2

β

N

23

K

1

P

22

–P

22

–

K

1

K

1

K

1

K

2

K

2

ν

bP

22

–N

1-2

N

1-2

Q

4

Q

4

P

22

–P

22

–

Q

1-2

K

2

J

3

–K

1

Q

1-2

N

1-2

pC–

P

Z1

–-

P

22

–

P

22

–

β

N

23

β

N

23

Q

2

Q

2

Q

4

Q

1-2

Q

1-2

Q

4

Q

1-2

Q

1-2

Q

1-2

J

3

–K

1

K

2

P

22

–

P

22

–

P

22

–

N

2

N

2

N

2

N

2

N

2

N

1-2

N

1-2

ν

bP

22

–

P

22

–Q

3-4

Q

3-4

Q

3-4

K

2

K

1

K

2

J

3

–K

1

J

γ

J

2-3

K

2

K

2

K

2

K

2

J

3

–K

1

K

2

P

22

–

P

22

–

P

22

–

P

22

–P

2-3

–

P

22

–

P

2-3

–

N

P

22

–

γ

bP

22

–

ν

bP

22

–P

22

–K

2

Q

3-4

P

2-3

–

λ

P

22

–

P

2-3

–

P

22

–

P

22

–

P

31

–

Q

3-4

P

31

–

K

1

P

22

– P

23

– P

22

–

γbP 2

-3–

K1K

2

γP2-

3–

P 22

–

P 22

–

P 3–

P 22

–

K2

K2

γK1

K2

K2

νbP 2

2–

K2

P 22

–

P 22

–

K2

P 22

–K1

K1

K1

K1

P 22

–

P 3–

P 22

–

P 2–

K2

K2

P 22

–

P 22

–P 2

2–

Q4

K

βN23

βN23

P 22

–

IZVESTIYA, PHYSICS OF THE SOLID EARTH Vol. 40 No. 8 2004

SEISMIC TOMOGRAPHY CONSTRAINTS ON THE DEEP STRUCTURE 649

by young lavas of the Kars plateau on the territory ofTurkey. To the east-southeast, the synclinorium extendsinto the northern shore of Lake Sevan and joins theSevan ophiolitic belt [Milanovsky and Khain, 1963;Adamiya et al., 1989]. In the south, the Sevan–Akerazone is adjacent to the Miskhan crystalline inlier of thepre-Alpine Miskhan–Zangezurskii basement [Mil-anovsky and Khain, 1963].

The Sevan–Akera zone and the Miskhan inlier differbasically in geological structure. The former is charac-terized by narrow compressed linear folding inEocene–Oligocene volcaniclastic rocks with the “Cau-casian” strike of fold axes and numerous fold-relatedfaults, whereas the Miskhan inlier is distinguished by anonlinear brachyomorphic shape of folds in UpperJurassic, Cretaceous flyschoid, and Paleogene–Neo-gene volcaniclastic rocks (Fig. 1). Faults in the inlierhave different strikes and form a broken-plate structurerather than a system of linear parallel fault sets.

The core of the Sevan synclinorium is complicatedby an en echelon chain of young (Holocene) superim-posed Pambak basins overlying the main tectonicsuture of the Lesser Caucasus, namely, the Pambak–Sevan deep fault [Milanovsky, 1968; Bal’yan et al.,1989]. Lower–Middle Quaternary lava flows of youngvolcanoes (in particular, Aragats Volcano) are devel-oped on the territory of the northern basement ridge inthe Miskhan–Zangezurskii median mass.

The deep structure of this part of the Lesser Cauca-sus is known from results obtained on a deep seismicsounding and reflection profile 14–16 km west of theepicentral zone of the Spitak earthquake [Shchukinet al., 1998]. This regional profile of a NW orientationcrosses the majority of Lesser Caucasus tectonic zones,including the Akhalkalaki (Dzhavakhet) plateau, super-imposed on structures of the Somkhit–Karabakhmedian mass, Sevan–Akera structural–lithologic zone,and Miskhan–Zangezurskii median mass. The profileextends from the town of Akhaltsikhe (in the northwest)to the town of Vedi (in the southeast). The main featureof the resulting velocity structure is the presence ofsmall crustal blocks in this part of the Caucasian mobilezone, as well as a distinct reflectivity and clearly recog-nizable long horizons at various depths.

The upper part of the crust (to a depth of 15 km)including the Spitak source consists of two layers. Theupper layer is characterized by a sharp increase in the Pwave velocity with depth (from 3.5–4 to 6 km/s) and byweak lateral variations in its thickness (3–5 km). Thislayer is identified with the Mesozoic–Cenozoicsequence of volcaniclastic rocks.

The next, deeper layer occupies the whole upperpart of the solid crust and is characterized by a mosaicstructure of its seismic velocity distribution and by aninsignificant increase in the velocity with depth. Thenorthwestern, central, and southeastern segments of theprofile yield sharply contrasting distributions of veloc-ity parameters in this layer. Lower velocity zones are

widely developed in the northwestern and southeasternparts of the layer at depths of 5–15 km. Data from thecentral profile segment, crossing the source area, indi-cate a mosaic, finely fragmented structure character-ized by the alternation of thin discontinuous layers oflower and higher velocities. According to these data,the Spitak earthquake hypocenter lies in an anomalouszone that has velocities lower than in the surroundingcrustal blocks by 0.2 km/s and is controlled by theSevan–Amasiiskaya (Sevan–Pambak) fault zone. Ahigh-velocity body (VP = 6.8–7.0 km/s) revealed atdepths of 13–14 km in the source zone coincides withthe root zone of the Sevan ophiolitic complex, whoserocks penetrate upper crust layers in the form of fault-line protrusions. This anomaly is distinguished by thefact that it is underlain by a lower velocity layer (VP =6.5–6.6 km/s), thereby producing inversion relationsbetween the middle and upper crust.

