Heterogeneity of the uppermost mantle beneath Russian ... - Wiley

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 104, NO. B9, PAGES 20,329-20,348,SEPTEMBER 10, 1999

Heterogeneity of the uppermost mantle beneath Russian Eurasia from the ultra-long-range profile QUARTZ ElenaA. Morozova,Igor B. Morozov,1 and ScottB. Smithson Departmentof GeologyandGeophysics, Universityof Wyoming,Laramie

Leonid N. Solodilov

Centerfor RegionalGeophysical andGeoecological Research (GEON),Moscow,Russia

Abstract. The3850-kmlongDeepSeismicSoundingprofileQUARTZ crosses sixmajor geologic provinces in Eurasiaandis sourced by 3 nuclearand48 chemical explosions. We present thefirstinterpretation of theentiredataset,usingtwodimensional (2-D) raytracing andinversion, resolution analysis, and1-D amplitude modeling.Ourinterpretation showsa 42-km-thick,high-velocity crustundertheBalticShield,a 29-km-thickcrustandhighvelocityuppermantleundertheMezenskaya depression, 52-km-thickcrustwithhighvelocitylower crustanduppermostmantleunderthe Urals, and40-km-thickcrustunderthe

WestSiberianbasindeepening to 45 km undertheAltay-Sayanfoldbelt. High-velocity(8.4 km/s)uppermost mantleis foundunderthe Mezenskayadepression andunderthe eastflank of theUrals. Onealmostcontinuous uppermantleboundaryoccursat 65- to 80-kmdepth, andanotherwith an approximately 40-kin-thickLVZ occursat 120-to 140-kmdepth.The shallowuppermantleblocksandthetwo extensive interfaces indicatestronguppermantle heterogeneity.Resolutionanalysisbasedon directmultivariatemodelperturbations, artificial neuralnetworkandprincipalcomponent analysis,indicatethe depthuncertainty of the410km discontinuity within+6 km, andalsoitstrade-offwith dip andvelocitiesaboveandbelow the discontinuity.Decreasednear-criticalamplitudesof reflectionsfrom the 410-km and 660-kmdiscontinuities indicatethattheseboundaries aremostlikely represented by gradient zonesabout15-20 km thick. Lithospherethins,asthenospheric velocitydecreases, andthe 410-km discontinuitydipsto the SE approaching the Himalayanorogenicbelt. 1. Introduction

The ultra long Deep Seismic Sounding (DSS) profile QUARTZ, which crossessix major geologic provinces in Eurasia (Figure 1) and is sourcedby 3 nuclearand 48 chemical explosions,provides unusually detailed and continuous coverageof the uppermostmantle. The 3850-km-longprofile is recordedby 400 three-componentrecordersat a nominal spacingof 10 km and showsgood energy out to more than 3100 km from peacefulnuclearexplosions(PNEs) and 300 600 km from chemicalexplosions.This data set is uniquein the length of continuousprofiling, the number of source points, and in PNE recordingin two directionsfrom each sourcepoint. The profile has been interpretedby various groups for crustal and uppermostmantle structureusing chemicalexplosions[Egorkinand Mikhaltsev,1990;Egorkin, 1991; Schueller et al., 1997], for one-dimensional(l-D) [Mechie et al., 1993] and 2-D [Ryberg et al., 1996] upper mantle structureusing only first arrivalsfrom the PNEs, and

by applyingtravel time tomographicinversionto the PNE records [Lorenz et al., 1997]. A review of earlier interpretationsof otherDSS profileswas given by Pavlenkova[1996a]. Two general conclusionshave been made from DSS studies. First, a number of 1-D mantle models derived for north-

ern Eurasia[Mechieet al., 1993;Priestleyet al., 1994;Pav-

lenkova, 1996b]exhibita prominent low-velocity zone(LVZ) belowabout200-km depthand probablyreducedvelocity contrastat the 660-kin boundarycomparedto the IASP91 model[Ryberget al., 1998].Thereis no generalagreement aboutthe sharpnessof the mantletransitionzone discontinui-

ties. Mechieet al. [1993] andRyberget al. [1998]modeled the 410-km discontinuity asa sharptransitionconsistent with the IASP91model,but Priestleyet al. [1994] arguedthatthe 410-km discontinuity mightactuallybe a 35-km-thicktransition zone.

Another importantresult of DSS studiesis the demonstra-

tion of stronglyheterogeneous, 2-D structure of the uppermost mantle within the regional scale and down to at least ADD

lorn

[]•/[o,•I.,;o

of r•l

1 OOq.

P•.;o,•tlo•,

of .•,l

100A.

D,,A ....

t

a/., 1995;Pavlenkova,1996b;Morozovaeta/., 1997]. This heterogeneity manifestsitself in the velocityand attenuation •Nowat Department of Geology andGeophysics, RiceUniver- structure[Morozoveta/., 1998b],aswell as in the properties sity, Houston,Texas. of regionalandglobalmantleboundaries [Pavlenkova,1996a; Morozova et al., 1997]. The variability of the uppermost Copyright 1999bytheAmerican Geophysical Union. mantlepartly limits the significanceof 1-D regionalmodels Papernumber1999JB900142. andpartlyminimizessomeof the controversies in interpreta0148-0227/99/1999JB900142509.00 tions;on the otherhand,thisvariabilitydemands (at least)220,329

20,330

MOROZOVA

ET AL.: UPPERMOST MANTLE HETEROGENEITY

cratons,ancientplatformregions,Hercynianorogens,extensionalbasinswith thick sedimentarycover,and active fold basinsof northernEurasiawere In this paper,we presenta combinedinterpretation of all belts. The largesedimentary and QUARTZ PNEs and chemicalexplosionsusingboth primary formedby deeperosionof the underlyingPrecambrian dueto the extenand secondaryarrivalsfor the detailedstructureof the upper- Paleozoicbasementfollowedby subsidence or rifts [Zonenshain et al., most mantle after taking into accountthe effect of crustal sionof the crustwithin aulacogens crossed by the structure. SinceQUARTZ datahave apparentlybeenstudied 1990]. FromNW to SE,thetectonicstructures et al., 1990]: morethan any otherDSS profile,the naturalquestionis, Why profileareasfollows(Figure1) [Zonenshain 1. The EastEuropeanplatform,includingtheBalticShield, is it necessaryto proposeyet anotherreinterpretation?The D interpretations,of which very few exist at the regional scale.

answerlies in the complexityand ambiguityof the traveltime inversionof data setsof sucha scaleand in the significant nonuniformityof coverageof the subsurface.Thus an integratedapproachincorporatingall seismicphasesis required in orderto createa consistentimageof the crustand uppermost mantle. However, in all previousstudies,only partsof the available data were employed. Thus, in the original analysisof the QUARTZ crustaldata set lEgorkin and Mikhaltsev,1990], only the chemicalexplosiondata were used. As we showbelow, by includingthe PNEs in the analysis,we are able to put betterconstraintson a numberof criticalfea-

consolidated at the end of the middle Proterozoic after accre-

tion of smallercontinentalblocksthatjoined to form a stable craton. The craton remainedvirtually intact since the last

graniticintrusions at 1.5 Ga. The Precambrian BalticShield is the oldestand moststablepart of the East Europeanplatform.

2. The late PrecambrianTiman fold belt is an uplift in the

Ripheanbasement of the TimanPechoraplatform. It separatesthe proposed BarentsmassiffromeastEuropeandmay be considered as a suturebetweeneastEuropeandthe minor Baremsiacontinent,resultingfrom a collisionat the end of

tures of the Moho and sub-Moho structure. Also, the inter-

the late Precambrian.

pretationby Egorkin and Mikhaltsev[1990], like many RussianDSS interpretations, employsratherstrongnonlinearvelocity filtering that may lead to the appearance of spurious aliasedeventsandmorecomplexityin the resultingmodel. In the previousstudiesof the mantlestructure[Mechieet al., 1993; Ryberget al., 1996], in both 1-D and 2-D models only the first arrivalswere used. Suchan approach,however, leadsto strongambiguitiesin the resultingmodels,evenin 1D cases[Aki and Richards,1980; Mechie et al., 1993;Ryberg et al., 1996]. Therefore,to betterconstrainthe mantlevelocity/interfacestructure,one mustemploythe informationcon-

3. The southernpart of the Timan-Pechora basincomains earlyDevonianto lateCretaceous andearlyPaleogene sedimentaryrocks. Thethickness of the sediments in thisbasin rangesbetween3 and7 km. 4. The Uralian fold systemrepresents a linearcollisional

tained

in

the

numerous

mantle

reflections

observed

in

QUARTZ data. Moreover,aswe showbelow,a bettercrustal modelthat is consistent with PNE recordingsis of criticalimportancefor the interpretationof the deepermantlestructures. Velocity structureof the uppermostmantle is affectedby travel time anomalies in the crust as determined from chemi-

cal explosions, andimportantly,uppermost mantlevelocityis also determinedfrom these chemicalexplosionscomplemented by PNEs for greater detail than can be determined from one source alone.

