Mar 10, 1993 - reflection points under North Korea, which has laterally varying 31 to 36 km ... wave impedance contrast around this depth under North Korea.
JOURNAL OF GEOPHYSICAL
RESEARCH, VOL. 98, NO. B3, PAGES 4389-4405, MARCH 10, 1993
Investigationof Upper Mantle DiscontinuitiesNear Northwestern Pacific SubductionZones Using Precursorsto sSH ZHI ZHANG and TtIom•
LAY
C. F. RichterSeismological Laboratory andInstituteof Tectonics, University of California,SantaCruz
Teleseismic long-period Word-WideStandard Seismograph Network(WWSSN)tangential component
recordings of deepandintermediate depthearthquakes areanalyzed for thepresence of sS precursors, denoted sxS, signifying underside reflections fromdiscontinuities at depth x belowthesS reflection point abovethesource.ThesesS precursors canbe usedto placeconstraints on uppermantlediscontinuities in
thevidnityof subducfion zones.Theclearest precursor is usually thesinsphase, theunderside reflection from the Moho, whichis observed for a numberof eventsin the northwestem Pacificfor pathsreflecting
undertheSeaof Okhotsk andNorthKorea.Theamplitude andtimingof s,•S relativeto sS aremodeled usingsynthetic seismograms to determine theshearwaveimpedance contrast at theMohoandthecrustal thickness in theseregions.Theresults arecompared withprevious workonpP precursors reflected from
theMohofor similar paths,withconsistent crustal thickness beingfound.s,•S is strong for continental reflection points under NorthKorea,which haslaterally varying 31 to36 kmcrustal thickness and31.2%to 18%shearwaveimpedance contrast at theMoho.In thenorthem Seaof Okhotsk region thecrustal thicknessis around 21 to 29 km with26%to 28%shearwaveimpedance contrast at theMoho.Stacked sins waveforms for theSeaof Okhotsk varyfromstrong, isolated arrivalsin thenorthto weak,poorlydefined
precursors toward thesouth, reflecting a transition fromcontinental to oceanic crust.s,•S is notclearly separated fromsS foroceanic paths under theIzuandJapan regions, andourprocedure cannot resolve oce-
anicMohoproperties unless broadband dataareused.Other precursors, sxS,aremuch weaker thansins
andaredifficultto identifyin individual waveforms. We useslantstacking to enhance the signalto noise
andsearch forprecursors withvarious slownesses, using recordings from13deepevents. FourSII wave reflectors mayexistbelowtheMohoin ourstudyarea.Theshallowest is the"80-km"discontinuity, which varies in depth laterally, beingnear80 to 85kmbelowtheSeaof Okhotsk, near66 kmbeneath IzuJapan, and near90 km underNorth Korea. A reflectornear200 to 210 km depthis indicatedby datafrom
Japan, butthereisnoevidence foranyshear waveimpedance contrast around thisdepth under NorthKorea or the Sea of Okhotsk. Weak evidenceis found for a "330-km" discontinuityin theseregions,with the
depth of a fewpercent impedance contrast varying from325to 335km. Relatively strong andconsistent arrivals indicate thepresence of a "400-km" discontinuity, withthedepthvarying from380 to 400 kin, perhaps indicating slight elevation oftheolivine•>• phase transformation neartheslabs. INTRODUCTION
impedancecontrastsat the Moho are about 29.2% and 32.6%, respectively. Lerner-Lam and Jordan [1987] constructed
Investigation of uppermantlediscontinuities is an important models EU2 and PA2 for northern Eurasia and the western step in understanding fine structure,dynamicsand chemical PacificOcean,respectively, usingfundamental andhigher-mode evolutionof Earth. Of the variousremotesensingtechniques Rayleighwaves. The shearandcompressional impedance conemployed,seismology playsa prominentrole in studyingman- trasts at the Moho are 26.0% and 28.8% in EU2 and 36.8% and tle structure. By the 1930s, the gross radially symmetric 38.4% in PA2, respectively.These modelsimply that the seismicvelocity distributioninsidethe planet was established, impedance contrastat the Moho is significantly lower in conshowingregionsof different seismiccharacter(crust, upper tinentalthan in oceanicregions. However,there are many mantle, lower mantle, inner and outer core) corresponding to regions for which Moho properties,particularlyfor shear gross compositionaldivisions of Earth. With the advent of waves, remain undetermined. digital data and refinement of interpretationaltechniques, The 410- and 660-km discontinuitiesare global features modernseismologyhasrevealedthat the uppermantlehasgloexploredby precursorsto tvP ' [Whitcomband Anderson,1970; bal layering associatedwith the Moho, "410-km", "520-km" Husebye et al., 1977; Nakanishi, 1986, 1988], waveform and "660-km" discontinuitiesand other regionally varying modeling[Burdick and Helmberger,1978; Grand and Helmstructuressuch as the "60-km", "80-km", "220-km" and "330-
km" discontinuities.
berger, 1984a,b;LeFevre and Helmberger,1989], stackingof reflectedand convertedphases[Bockand Ha, 1984; Wajeman,
The seismological Moho, definingthe upperboundaryof the 1988; Paulssen, 1988; Bowman and Kennett, 1990; Richards mantle, has been widely mappedbasedon the abruptincrease and Wicks, 1990; Shearer, 1990, 1991; Vidale and Benz, 1992], of seismicvelocity with depth inferred from earthquakeand andmigrationof manfiereverberations ScSn[Revenaugh, 1989; controlled source seismology,mainly using PmP (reflections Revenaughand Jordan, 1987, 1989, 1991a,b]. Basedon these from the top of the Moho) or Pn (Moho head waves). In the previousstudies,averageP velocityjumpsof 5-6% at 410-km preliminary reference Earth model (PREM) [Dziewonskiand and 4% at 660-km, and S velocityjumpsof about5% at 410Anderson, 1981], the average shear and compressionalkm and 7-8% at 600-km are inferred all with a + 1-2% uncerCopyright1993by theAmericanGeophysical Union. Papernumber9ZIB02050. 0148-0227/93/9ZIB-02050505.00
tainty. The long-wavelength(>50 km) topographyfor these two discontinuities is on the order of + 20-30 km. Ringwood [1975] suggested that the 410-km discontinuity is causedby the phasechangefrom olivine to •-phasefor (Mg,Fe)2SiO4. The 4389
4390
ZHANaANDLAY: UPPERMAm'LEDiscosTiNUrrmsABOVEDEEPSLABS
670-km discontinuitycan be explainedby the transition),-spinel such as partial melting or anisotropicfabric. Resolvingthe --> perovskite + magnesiowttstite[Liu, 1976, 1979; Jackson, natureof thesediscontinuitieswill undoubtedlycontributesub1983; Ito et al., 1984; Ito and Takahashi, 1989]. Gamets also stantiallyto understanding the dynamicsof mantleconvection. transformto perovskitestructurenear this depth with a two- Of course,no singlemethodor data set can hopeto resolveall phase region several gigapascalsin extent (not an efficient propertiesof a given discontinuity, and a varietyof methods reflectorat high frequency). mustbe usedto fully characterizethe uppermantle. The 520-km discontinuityis potentially a global feature In this paper we focus on analysisof precursorsto the [Whitcomband Anderson,1970; Helmbergerand Engen, 1974; transversecomponemsSH phase(below we write S only for Jones and Helmberger, 1990; Revenaugh,1989; Revenaugh the SH component,which we use exclusively)on long-period and Jordan, 1991b; Shearer, 1990, 1991] but is very weak. recordingsfor deep earthquakes to analyzethe regionalmantle The shearimpedancecontrastat 520-km discontinuityis about structureabovethe sources. Our procedureis mostsensitiveto 3% [Revenaughand Jordan, 1991b]. The phasetransitionof mantle structure above and near to subductingslabs, where [l-phase--> ),-spinelis a possibleexplanation for the 520-km strong thermal and chemicalperturbationsmay affect mantle discontinuity[Whitcomband Anderson,1970, Ringwood,1975; discontinuities. We determine crustal thickness and shear wave
impedancecontrastat the Moho usingsinS, the Moho underOtheruppermantlediscontinuities havebeenproposed based side reflection; and investigateupper mantle discontinuities at x km on seismological observations,such as the 60-km, 80-km usingsxS, undersidereflectionsfrom discontinuities Weidner et al., 1984].