The source zone of the earthquake is also character-ized by a higher degree of stratification and fragmenta-tion of the upper lithosphere material. Refractors andreflectors are laterally inconsistent, oblique, and oftencomplicated with faults, which indicates that themedium in the source zone is more fragmented com-pared to adjacent crustal blocks [Shchukin et al., 1998].

According to the distribution of velocity character-istics, the lower parts of the crust can be united into ageneral lower layer. Main strata of the latter are thickand laterally consistent, with their seismic velocitiesvarying rather moderately (within 7.0–7.2 km/s). A vastbody of lower velocities (VP = 6.6–6.8 km/s) is presentin the lower crust of the Sevan–Akera fold zone. Thisbody, having a thickness of 7–10 km and an along-pro-file length of about 25 km, completely thins out north-west and southeast of the Sevan–Akera zone, in thelower crust beneath the adjacent median masses. It hasvelocity inversion relations with respect to overlyinglayers of the lower crust. In particular, it underlies theearthquake source zone.

Coseismic deformations of the Spitak earthquakeinclude a number of phenomena in the near-field zone,such as an Earth’s surface uplift in an area 50 km in sizecentered at the town of Spitak (according to data of therepeated leveling surveys conducted by the MainDepartment of Geodesy and Cartography along theLeninakan–Spitak–Kirovakan railroad and along theSpitak–Yerevan highway) [Milanovsky, 1968; Karakh-anyan, 1989]; disturbances in the Earth’s surface incon-sistent with the topography (primary coseismic rup-tures); fractures of shaking and subsidence of upperslopes; and seismogravitational phenomena such aslandslides, downfalls both descended and incipient(with insignificant downslope movements), taluses,subsidences in loose alluvium and artificial grounds,and so on (Fig. 1). The overall length of the zone oftraceable fractures amounts to about 35 km. Three long(8–9 km each) segments of intense primary coseismic

650


DORBATH et al.

deformations, separated by undeformed areas, are rec-ognizable.

The first, southeastern segment embraces the north-ern slope of the Pambak Ridge between the village ofAlavar and the Spitak River valley. Right-lateral (witha possible reversed component) and left-lateral faultsform an en echelon system. Each individual fault is500–1000 km long. The amplitude of the reversed faultmotion (with upthrown northeastern sides) is highest inthe central part of the segment (1.5 m) and graduallydecreases to zero toward its ends. The same variationpattern is characteristic of the strike slip amplitudes(from 0.7 to 0). Four isolated en echelon systems ofsmall fissures discovered at the northwestern end of thissegment of the coseismic deformation field are pro-duced by left-lateral slips of small amplitudes (a fewcentimeters). These systems are oriented N–S, and theirjoints strike SE. A distinctive feature of the first seg-ment (generally striking NW) of the coseismic defor-mation field is the presence of strike-slip faults produc-ing en echelon fissures with heave pressure ridges per-pendicular to the latter.

Coseismic deformations are most pronounced in thecentral segment, which extends from the town of Spitakto the village of Gekhasar and includes reversed–strike-slip motions of NW and E–W strikes. The vertical slipcomponent can reach 1.5–2.0 m, with northeastern (ornorthern) sides being upthrown. Maximum right-lateralslips reach 1.5–1.8 m. Here, coseismic deformationsmainly belong to a general seismogenic fault of a fes-tooned shape in plan view that separates in some placesinto two or three subparallel branches. The fault planedip varies from vertical to very gentle (10°–15°).Trench investigations showed that shallow portions ofseismogenic faults in this segment of the deformationfield vary in structure, but they are undoubtedly associ-ated with an older fault with large amplitudes (up to9 m) of Holocene reversed motions. The presence oftwo fragmentary soil horizons buried under paleosolcolluvium in the section of the downthrown side of theseismogenic fault is evidence for an impulsive charac-ter of these motions [Rogozhin and Filipp, 1991;Albarede et al., 1980].

Two samples were taken from buried soils of thedownthrown wall of the fault for radiocarbon age deter-minations. Their respective ages were estimated at17 565 + 770 and 19 960 + 225 years [Albarede et al.,1980]. A small humus inclusion was discovered in thecolluvial apron in the upthrown side. Its analysisyielded an age of 24 765 + 770 years. Based on thepaleosol age determinations (17–20 ka) and the forma-tion time of the full profile of the contemporary soil (5–7 ka in northern Armenia), one may state that an earth-quake occurred here at the Upper Pleistocene/Holoceneboundary. The rupture of this earthquake reached theEarth’s surface, and the northern wall of the fault wasupthrown. The vertical motion amplitude associatedwith this ancient event amounted to more than 60 cm;

i.e., the event was apparently stronger than the catastro-phe of 1988.

The fact that the downthrawn wall of the fault con-tains fragments of a deeper paleosol horizon buriedunder the colluvial apron of an absolute age of 24–25 kasuggests that at least one earthquake of the same mag-nitude took place here at an earlier time [Rogozhin andFilipp, 1991]. Paleoseismogeological studies haveshown that the Spitak source is a stable crustal structureand was activated in the Late Pleistocene and theHolocene, which allows one to estimate the recurrenceperiod of the strongest earthquakes within this source.

The northwestern segment of the coseismic defor-mation system is characterized by the presence of short(a few tens of meters) no-slip fractures dissectingmountain tops and divide crests of the Chichkhan andPambak rivers. Apparently, these surface deformationsarose due to strong shaking motions. Also noticeableare two primary seismogenic faults (right-lateral–reversed motions) up to 200 m long, with amplitudes ofvertical and horizontal displacements of 20−30 cm.

