langston 79

8/7/2019 Langston 79

1/14

VOL. 84, NO. B9 JOURNAL OF GEOPHYSICAL RESEARCH AUGUST 10, 1979

Structure Under Mount Rainier, Washington, InferredFrom Teleseismic Body Waves

CHARLES A. LANGSTON

Department of Geosciences, Pennsylvania State University, University Park, Pennsylvania 16802

Teleseismic long-period P waves recorded at the World-Wide Standard Seismograph Network stationLON (Longmire, Washington) are shown to exhibit strong anomalous particle motion not attributable toinstrument miscalibration or malfunction. In particular, a large and azimuthally smoothly varyingtangential component is observed after vector rotation of horizontal P waves into the ray direction andafter application of a deconvolution technique which equalizes effective source time functions andremoves the instrument response. These tangential waves attain amplitudes comparable to the radialcomponent and demonstrate wave form antisymmetry about a NNE azimuth. A model which contains asingle high-contrast interface dipping toward the NNE at a depth of 15-20 km can explain most of thecharacteristics of the long-period P wave data, provided dips are greater than about 10 and only theinterference of P and Ps generated at the interface is considered. The model breaks down for later arrivalswhich are presumably multiples or scattered waves. Examination of long-period S waves from severaldeep teleseisms shows a prominent Sp arrival 18 s before S. The timing of this phase conversion suggestsan interface at about 145-km depth, and its sense of polarity suggests that the velocity contrast is from

higher to lawer velocities as depth decreases. This interface may correspond to the bottom of the uppermantle low-velocity zone in the area.

INTRODUCTION

From a geologist's point of view, the top of the earth's crustis composedof a myriad of differingrocktypeswithreadilydiscernable lateral variations in lithologies and structures. Onthe basis of these surface observations it is easy and probablyjustified to assume that the entire crustal and upper mantlecolumn behaves in a similar complicated way. Recent high-resolution reflection profiling and other seismological studiesare beginning to support this view [e.g., Smithson, 1978; ,4ki etal., 1977]. However, when seismic waves are used to deduceearth structure, the usual state of affairs is to assume verticallyinhomogeneous earth models with no lateral variations. Thisassumption is good in many cases because, to first order, the

most dramatic aspect of the data from LON is that long-period P waves exhibited a large tangential component ofground motion. The simple existence of 'tangential P waves'automatically precludes any explanation or interpretations us-ing honogeneous plane layered earth models, since the P-SVand SH wave systems are decoupled.

The study of these waves will be based on a timid extensionof standard seismological practice. Using theory presented in aprevious paper [Langston, 1977b], earth models will still haveplanar interfaces, but these interfaces will be allowed to dip. Athree-dimensional ray-tracing formalism is used to find thepath of rays in the structure. Ray amplitudes are computed byfirst decomposing the incident ray into P-SV and SH com-

earthappearstoberadiallystratifiedandseismicwavesfortu- ponentsinlocalinterfacecoordinates,multiplyingeachampli-natelyaverageoverdistancescalescomparabletotheirwave-.tudebytheappropriateplanewavereflectionortransmissionlength. coefficient,andthenpropagatingthewaveto thenextinter-

This paper examines an interesting case where the seismicdata cannot be explained with any vertically inhomogeneousnonlaterally varying earth model. Teleseismic P and S wavesas recorded at the World-Wide Standard Seismograph Net-work (WWSSN) station LON (Longmire, Washington) wereexamined to determine crustal and upper mantle structurebeneath the station. A previous study of teleseismic bodywaves at COR (Corvallis, Oregon) and of P wave forms re-corded teleseismically from the 1965 Puget Sound earthquakeindicated that the area of western Washington and Oregonmight be underlain by a distinct low-velocity zone (LVZ)[Langston, 1977a; Langston and Blum, 1977]. The base of thisLVZ was inferred to be at 55-km depth under Puget Soundand at 45-km depth under COR. LON, which is located on theflank of Mount Rainier, a Quaternary stratovolcano, was cho-sen for study in an attempt to define the limits of this LVZ.