[Revenaughand Jordan, 1991c], 220-km [Jordan and Frazer, depth in the mantle above deep sourcesin the northwestern 1975; Hales et al., 1980; Drummondet al., 1982; Bowmanand Pacificregion. Kennett, 1990; Revenaughand Jordan, 1991c; Vidale and DATA Benz,1992], and330-km [Wajeman,1988;Revenaugh andJordan, 1991c] discontinuities,but the universalityof these Seismic waves that encounterdiscontinuitiesin the upper featuresis not yet established nor is their underlyingphysical mantle spawn many phases. We use only transversecomexplanation. ponents(SH waves) in this study since they have particularly It is critical to determinethe propertiesof all uppermantle simpleinteractions,with no conversionof S to P energy. Data discontinuities and to establishwhether they are global or from 23 intermediateand deep eventsare analyzedin this study regional structures. Propertiesthat seismologycan resolve (Table 1). Figure 1 showsthe locationof these23 earthquakes include the sharpnessand size of velocity, density, and in the northwestern Pacificregion. The depthsof theseevents impedancecontrastsof these discontinuities.This information range from 115 to 580 km, and the magnitudesvary from 5.6 is neededfor mineral physicsexperimentsto determineif the to 6.0. Table 1 lists the Intemational Seismic Centre (ISC) discontinuities can be explainedby phasechangesor, failing sourceparameters,along with focal mechanismsfrom Gaherty that, must be associatedwith compositionalor other changes and Lay [1992]. TABLE 1. Earthquake Parameters (ISC) for EventsUsedin This Study No.
1
2
Date
March 18, 1964
Time, GMT
Latitude, Longitude, Depth, deg deg km
mb
Strike, deg
Dip, deg
48
84
04 37:25.7
52.56
153.67
424
5.6
28.96
128.23
195
6.0
Rake,
deg -76
3
Sept.21, 1965 July 4, 1967
01 38:30.3
23 42:12.9
43.10
142.58
157
5.6
274
88
8O
4
Dec. 1, 1967
13 57:03.4
49.45
154.40
144
5.9
50
87
109
5
May 14,1968
14 05:05.4
29.93
129.39
162
5.9
March 31, 1969
19 25:27.0
38.49
134.52
397
5.7
38
75
230
52.28
151.49
560
5.7
3
72
-90
51.69
150.97
515
6.0
34
72
-110
-94
6
Sept.5, 1970
07 52:27.2
8
Jan. 29, 1971
21 58:03.2
9
May 27, 1972
04 06:49.0
54.97
156.33
397
5.7
24
85
49.47
147.08
573
5.9
15
17
47
28.22
139.30
508
5.9
317
72
-74
24
17
107
205
79
8O
7
lO
Aug. 21, 1972
06 23:48.6
11
Jan 31, 1973
20 55:54.2
12
Sept. 10, 1973
07 43:32.3
42.48
131.05
552
5.8
13
Feb. 22, 1974
00 36:54.6
33.17
136.98
391
5.9
14
Sept.21, 1974
15 54:59.1
52.19
157.44
119
5.7
15
Dec. 21, 1975
10 54:17.2
51.93
151.57
546
6.0
16
July 10, 1976
11 37:14.0
47.31
145.75
402
5.8
41
89
-87
17
Dec. 12, 1976
01 08:51.1
28.04
139.67
503
5.8
328
72
-74
18
May 23, 1978
07 50:28.3
31.07
130.10
160
6.2
19
June 21, 1978
11 10:38.7
48.27
148.66
380
5.8
288
32
32
20
Sept.2, 1978
01 57:34.2
24.81
121.87
115
6.0
34
28
138
21
Aug. 16, 1979
21 31:24.9
41.85
130.86
566
5.8
56
24
133
Feb. 1, 1984
07 28:28.7
49.05
146.63
580
5.9
231
85
83
April 20, 1984
06 31:10.6
50.12
148.75
572
5.9
257
16
-74
22 23
ZHANGANDLAY: UPPERMAs'n_•DISCONTINUITIES ABOVEDEEPSLABS
[
} \
•1
500
d 12/01/67
09/1
400
01/29/71
\ o4nom•,
4391
__-
I'•08/16/79 •
.,•
30 o•
7
05/23fi8•-'
3 09/2v6s ß.
iit3
200
I 120 ø
I
I
140ø
I 160ø
Fig.1.Earthquake locations (triangles) for23events inthenorthwestem Pacific forwhich weanalyze data. Thenumbers beside each location gives thedateoftheevent (month/day/year). Table1 indicates theother source parameters.