In addition to the deformations described above,intensity-8 contours [Karakhanyan, 1989] within theareas of all three segments include numerous seismo-gravitational disturbances (landslides, rockfalls, sub-sidences, and so on). The largest (1.0 × 0.5 km) land-slides are developed in the area of the village of Dzo-rashen on the right-hand side of the Chichkhan River.

The near-field area of the earthquake includes alarge disjunctive node formed by the intersection of thePambak-Sevan (E–W) and Aragats-Spitak (N–S) deepfaults. The latter is one of the major tectonic lineamentsforming the Trans-Caucasian transverse uplift [Mil-anovsky and Khain, 1963; Albarede et al., 1980]. Thediagonal (SE trending) Alavar (Garniiskii) fault alsojoins the node. All three structural trends variouslyaffected the orientations of the primary coseismicdeformations. The diagonal right-lateral strike slipswith possible reversed components of the southeasternsegment are associated with the Alavar fault, and the E–Wsegment of coseismic fractures west of Spitak wasapparently controlled by the motion on a fault of thePambak–Sevan zone. A small dextral slip following theearthquake was recorded on the Pambak-Sevan activefault in the Chichkhan River valley [Trifonov et al.,1990]. Short N–W series of sinistral slips on the north-western slope of the Spitak River valley can beregarded as coseismic motions on faults of the Aragats-Spitak lineament. Two of these structural directions(the Pambak-Sevan and Alavar trends) have their signa-tures in the configuration of the aftershock epicentralfield (see Fig. 5 below) [Arefiev et al., 1991]. Mostlikely, the Spitak earthquake source was confined to thejunction zone of these two fault zones (of E–W anddiagonal strikes) under E–W compression conditions.This is decisively indicated by the results derived fromstudies of coseismic deformations, the aftershock activ-ity, and the focal mechanism of the main shock



[Rogozhin and Filipp, 1991; Arefiev et al., 1991; Kon-dorskaya et al., 1991].

METHOD OF LOCAL EARTHQUAKE TOMOGRAPHY

Seismic tomography is a method for the reconstruc-tion of internal structures of the Earth from a large setof observations on the periphery of the object studied.Traveltimes of seismic rays rather densely covering thisvolume are used for the localization of velocity anoma-lies in space and determination of their amplitudes. Onthe whole, seismic tomography methods can be classi-fied in accordance with the type and distribution ofsources and receivers.

Sources used in the local earthquake tomography(LET) are natural earthquakes that occur inside theregion studied. Their arrival times are recorded by afairly uniform network of seismic stations covering thewhole region. As compared with controllable sources(e.g., explosions), local earthquakes are advantageousin that they excite both longitudinal and transversewaves and are distributed in a 3-D volume. However,the spatial distribution of local earthquakes is not wellconstrained and their coordinates (primarily, depths)are poorly known. As compared with the teleseismictomography, LET usually provides a better structuralresolution due to a higher density of seismic rays,higher wave frequencies, and smaller epicentral dis-tances. On the other hand, the maximum depth of hypo-centers limits the size of possible models.

If any finite parametrization of the velocity structureis acceptable, all LET methods are based on the equa-tion

where rij is the residual, i.e., the difference between theobserved and predicted traveltime from the ith earth-quake to the jth station; xk are source coordinates; τi isthe earthquake time; and ml are L parameters of thevelocity model (δTij/δxk and δTij/δml are the partialderivatives with respect to the hypocentral coordinatesand velocity model parameters, respectively). The LETmethods deal with various aspects of the problem. Inthis work, we apply the approach developed by Thurber[1983] and briefly described below.

The velocity structure is represented in terms of avariant of the discrete block parametrization [Aki andLee, 1976]. The model velocity, specified on a 3-D grid,is continuous in space with the use of a linear B-splineinterpolation between nodes. The nodes are generallymore closely spaced in the central part of the objectstudied, with their density being lower on its periphery.Nodal planes should also be placed at infinity on allsides of the model (i.e., in such a way that all rays liewithin the model).

rij

δTij

δxk

---------∆xk ∆τi

δTij

δml

---------∆ml,l 1=

L

∑+ +k 1=

3

∑=

The chosen 3-D structure predetermined the methodmost suitable for the calculation of ray paths and trav-eltimes that are necessary for finding time residuals andpartial derivatives from hypocentral coordinates andvelocity model parameters. Methods for the approxi-mate representation of ray paths by segments are oftenused to decrease the computation time. In particular, themethod proposed by Thurber [1983, 1993] admits acurvature of a ray path, making it similar to the true ray.

Analysis of the relationship between parameters ofhypocenters and the velocity model also reveals certaininconsistencies. The simultaneous inversion using theabove method takes into account this relationship andinvolves the inversion of the complete matrix of theequation. To decrease the dimension of this matrix[Thurber, 1983, 1993], Pavlis and Booker [1980] pro-posed the construction of a system of equations incor-porating seismic velocity parameters alone, withoutregard for the formal mathematical relation betweenhypocentral and velocity parameters.

For some LET problems, the direct solution of theseequations can be impossible due to a large dimension oftheir matrix. An alternative method proposed byThurber [1983, 1993] is combination of resolutionparameters with the construction of normal equationsand the use of a damping parameter. In this case, thematrix dimension is determined by the number ofvelocity model parameters and does not depend on thenumber of earthquakes involved in the inversion. Thesolution is very sensitive to the choice of the dampingparameter. To choose a reasonable value of the latter,Eberhart-Phillips [1986] proposed the construction oftradeoff curves of the current velocity model resolutionas a function of the error of the model for one-stepinversions at various values of the damping parameter.

The joint interpretation of P and S velocity struc-tures provides the best constraints on mechanical prop-erties and geological identification of rocks. In practice,S wave observations are less numerous and inferior inquality as compared with P observations. Therefore,the S wave inversion has a lower resolution and doesnot allow one to analyze the VP/VS ratio; in this connec-tion, Thurber [1993] developed a method for convertingthe S–P traveltime differences into 3-D VP/VS varia-tions.