Our interpretationis based on 2-D ray-tracing,forward modeling of all observed refracted and reflected seismic phases,includingthe whispering-gallery (multiplyrefracted) phasesrecentlyrecognizedin QUARTZ records[Morozovet al., 1998a]. In our modelingwe primarily employthe forward travel time modelingprogramby Zelt and Smith[1992] complementedwith our own tools for model editing,travel time picking and management,and plotting. After the inversion, we analyze the resolvingpower of the model by performing a sensitivitykernel test describedin the appendix. Sharpness of the transitionzone discontinuities is studiedusing 1-D dynamicmodeling[Fuchsand Mtiller, 1971]. The mantle structurealong the QUARTZ transectdevelopedin this paperis furtherinvestigatedby Morozovet al. [1998b].In that paper we incorporateshort-periodP wave attenuation estimatesobtainedfrom the spectralanalysisof the first arrivals from the PNEs.

2. Tectonic and Geologic Setting

belt formed at the end of the Paleozoicand the beginningof the Mesozoic. The Urals developedout of a sequence of oceanic closing,subductionof oceaniccrustunder islandarcs, convergence of lithosphericplates,and continentalcollision. They representa classicobductionbeltwith a complexaccretionaryhistory. 5. The West Siberian basin, which is one of the world's

largestplatformstructures.The basementof this platform was formedby Paleozoicfold structuresand older Precambrianblocks. The profile passessouthof the failed Triassic (235 to 215 Ma) WestSiberianrift (Figure1). 6. The northernAltay-Sayanfold belt representsa very complicatedstructureformed duringthe whole Paleozoic, originatingfrom a middle Cambrianaccretionarymosaic formedfromfragments of continents andislandarcs. Priorto the late Cambrian,this accretionarymosaiccollidedwith the passivemarginof Siberia,wasthrustoverit, andunderwent an Alpineuplift. Earlier interpretations of DSS data [e.g., Volkovet al., 1984;Egorkinand Mikhaltsev,1991;Pavlenkova,1992] describethe crust of northernEurasiaas an extremelynonuni-

form,block-layered structure.The crustalblocksarebounded by crustal-scale faultsand shearzonesthat oftenoffsetthe Moho and extend into the mantle. The fault zones are associ-

ated with the sites of repeatedand intensivetectonicand magmaticactivitythat is relatedto the deformation of the originallyplasticlithosphere [Nalivkinaet al., 1987]. 3. Data and Indications of the Mantle

of Structural

Features

Three QUARTZ PNEs were locatedin the Pechorabasin, in the West Siberianbasinand in the Altay Mountains,and 48 chemicalshotswere spacedat about80 km alongthe profile (Figure1). Two uniqueaspectsof the profileare (1) continu-

ousprofilingof the crustand uppermost mantleover sucha seismic The profile QUARTZ extendsacrossa wide variety of great distance,and (2) crustalrefraction/reflection continentalstructures of northernEurasia(Figure1): Archean coverageallowingtwo-sided?m? reflectionsto be recorded

MOROZOVA

ET AL.' UPPERMOST MANTLE

•0 u

HETEROGENEITY

20,331

40ø 50 ø60 ø70" BOø øø•

•O o

•Fold

aøø

9d 40"

•0'

50'

60'

70'

km •

0

500

Figure 1. Map of the NW part of the formerUSSR crossedby the profile QUARTZ. Trianglesindicatethe locationsof chemicalexplosions; starsshowthe locationsof thethreepeacefulnuclearexplosions (PNEs)recordedby the profile, labeledwith their corresponding shotnumbers.Major tectonicstructuresare indicated. The insetshowsthe locationof the map area.

alongthe entireprofile and providingempiricalevidenceof crustalvelocity and Moho variations. The chemicalexplosionsare recordedinto the samereceiverson the groundas the nuclearexplosions.Travel time curvesfrom chemicalexplosionsalongthe entireprofile showgood coverageof the Moho corresponding to a depthof about40 km. Deviations from this time occur in the Mezenskayadepression,Urals, and Altay-Sayanbelt. The chemicalrecordingstypically

crossoverpoints, amplitudevariations)we did not use automated inversionprocedures. In our inversionwe took advantageof the previouslypublishedresultsof forward modeling [Egorkinand Mikhaltsev,1990] and of the tomographic inversionwithin the NW part of the profile [Schuelleret al., 1997]. By progressivelymoving deeperinto the model, we constrainedthe deepermantle structures.However, this procedurewas not reducedto a pure "layer-stripping"approach, showPg,oneortwodeepcrustal refractions andreflections,sincewe observedthat in a numberof cases(suchasthe highProp and somePn arrivals(Figure2). A consistentupper velocity mantle body under the Mezenskaya depression), mantle event, PN, may also be observed(Figure 2). The distantPn arrivals from the PNEs contributedimportantincrustaland uppermostmantle structureis further constrained formationto the crustalmodel. Thereforethis inversionprocby datafromthe PNEs (Figure3). esswas repeatediteratively,in somecasesleadingto reevaluData processingbegan with extensiveediting (primarily ationof seismicphaseidentification. selectionof the higher-qualityrecordsout of the two setsof Plate1 showsour resultingcrustaland uppermostmantleP three-component recordingsat differentamplificationlevels), wave velocity and interfacemodel obtainedusing the infordatavisualization,identificationof seismicphases,andtravel mation from all seismicphasesidentifiedin 51 shot gathers. time picking. Only the most reliable phaseswere picked. In the following sections,we summarizethe observations that The traveltime pickswerecheckedfor reciprocitywherepos- constrainthe generaland specificfeaturesof this model;folsibleand input into a forwardtraveltime modeling. All seis- lowed by furtherdiscussionand interpretationof the model.

mic processing was carriedout usinga processing system DISCO enhanced with a speciallydesigned interfacefor handlingultralongDSSrecords.At laterstages of thiswork,the processing system[MorozovandSmithson, 1997]. Traveltime inversionwasperformed usinginteractive forwardmodeling[ZeltandSmith,1992],startingwith 1-D velocitycolumnsmodeledin differentpartsof the profile,and includingan appropriate spherical-Earth correction.Sincethe ray coverageof crustalandmantlereflectorsis still relatively sparseand heterogeneous, and sincewe employedcomplex modelfit criteria(suchastraveltime deviations, positions of

3.1. General Character of the Records:Upper Mantle Heterogeneity

The generalcharacterof QUARTZ records(Figures2, 3, and 5) indicatesthe presenceof significant2-D structureof the upper mantle. The crustand the more denselycovered from 130 - 160 km of the uppermantleexhibitstrongvertical and horizontalvelocity contrastsilluminatedby numerousreflectionsand refractions. Our interpretationidentifiesthree regional uppermostmantle boundarieswith distinct lateral

20,332

MOROZOVA

ET AL.: UPPERMOST MANTLE HETEROGENEITY

Baltic Shield

Baltic Shield

o •:•?;,,[,•,},'•,, .... •,-,,,,>,,?,},,-',,,,'W,•, 1•

150

2•

250

3•

MezenskayaDepression

1-q--• r-[-l-l-n'l-j [ r 1-1-1,

350

4•

o

450

o

,,•,,,,,• .... , .... , .... , .... ,,•,,,•,,,,, .... , .... ß o 0

50

1•

150

Timan

200

250

3•

350

PechoraBasin

r i-l-l-i-r [ •r-l-l' l i r-r-•]-1 mm -r-•l • • 'l j r r 11 ' *

' I ....

I .... I ' '-' ' I '" ' ' I .... I .... I ....

I ....

I ....

[ ....

I ....