contrast abovethe The sensitivity of a reflected seismicwaveto the vertical be anydepthat whichthereis animpedance source. Underside reflections from the Moho are given the speextent(sharpness) of thevelocityor impedance increase across if thesourcedepthis 560 km, a discontinuity is a function of thesignalwavelength [Richards,cial names,•S. For example, maybe generated by theMoho,80-km,2201972].Largewavelength, long-period wavesvertically incidentthenprecursors if theyexist, ona boundary will bereflected effectively onlyif thechange in km, 330-kin,400-km,or 520-kindiscontinuities and are then labeled as s,,S, ssoS, s22oS, s33oS, s4oosand material properties is distributed overa depthinterval lessthan In the teleseismic distance rangeof our -1/4 wavelength. Nearverticalincidence reflection coefficientsss20S,respectively. areonlysensitive to impedance contrast (A(velocity xdensity )). data, the core-mantleboundaryreflectionScS may arrive in thes•S interval.Figure2b is a Thus,if we analyzelong-period wavestravelingnearlyverti- betweenS andsS phases sketch of a teleseismic waveform correspondingto the ray callythrough theuppermantle, we canconstrain gross characof Figure 2a,suggesting thatat some distances eachs•S teristicsof uppermanfieimpedance structure.We examine paths be wellseparated fromScSandthusobservable. Figure long-period SH wavespropagating upwardfromdeepearth- should 2c showstraveltime curvesfor S, sS, ScS, ands•S phasesfor quakes thatreflect fromdiscontinuities andtraveltoteleseismic
a sourcedepthof 560 km. distances. By usingonlylong-period WorldWideStandardized
SeismicNetwork(WWSSN)recordsin this study(dominant
As an exampleof our waveformdataquality,we showthe
earthquake,that periods closeto 20 s andwavelengths near100km)we are entire set of waveformsfor one representative ableto imagemantleshearvelocityimpedance contrasts that of August16, 1979,in Figure3. The dataare globallydistri-
are distributedover as much as 25-30 km in depth. The cri- buted;thusthe wavessamplea coneof mantlematerialabove teriafor choosing ourdataarea highratioof signalto noise the source. The horizontal componentswere digitized and andpolarityreverand clear directS (to use for a sourcepulse)and surface rotatedto give thesetangentialcomponents, reflection sS phases. Theepicentral distance rangein ourdata salsdue to radiationpatternhavebeencorrectedfor. The sS
is from30ø to 95ø to avoiduppermantletriplications andcore phaseon all tracesare aligned,andwe canseetheclear,large amplitudearrivals of sS and S phases. The lines give
diffraction.
We defineprecursors to sS as sxS,the upgoing S wave
differential travel time curves calculated with the PREM velo-
showingthe arrivaltimesexpectedfor possible reflectedfrom the undersideof a discontinuityabovethe source city structure, s•S precursors s,,S, ssoS,S22oS, andS4oos between S andsS. at a depthof x km. Thesephases arriveat teleseismic diss,,S traveltime for a 36-km crustalthickness. tancesafter directS but beforethe surfacereflectionsS. Fig- We computed ure 2a showsthe geometryof ray pathsfor &S, wherex can Note that these curves are not based on actual arrivals. The
4392
ZuxsoANnL•Y: UPPER M•n•
D•scoswm-m•Es ABOVE DEEP SLABS
Surface
a)
1600 ;s
1400 S{
cS
o 1200 op,.•
S
- 1000
b)
400 c)
860 Distance (deg.)
Fig.2.(a) Geometry ofraypaths fors,•Sands•Sprecursors tosSHalong withS,sSand ScSpaths and(b)acartoon of a corresponding waveform. (c) Travel timecurves forS, sS,ScSandsxSphases fora source depth of560km. traveltimecurvefor ScSis alsoshownandclearlyhasassoci- enhance the signalto noiseratiobeforedrawinganyconcluatedarrivals.In theplotwe canseea smallwiggleabout18 s sions. Sincethe Moho produces the strongest and most
beforethe sS phasein mosttraces,whichis the Moho under- coherentreflectionof the uppermantlediscontinuities, we are sidereflection s,•S. Identifyings,•S in eachindividual traceis ableto identifys,,,Sin thestacked tracesreadilyandto actu-
sometimes difficult,butit is evenharderto seeanyotherpre- ally model the waveforms to determinecrustal thicknessand cursors between S andsS. Thenoiselevelsarequitelow for shearwaveimpedancecontrastat the Moho. thisevent,andthedatacanbe usedto placebounds on thesize As shownin Figure3, underside Mohoreflections smScan
of anyuppermantlediscontinuities. However,sincethemantle
beseenin sometraces, butit is hardto identify thephase in
discontinuities (abovethe sources) areexpected to be lessthan others.Theonlysignificant difference between sS andsmSis
half the size of the Moho arrival, we will stack the traces to
thepaththrough thecrustforsS. If thecrust haslaterally uniform propertiesabove the source,then all of the arrivalsat differentazimuthsshouldhave similarwaveforms.Our events
havestable upgoing SH radiation patterns, sothewaveshapes areverysimilar. To increase thesignalto noiseratio,we stack the signals.We alignthe tracesfor a giveneventon thesS phaseandnormalize thepeakamplitudes to unity,thenstack the arrivals.This is validbecause we expectlittlemoveout
between s,.S andsS at teleseismic distances. Figure4 shows stacked waveforms forall 23 events, someof whichhavevery clearandisolated s.,S, suchastheevents onFebruary 1, 1984, April20, 1984,August 21, 1972,August 16,1979,September 5, 1970,andSeptember 10,1973,whilesomelacka clearSmS. Notethatalignment onsS causes earlierphases likeS andScS or deeper &S phases to stackincoherently giventhelargedisrance range spannedfor each event. A total of 451 individual
seismograms is includedin Figure 4. Since oceaniccrust is expected to be less than 10 km thick, the Moho underside
reflections.,S will overlap the sS arrival in long-period waveforms;therefore,it is very hard to identifys,,,S from oceanic Moho, as for events on March 31, 1969, December 12, 0
150
300
Time (s)
1976,andFebruary22, 1974,(seeFigures1 and4).
Otherprecursors, sxS,axemuchweaker thans,•S,andthey haveslightlydifferentslownesses as indicated by the travel
Fig.3. Waveform distribution foranevent566kmdeeponAugust 16, time curvesin Figure 3, so later we will slant stack the 1979.Thecurves aredifferential traveltimesindicating timesat which at variousslownesses to seekevidencefor any sxS, S, andScS phases wouldbe expected ff thereare mantle waveforms impedance contrasts at depths x. uppermantlediscontinuities in thesS precursor window.