It was immediately apparent upon examination and com-parison of the long-period P wave form data that earth struc-ture at LON is radically different from structure under COR.As is the usual case when new data are considered, new ques-tions and problems immediately arose with respect to both theregional structure picture and seismic wave propagation. The

Copyright 1979 by the American Geophysical Union.

face, where the procedure is repeated. Diffracted waves fromstructure edges or corners are not included. Hence all 'tan-gential P' wave effects will be explained by off-azimuth arrivalsof primary, converted, and reverberated rays. This is a hypoth-esis which was only partially successful in explaining the ob-served P wave data, whether because of a breakdown in theory(existence of diffracted waves) or in assuming planar interfaces

(other geometries or topography).After a short description of regional tectonics and setting,this study will be presented in four parts. First, the long-periodP wave form data recorded at LON will be presented with adiscussion of instrument calibration. This is crucial, since in-strument miscalibration or malfunction is the first likely causeof apparent tangential P waves. Next, a general method forequalizing the wave form data for arbitrary effective sourcetime function will be discussed. Afterward, the equalized Pwave form data will be interpreted by calculating theoreticalray amplitudes and synthetic seismograms from earth modelswith planar dipping interfaces. The final section presents re-sults concerning long-period S wave form data. While notadding significant constraints to earth models deduced fromthe P wave data, there is an indication of an interface atapproximately 145-km depth under LON from Sp precursorobservations.

Paper number 9B0335.0148-0227/79/009B-0335501.00

4749


2/14

4750 LANGSTON: STRUCTURE UNDER MOUNT RAINIER

49 o

48

46

44

42

123 CANADA

/

729/5

g

0 50m

TNS. Oregon....California

Fig. 1. Index map of the Pacific Northwest region. Triangles in-dicate locations of the WWSSN stations COR (Corvallis, Oregon) andLON (Longmire, Washington). Also shown is the location of the 1965Puget Sound earthquake and major physiographic provinces of thearea. The dashed line represents the inferred limit of a Tertiary eu-geosyncline [after Snavely and Wagner, 1963].

REGIONAL SETTING

Longmire station is located on the south flank of MountRainier, which is the highest member of the Cascade Range ofnorthern California, Oregon, and Washington (Figure 1).Mount Rainier is a relatively young feature with vulcanisminitiating in early Pleistocene and continuing into recent times[Fiske et al., 1963]. In the plate tectonics scheme the Cascadesare part of a continental arc of andesitic vulcanism related tosubduction of oceanic lithosphere from Cape Mendecino, Cal-

ifornia, to Vancouver Island, British Columbia. An oceanicridge system consisting of the Gorda and Juan De Fuca rises islocated a few hundred kilometers offshore. The small Gorda-

Juan De Fuca plate, lying between the continent and theoceanic ridge system, is the last remnant of the previouslymore extensive Farallon plate which was subducting underwestern North America throughout the Tertiary [Atwater,1970]. For a more comprehensive summary of tectonics in thePacific Northwest, please refer to Riddihough [1978] or Lang-ston and Blurn [ 1977].

Geophysical studies in the Mount Rainier area have usuallybeen of a reconnaissance nature. Gross crustal structure is

essentially unknown; the station LON is almost always used asthe last station of long-range refraction profiles [Johnson andCouch, 1970; Berg et al., 1966; Zuercher, 1975]. However, theirtudies and others [e.g.,Dehlingeret al., 1965] indicatethat theCascades are a region of crustal transition between highermantle P velocities and thicker crustal sections to the east

compared to low Pn velocities and thin crust to the west.Studies of teleseismic P wave residuals indicate that LON is

not a particularly anomalous station, at least for average tele-seismic travel times (for example, see Sengupta and Julian[ 1976] for a synopsis). However, Lin [ 1974] suggested that therelative travel time residual of LON with respect to the rest ofthe University of Washington Puget Sound array was consis-tent with an eastward-dipping high wave velocity slab or aneastward thinning low-velocity zone.