I' ''I

0

-200 -150 -100 -50 0 50 1001502• 250300SE NW -350 -300 -250 -200 -150 -1• -500 50100 150 200250 Urals

West Siberian Basin

....... SP ...... 142•.• •- • .... ,......... ,................... ,......... ,......... ,......... ,......... ,......... ,......... ,..... 50

NW

100

150

200

-300

-200

-100

Offset(km)

0

100

200

300

400

500

600

SE

Figure 2. Vertical component recordsectionsfrom six chemicalexplosions alongthe profile QUARTZ showingthemajorphasesandcorresponding traveltimescalculated usingourmodel(Plate1). Mostrecords

showsequences of 1-2crustal reflections following Pgtraveltimecurves (notlabeled). Noteconsistent reflectionsfrom the depthof about70 km, labeledasPN.

depthvariationsandlow- andhigh-velocityzoneswithinthis eventsto the SE from PNE 123 have lower apparentvelocities and largertraveltimescomparedwith the reverseshotto part of the mantle(Plate1). A large-scalefeatureof the heterogeneous mantlestructure the NW from PNE 323. Suchstrikingasymmetryof the arriclearlyobservedfrom QUARTZ PNE recordsis the promi- vals from the reversedPNE records(Figure 6) can only be nent LVZ 1 extendingfrom about200 km to a greatdepth interpretedas an indicationof a dip of the 410-km disconti(Plate 1). The identificationof this structureas an LVZ nuity and of a SE velocitydecreasewithin the uppermantle. (ratherthana low-gradient zonethatwouldalsocausea drop This heterogeneityextendsdown to about500 km, which is of energyin firstarrivals)is basedonanobservation of strong the maximum depth of reversedcoverageprovidedby the reflectionsfrom its top (Figure4) andon the observedmove- PNEs. Below this depth,the datado not allow reliableidentioutsof reflectionsfrom the 410-km discontinuity providing fication of a horizontal variation of the mantle structure,and constraints on the averagevelocityin this region. Although thusour modelis 1-D in this region(Plate 1). the shadowzonecausedby this LVZ1 concealsthe structure

downto the 410-km discontinuity (Figures3 and 5), travel time modelingsuggests that the velocityreversalextendsto about300-kmdepth[Mechieet al., 1993]. A generalconclusionfrom our phaseinterpretationof the waves penetratingthe mantle transitionzone is that all the

3.2. The Crust and Mantle to Depth of 60-80 km

The Baltic Shield is characterizedby high-velocity(7.2 7.4 km/s) lower crustindicatedby a high-velocityfirst arrival and by the moveout of post critical PmP reflections(shot

MOROZOVA

ET AL.' UPPERMOST MANTLE

HETEROGENEITY

20,333

10

I

Pechora Basin

-800

-5

-400

PNE 213 x

400

'

800

•

1600

1200

P

2( •00

''

10

i

g

0 LVZ 2

--5 -- --r1600 -1200 -800 -400 ....... 0r 400800 1260' 1600,

5

-5 -240'0 ..... -2000 ..... -i600 ..... -i20'0 ..... 2800 ..... 460'

NW

-5

Distance (km)

SE

Figure 3. Vertical-componentrecordsfrom QUARTZ PNEs. Prominentfeaturesof the crustaland Moho

structureconstrained by theserecordsare indicated. Stepsin the first arrivalsdue to the interpretedlowvelocity zones(LVZi_3, also shownin Plate 1) are indicated.Lines showtravel times calculatedin our velocitymodel(Plate 1). Note thatthe Pn arrivalSE from PNE 123 is blockedby the Urals andthe reflection fromthe 70-km discontinuity (PN) is observedin the firstarrivalsto offsetsof 900 km.

points0 and9 in Figure2). The resultingcrustalthicknessis

about42 km, in agreement withanearliertomographic study [Schuelleret al., 1997]. The Mezenskayadepression is markedby sharp,localadvances in traveltimeforProp(Figure2, shotpoint77) andfor firstarrivalsfromPNE 123(Figure 3). Both forwardandreversed shotsshowPn at 6 s reducedtimeversus8 s for the adjacent crust(Figure2). Our modelaccounts for thiswith a Mohoupliftto 29-kmdepth

underthe Mezenskayadepressionand an uppermantlevelocity increaseto 8.4 km/s, which is necessaryto explain the travel time advance in the refracted wave from PNEs 123 and

213 (Figure 3). The modeling showsthat the travel time curvescannotbe fit with only an uplift in the Moho. Also, when

we

include

events

from

PNE

123

in

the

crus-

tal/uppermostmantle interpretation,local low-velocity zones are interpretedin the mantleNW of the Mezenskayadepre-

20,334

MOROZOVA ET AL.: UPPERMOST MANTLE HETEROGENEITY

ca

•

z.

•

West SiberianBasin

Altay

NW

SE

Gravity Anomaly

E

0-'

,•

10 ' 6.3^

20 ' 6.7•' 130-

40 7.2

50 8.! 60 i_[ a

..

123. .

6.0

6.2

""'"x.6..5

7.2"-.

.

6.8

8.0

8.t ß

M

5.0

500

1000

1500

____ • ......•'-,...,.-•_• 8.4

2000

80

8q

••

8,

....... ,,t

200

2500

•'L,• ., I I ,,,

8.2

8

lOO

8,5 8.0 7.5 7.0 6.5 6.0

64

6.7

8.4

'

-500 0

6.4

•.

3000

I ,:

3500

, ! ....

•

I

11

M

8.•........

,

10

•

__.__•8.5

8.3

300 8.5

400 500

9.8

b

600

700

-500

500

1000

i

0 _•

1500

i

2000 ,

2500

3000

i

35O0

i

0.6

M

•

100 200

• •

-

E

LVZ,

"• 300 -

700

-

._>.

- -0.2 • 520-km

-

•

o.o .,9

-

600

0.2

410-km

cz.400 500

-"

C t 660-km

._.g, -0.4

o o

-0.6

Plate 1. (a) Crustal and (b) uppermostmantle velocity-interfacestructurealong the profile QUARTZ obtainedusing3 PNEs and all 48 chemicalexplosions.Locationsof the PNEs are indicated. Thick black line segmentsindicatethe reflectingboundariesimagedwithin the crustand mantle. P wave velocity valuesare given in the labels. Free-airgravityprofile is shownabovethe crosssections.Note the Moho relief, especially the rootsunderthe Ural and Altay Mountains,andthe Moho uplift underthe Mezenskayadepression. (c) Deviation of the 2-D velocity model (Plate lb) from the averagecontinentalcrust [Christensenand Mooney, 1995] abovethe Moho and from the IASP91 model [Kennettand Engdahl, 1991] below the Moho and betweenthe transitionzone discontinuities.Mantle transitionzone discontinuitiesimaged in this study are indicated.Note the low- and high-velocityzones,horizontalvelocity variationswithin the uppermantle, and mantle reflectorsat the depthsof 60-90, 120-140 and 129-170 km, interpretedas lithosphericdiscontinuities and labeled N, 8ø, and L. Also note the low-velocity zones interpretedwithin the asthenosphere (LVZ 1) andwithin the baseof the lithosphere (LVZ2 andLVZ3).

MOROZOVA

ET AL.' UPPERMOST MANTLE HETEROGENEITY

20,335

Reflection 12

10

•

6

2

-2

600

800

.... 8

1000

1200

1400

1600

1800

2000

of LVZ,

12

,•.,••, •••,.vRefi.•,5tio nfrom too•LV5 lO

•

o

-2

-1800 -1600 -1400 -1200 -1ooo-8o7• -600

lO

lO •-....?'•,':,•: ._•-,r> >•}5' :-; .•: -" ,7 topofLVZ 1

-2000 -1•00

-1600 -1400 -1200 -1000

-800

12

-600

Figure 4. Segmentsof the recordsfrom the PNEs (Figure3) showingindicationsof the lithosphericLVZ2 andLVZ3, strongrefractionsfrom beneathLVZ2, andreflectionsfrom the top of LVZ1 (Plate 1). Note the differencein the charactersof these seismogramsindicatingstronghorizontalvariationswithin the lithosphere. The largetravel time gap between1000 and 1200 km in the recordsfrom PNE 213 is partly due to the effectof a high-velocitymantleunderthe recordingspreadacrossthe easternflank of the Urals (Figure 3 andPlate 1). Also notea strongand coherentmultiplerefraction(whispering-gallery mode,labeledWG) indicatinga significantpositivevelocitygradientaboveLVZ 2.

20,336

MOROZOVA 70

ET AL.: UPPERMOST MANTLE

HETEROGENEITY

- -

65

60

- -

_

......................