LAY: UPPERMmqTLE DISCOS'rINurrIESABOVEDEEP SLABS
4393
isolateddowngoingS phasecontainsthe attenuationcontribution from the lower mantle and near the receiver,along with any receiverreverberations.Upgoingphases,such as sS, have additionalattenuationand accompanying multiple arrivalsfrom undersideand topsidereflectionsfrom any discontinuities above the source.To model the sS phaseand its precursors,we use the S phaseas a sourcewaveletcontainingthe deepmanfie and receiver effects, the source radiation, and the instrument
response.The additionalfactors that have to be accountedfor thus include the extra attenuationof the sS phases and the interactionswith any velocity discontinuitiesabove the source. We determineprecursorcharacteristics relative to sS, so scalar radiation pattern differencesof the downgoingphase are not important,being eliminated by normalization. The paths of s,,S and sS in the mantle are very similar, the difference between the two arises near their reflection points. For deep foci, the incident wave at the base of the crustis nearly planar, and a simpleplane wave reflectivitymethodis sufficientfor the calculationof synthetics,,S and sS seismograms.There are many parametersthat affectthe syntheticmodeling,suchas ray parameterp; sS differential attenuationrelative to S, Atss*; crustalthicknessh; and shearvelocity and densitycontrastat
05.23.?8
I
I
I
I
I
I
the Moho, AI$andAp; but whichplaysan importantrole in the
1
2õ0
Time (s) Fig. 4. Event-stackedwaveformsfor all 23 events,alignedand stacked
computation? We discussthis at some length since our procedurediffers from previousmethodsused to study underside Moho reflections[e.g., Revenaughand Jordan, 1989].
onthesS phase. Parameter
Search
We first test the sensitivityof syntheticwaveformsto the five
ANALYSIS OFs,nS PHASES TO DETERMINE PROPERTIESOF THE MOHO
parameters: p, At•s*, h, A•, and Ap. For eachtest,only one parameteris varied. For the rangeof epicentraldistanceof our We use the September5, 1970, deepKurile slabearthquake sS data (September5, 1970 event)from 50ø to 90ø, the param(Table 1) to developour modelingmethodand to estimatethe averagecrustalthicknessand shearimpedancecontrastat the etersp and Atss* were computedfor PREM: p changesfrom Moho near the s,nS reflectionpoint under the northernSea of 9.5 to 14.5 s/deg,and Atss*variesfrom 1.5 to 3.0 s. The averOkhotsk.We have a total of 58 digitizedWWSSN transverse agep and At•s* for our data samplesare 11.81 s/degand 2.56 componentrecordsavailablefor this event. Twenty tracesof S phasesand 34 tracesof sS phasesare chosenfor stackingon each phasefor this event, at distanceswhere the phasesare well isolated.We separatelystack the data alignedon the S and sS phasesto enhancethe ratio of signal to noise. The stackedS pulseis usedas a sourcepulseand the averagedsS waveform is modeledwith syntheticwaveforms. Clearly, this stackinginvolveslateral averagingover the regionsampledby sS, as well as the intrinsicaveragingof the long wavelength waves. However, the gain in signal stability gives us confidence that the average structure can be determined robustlyfrom the event-averaged data. Later we will consider subdivisions of the databasedon reflectionpoint grouping. In syntheticmodeling, the computedseismogram,S(t), is a convolutionalproductof time series: S (t)=X (t)*E (t)'I (t)
(1)
where X(t) is the sourcetime history,E(t) is the Earth transfer function,andl(t) is the instrumentimpulseresponse.For a simple rupture,it is sufficientto treat the sourceas a point source for which the spectrumof upgoing and downgoingradiation from the sourceis the same,with separatescalingfactorsdue to the differencein radiationpatternfor eachray parameter.At
s, respectively. For the numericaltests,we prescribea shear wave surfacevelocity • = 3.2 km/s and surfacedensitypo=
2.6x103kg/m3, uppermost manfieshearvelocity•,,, = 4.7 km/s anddensity p,,,=3.38x103 kg/m3, crustal thickness h= 25 km, At•s* = 2.56 s,p = 11.81 s/deg,A• = 0.7 kin/s, and
Ap= 0.4x103 kg/m3,wheneachparticular parameter is not being varied. The crusthas a linear velocity m•d densilygradient from the surfaceto the Moho, approximatedby many thin layersin the reflectivitymodeling.
In Figure5a the effectof variableray parameter p is shown in superimposed waveformsfor the rangep = 9.5-14.5s/deg; the variation interval is 0.5 s/deg. All other parametersare fixed at the values listed above. These signalsshow only a small changein waveformas ray parametervaries,indicating
the possiblesmoothing effectof stackingthe data. Sincethe dataare actuallyconcentrated in a limitedrangenear75ø, this is a minor bias. Thus, in the following analysiswe fix an average value of p = 11.81 s/deg,which is an appropriate average for our shacked sS data.
The waveformsensitivityto At•s* is shownin Figure 5b. At•s* variesfrom 1.5 to 3.0 s, with intervalof 0.1 s. Although
teleseismicdistancesthe Earth transfer function is essentiallya
the changeis visible, it is quite small. Comparisons with
sequenceof arrivals with different geometricspreadingand
observedwaveformswill only be affectedif we comparewith the entire sS waveform, includingits instrumentalovershoot.
attenuation
filters convolved
with
the receiver
function.
The
4394
ZHANGANDLAY: UPPERMANTt• DISCONTINUm•ABOW-DEEPSLABS
Varying Ray Parameter
c)
1.0
Varying Crustal Thickness
1.0
0.5
0.5
p=14.5s/deg ] 0.0
-
0.0 l=30
-0.5
-0.5
-1.0
-1.0
0
10
20
b)
30
40
50
60
km h=15
I
I
I
I
I
I
I
0
10
20
30
40
50
60
Varying At.S*
d)
1.0
km
Varying Density Contrast
1.0
At•s*=3.0s
Atss*=1.5 s 0.5
0.5
0.0
0.0
Ap=50 kg/m
-0.5
-1.0
-0.5
I
I
I
I
I
I
I
0
lO
20
30
40
50
60
-1.0
Time (s)
Ap=750 kg/m 3
I
I
I
I
I
I
I
0
10
20
30
40
50
60
Time (s)
Fig.5. Synthetic waveforms of sS superimposed waveforms (a) forp =9.5to 14.5s/deg withaninterval of 0.5s/deg; (b) for Atas* = 1.5 to 3.0 s with an intervalof 0.1 s; (c) for h = 15 to 30 km with a 1-kminterval;notethattraveltime
difference between sS andsinsincreases withincreasing h' and(d) forAp= 0.05to0.75X103 kg/m3 withaninterval of
0.05x103 kg/m3; note that Apinfluences theratio ofsmS/sS. Other model parameters areheld fixed atthevalue given in the text.