TOMOGRAPHIC INVERSION RESULTS

The inversion was performed with the help of theSIMULPS12 program [Evans et al., 1994], using as ini-tial data the traveltimes TP, traveltime differences S – P,and ratios VP/VS. Initial data were obtained from epi-central observations in the Spitak earthquake zone. Thenetwork included 30 stations operating from December22, 1988, to February 26, 1989 [Arefiev et al., 1991].Initial coordinates of hypocenters were determinedwith the Hypoinverse program [Klein, 1978]. From theentire dataset (about 3000 events), we selected events

652


DORBATH et al.

for which no fewer than 10 phases were available,including at least two S phases, and the rms accuracy ofthe hypocentral solution was no worse than 0.25. In all,1205 events met this criterion, providing more than15 000 P and 7300 S traveltimes.

In accordance with the LET method, the studyregion was parametrized via a 3-D grid (with a step of4 km); its projection onto the Earth’s surface is shownin Fig. 2. Boundary nodes are not shown in the figurebecause they lie far beyond the region studied. Alsoshown are the seismic stations in use and, schemati-cally, active faults mapped in Fig. 1. Note that thechoice of the grid covering the study region is largelyarbitrary and was done by optimizing the relationbetween the resolution attainable with the available ini-tial data and a reasonable accuracy, on the one hand,and the maximum detailedness of the velocity struc-ture, on the other hand. The following depth levels werechosen: 0.0, 2.0, 4.0, 6.0, 8.0, 12.0, and 20.0 km. Weshould note that not all of these levels were constrainedby a sufficient amount of data. For this reason, the res-olution parameter was mapped for each layer. Based onthese maps (not presented here), velocity values on agray scale are shown only in those areas of the studyregion where the resolution was no worse than 0.1.However, velocity contours were constructed for theentire region. Note that the resolution is 0.5 and morein central parts of the model, where the data coverage isdensest. The uncertainty of the model averaged over allnodes of the 3-D model amounts to 0.05 km/s. Theinversion was accomplished through six iterations.

The initial velocity structure, based on the modelused for the calculation of the main Spitak earthquakecatalog [Arefiev et al., 1991], has the following param-eters:

Layer top, km –5 0 2 4 6 8 12 20 60

VP, km/s 4.7 5.0 5.3 5.8 6.1 6.3 6.5 6.9 8.2

An important regularization parameter in themethod of Thurber [1983] is the damping. We selecteda value of 20, using a method similar to that proposedby Eberhart-Phillips [1986]. The resulting P wavevelocity model is presented in Figs. 3 and 4.

Analysis of the distribution of seismic velocitiesindicates their gradual, albeit irregular, increase withdepth. Areas of relatively high and relatively low veloc-ities are present at the same depth levels. Also notewor-thy is a linear–mosaic pattern of velocity heterogene-ities at each of the depth levels considered. Thus, thesurface layer (0 km, Fig. 3) reveals the predominance ofcomparatively low velocity structures in the upper partof the volcaniclastic layer. Velocities of P wavesamount to 4–4.5 km/s, and the wavefield is evidentlystructured. A N–S trending linear band 5–10 km widecharacterized by higher values of VP (5.0–5.5 km/s) isrecognizable in the central part of the source region.The remaining part of the source region is mostly char-acterized by relatively low values of VP (4–5 km/s). Anarrow (about 10 km in width) N–S trending band oflower velocities (4–5 km/s) nearly coincides with thetown of Spitak meridian. Velocity contours in thisrange are distinctly traceable from north to south.Therefore, this shallow layer clearly reflects struc-tures consistent with the Trans-Caucasian transversetectonic zonation.

The picture changes dramatically at depths as shal-low as 2 km (Fig. 3). The Caucasian zonality of theWNW strike is clearly distinguishable here, althoughstructural features of the Trans-Caucasian N–S orienta-tion are still recognizable. Figure 3 shows that the west-ern part of the source zone (west of the Spitak merid-ian), represented by more numerous aftershock epicen-ters, has lower values of VP (4–5 km/s), closely followsthe Pambak-Sevan fault, and underlies the chain ofHolocene depressions of the Pambak River. The easternpart of the epicentral zone, 10–15 km wide and 25–30 km long, is distinguished by relatively high veloci-ties (5–6.3 km/s) and trends NW, parallel to the Alavarfault. A narrow (3–4 km wide) N–S trending band oflower velocities is fixed on the Spitak meridian and isparallel to a wider band of higher velocities recogniz-able east of the town of Spitak. Thus, the 2-km layerexhibits seemingly interfering north–south, east–west,and diagonally trending bands of lower and highervelocities.

Against this background, the 4-km layer (Fig. 3) isdistinguished by a weakly disturbed structure of thevelocity field. The VP are 5–6.5 km/s within the nearlyentire region studied. There are no structurally pro-nounced velocity heterogeneities. On the whole, theeastern half of the source zone (east of the Spitakmeridian) is characterized by higher P wave velocitiescompared to its western half. However, neither longitu-

41.0°

40.8°

43.8° 44.0° 44.2° 44.4°

10 km

Leninakan

Spitak

Stepanavan

Kirovakan

Fig. 2. Region studied, main inhabited localities, gridnodes, the Pambak-Sevan fault (the broken line), and thesurface coseismic rupture.



dinal (Caucasian) nor transverse (Trans-Caucasian)structures are observed.