50

tiiø-45

PNE 35

6

123 35

- -

30

30

1200

1600

2000

2400

2800 70

65

60 o

*

m4,om i i [t,i

55

60

I -!•

-

• 50

55

50

I

45

40

40

-

35

PNE

323 30

-2800

NW

-2400

-2000

Range (km)

- 1600

- 1200 SE

Figure 5. Far-offsetvertical componentrecordsfrom PNEs 123 and 323. Deep mantle refractionsand reflections,andthe whispering-gallery(WG) modesare indicated.Note the differencein apparentvelocitiesof the phasesP410, P520, andP660 fromthe two reversedPNEs. This differenceis interpretedas an indication of a SE dip of the 410-km discontinuity andof a horizontalmantlevelocityvariation(Plate 1).

sion (Plate 1). A large gravity high occursover the Mezen-

the Urals and a sharplocal anomalyon the eastflank of the skayadepression confirming theMohoupliftandhigh-mantle Urals (Figure3, PNE 213). Also, Pn from PNE 123 appears velocities(Plate la). Basedon PmP, the Moho depthis 39 to be blockedby the Urals,andfirst arrivalsat 300 to 800 km km under the Timan belt and 33 km under the Pechora basin

offsetto the SE are modeledas PN, a reflectionfrom within the uppermantle(Figure3, PNE 123). Thesedataindicatea For the Ural Mountains, chemicalshotsand PNE 123 show 30-km-thick, high-velocitylower crust with a 52-km-deep by high velocities(up to 7.9 a lowercrustalrefractionwith a velocityof 7.1 km/s(Figure root particularlycharacterized 2, shotpoint142,andFigure3, top) andPmP fromPNE 123 km in the lowermost crust and 8.4 km/s in the uppermost (Figure3, top, SE branch)has a distinctlylongerreduced mantle(Plate1). Thus a large amountof anomalouslyhightraveltimeof 8.5 s undertheUralscompared withPrnPto the velocitymaterialmustbe presentunderthe easternflank of NW under Pechorabasin. Upper mantle refractedarrivals the Urals. However, the availableray coveragedoesnot alfrom PNE 213 have a broadtraveltime advancejust eastof low unambiguousjudgmentwhetherthis high-velocitybody

(Platel;shotpoint119 in Figure2).

20,337

MOROZOVA ET AL.' UPPERMOST MANTLE HETEROGENEITY SE

PNE 323

PNE 123

NW -100

-500

2300

3100

2700

300 700 110015001900

3500

0

ß'' ' ' '

ß'

'

...... "": ..... "' •'•"•i• '•";•'"P5•oP

'

841o

•"•'%; ' P660

P•7• ........ ;.::" :::?:"" P•7o ..,.' .., ,

ß• 35[""

:ø• P 660

30-

-5oo -lOO 300

•

PNE323 700

...... ß

" .'.......

PNE.123 •

" ."1900

1100 1500

2300

2700

3100

3500

Distance (km)

Figure 6. (top) Rays traced through our velocity model from PNEs 123 and 323 (Figure 1), with (bottom the corresponding travel time fit. In the bottom plot, large dots represent the calculated travel times, small dots correspond to our travel time picks. Note the reversed coverage of the 410-kin discontinui and of the lid of LVZ 1 by the reflections from the two PNEs. Also note the asymmetry of the refractions under the 410kin discontinuity, indicating the horizontal velocity variation beneath the discontinuity (compare Figure 5).

a mantle event refracted under the 70-kin depth boundary PNE 323 (Figure 3) shows a short reflection from 65-kin is eraplaced into the crust or immediately below the Moho. depth followed in offset by a long phase refracted under this We do not observe Pm P reflections from the base of the Uraboundary. The first arrival from 850 to 1600 km occurs in lian root and constrain the root by the reflections from its two branches with apparent velocities of 8.5 and 8.7 kin/s, reconverging flanks, similar toSchueller et al. [1997]. Similar observations of the root under the middle Urals, about 700 kmspectively.

south of profile QUARTZ were made from wide-angle observations by Thouvenot et al. [1995] and under the southern Mantle Structure Between 80 and 410-kin Depth Urals from vertical incidence reflection data by Knapp etal.3.3. [1996]. Beyond 700-kin offset inPNE records, steps in the first arInthe West Siberian basin, Pm Ptravel times are almost rivals indicate the presence of two low-velocity zones in the constant (Figure 2,shot point 277), and no travel time uppermost mantle (Figure 4). The first of these zones is anomalies are observed in first arrivals from PNEs. Crustal marked by gaps between 700 and 900 km in the arrivals from thickness isconstant around 41 kin. Inthe Altay-Sayan belt, PNE 123 and 213 but isnot sampled by the energy from PNE pm P travel time increases and crustal thickness of about 45 323 (Figure 4). We interpreted this structure as a localized kmiswellconstrained. LVZ3 beneath ahigh-velocity mantle body under the eastern A major uppermost mantle reflection from an uppermost flank of the Urals (Plate 1). mantle boundary labeled N inPlate 1is traced nearly con-Between the offsets of1100 and 1500 kin, gaps inthe first tinuously across the entire profile. This reflection isrecorded

arrivals indicate an LVZ ofabroader horizontal extent. This from 10 chemical shots (labeled PN in Figure 2) and from all LVZ2 begins at 120to 140-kin depth and extends to about three PNEs (Figure 3). To the of PNE 123, the first arrival 170 km based on asmall velocity decrease and on imaged refrom 300 to 850 km offset isSE correlated as areflection from flections t•om itsbottom (Plate 1). the PN interface on the basis of its move-out (right branch of Atoffsets greater than 1400 -1500 kin, prominent events PNE i23, Figure 3). Here our interpretation differs f•om that are observed as first arrivals from the PNEs and sharply drop ofMechie etal. [1993] and Ryberg etal. [1996], who interin amplitude near 1550 1750 km although they can be traced preted the same event as an uppermost mantle refraction atin the records to 1800 -2000 km (Figure 4). These events offsets from 500 to 850 km with an apparent velocity of 8.1 have apparent velocities of about 8.7 km/s in all three PNE kin/s, ignoring the kinematics of this event at shorter offsets. records. Because these events are so strong and show some Similarly, the PN arrival from PNE 213 tothe NW resemwe interpret them as refractions from avelocit bles that from PNE 123 to the SE (Figure 3). To the SE, thecurvature,

Pn arrival from PNE 213 is very v•eak, and PN isfollowed by

20,338

MOROZOVA

ET AL.: UPPERMOST MANTLE HETEROGENEITY

gradientzone below 170-km depthand reflectionsfrom the top of LVZ1 (Plate 1). Turningwavesfrom 170- to 200-km depthsare followed

2450-2550 km offse• from PNE 323 and between 2330-2500

km fromPNE 123 are interpreted asrefractions from530 km depth(Figure5). We associate this boundarywith the dis-

by a strongdecreasein energyassociated with the major continuity firstobserved by Shearer[1991,1996]by stacking LVZ1 (Plate1). In the recordsfromPNE 213 (Figure3), the long-periodbottom-side ? and $ reflectionsand placedat effectof thisLVZ appears to be not so stronglypronounced. 520+4 km depth. After a reductionat 10 km/s, travel times of the 530-km reTraveltime modelingshowsthat this observation, although somewhat puzzlingfromtheviewpointof a 1-D interpretation fractionfromPNEs 123and323 areabout53 and49 s, re[Mechieet al., 1997],corroborates our 2-D model(Plate1). spectively(Figure 5). Apparentvelocityof this eventfrom The detailsof the mantlestructurebetweenapproximately PNE 123is about11.0km/s,andfromPNE 323 it is slightly 200- and 410-km depthcannotbe established because they but distinctlyhigherat 11.1 km/s. A strongeventobserved are concealed in the shadowzonecreatedby LVZ1. Never- only from PNE 323 and immediatelybehindthe refraction theless,as we show below, large-scaleconstraintson the structure of thisregion,suchasits averagevelocity,dip of its bottom,and horizontalvelocity gradient,can be established fromreversedrecordings of the deeperphases. 3.4. Mantle Transition Zone Discontinuities

Followinga gapof about7 s afterthe refractions from 170km depth,the next first arrivalsare associatedwith the 410km discontinuity (Figure5). The 410-kmrefractions are observed as first arrivals between offsets of about 2000 to 2400