Thereforewe choosean average Atss * = 2.56s in themodeling below
and use a short window
waveform
which excludes the sS
We alsotestAll for waveform sensitivity. If wekeepthe averagecrustal velocity constant,identicalwaveformscan be
overshoot.
produced by varyingAll andAp sincethereflection coefficient Figtire5c showsthe waveformsensitivityto crustalthick- at near-verticalincidence is primarily sensitiveto the
nessh at the reflectionpoint, which variesfrom 15 to 30 km.
impedance contrast A(p[3).Thes,•S/sSamplitude ratiopro-
The substantial waveformvariationshownin thisfigureis due videsour estimateof the impedance contrast.
to the travel time difference between sS and s,.S, which increaseswith increasingcrustalthickness.The ratio of sS and Best Model Search s,.S amplitudealso changesdue to the variable interference, but this is less pronotmcedthan the time difference effect. The testsaboveshowthattheparameters h, A•, andAp
Crustalthickness is mainlydetermined by the difference in affectthewaveform modeling muchmorethantheparameters travel time betweensS and s,.S. It is clear that we should At•s*andp. Thuswe useaverage valuesof p = 11.81s/deg dataanalysis. As Ap and avoidstacking waveforms thatsample regions withvarying cru- andAt•s*= 2.56s in thefollowing we combine thesetwoto stalthickness. Thisprevents usfromstacking multiple events, All tradeoff directlyin themodeling, astheeffective footprint of sinsbecomes toolarge. determinethe impedance.contrastA(p•) (A(p•) =
in thefollowing. Sincethevelocity Figtire5d is anexamination of thewaveform sensitivity to 2(p1•1-P2•2)/(p1•+P2•2)) density variation ApattheMohoreflection point,which rangesanddensityarelikelyto be relatedto eachother,we specify from 0.05x10 • to 0.75x103 kg/m3, with a step of the absolutedensityassociated with a givenvelocitymodel 0.05x103kg/m3. The surfaceandMohodensities arefixedat usingtheNafe-Drakerelation[GrantandWest,1965]'
po= 2.6x103 kg/m3 andp,. = 3.38x103 kg/m 3,respectively, andthecrusthasa lineargradient from2.6x103 kg/m3 at the
pc---0.1730cc +1.695
(2)
stirface to thedensity at thebaseof thecrustgivenbypm-Ap wherePc is densityin thecrust,andOCc is theP wavevelocity.
kg/m 3. Theamplitude ratios,.S/sSincreases asthedensityWe assume oct= '/J[[,where[1is theshear velocity in the contrast Ap increases as shownin the superimposed tracesfor crust.We asstune a linearcrustalvelocitygradient throughout different Ap models. There is no travel time difference the crust and do not have any midcrustaldiscontinuities.
between sS ands,.S dueto varyingaverage density, sincewe Therefore, thefreeparameters in thesynthetic computations are do notperturbthevelocities, although theapparent risetimeof thes,,,S phasevariessubtly.
the surfacevelocity{]o,the Mohovelocity{3,,,,Atss*,thecrustalthickness h, andtheimpedance contrast A(p{])at theMoho.
LAY: UPPERMAN"ILEDIsco•m•
0.0
ABovE DEEP SLABS
4395
different At•s*. For the best model, the crustal thicknessis 24 km, and the shear wave impedancecontrastat the Moho is 27%. The correlation coefficient is slightly lower for At•s*=1.98 s than for At•s*=2.56 s.
20
Comparisonfor IndividualStations 22
Since some individual records have clear s=S arrivals, we 24
can assessthe variabilityin the Moho parametersby fitting single stations,while recognizingthat thesehave a much higher noise level but involve less horizontalaveragingof the Moho.
26
We
28 ß
.
chose a data set from four individual
stations with con-
..
30
..
.. :::. :.
38.6
1.0
44.
Impedance contrast (%)
Fig. 6. Contourof cross-correlation coefficients betweensynthetics and the stackeds,•S +sS waveformfor September5, 1970, for a model
sistentlyhighsignalto noiseratiosfor detailedcomparison with synthetics. Thesefour WWSSN stationsare locatedin Europe (KON, TRI, STU) and North American (SCH). Since they have differentpath effects,we use the corresponding S phase at each stationas input signalsfor the modeling.We use the same search method as described above, i.e., for each station,
we compute400 tracesto comparewith the observeddata at with [•0= 3.2 km/s,[•,, = 4.7 km/s,andAt•s* = 2.56 s, for varying fixed (p[3)o and (pl3),,, and choosethe best fitting one. The crustalthicknessand Moho shearimpedancecontrasts. correlationcoefficientsdiffer for variouschoicesof (pl3)oand
(p[3)=. Table 3 gives the resultsand Figure 7 showsthe We allow •o, •=, andAtss*to vary over a limitedrangecom- observedand syntheticwaveformcomparisons for eachstation. patiblewith reasonablecrustaland mantlevelocities,perform- The waveformcomparisonfor the event stack is also shown, ing completesearches of h andA(p[•)for eachcase. with the syntheticparameterscorresponding to the best fit First, we vary h over a large range,findingthat the correla- foundin Figure 6. The resultsshowthat iX(p[5)variesfrom tion coefficientsbetweendata and syntheticsfor crustalthicknessless than 14 km or more than 33 km for the September5,
20% to 36%, and h varies from 23 to 25 km for optimal cru-
1970 test event are less than 0.2.
stablebecausethe waveformcorrelationprocedurein the pres-
For each model suite in the
final search h varies from 14 to 33 km at a 14:m interval.
For
stal structures. The crustal thickness estimates tend to be more ence of noise is more robust for differential times than for rela-
someeventswe explorethickercrusts.A(p[3)has20 intervals tive amplitudes.As expected,the parametersfound for the between 1% and 46%. Our absoluteMoho depth determina- stacked trace are intermediate to the individual observations.
tionsthusdependon how well this simplemodelrepresentsWe feel that the parametersdeterminedfor the stackedtraces averagecrustalvelocity. For eachmodelsearchwith a arethe mostreliable,giventhehighersignalto noise.