On the contrary, the 6-km layer exhibits distinctzonation of velocity heterogeneities correlating with

structural elements of the tectonic zonality in the region(Fig. 3). Here, an E–W trending zone of lower veloci-ties (VP = 5.5–6.2 km/s) is observed in the central partof the source region (in a 20–30-km wide band includ-

10 km 10 km

43.8° 44.0° 44.2° 44.4° 43.8° 44.0° 44.2° 44.4°

41.0°

40.8°

41.0°

40.8°

41.0°

40.8°

10 km 10 km

10 km 10 km

41.0°

40.8°

41.0°

40.8°

41.0°

40.8°

H = 4 km H = 6 km

H = 0 km H = 2 km

H = 8 km H = 12 km

4 5 6 7VP

Fig. 3. Layer maps of P velocity contours. Shaded are areas of a velocity resolution no worse than 0.1 km/s. Each map shows earth-quakes with hypocentral depths within its layer. Other notation is the same as in Fig. 2.

654


DORBATH et al.

1535

1050

4045

5055

20

15105015105015 10 5 0 15 10 5 0 15 10 5 0

2530

3540

4550

5560

6570

2025

3035

4045

60

45

67

Vp

1 2 3

4 5 67

6

5

43

21

7

Fig

. 4. V

ertic

al c

ross

sec

tions

alo

ng li

nes

1–7.

Sha

ded

are

area

s of

a v

eloc

ity r

esol

utio

n no

wor

se th

an 0

.1 k

m/s

. Oth

er n

otat

ion

is th

e sa

me

as in

Fig

. 2.



ing the Pambak-Sevan active fault); this zone isbounded to the north and south by small blocks ofhigher velocities (6.2–6.8 km/s).

On the whole, the central low-velocity zone corre-lates with the Sevan synclinorium in the southern partof the Sevan-Akera fold zone. As mentioned above, thelatter is structurally expressed by folded rock masses ofthe Sevan synclinorium and Bazumskii anticlinorium.The southern high-velocity zone correlates with theMiskhan-Zangezurskii median mass. The northernmosthigh-velocity zone, underlying the Bazumskii anticli-norium, is represented in the study region by only asmall isometric fragment. Importantly, the crystallinebasement surface appears to occur precisely at depthsof about 5–7 km in this part of the Lesser Caucasus. Theaforementioned velocity heterogeneities are likely tocharacterize variations in the basement compositionbeneath the tectonic zones noted above. A higher veloc-ity and probably denser basement is shown to existbeneath stable tectonic zones, whereas a lower velocitylower density basement is established under the mobilefold zone of the Sevan synclinorium. The mosaic pat-tern of high-velocity blocks is evidence for a frag-mented basement of stable zones; this pattern is domi-nated by narrow N–S trending bands of lower values ofVP dividing the basement into separate small high-velocity blocks. A similar lower velocity band corre-lates with the Alavar fault, which was activated duringthe earthquake and is represented at the surface by thesoutheastern branch of the Spitak seismogenic fault. Onthe whole, the Pambak-Sevan active fault, the Alavardisjunctive structure, and the central segment of pri-mary coseismic deformations are associated at thisdepth with lower velocity bands.

The tectonic control of structural velocity heteroge-neities is even more evident in the tomographic layer ata depth of 8 km (Fig. 3). As seen from the figure, thelow velocity zone (VP = 5.5–6.5 km/s) of the Caucasianorientation correlating with the Sevan synclinorium ofthe Sevan-Akera fold zone becomes here much nar-rower (5–10 km) compared to the overlying slice (20–30 km). The Pambak-Sevan active fault is traceable inthe axial zone of this linear band; however, taking intoaccount its steep northern dip, this fault gravitatestoward the northern boundary of the band, i.e., boundsthe low-velocity zone to the north. The northeastern ter-mination of the Alavar fault also lies within this zone,somewhat widening eastward (east of the town of Spi-tak). Interestingly, the central E–W segment of the Spi-tak seismogenic fault and its NW segment bound at thisdepth the low-velocity band to the south, separating itfrom three high-velocity blocks (VP = 6.5–7.0 km/s)underlying the northern Miskhan-Zangezurskii medianmass. These blocks are located beneath the Bazumskiianticlinorium north of the low-velocity band of theSevan synclinorium.

Overall, the distribution of high- and low-velocitybodies in the 8-km layer is characterized by a banded–

mosaic pattern. The high-velocity bodies (V = 6.5–7.0 km/s) are nearly isometric in plan view, whereas thelow-velocity bands are, rather, linear. The generalbanded–mosaic structure clearly exhibits lower veloc-ity bands of both the Caucasian (WNW to E–W) andTrans-Caucasian (N–S) orientations. It is the N–Sbands that divide the high-velocity bodies of the south-ern (Miskhan–Zangezurskii) and northern (Bazumskii)zones into separate quasi-isometric blocks.

The 12-km layer is the deepest. Tomographic datacover here a very small area in the northwestern part ofthe source region. Two zones of different P wave veloc-ities are observed within the field studied. The southern,high-velocity zone (VP = 6.6–7.0 km/s) correlates withthe Miskhan-Zangezurskii stable massif, and the north-ern, low-velocity zone (VP = 6.0–6.5 km/s) underliesthe Sevan-Akera mobile fold zone. It is interesting that,in a shallower layer (8 km), the latter is underlain byhigher velocity bodies. Apparently, a velocity inversiontakes place at depths between 8 and 12 km beneath theSevan-Akera tectonic zone.

As regards the structural relations between thehigh- and low-velocity volumes of the crust, weshould note that the distribution of high-velocityblocks of the Miskhan-Zangezurskii zone widensnorthward in deeper layers. Thus, the boundarybetween low- and high-velocity blocks lies at40.85°N in the 6-km layer, at 40.9°N in the 8-kmlayer, and at 40.97°N in the 12-km layer; i.e., thisboundary steeply dips northward.