(Figure 5) betweenthe offsetsof 2350 - 2550 km has been

interpreted asa reflectionfromthesamediscontinuity. Other PNEs do not showany similarevents,probablybecauseof the lowerenergiesof theseshots,andowingto the presence of complicatedwave field createdby the 660-km reflections [Grad, 1997]. The most distant travel time branches occur in first and second arrivals at offsets between 2550 and 2850 km from PNE 123 and between 2800 and 3150 km from PNE 323

(Figure5). Reduced(at 10 km/s)traveltimesareagainlower for PNE 323 (45 s at theoffsetof 2750km, ascompared to 50

km fromPNE 123andPNE323 (Figure5) andat thevery s for PNE 123) and apparentvelocitiesare 12.1 km/s for both

endof the records(1700 km) fromPNE 213 to theNW. The firstarrivalfromPNE 323 hasstronger amplitude thanfrom PNE 123 that is probablydueto the greatereffectiveness of theshot.The apparent velocityof thiseventfromPNE 323 is 10.2 km/s,andfromPNE 123 it is significantly lower,9.8 km/s, with measurementerror of not more than 0.05 km/s

(Figure 3). Also, prominentreflectionsfrom the 410discontinuityare observedbetweenoffsetsof 2300 to 2700

PNE 123 and323. Thesebranchingeventsare interpreted as a refractionandreflectionfromthe 660-kmdiscontinuity, respectively. As in the caseof the 410-km and 520-km discontinuities, the about 5 s difference in the travel times and the

corresponding differences in the offsetsrangesareinterpreted as due to a horizontaldecreaseof the velocity to the SE withinthe 200- to 520-kmdepthrange(Figure4).

kmfromPNE323and1900-2800 kmfromPNE 123(Figure 4. Two-Dimensional 5). These differencesin the arrivals from the two reversed Upper Mantle

Model of the Crust and

PNEsindicatea horizontallyvaryingstructure below200-km depth(Figure6).

The P wave velocity structurebasedon the observations

Reversedcoverageof the 410-km discontinuity by re- discussed aboveandpresented in Plate1 is characterized by a fractedandreflected wavesfromQUARTZPNEsis a unique numberof significantfeatures: 1. A prominentasthenospheric LVZ1 is observedbeneath featureof this profile,allowingobservations of laterally varyingstructure aboveandalongthisdiscontinuity. Theray about200-km depthin the centralpart of the profile and pathsand subsurface coveragebelowthe discontinuity are shallowingin the SE direction.A numberof smallerlow- and similarfor boththeseshots(Figure6), andtherefore, the ob- high-velocityzonesareidentifiedwithinthe upper100 - 150 servedtraveltimeasymmetry indicates a moderate SE dipof km of the mantle. thediscontinuity (Plate1 andFigure6). Moreover, asymme2. A high-gradient zone is foundaboveLVZ1, between try of thetraveltimecurvesof thedeeperevents(Figure6) about170- and 190-kmdepthin the middleof the profile. impliesthatthevelocitywithintheoverlying LVZ1 decreases This zonehasnot beenrecognizedbefore,but it is required in SE direction,and mostlikely, this horizontal velocity by the strongamplitudesof the refractionsfromthePNEs. variationextends beneath the 410-kmdiscontinuity, modeled 3. Three consistentsequences of reflectorsare present

withaNW velocity gradient of about 2.5x10-4 s-1(Plate1).

betweendepthsof 70 and 180 km. Thesereflectorsare ob-

Abovethe level of the 410-kmdiscontinuity, we placea 15-to 20-km-thickvelocitygradientzone. Althoughwe cannot resolverefractions from sucha thin layerat this depth, thiszoneis indicatedby the absence of precritical reflections from the 410-km discontinuity, as is describedin section6

servedfrom manychemicalexplosions and PNEs,spana considerable lateralextentof themodelandapparently areas-

below.

Observationof topsidereflectionsfrom the 520-km discontinuityrequiresdensesampling withina narrowoffsetinterval[Grad,1997],andthusthisboundary maybeelusivein wide-anglerecordings.Nevertheless, denseinstrument coverageof the DSS datameetsthis criterion,and the 520-km discontinuityhasbeenidentifiedin the recordsfrom a number

of DSS profiles[Ryberget al., 1996;Egorkin,1997]. In

sociatedwith regionalmantleboundaries. 4. The 410-km discontinuitydips southeastward with

mantlevelocities aboveandprobablybelowit decreasing in theSEdirection.Belowthe520-kmdiscontinuity, ourmodel becomes one-dimensional because of limitedsampling, and belowthe 660-kmdiscontinuity, our modelcorresponds to theIASP91model(Plate1). Although"local"LVZs arepresentin theuppermost mantle, the (interpreted) asthenospheric LVZ1 (Plate1) occursat shallowerdepthin the SE partof the profileunderthe West

Siberianbasinand Altay Mountainswhereit shallowsfrom QUARTZ records,weak, shorttraveltime segments about 180-kmdepthto about130 km. Within the lithosphere, ve-

MOROZOVA ET AL.: UPPERMOST MANTLE HETEROGENEITY

locitiesin the upper40 km of the mantleare generallyhigher to the NW underthe Baltic Shieldand East Europeanplatform. (Plate 1). The variabilityof the velocityof the uppermantleis com-

20,339

and 323 (Plate 1). We examinedthe trade-offof six parameters that controlour velocity model in this region:(1,2) two velocity values abovethe discontinuity,(3) velocity and (4) gradientbelow it, and(5) depthand(6) dip of the discontinuparable or exceedsthe differencesbetween the 1-D models ity (Plate 1 and Table 1). A more detailedsamplingof fineproposedpreviously[King and Calcagnile, 1976] (models scalelayeringwithin LVZ1 would apparentlyadd little to the obtainedfrom individualQUARTZ PNEs by Mechie et al. analysisof the 410-km discontinuity,only providing addi[1993]) andthusillustrates the limitationsof 1-D approxima- tional degreesof freedomwithin the low-velocityzone above tions. At the sametime, in agreementwith thesemodels,our the discontinuity. This structure,however,is constrainedby 2-D uppermantle structurerevealsa characteristic velocity the data only in an averaged,large-scalesense,and thuswe profile in northemEurasia,with velocitiesbetween100 - 200 prefer to maintain the same level of parameterizationin our and 530 - 650 km depthhigherby 0.2 - 0.3 km/s than in the testing. IASP91 model(Plate 1). The resultingscatterplotsin Figure 7 showthe acceptable modelsextractedin such a way out of 8000 randommodel perturbations. By measuringthe maximum extents of the 5. Model Uncertainty sampledareasin the plots of Figure 7, we obtainthe variaSincethe modelwasderivedby forwardmodelingandin- tionsof parametersallowedby our traveltime data(Table 1). terpretativeinversion using multiple and heterogeneous Note that for each of these parameters,this uncertaintyinmodelfit criteria,its uncertainty is difficultto evaluatein full cludesthe variationsof the other five parametersand is sigat present. Tomographicinversion(e.g., usingthe damped nificantly larger than the uncertaintiesthat would have been leastsquares approach by ZeltandSmith[1992])wouldpro- obtainedby varyingoneparameterat a time (Figure7). vide an estimateof the resolutionand modelvariance;howOur multi-parametricuncertaintyestimatesin Table 1 indiever, suchan estimatewould be highly dependenton the cate that the dip of the 410-km discontinuityand velocity dampingscheme employed anddifficultto interpret.Thusin gradientbeneathit are constrainedwith about 40 and 50% a recenttomographic studyof first arrivalsandPmP reflec- relative uncertainty, respectively. Velocity uncertainties tionsfromthe NW two thirdsof QUARTZ dataset,Schueller abovethe discontinuityare within about25% of the horizonet al. [1997] reported_+3to 5% variancesof crustalvelocities tal velocity variation in the model (Plate 1). Althoughthese and_+600m variancesin theMoho depth. At the sametime, uncertaintyvalues are significant,they are still within the Schueller et al. [1997]observed thatbecause of highly non- principal structuralcharacteristicsof the model. Thus the uniformray coverage,a modelthat is correctlydampedon conclusionfrom our model sensitivitytestingis that the dip averageappearsstronglyoverdampedin its bottomparts,re- and horizontal variation of mantle structure observed near the sulting in surprisinglylow (_+50m/s) velocity uncertainties 410-km discontinuity(Plate 1) cannotbe attributedto an amandlargespatialuncertainties (-300 km) in the regionswith biguityof traveltime inversion. low ray coverage.This problemshouldbe apparentlyaggraAlong with the travel time model sensitivity discussed vatedwhenusingstill sparserand lessuniformcoverageof above,phasecorrelationand picking representsanothersigthe deepermantle. nificant yet difficult to control sourceof uncertaintyin the In an attemptto overcomethis dependence of uncertainty interpretation. Phase correlationand cycle skipping in a measureson the intricaciesof inversion schemes,we develcomplicated,secondaryarrival wave field may be a difficult opeda directapproachallowingassessment of modeluncer- task requiring iterative refinementof the picks during the taintyusingforwardmodelingandMonte Carlo samplingof modeling using the complete set of available information. the modelparameterspace. The approachconsistsof multi- Note that as was discussedabove,phasecorrelationaccounts ple perturbations of the model followedby calculationof a for the major differencesof our model from the previousinnumber of travel time fit measures and assessment of the terpretations.Althoughnot tractableat present,the problem overalldatafit usinganartificialneuralnetwork (ANN). A of an adequateassessment of this sourceof modeluncertainty detailed description of thisprocedure is givenin theappen- (and errors)couldbe resolvedwith the developmentof autodix. mated,nonlinearinversionapproachesallowing generationof We appliedthis uncertaintyestimationapproachto an representative setsof alternativemodels("movies"[cf. Moseanalysis of the velocity/interface structure in thevicinityof gaard and Tarantola, 1995]). The Monte Carlo searchtechthedipping 410-kmdiscontinuity in ourmodelconstrained by niquedescribedin the appendixcouldgive riseto suchan apfour refractions and reflections from the reversed PNEs123 proach.