specified 13o, [3=,andAt•s*,wecompute 400waveforms which we comparewith the observed one by computing the cross AverageMoho PropertiesUnderneathsS ReflectionPoints correlationbetween the observedstackedsS waveform and the
synthetics. We correlate overan intervalcontaining thes,,S
From the discussion of our method, it is dear that we can
arrivaland the first half cycleof the sS arrivalto reducethe determine laterally averagedcrustal thickness(for specified averagecrustalvelocity) and shearwave impedancecontrastat sensitivityto At•s*. the Moho from clear and isolateds=S phasesprecedingthe sS A contourplot of the correlation coefficients obtained for reflection. This is becausewe use waveform modelingto fixedvaluesof (p[3)oand(pl3)=andnumerous combinations of determine the structure. Figure 8a shows the surface sS h andA(pl3)(Figure6) indicates thatthesearch algorithm does reflection point locationsfor the event of September5, 1970. definean optimumchoiceor rangeof choices of parameters. The actual reflection for s=S involvesa horizontalregionwith There is some couplingbetweencrustal thicknessand scale defined by the Fresnel zone, which dependson the impedance contrast because we assume a linearvelocitygradientin the crustbetweenthe surfacevelocity13oandthe velo- wavelength and the source depth. The inversionresults for
A(p[3)and h are thus averagesover this zone and not local city at thebaseof thecrust,I3,,-AI3.Thisparameterization avoidssolutionswith unrealistically high mantle velocities. values at any reflectionpoint. The Fresnelzone radiusfor our Table2 showsthebestfittingmodelfor theSeptember 5, 1970, long-periodS wave data is about170 km for a sourcedepthof
event,with velocities[3o=3.1km/s and [3,,,--4.65 km/s and 560 km and wavelengthsof about 100 km. Most of the data TABLE 2. Parameters for BestFittingModelsfor the Eventof September 5, 1970
•3o' era' km/s
km/s
A½, A(la[•), Thickness, At•s* Correlation % km Coefficient
km/s
3.1
4.65
0.81
27
24.0
2.56
0.993520
3.1
4.65
0.81
27
24.0
1.98
0.992514
4396
ZUANOASOLAY: UI,PF• MineroE D•SCONT•NUmES ABOVEDEEPSLABS
TABLE 3. Parameters for BestFittingSingleStationModelsfor the Eventof September 5, 1970 NalTle
150, 15 m,
km/s
km/s
AI3, A(p[5),Thickness, at,s* Correlation
km/s
%
km
Coefficient
KON
3.4
4.4
0.85
30
25.0
1.98
0.973723
SCH
3.2
4.7
1.10
36
25.0
1.98
0.969731
STU
3.9
4.4
0.60
20
24.0
1.98
0.971232
TRI
3.2
4.7
1.05
34
23.0
1.98
0.983613
To explore for lateral variations on a smaller scale, we for the event of September5, 1970, samplethe crustnorth of the source,over an area spanning4.5ø. Only a few tracessam- separatesignalswith reflectionpointsinto threegroups(Figure ple structuresouth of the event. Thus, stackingthe data and 8a): groupA with reflectionpointsat azimuthsto European stathe intrinsicFresnelzone limitationsinvolve smoothingover a tions (shadedtriangles),group B with reflectionpoints at region at least 900 km across. Therefore, the estimatedcrustal azimuthsto North Americanstations(opentriangles)andgroup thicknessand impedancecontrastare appropriateaveragesover C with reflectionpointsat azimuthsto Southeast Asia andAusthe entirereflectionpoint area. tralianstations(solid triangles). Then we separatelystackdata for eachgroupand performthe waveformmodeling. GroupC a) only has four tracesall with oceanicreflectionpoints,and can0.4
not be modeled
0.0
from modelingof groupsA and B. The crustalthicknessestimatesrange between23 and 26 km, while the shearimpedance contrastsat the Moho vary from 30.8% to 27% from groupA to groupB. Other modelingparametersare shownin Table 4. Explorationof model trade-offs suggestsabout + 2 km uncer-
-0.4
K0N
-0.B
0
5
10
15
20
tainties
0 0
-0 -0
C'
0
5
10
15
20
25
STU
0
5
10
15
20
0.4
0.0
\\
in
due to the thin crust.
crustal
thickness
and
+
Table
5%
4 shows results
uncertainties
in
impedance contrast for the two groups separately. There appearsto be a slight thickeningof the crust and decreasein Moho shearimpedancefor paths under the Kamchatkapeninsula (groupB); however,thesedifferencesare only marginally resolvable,given the decreasein numberof tracesavailablefor the separatestacks. From this analysisof s,,S, we infer that when largedatasets are availablewith sufficientsignal to noiseratio, we can determine average crustal thicknessand impedancecontrastat the Moho under sS reflectionpoints for reasonablechoicesof average crustalvelocity, with uncertaintiesof about+ 2 km for crustal thicknessand + 5% for impedancecontrast. We will now considerhow well we can hope to resolvepropertiesof deeper discontinuities,which we expect to be far weaker than s=S, and then we will presentresults.
// SLANTSTACKING Sx TO SEEKEVIDENCE FOR UPPER MANTLE
-0
e
0
5
10
15
20
0.0-0.4
Event
Stack
-0.[3 I
i
5
10
15
2O
Time (s)
DISCONTn•.nTmS
As shown in Figure 3, it is very hard to identify ,any coherentundersidereflectionphasesotherthan s,,,S betweenS and sS phasesin the raw data, suggestingthat only weak impedancecontrastsexist above the source. Isolated arrivals observedat only a singlestationcannotbe safelyinterpreted in terms of mantle layering. ScS arrives betweenS and sS phasesover much of the distancerange,thus we must avoid time intervalsin which the strongScS arrival would obscure
any sxS phase. After someeditingof the datato accomplish
Fig. 7. Comparisons betweenthe bestfittingsynthetic(solidline) and this, we stack all the waveforms on different slownesses. We observed(dashedline) sS and s,•S signalsfor individualstations: considerall possibledepthsfor a reflectingboundary.
KON, SCH, STU, andTRI andfor the eventstackfor September 5, 1970,eard•quake.The structural parameters for eachcasearegivenin Table 3.
Becauseof possiblelateralvariationsin reflectordepth,the sharpness andimpedance contrastof someuppermantlediscon-
ZI-IANGAND LAY: U•
MANTt• DISCO•rnNUmES ABOVV.DEEP SLABS
4397
b)
,../•/10P3
40 ø
50 ø
160 ø
120ø
130ø
140ø
Fig. 8. (a) Locations of sS reflection pointsfor theeventof September 5, 1970.Theasterisk indicates thesource locationand trianglesare reflectionpointsfor pathsto the labeledstations.We separate reflectionpointsinto threegroups:groupA, reflectionpointsat azimuths to European stations (shaded triangles); groupB, reflection pointsat a•muthsto NorthAmerican stations (opentriangles); andgroupC, reflection pointsat azimuths to southeast AsiaandAustralian stations (solidtriangles). (b) Similarplotfor theNorthKoreaneventof September 10, 1973.
tinuities may be o•scured in a global analysis like that of Shearer [1990, 1991]. Our stackingexplorationfor sxS precursors reveals relatively localized upper mantle propertiesabove the deep sourceregionsnear the northwesternPacific subduction zones. The same Fresnel zone and laterally averaging effectsas describedfor the s,,,S analysisapply for this portion of the study,so if the boundaryvariesover a 1000-kmscalewe will obtainedonly locallyaveragedproperties of the structure. Differing distance ranges and source depths of the earthquakesaffect the arrival times for a certainsxS phase.We line up all the data for a given event on the sS phaseto provide a clear referencephaseand to eliminatestationstatics. We also separatethe data into three groupsto make the depthinfluence as small as possible. These three groupsare for differentfocal depthranges:(1) 115 km < H < 195 km, (2) 348 km < H < 424 km and (3) 503 < H < 566 kin. Given possiblelateral variations in our study area, as indicated in analysis of s,,,S phases,we divide the data into three regions for each depth group;(1) the Sea of Okhotsk,(2) North Korea, and (3) Japan, RyukyusandIzu.