To gain insights into the vertical structure of the Spi-tak source zone, we address the tomographic profilescrossing the zone in the E–W and N–S directions andshown in Fig. 4.

The longitudinal profile (cross section 7 in Fig. 4)exhibits significantly different velocity distributions inthe western and eastern parts of the cross section. In thewestern part, seismic velocities are clearly seen to grad-ually increase with depth (from 4.0–5.1 km/s at the sur-face to 6.5–7.0 km/s at depths of 7–10 km). This part ofthe profile crosses the Miskhan-Zangezurskii medianmass. The eastern segment of the profile crosses thezones of the Spitak seismogenic fault and Pambak-Sevan fault, and its eastern end lies within the Sevan-Akera fold zone. Here, in the zone of active faults, thedeep structure of the wavefield is markedly disturbedand higher velocity regions (VP = 5.5–6.0 km/s) reachthe Earth’s surface. They are underlain at depths of 5–7 km by small bodies, isometric in cross section, whichhave high seismic velocities (VP = 6.5–6.7 km/s). Seis-mic velocities decrease to VP = 5.5–6.4 km/s at greaterdepths (8–12 km); i.e., a velocity inversion takes placein the middle crust. Moreover, the greatest number ofrepeated shocks was recorded in this area, and many ofthem were deep (7–13 km). East of the Pambak-Sevanand Alavar active fault zones, the P wave velocity grad-ually increases with depth (from 4.0 km/s at the surfaceto 5.5–6.2 km/s at a depth of 15 km). Therefore, no

656


DORBATH et al.

velocity inversion relations are observed at these depthswithin the Sevan-Akera zone east of the source area.

The structure of the source zone is observable incross sections of all N–S profiles. The westernmost sec-tion (1, Fig. 4) reveals a gradual increase in seismicvelocities with depth. Higher velocity horizons occur ata greater depth (15 km) in northern part of the sectionbeneath the Sevan synclinorium as compared with itssouthern part beneath the Miskhan basement protrusion(7 km). The majority of aftershocks group in the gradi-ent zone on the southern side of the Pambak-Sevanfault, where the high-velocity layer steeply dips. In the

upper crust, the zone of the Pambak-Sevan active faultis associated with a wide region of lower velocitiesextending to depths of 5–7 km.

Next to the east, the N–S cross section (2 in Fig. 4)exhibits a considerably more complicated structure ofcrustal seismic velocities. In the northern part of theprofile at depths of 6–9 km, a higher velocity layer(VP = 6.5–6.8 km/s) is under- and overlain by lowervelocity bodies (VP = 5.5–6.4 km/s). The cloud of after-shock hypocenters is divided into two separate groups.The northern group is located on the northern,upthrown side of the Pambak-Sevan fault, which is

2043.8

15

10

5

0

40.7°

43.9 44.0 44.1 44.2 44.3 44.4 44.5

40.8°

40.9°

41.0°

(a)

(b)

Fig. 5. Comparison between initial (circles) and revised (opposite ends of segments starting at the circles) hypocenters of earth-quakes: (a) map of epicenters; (b) E–W cross section. Other notation is the same as in Fig. 2.



associated with a lower velocity zone extending to adepth of 10 km. The second, southern group of hypo-center concentration, similar to the profile west of it,is confined to the steeply dipping zone of the high-velocity layer (VP = 6.6–7.0 km/s) at the boundarybetween the Sevan synclinorium and the Miskhaninlier.

The cross section along profile 3 (Fig. 4) displaysthe structure of the source zone in its central part. Low-velocity pockets are present at shallow depths (up to5 km) along the zones of the Pambak-Sevan fault andthe Spitak seismogenic fault. At greater depths (up to12–13 km), these two pockets join to form a narrowchannel filled with low-velocity material and separat-ing high-velocity blocks located at depths of more than7–8 km north and south of these active fracture zones.The low-velocity channel lies beneath the Sevan syncli-norium, and the high-velocity bodies, beneath theBazumskii anticlinorium and the Miskhan basementprotrusion, respectively. The cloud of aftershock hypo-centers mostly concentrates within this narrow low-velocity channel and steeply dips northward. On thewhole, the cloud is located on the northern, upthrownside of the Spitak seismogenic fault.

Profile 4 (Fig. 4) reveals a somewhat different struc-ture of the source region in its central segment. Twodeep low-velocity pockets, each about 5–8 km wide,are clearly present along the zones of the Pambak-Sevan active fault and Spitak seismogenic fault; thepockets are separated at depths of 5–13 km by a localhigh-velocity uplift about 5 km wide. The cloud ofaftershocks in this section steeply dips northward; isconfined to the northern, hanging wall of the Spitaksource fault; and immediately overlies the high-veloc-ity uplift. The cloud of hypocenters is separated fromthe Pambak-Sevan fault and appears to be unrelated toit. South of the low-velocity pocket confined to the Spi-tak seismogenic fault, a high-velocity body associatedwith the northern margin of the Miskhan basement pro-trusion is observed at depths of 6–7 km and more.

Section 5 crosses the southwestern margin of thesource zone. A relatively uniform increase in seismicvelocities with depth is observed here (from 4 km/s atthe surface to 6.0–6.5 km/s at depths of 5–10 km). Adistinct low-velocity pocket 5–7 km wide extending todepths of 5–6 km is associated here with the Pambak-Sevan fault. However, the Alavar active fault, alongwhich the southeastern segment of the Spitak earth-quake rupture arose, is not expressed here so clearly bya fault-line zone of relatively low velocities. Only at adepth of about 5 km is a thin high-velocity layer brokenby this rupture into two separate segments. Mainrecorded hypocenters of aftershocks are also concen-trated here.