Table 1. Model Parameters, TheirVariationRangesandUncertainties Established by SIX-Parameteric MultivariateAnalysisof the Structureof Our Model Near the 410-kmDiscontinuity(Plate 1) Modelparameter

Observed value

Testedvariation

Uncertainty

Depth,km Dip, deg

400-410 2.8

+ 10 +3

+6 +1.2

Vabove410, at knl 800, km/s V•,ove410, at km 2500, km/s Vbelow 410, km/s S-I Gradient Vbdow 410

8.3 - 8.6 8.3 - 8.6 9.4 2.5'10'4

+2% (+0.16-0.18) +2% (+0.16-0.18) +2% (+0.19) + 1.4'10-4

+0.11 +0.11 +0.06 + 1.2'10'4

20,340

MOROZOVA

ET AL.' UPPERMOST MANTLE HETEROGENEITY

-2 2

-2

-2

-i .... 0.... i .... 2 bVnvz,left (%)

2

-2 2

+

-2

+

15Depth(km) 8 Dcpth(km) Figure7. Resultof modeluncertainty testingin thevicinityof the410-kmdiscontinuity (Plate1). The scatterplotsrepresent projections of thesix-parameter testdatasetcorresponding to theartificialneuralnetwork (ANN)-determined traveltimemismatches thatareconsidered acceptable (witha normalized traveltimemisfit valueof not morethan0.9; seeappendix).The uncertainties in modelparameters (Table 1) areestimated fromthe total extentsof the scatterplots,andthe mutualtrade-offrelationsbetweenall six parameters are indicatedby thecorrelations in thedistributions. Notethattheprimarytrade-offat wide-angleincidence comes fromanticorrelated velocityvariationsabovethe discontinuity.Alsonotethepresence of the secondcluster of modelswith reducedvelocityunderthe410-kmdiscontinuity, apparently associated with nonlineareffects of travel time modeling.

The velocity and depthuncertaintiesshownin Figure 7 are relatively low and certainly do not imply that, e.g., the velocities everywherewithin the imaged area deviatefrom the modeledvalues(Plate 1) by lessthan 0.1 - 0.15 km/s. On the basisof our analysis,we can only statethat the mantlevelocity averagedin such a way that it representsa smooth2-D patternis constrainedwithin such limits. However, geodynamic interpretationof the mantlestructureis basedon these averaged,regional-scalevelocity anomalies,and thus our model uncertaintyanalysisis consistentwith the target of suchinterpretation.The averagedvelocity valuesare, on the one hand, less subjectto fluctuationscausedby localized variations of mantle velocity and, on the other hand, well constrained by the traveltime data. Similarinterpretation appliesto the othermodelparameters includedin the preceding analysis.

6. Discussion

The determinationof a consistentshallow structureusing all 51 shotswas most importantfor our interpretation,becauseany discrepancies in the crustalstructurepropagated deepintothe mantle.We foundthatphasecorrelationandinterpretation of the crustalmodelhasthe mostprofoundeffect on the resultingstructure of the uppermost 20 - 50 km of the mantle (such as the high-velocitymantle body under the Mezenskayadepression).Also, identificationand modeling of mantlereflectionsare mostimportantfor the interpretation of the mantle structure.

As a result of a combinedand detailed analysisof the crustalanduppermostmantlestructureand of the useof numerousmantle reflections,we were able to resolvesignificant

uncertaintiesin the previous2-D study by Ryberg et al.

MOROZOVA ET AL.' UPPERMOST MANTLE HETEROGENEITY

20,341

[1996]. At the sametime,thegeneralfeaturesof ourresulting 2-D velocitystructuresupportthe modelsderivedpreviously lEgorkinand Mikhattsev,1990;Egorkin,1991;Mechieet at., 1993]. Our modeldemonstrates the strongheterogeneity of the uppermantle,modifiesandreconciles the contrasting details of the mantlelid presentedin the three 1-D modelsderived by Mechie et at. [1993] for the three QUARTZ PNEs. Below, we discussthe importantfeaturesof our interpreta-

sent a systematicpatternthat they can be tracedthroughout most of the imagedareaof the mantle(Plate 1). This observation suggestsassociationof thesereflectorswith (at least) regionallithosphericboundaries.All threereflectingboundaries becomedeeperunder the sedimentarybasins,and they

mic data.

indicative

become shallower

under the Baltic Shield and toward the Al-

tay Mountains. Also, the low-velocity zone LVZ2 in our model is apparentlycloselyassociated with the lower two of tion. theseboundaries(Plate 1). The upperof theseboundariesmay be associated with the 6.1. The Crust and the Uppermost Mantle regionalN discontinuityalso found from other DSS profiles Probablythe mosttediousandthemostdifficultpartof the (Plate 1 andPavtenkova[1992]). The depthof this discontiinterpretation is the analysisof the crustaldataset(48 shots) nuity and velocity contrastacrossit (about 8.1 to 8.4 km/s; in which we includedonly the clearerevents,without their o,• mo,• •) ....... *• •*...... •*•,,, with a lithosphericdisenhancement by nonlinearvelocityfiltering. As a result,our continuityidentified by Hales [1969] in the central United crustalanduppermost mantlemodel(Plate 1a) is simplerthan States,at a depth of 80 - 90 km and with velocity increase the modelsby Egorkin and Mikhattsev[1990] and Egorkin from 8.05 to 8.3 - 8.45 km/s. Hales [1969] suggestedthat such seismicvelocity increasemight be due to a transition [1991]. from spinelto garnetperidotite. Our studyconfirmsthe existenceof the Uralian crustalroot We interpret thelowertwoboundaries bordering LVZ2 observedearlier in the southern[Knapp et at., 1996] and middle Urals [Thouvenotet at., 1995], which is the main (Platel) as the 8ø [Thyboand Perchuc,1997a]andLehmann As proposed by Thyboand controversyconcerningthe deep structureof the Urals. The [Lehmann,1959] discontinuities. marksa changein the significant,high-velocity crustal root makes the Urals the Perchuc[1997a] the 8ø discontinuity only Paleozoicorogenin the world that showssucha feature. natureof seismicphasesat about8ø rangesin continentalreThe 14-kmrootcrossed by the QUARTZ profileis morepro- gionsand might be a global featureof the continentalmantle. nouncedthanthe rootunderthe middleUrals(6 km [Thouve- Seismically,the mantle layer beneaththis discontinuityis not et at., 1995]). The deeperroot is probablyrelatedto the characterizedby lower velocitiesand increasedscattering highertopography(1500 m, as comparedwith 500 m in the [ThyboandPerchuc,1997a]. In QUARTZ data,thisLVZ 2 is clearlyindicatednot only by the traveltime delaysin the arrimiddleUrals). by Anotherstrikingfeatureof the Urals is the vast amountof valsfrom PNEs (Figure4) (as hasbeenalsodemonstrated as well [Mechieet at., 1993;Ryberget high-velocitymaterial(>7 km/s)in thecrustanduppermantle earlierinterpretations on the easternflank of the belt. The Uralian root is, at least at., 1996]) but also by a consistentsequenceof reflections from its top. partially,supportingthis abnormallydensecrust. The layer below the 8ø discontinuitycorresponds to a parOtherprominentfeaturesof ourcrustalmodelarethe highvelocity Archcancrust of the Kola Peninsula,Moho uplift tially moltenzoneat the baseof the lithosphere.Perchucand Thybo [1997] also suggestedthat this partially molten zone underthe Mezenskayadepression, and the approximately5extends to the Lehmann discontinuityin tectonicallystable km root of the Altay-SayanMountains. Gravity measurecontinental areas,while it may continueto the 410-km dismentsalongthe profile (Plate l a) supportthe sharpcontrasts in the Moho depth and lateralvelocity (and density)varia- continuityin hot (tectonicallyactive) regions.Thus the variationsin the lower crustanduppermantleobservedfrom seis- tionsof the thicknessof this zonelike thosein Plate 1 may be of the thermal state of the continental

mantle.