By aligningthe traceson sS, we set the corresponding slowness of the sS
arrivals
to zero.
The
slowness
for different
undersidereflectionphasesprovidesa meansfor identifyingthe s•S arrivals, if there are any, relative to random or signalgeneratednoise. Slant stackingon a particularslownesscreates interferencebetweenthe signals,which is destructiveexceptfor a phase with the correspondingslowness.Such phaseswill appear with a maximum amplitudeat a certain time on the stackedtrace. The confidencethat one has in the stackedsignal dependson the degreeto which noiseis suppressed in the overall stack and consistencyof the timing of the arrival with the depthexpectedfor a particularslowness.Figure9 givesan exampleof slant stackingat differentslownessvaluesfor the event of August 16, 1979. For slowness As = 0.00, which corresponds to a stackwith the sS phaselined up, we can see a very strongsS phase and a weak incoherentS phase. As As decreasesfrom zero, the sS phasebecomesweaker and distotted, while the S phasebecomesstrongerand more coherent. The energybetweenthe S andsS phasesalsochanges.For As = -0.0065 s/km, the S phaseis aligned and a singlepulse is Upper mantle discontinuities can have differentreflection retrieved. Figure 3 showsthat a singledifferentialslownessis coefficients,but most of them are expectedto be very weak. not ideal for aligningS, but will be an excellentapproximation For example, shearimpedancecontrastsof 2-15% have been for any s•S phases. Therefore,when slant stackingover a suggestedfor the 80-, 220-, 410-, 520-, and 650-km discon- range of slowness,any phase with a particularslownesswill stacks. tinuities,explainingin part why it is difficult to identify s,S have maximum coherencefor the corresponding Since we exclude traces with ScS between S and sS phases, precursorsin the individualtraces. We do know that different the traces which can be used for slant stacking are fairly sparse phaseswill have differeraslowness.
TABLE 4. Parameters for BestFittingModelsfor DifferentReflection RegionsUndertheSeaof Okhotsk
Sept. 5,1970
•m,
A•,
km/s
P0,
km/s
km/s
Thickness, A(Q•), km
A
3.15
4.8
1.05
23
30.8
0.993747
B
3.1
4.8
0.9
26
27
0.992433
%
Correlation Coefficient
4398
ZHANOANDLAY: UPPERMAmLE DISCOmUNUmES ABOVEDEEPSLABS
0.0000 s/km
2, 1978, becauseof their shallow sourcedepths. Since the periodof our data is about20 to 25 s, it is difficultto isolate the s,nS phasebetweenS and sS for theseevents. Allowing for substantial uncertaintyin the s,•S phasefor eventson Septe;aber 21, 1974, and September2, 1978 (Figure 4), the waveformmodelingindicatescrustalthicknesses of 19 and 30 kin, respectively,and 21% shearimpedancecontrasts for these
• 0.0007 s/km
two events. The crustappearsto be thickerunderTaiwan than - 0.0020 s/krn
under Kamchatka
if these numbers are reliable.
It should be
possibleto improve the resolutionwith broadbanddata as it becomesmore extensive. The depthsof eventson May 23,
...... I
,
,
0
,
•-_ I
100
,
,
,
I
,
200
i
1978, May 14, 1968, and September21, 1965 along the Ryukyuarc rangefrom 160 to 195 km (Table 1). The modeling (Figure10 andTable5) givescrustalthicknesses of 25, 20, and 20 km and shearimpedancecontrastsof 21%, 29% and 29%, respectively. Sincethe sins phasesare not well isolated, relatively low correlationcoefficientsare found, as listed in
0.0065 s]km ,
I
300
,
i
,
I
400
Time (s)
Fig. 9. Slantstacked waveforms for relativeslowness variations of As
Table 5. Nonetheless,there is some indication of crustal thick-
= 0.0, -0.0007, -0.002, -0.0065 s/kmfor the eventof August16, 1979,
erring along the arc toward the Japaneseislands. Event on March 31, 1969 underthe Sea of Japanand eventson February 22, 1974, January31, 1973 and December12, 1976, underthe
whichhasa depthof 566 km. thirty-fourwaveforms areincluded in the stacks. The relative slownessfor sS is 0.0, while for S it is -0.0065 s/km.
for eachevent. Therefore,we combinenearbyeventsthathave similardepthsinto groupstacks.This is possible because the
North
eventsall have simple,similar waveforms. In the slant stack-
o 1679••
ing, we firstline up all the traceson thesS phases, flip any tracesthat have reversedsS polarity, and then normalizethe
tracesto the peakof thesS phase.Giventheveryweakprecursors sxS, we plot the logarithmof the envelope of thetrue slantstacksusingthesamemethodasVidaleandBenz[1992]. Very weakvariations canbe seenin thelogarithmic scale.In our slantstacking,we considerrelativeslowness variationfrom
-0.015 to 0.015 s/km, a rangespanning any plausiblesxS
Korea
Sea of Okhotsk
031864
08
/
090570
Taiwan
122175
arrivals.
For shallowerfocaldepthevents(115 km < H < 195 km), it
O9O
012971
is very difficult to separateprecursorsfrom S and sS in the long-period data. We did testson shalloweventson September
21, 1965,July4, 1967,May 14, 1968,May 23, 1978.Because of the shallow depth for those events,it is very hard to confidently identifyany energybetweenS andsS phases,so we only discusseventsin the depthrangeof 348 to 566 km in the investigation of sxS phases.