Finally, the easternmost section, located beyond thesource region, is characterized by a relatively undis-turbed pattern of the seismic velocity distribution in theupper crust. The velocities are seen to gradually

increase with depth, from 4.5 km/s at the surface to6.0 km/s at a depth of about 5 km. The hypocenters ofthe few aftershocks are irregularly distributed at depthsof 2–5 km, concentrating on the southwestern side ofthe Alavar fault zone striking NW.

In addition to the construction of a velocity model inthe region studied, the LET method provides additionalconstraints on the position of hypocenters of earth-quakes used. New hypocentral coordinates are com-pared in Fig. 5 with previous estimates derived fromdetailed studies of the seismic regime in the sourceregion. With the presence of large individual diver-gences, the overall pattern of the spatial distribution ofaftershocks in the source region is seen to remainunchanged.

DISCUSSION

The above tomographic analysis and geological evi-dence on both general and detailed scales revealed ananomalously complex deep structure of the crust in thesource region of the Spitak earthquake. Note that thisfact was also noted in previous attempts to performtomographic analysis. A gradual increase in seismicvelocities with depth is characteristic of the relativelystable undisturbed crustal blocks bounding the sourcezone to the north and south, the Bazumskii anticlino-rium of the Sevan-Akera fold zone and the Miskhanbasement protrusion of the Miskhan-Zangezurskiimedian mass, respectively. Neither significant inver-sions of the P wave velocity nor large lateral variationsin seismic velocities across main tectonic zones of theLesser Caucasus are observed. Significant heterogene-ities are related to the transverse, Trans-Caucasianzonation. The distribution pattern of seismic velocitiesis substantially disturbed, both laterally and vertically,along several N–S trending lineaments and the Aragats-Spitak transverse fault zone. Laterally, these distur-bances appear as local N–S striking bands of higher andlower velocities. These trans zonal bands are clearlyrecognizable in slices at depths of 0, 2, 6, and 8 km.

The mobile Sevan synclinorium of the Sevan-Akerastructural–lithologic zone, including the source regionof the earthquake, is underlain by crustal layers of avery complex structure, as is evident from both hori-zontal and vertical seismic tomography patterns. Seis-mic velocity inversions and sharp lateral variations invelocity, often involving middle crust layers, areobserved here. The Sevan synclinorium is bounded bythe active Pambak-Sevan fault in the north and by theSpitak coseismic rupture in the south; the latter arose atthe northern termination of the Alavar diagonal fault.As seen from tomographic images at depths of 0, 2, 6,and 8 km, these fractures are associated with narrowzones of relatively low seismic velocities. The widthand configuration of these fault-line zones at variousdepths can be estimated from transverse tomographiccross sections. The width of the zones of fault-line vari-ations is typically 1–2 km or somewhat more. These

658


DORBATH et al.

zones extend, in the form of narrow pockets, into mid-dle crust layers and sometimes are inclined to the northin accordance with the general dip of the fault planes,established from geological observations at the surface.

Based on laboratory tectonophysical experiments,Sherman et al. [1983] established that zones ofdynamic influence of fault’s arise along active faultsexperiencing strike-slip, reversed, normal, and com-bined motions. The width of these zones depends on thekinematics of the motions. The influence zones in lab-oratory experiments are distinguished by the presenceof numerous feathering cracks that fracture the mediumaround the main rupture. By analogy with these exper-imental data, one may expect that, under natural geo-logical conditions, active recent faults also give rise todynamic influence zones of various widths. Geologicalobservations in large fault zones indicate the presenceof numerous feathering faults severely fracturing theirwall rocks. In particular, detailed mapping of primarycoseismic tectonic deformations related to the Spitakearthquake fault showed that the seismic rupture zonehas nearly everywhere a very complex structure, beingcomplicated by numerous fractures of various ranksand by small anticlinal folds [Rogozhin and Filipp,1991]. The width of the surface zone of such distur-bances exceeds a few tens (up to one hundred) ofmeters; however, taking into account very small frac-tures and fields of local stresses, which are invisible atthe surface, this width can amount to a few hundredmeters or even a few kilometers.

The bands of anomalously low velocities related atdepth to longitudinal and transverse faults in the sourcezone of the Spitak earthquake are likely to be the afore-mentioned dynamic influence zones revealed by seis-mic methods. Apparently, numerous fractures and rup-tures in fault-line zones hinder the propagation of seis-mic waves, so that P waves travel within these zones ata lower velocity compared to the undisturbed geologi-cal medium.

It is probable that recently inactive faults do nothave deep low-velocity fault-line zones of dynamicinfluence because inactive disjunctive tectonic distur-bances are rapidly filled at depth with mineral material,are healed up, and do not hinder the propagation of seis-mic waves. Actually, as is evident from our study, theAlavar fault southeast of the source region of the Spitakearthquake, outside the area of coseismic deformations,is not associated with a low-velocity zone. On the otherhand, the Pambak-Sevan and Aragats-Spitak faults, aswell as the Spitak seismogenic fault, have such zones ofdynamic influence and can be regarded as recent activedisjunctive structures.

CONCLUSIONS

(1) The method of local seismic tomography iseffective for the construction of a velocity model in theepicentral area of a strong earthquake from aftershock

data; the velocity structure derived from the study of theSpitak earthquake is distinguished by a high degree ofheterogeneity.

(2) The study revealed the presence of low-velocitybands associated with active faults in the source regionof the Spitak earthquake.

ACKNOWLEDGMENTS

We are grateful to all staff members of the Epicen-tral Expedition for their help in gathering experimentaldata. Most figures were prepared with the help of theGMT software package. This work was supported bythe Russian Foundation for Basic Research, projectnos. 02-05-64946 and 02-05-64894, and Russia–NATOgrant JSTC.RCLG.978401.

REFERENCES

1. Sh. A. Adamiya, G. L. Gabuniya, and Z. A. Kuteliya,Characteristic Features of the Caucasus Tectonics(Nauka, Moscow, 1989), pp. 3–15 [in Russian].