At the bottomofLVZ2, a 30-km thick (in the middle of the An additionalsupportfor the observedverticaland lateral variation of the deep lithosphericstructurecomesfrom the model) layer of increasedvelocity and gradientmay be assoanalysisof the long-rangePn phasethat has also been ob- ciatedwith the baseof the lithosphere.The top of this layer servedin QUARTZ records[Ryberget at., 1995; Morozovet is imaged by reflectionsfrom the PNEs nearly continuously at., 1998a]. As we recentlyobserved[Morozovet at., 1998a], betweenkm 300 and 1800 of the profile (Plate 1). NW of km the amplitude-frequency pattern of the QUARTZ PNE rec- -1000 of the profile, the base of LVZ2 tilts upward, apordsrequiresa relatively sharpincreasein the attenuationat proachingthe 8ø discontinuity.At the sametime, the top of LVZ1 indicatedby the reflectorat about the level of the 8ø discontinuity(about 150 km in the middle the asthenospheric 220 km depth can still be tracednearly horizontallyin a NW of our model;seePlatel). Also, Morozovet at. [1998a] sugdirection shallowing to the SE after km 1600 on the profile gestedthat the relatively high frequencymultiple refractions 2xPn and 3xPn propagatemore efficientlyin the NW direc- (Plate1). Althoughthe experimentdoesnot providereversed tion, where the deeperbranchof the latter phaseis observed. PNE ray coveragefor a more detailedstudyof the marginsof This impliesthat thesewaves shouldbe able to dive deeper the model, it can still be concluded:1) the lateral extent of LVZ2 is limited from aroundthe Pechorabasin to the West than about 150 km depth without significant attenuationin the NW part of our crosssection,a featurethat is positively Siberianbasin,or it is thinningoutsidethis region(Plate1), 2) the top of LVZ 1 is at about 220 km depth under the East ruled out in the middle of the model. Europeanplatform and shallowsto 160-150 km under the Altay fold belt. 6.2. The LithosphericDiscontinuities

Utilizingthe largeextentof the presentstudyandits reversedcoverageof the mantle,we demonstrate that the re-

6.3. Transition

Zone Discontinuities

Observations of the mantlediscontinuities are of key im-

flections withintheupper65 - 150kmof thelithosphere pre- portancefor the interpretationof the mantle transitionzone.

20,342

MOROZOVA

ET AL.: UPPERMOST MANTLE HETEROGENEITY

QUARTZ recordsallowed us to identify all three transition zone discontinuitiesusing both refractionsand reflections from the reversedPNEs 123 and 323 (Figure5). As an only exception,a reflectionfrom the 520-km discontinuitycould not be found in the recordsfrom PNE 123. However, refractions and topsidereflectionsfrom this discontinuityare very sensitiveto the overlyingstructureandto the spatialsampling [Grad, 1997]; the arrivals from this relativelyweak discontinuity may be lurking within the recordsection[Priestleyet at., 1994]. Thus althoughthe universalityof this featureis currentlyan issueof debate[Cumminset at., 1992; doneset at., 1992], our data contributean evidencefor a 520-km discontinuityundernorthernEurasia. Note that the weaknessof the impedancecontrastacrossthis boundary,possiblevariations of its depth,the complexityof the overlyingstructure lead to a failure to observethis boundaryon regionalDSS stacksanalyzedby Ryberget al. [ 1998]. The sharpness of the 410-km discontinuitycanbe bestestablishedthroughthe observationof short-periodprecritical reflections,at rangesbetween1200 and 1400 km. Using this technique,Fidale et at. [1995] analyzedshort-periodrecordingsfrom NTS nuclearexplosionsand observedprecriticalreflectionsfrom the discontinuityunder the basin and Range Province, thus estimatingthe sharpnessof the impedance contrastas lessthen 6 km. Wide-angleDSS recordingsalso allow suchobservations, althoughthe signal/noiseratio of the DSS recordingsis poorerthan that of the permanentarrays usedby Fidale et at. [1995]. Based on the absenceof precriticalreflectionsfrom the 410-km discontinuityin PNE data from the DSS profile RIFT in centralSiberia,Priestleyet at. [1994] suggestedthat this discontinuitycouldbe as broadas 35 km.

Using an "averaged"1-D velocity column (Table 2) derived from our velocitymodel(Plate 1), we calculateda series of reflectivitysynthetics(Figure8) [Fuchsand Miilter, 1971; Mailer, 1985] and comparedthem with the recordsfrom the PNEs 123 and 323. Since the existing techniquesof 1-D modelingof amplitudecharacteristics of PNE recordsdo not reflect suchimportantcharacteristics as crustalscatteringand effective sourcedirectivity, amplitudecalibrationthroughout the entire lengths of the records still presentsa problem. However, the waves illuminatingthe transitionzone discontinuitiesspana very limited rangeof take-offangles(Figure6), andthuswe can safelyassumethat the aforementioned effects have a small influenceon the amplitude-offsetdependenceof the reflections from these discontinuities.

Calculationsshow that for a sharp410-km discontinuity, high-amplitudereflectionsshould be observedwithin the precriticaloffset rangeof about 1300 - 1500 km (Figure 8). This is clearlynot the casein the data (Figure3), wherethe reflection from the 410-km discontinuity cannot be seen above the noise level until beyond the critical distanceof 1500 km. On the basis of this observation,we conclude that

the 410-km discontinuitycan be more likely represented by a zone of velocity gradientof at least 15-km thickness. This gradientzone is similarto the observations by Priestleyet at. [1994], althoughwe set a more moderateconstrainton its thickness,but it contrastswith the resultsof the observations under the tectonically active areas of the western United States[l/idate et at., 1995]. Similar amplitudeconsiderations usingreflectivitymodeling canbe appliedto the 660-km discontinuity.As Figure8a shows,the 1-D velocitystructurewhich is kinematicallycor-

Table 2. Velocity and attenuationmodelused in our syntheticsimulations(Figure8). Depth,km 0 38 38 88 118 136 136 151 194 194 388 412 503 519 616 631 685

Vp,km/s

Qp

6.15 6.15 8.05 8.10 8.44 8.50 8.30 8.25 8.62 8.23 8.93 9.40 9.78 10.10 10.40 10.80 11.00

800 800 1400 1000 1000 1000 600 600 800 800 1300 1300 1300 1300 1300 1300 1300

The velocitycolumnis extractedfrom the middl of our 2-D model(Plate 1). The attenuationis derived from 2-D model by Morozovet al. [1998b] and represents an increasein the attenuation within the lower lithosphere, as

suggested by Morozovet al [1998a](136 km in this model). The corresponding S wave parameterswere set as Vs=0.577Vp,andQs=0.75Qp.

rect and consistentwith the IASP91 model includinga sharp velocity contrastat the 660-km discontinuity,resultsin relative amplitudesof the reflectionsmuchhigherthan observed in the records(Figure3). In particular,this discrepancy is indicatedby the presenceof strongnear- criticaland precritical reflectionsthat are not found in the data(Figure 5). The observationsare betterapproximatedby a model with the 660km discontinuityrepresented by a 20-km wide gradientzone

(Figure8b), resultingin shiftingof the peakreflectedenergy beyond 2200 - 2300 km. This conclusioncorroboratesthe GNEM modelof Ryberget at. [ 1998],who suggested thatthe velocity contrastacrossthe 660-km discontinuityshouldbe abouthalf of the one presentin the IASP91 model. Note, however, that Ryberg et al. [1998] inferred their GNEM

model from a stack of several DSS records (including QUARTZ) corresponding to the regionsseparatedby a few thousandkilometers. Such stackingprocedureintroduces smearingof the discontinuities(especiallyin the precritical partsof the reflections)andmixing of differentstructures, and thereforeit may be misleadingfor an analysisof local characteristics of the boundaries.