042084 /
Ryuk•_L.Arc
092165-•
051468•
RESULTS
ResultsFrom Analysisof smS
020184
092174
05237•
Figure10 showscomparisons betweenobserved (dashedline) and synthetic(solid line) waveformsfor all eventsthat have identifiables,•S phasesthat could be modeled. Table 5 lists resultsfrom the modeling.Table 5 indicatesthat crustalthick-
082172
I
I
I
I
I
J
0
10
20
30
40
50
Time (sec) impedance contrastat the Mohois about18%. The relatively Fig. 10. Comparisonsbetween synthetic(solid line) and observed for eventswhichhavemodelable s,,,S phases. earlys,•S arrivalsin thisregioncanbe seenin Figure10. As (dashedline) waveforms
ness near North Korea is about 36 km, and the shear wave
shownin Figure 4 and Table 1, the durationbetweenS andsS
The eventsare arranged in the NorthKorea(September 10, 1973,and August 16, 1979), Taiwan (September2, 1978), Ryukyuarc (Sep-
is lessthan50 s for the Kurile Islandeventson September 21, tember21, 1965, May 14, 1968, and May 23, 1978) areasand the Sea 1974, December1, 1967, and the Taiwaneventon September of Okhotskby sourcelocationfrom northto south.
Znn•o ANDLAY: UPPERMmCTLEDISCONT•NUmEs ABOVEDEEPSLABS
4399
TABLE 5. CrustalThickness andImpedance Contrast for SomeEvents Latitude, Longitude, (deg) (deg)
Date
Depth, (km)
mb,
Thickness, A(p•), (%) (km)
Correlation Coefficient
North Korea
Sept.10, 1973 Aug. 16, 1979
42.48
131.05
552
5.8
36
18
0.99528O
41.85
130.86
566
5.8
36
18
0.994874
Taiwan
Sept.2, 1978
24.81
121.87
115
6.0
30
21
0.973122
Sept.21, 1965
28.96
128.23
RyukyuArc 195
6.0
20
29
O.975749
20
29
O.979251
6.2
25
21
0.938021
5.6
29
12
0.989448
24
26
0.993520
22
46
0.985712
0.991975
May 14, 1968
29.93
129.39
May 23, 1978
31.07
130.10
March 18, 1964
52.56
153.67
162
5.9
160
Sea of Okhotsk
Sept.5, 1970 Dec. 21, 1975
52.28
51.93
424
151.49
560
151.57
5.7
546
6.0
Jan. 29, 1971
51.69
150.97
515
6.0
23
28
April 20, 1984 Aug. 21, 1972
50.12
148.75
572
5.9
22
28
0.995713
49.47
147.08
573
5.9
21
26
0.991805
Feb. 1, 1984
49.05
146.63
580
5.8
22
26
0.996441
Sept.21, 1974
52.19
157.44
119
5.7
19
21
0.966426
Izu arc, all reflectunderoceanicstructureas shownin Figure4 contrast.This eventappearsto be unreliableand is the only whena 30-s and have unmodelableweak s,•S waveformeffectsso we can- eventfromtheearlydaysof theWWSSNsystem pendulum periodwasbeingused.Thesetwoanomalous results not quantifythe structure there. The crustal thickness beneath the northern Sea of Okhotsk
may represent locallycomplexuppermost mantlestructure or
varies from 21 to 29 km, and the shearwave impedancecon- simplynoisefor theseevents.Figure11 plotsstacked trast varies from 26% to 28% for all events except events on waveformsfor differenteventsin the Sea of Okhotsk,showing s,•S phases areobserved in thenorthof December21, 1975, and March 18, 1964, which give extreme thatclearandstrong
5, 1970,December 21, 1975, estimates of shearwaveimpedance contrast of 46% and 12%. thisarea,for eventson September 29, 1971,April2, 1984,August 21, 1972andFebruary Lookingat the datafor eventon December 21, 1975(Figure January at eventsfartherto the 10), we seethattheonsetof s,nS is unclearandthe synthetics 1, 1984,while weaks,•S are observed
givea moreimpulsive arrivalthanobserved, whichmightbe south:June 21, 1978, and July 10, 1976. We do not report for the lattertwo eventsbecause they are one reasonfor the high valueof A(p[•). It is hard to identify Moho parameters fromnorthto southprobably s,•S for eventon March18, 1964(Figures 4 and10) because poorlyresolved.Thisvariation to oceanic the amplitude is small,yieldinga low estimate of impedancereflectsthe crustalvariationfrom quasi-continental
60 ø
02/0/84 ., AVll \F
,, a•
120ø
130ø
140ø
rlII\F t'•06•1.8 /
150ø
0710/76
160ø
170ø
Fig.11.Stacked sS waveforms fortheSeaof Okhotsk events ploued fromnorthto south.Thes,•S arrival weakens forthe southern events.
4400
•o
mgDLAY: UPPERMm½mEDISCON•NUmES AnOVEDEEPSLAnS
stxucturein this area. We only have six tracesfor the Kurile event on May 27, 1972, so the stack is less reliable for this event and we did not model it.
ResultsFrom Slant StackingsxS
Plate 1 shows resultsfrom slant stackingsxS phasesfor source depths varying between 503 and 566 km in three regions. The plot showsthe logarithmicamplitudeof the stack envelope as a function of relative slownessand travel time. The color scaleis a logarithmicamplitudethat variesfrom 0 to 2, which meansthe true amplitudescalewill variesby a factor of 100 relative to the stackedaS amplitude. Black indicates less than 1% amplituderelative to sS. Any arrivals with amplitude2% to 10% of aS can be readily detected. Some clear artifactsof slant stackingare apparentin the streaking. Plates la, lb and lc are slant stacksof deep eventsin North Korea, the Sea of Okhotsk, and Izu Japan,respectively. The
slownessof any sxS precursorsfrom horizontallayerswill vary between the slownessesof S and sS phases (Figure 3). If energy is found betweenS and sS with correctcorresponding slowness,it is likely to be an &S arrival, but the limited ray parameterresolutionmust be allowed for. In each case, sS is the strongestarrival at a relative slownessof zero. As the relative slownessvaries, the sS energy will become less coherent in both slowness variation
directions.
The extent to which the
sS arrival is localizedin time andray parameteris indicativeof how well any arrival can be isolated, and real sxS arrivals should streak as much as sS.
A very clear stackis foundfor two deepeventsbelow North Korea (Plate la). We can see that relatively strong energy occurs in front of the aS phase with very subfie slowness differencefrom aS. This correspondsto energyreflectedfrom the Moho and the amplituderatio of sins to sS is about 8%. Ahead of sinS, other energy is seen which is consistentwith reflectionfrom a 90-km-deepboundary,with about6% ampli-
500 < H < 570 km
a)N. Korea 0.015
.
.
"4
-0.015
,
.
b)Sea of Okhotsk 0.015
-0.015
c) Izu-Ja an 0.015
0.000
-0.015
0
200
Time o.o
300
s ......
2.0
Log(Amplitude) Plate1. Slantstacksections for eventswithfocaldepths of 503 < H < 566 km below:(a) Noah Korea,(b) theSeaof Okhotsk,and(c) Izu Japan.Seetextfor details.
300