2. K. Aki and W. H. K. Lee, “Determination of Three-Dimensional Velocity Anomalies under a Seismic ArrayUsing First P Arrival Times from Local Earthquakes: 1.A Homogeneous Initial Model,” J. Geophys. Res. 81,4381–4399 (1976).

3. F. Albarede, B. M. Bogachkin, A. E. Oleinik, et al., “Ona Prehistoric Strong Earthquake in the Spitak Area,”Dokl. Akad. Nauk SSSR 313, (2), 397–400 (1980).

4. S. S. Arefiev, Zh. Ya. Aptekman, T. V. Afim’ina, et al.,“Catalog of Aftershocks of the Spitak, December 7,1991, Earthquake,” Fiz. Zemli, No. 11, 60–73 (1991).

5. S. P. Bal’yan, D. A. Lilienberg, and E. E. Milanovsky,“Contemporary and Recent Tectonics of SeismicallyActive Orogens in Armenia and the Spitak EarthquakeArea,” Geomorfologiya, No. 4, 3–16 (1989).

6. L. Dorbath, S. S. Arefiev, and B. A. Borisov, “Deep Seis-micity Zone of the Spitak Zone,” Fiz. Zemli, Nos. 7–8,42–52 (1994).

7. D. Eberhart-Phillips, “Three-Dimensional VelocityStructure in Northern California Coast Ranges fromInversion of Local Earthquake Arrival Times,” Bull.Seismol. Soc. Am. 76 (4), 1025–1052 (1986).

8. J. R. Evans, D. Eberhart-Phillips, and C. H. Thurber,“User’s Manual for Simulps12 for Imaging Vp andVp /Vs: A Derivative of the “Thurber” TomographicInversion Simu13 for Local Earthquakes and Explo-sions,” in U.S.G.S. Open-File Report (1994), pp. 94–431.

9. A. S. Karakhanyan, “Land-Based and Aerospace Sur-veys of Active Faults and Coseismic Deformations of theSpitak Earthquake of 1988,” Izv. Akad. Nauk Arm. SSR,Nauki o Zemle, No. 3, 20–24 (1989).

10. F. W. Klein, Hypocenter Location Program Hypoin-verse (US Geol. Surv. Open-File Rep. No. 694, 1978).

11. N. V. Kondorskaya, M. B. Mkrtchyan, and N. A. Lagova,“Analysis of the Spitak, December 7, 1988, Earthquakeand Its Strongest Aftershocks from Observations at Sta-tionary Near and Teleseismic Stations,” in CombinedAssessment of Seismic Hazard (Engineering Seismology



Problems, issue 32) (Nauka, Moscow, 1991), pp. 25–35[in Russian].

12. E. E. Milanovsky, Neotectonics of the Caucasus (Nedra,Moscow, 1968) [in Russian].

13. E. E. Milanovsky and V. E. Khain, Geological Structureof the Caucasus (MGU, Moscow, 1963) [in Russian].

14. G. L. Pavlis and J. R. Booker, “The Mixed Discrete Con-tinuous Inverse Problem: Application to the Simulta-neous Determination of Earthquake Hypocenters andVelocity Structure,” J. Geophys. Res. 85, 4803–4810(1980).

15. E. A. Rogozhin and E. Filipp, “Geological–TectonicStudy of the Spitak Earthquake Source Zone,” Izv. Akad.Nauk SSSR, Fiz. Zemli, No. 11, 3–17 (1991).

16. Yu. K. Shchukin, A. K. Astakhov, A. A. Belov, et al.,“Geological and Geophysical Conditions in the SourceZone of the Spitak Earthquake (the Decade after theCatastrophe),” Geofizika, No. 5, 54–66 (1998).

17. N. V. Shebalin, “Estimation of the Size and Position ofthe Spitak Earthquake Source from Macroseismic evi-dence,” in Combined Assessment of Seismic Hazard(Engineering Seismology Problems, issue 32) (Nauka,Moscow, 1991), pp. 36–44 [in Russian].

18. S. I. Sherman, S. A. Bornyakov, and V. Yu. Buddo,Dynamic Influence Zones of Fault from ModelingResults (Nauka, Novosibirsk, 1983) [in Russian].

19. A. Sluis and H. A. Vorst, “Numerical Solution of Large,Sparse Linear Algebraic Systems Arising from Tomo-graphic Problems,” in Seismic Tomography, Ed. byG. Nolet (D. Reidel. Publ., Dordrecht, 1987), pp. 49–83.

20. C. H. Thurber, “Earthquake Locations and Three-Dimensional Crustal Structure in the Coyote Lake Area,Central California,” J. Geophys. Res. 88 (10), 8226–8236 (1983).

21. C. H. Thurber, “Local Earthquake Tomography: Veloci-ties and Vp/Vs—Theory,” in Seismic Tomography, The-ory and Practice, Ed. by H. Lyer and K. Hirahara (Chap-man and Hall, 1993), pp. 563–583.

22. J. Trampert and J. J. Leveque, “Simultaneous IterativeReconstruction Technique: Physical Interpretation Basedon the Generalized Least Squares Solutions,” J. Geo-phys. Res. 95, 12 553–12 559 (1990).

23. A. V. Treusov, S. S. Arefiev, and B. A. Borisov, “Tomo-graphic Study of the Spitak Earthquake Zone,” Fiz.Zemli, No. 5, 23–27 (1993).

24. V. G. Trifonov, A. S. Karakhanyan, and A. I. Kozhurin,“The Spitak Earthquake as Evidence for Recent TectonicActivity,” Geotektonika, No. 6, 46–60 (1990).

seismic tomography constraints on the deep structure of...

Documents