Nevertheless, individual

QUARTZ recordsalso suggesta weaker, and most likely, gradientimpedancecontrastacrossthe 660-km discontinuity. 6.4. Causesof the Uppermost Mantle Heterogeneity

With the observedstrongheterogeneityof the uppermantle, the problemis to explainthe causesof this heterogeneity in termsof physicalpropertiesof rocks,keepingin mind that, in general,contrastsin velocity-densityrelationships are distinctly smallerfor the uppermantlethanfor the crust. Coherent

mantle

reflections

similar

to discussed in this

studyhavebeenobservedin a varietyof continentalgeologic

MOROZOVA ETAL.:UPPERMOST MAN•?LE HETEROGENEITY

65

65

55

55

,• 30 •

3000

' I I.[[LLLI I I I IJ///J..!.I_LLLI I2000 I I IJ/J//L.LLLLI I I I1500 IJJJlJ. J./'LLJ t11l_• 2500

30

• 70

70

'• 65

65

55

, ., ...,..•-,., •_., ,

55

50

50

45

45

40

40

35

3000

2500

2000

20,343

1500

i]'•••,•_..: 1 35

Range (km)

Figure8. Resultsof synthetic modelingin a simplified1-D velocitystructure approximating our2-D velocity model(Plate1): (a) withsharpvelocitycontrast atthe410-kmand660-kmdiscontinuities, and(b) 15-km gradienttransition at the410-kmdiscontinuity, and20-kmtransition zoneat the 660-kmdiscontinuity.The velocityandattenuation columns usedin thesimulation aregivenin Table2. Notethatin the caseof sharp boundary(Figure8a) the pre criticalreflectionbetweenthe offsetsof 1400 - 1600km is comparable in amplitudeto the first arrivals. Also notethat becauseof an asymmetryof the observedreversedPNE records (Figure6), 1-D simulationis notableto matchthetraveltimesfrombothPNEsaccurately.

environments in wide-angle[Bairdet al., 1995;Pavlenkova The uppermostmantle in our model (Plate 1) is stratified et al., 1996,Pavlenkova andKvyatkovskaya, 1992;Perchuc and "blocky,"as was suggestedby Pavlenkova[1996b], with and Thybo,1997;Hajnal et al., 1997] andvertical-incidence lateralvariationsin velocity beneaththe Moho rangingfrom

[Calvertet al., 1995;Knappetal., 1996;Warneret al., 1996] 8.1 to 8.4 km/s. While velocitiesup to 8.4 km/s within the seismic experiments. However,theunderstanding of thena- top 20 - 30 km of the mantle(Plate 1) couldbe causedby aniby ?avlenkova[1996b] (velocityin ture of theseeventsremainscontroversial.In general,the sotropy,aswas suggested possible causes of the observed sub-Moho velocitycontrasts singleolivine crystalsrangesfrom 7.9 to 9.1 km/s according andreflectivity are as follows:(1) velocityanisotropy pri- to the direction),they could also be causedby compositional marilycausedby olivine,(2) compositional differences such variations. Velocities of 8.4 km/s are found in dunites,perias pyroxenite versusperidotite, eclogite,or garnet-rich peri- dotites,and eclogites[Christensen,1982] althoughthey are

dotite,(3) phasechange,(4) magmain the formof partial abovethe commonvalues for all theserocks. Since gravity melt,(5) temperature effects, (6) freehydrothermal fluids,(7) anomaliescould not be causedby olivine anisotropy,we prefaultzones,(8) piecesof delaminated lowercrustincluding fer phasechangeandcompositional variation(suchas garnet eclogite,and (9) zonesof fluid alteration. An aid to distin-

peridotiteor eclogiteemplacedin spinelperidotite)as an ex-

guishing thevariouspossibilities is thetectonicsetting,i.e., planationfor the highvelocityunderthe Mezenskayadepresfaultzones,subduction zones,partialmelting,delamination, sion and under the eastern flank of the Urals. etc. A primarycriterionin explainingvelocityvariationsis whethertheyaresharpor gradational, andanotherwouldbe if

reflectioncoefficient were plusor minus(or of a LVZ) whether it couldbereliablydetermined [Priceetal., 1996].

Our interpretation revealsa strongvelocitygradientbelow 60-80 km [also confirmingmodelsby Mechie et al.[1993] and Ryberget al. [1996]). This gradientis too large to be causedby pressureeffectsalone [Kern and Richter, 1981;

20,344

MOROZOVA

ET AL.: UPPERMOST MANTLE HETEROGENEITY

Dufj• and Anderson,1989]. The possibleexplanations are a gradational compositional variation,a phasechange,andanisotropy. Within this depthrange,garnetshouldappearsin peridotitc[Hales, 1969; Wyllie,1995] leadingto an increase in velocity. Likewise, recrystallization to form jadeitc and garnetin an eclogite,leadingto a morepiclogiticcomposition [Dufj• andAnderson,1989],wouldalsoform a velocitygradient. Anisotropycausedby olivinefabricin dunitcor peridotitc could yield velocities as high as 8.5 to 8.6 km/s [Christensen,1982], but such an origin would imply increasedolivine preferredorientationand thereforeincreased strainwith depth. Xenolith studiesdemonstrate greatuppermost mantlevariability, includingthe presenceof hydratedphases,but do not revealthe scaleof thisvariability. Someof the regionallyreflectinginterfacesmightbe represented by fine-scalelayering whichwould increasereflectivityin the presence of relatively smallcontrastsin acousticimpedance.Someof the observed reflectivediscontinuities could be causedby phlogopite-rich layersthat would representa reservoirfor fluidsknownto be

6.5. Tectonic Implications

Associatingthe baseof the lithospherewith the top of the low-velocity (and also attenuating[Morozov et al., 1998b]) layer LVZ1 in Plate 1, we observethat the lithosphericthicknessunderthe East Europeanplatformis about200 km, decreasingtowardthe Altay Mountainsto about120 km. This is consistentwith the agesof the lithosphere,rangingfrom Archcanto Proterozoicwithin the East Europeanplatformto a Paleozoicage of the accretedterranesat the SE. The lithospherethinstowardthe activeHimalayanorogenicbelt; also, underboththe EastEuropeanplatformandwesternSiberia,it is significantlythickerthan underthe Europe,whereits averagethicknessis about100 km [Ryberget al., 1996]. Our model shows a low-velocity layer LVZ2 locatedat varying depthwithin the baseof the lithospherethat also exhibits a significantincreasein seismicattenuation[Morozov et al., 1998b] (Plate 1 and Figure 9). Theseseismicmanifestationscorrespondto a layer of softerand probablypartially molten mantle material [Mavko, 1980] that has been presentin the uppermostmantle. suggestedin other stable tectonic regionsby Perchuc and The thickeningof LVZ2 in the middle of the profile beThybo[1996]. The increaseof the attenuationmay be partly tween the 8 ø and Lehmann discontinuities and the observed due to increasedheterogeneityof the hotter and softerrocks high attenuationwithin this zone [Morozovet al., 1998a,b] within LVZ2, leadingto increasedscatteringof seismicwaves suggestthatthis part of the lithospheremightbe mechanically [Thyboand Perchuc, 1997a, b]. The presenceof this feature weakened,possiblyhydratedand/or contain melts [Wyllie, is corroboratedby the observations of a moderateincreasein 1995; Pavlenkova,1996b; Thyboand Perchuc, 1997a]. As heat flow underthe West Siberianplatformcomparedto the an alternative, a broad, delaminated slab from the lower crust Baltic Shield area [Kutas and Smirnov, 1991]. Thyboand or uppermostmantle appearsless likely becauseof the obPerchuc [1997a,b] suggestedthat this partially molten zone servedincreaseof attenuationin this zone. The presenceof mightbe a globalcharacteristic of tectonicallystableregions, this featureis alsocorroborated by the observations of a mod- relatedto a small (