Microearthquakes have been monitored in the Mount Rain-ier area on an intermittent basis using a small portable array[Unger and Decker, 1970; Unger and Mills, 1972]. An uppercrustal P velocity of 6.1 km/s was determined using a well-timed blast recorded on this array. This relatively high velocityis consistent with propagation through the T atoosh pluton,which is thought to directly underlie the volcano [Fiske et al.,1963]. Gravity results indicate that this pluton may only be afew kilometers thick (Z. F. Dane, personal communication,1978).

Because there are few data to control average crustal prop-erties, this study will necessarily be limited to determiningtypes of structures consistent with the data. Therefore it willrepresent further reconnaissance on the nature of crustal struc-ture under Mount Rainier.

EVENT- IO/l;/d7 EVENT- 12/27/67BAZ= 127. 2 BAZ= 1 ]GO. qDEL=46, O BEL=;. 2

'''''''''''''1 ''''''''[''''1

E_n%.1- E . DRAB ;6

_

I . . I . I , I . I , I , I , I , I , I , I , IO O 20 ]GO 4'0 l;O 60 O 10 20 ]GO 4'0 ;O dOSEC SEC

EVENT - 2/21/71BAZ= 1]G1. ;BEL=;. ;

RAD O.20

I , I , ! , ! , I i I , I/O 10 20 ]GO 40 l;O dO

$EC

EVENT - 12/21/ff7BAZ= 1]G2. 7BEL =;2. 7

' I " I " I " I ' I

TAN 1_I , I , I , I , I , I ID 1D 2D ]GO 4D ;D llO

$EC

Fig. 2a. Display of P wave form digitizeddata and vector rotations into the theoretical ray back azimuth. Positivevertical, radial, and tangential ground motion is defined as up, away from the source, and clockwise around the source(looking downward), respectively.Note the relativelylarge tangential components. Events on October 15, 1967; December27, 1967; February 21, 1971; and December 21, 1967.


3/14

LANGSTON:STRUCTUREUNDER MOUNT RAINIER 4751

EVENT - ;/7/66B EVENT - 7/2F/6; EVENT - 3/23/74 EVENT - 3/17/65BAZ= 1F6. q BAZ-22F. 7 BAZ=2I; 1. q BAZ=2I;2. 1OEL=16. 0 OEL=q2. F OEL=7. q OEL=F. 0

Z 1.DD Z 1.DD Z 1.DD

RAO '

I , I , I , I . I , I , I/ I . I . I , I , I , I , I I , I , I . I . I . I , JD 1D 2D 3D 'I,D S;:D 6D D 111 2D 3D ,I, rl S;:D 6D g 1D 2D 3D 'l,g s;:rl ig D 1D 2D 3D ,I,D S;:D 6D

SEC SEC SEC SEC

Fig. 2b. Same scheme as Figure 2a for events on A ugust 7, 1966 B; July 25, 1968;March23, 1974;andMarch 17, 1966.

LONG-PERIOD P WAVE FORM DATA ANDINSTRUMENT CALIBRATION

Figure 2 shows vector rotations of P waves from 12 eventsdistributed in three quadrants of back azimuth. Back azimuthis defined as azimuth clockwise from north, to the source asseen from the receiver. Source parameters were obtained fromISC (International Seismological Center) and PDE (Prelimi-nary Determinations of Epicenters) bulletins and are listed inTable 1 for all events used in this study. Initially, data fromonly a few deep teleseismic events were examined in the hopeof observing P to S converted phases on the radial component.By modeling the time and amplitude of these converted Pwaves in the time domain, estimates of crustal thickness and

author in a previous work [Langston, 1977b]. Although theother three quadrants were covered, no events were obtainedwith northwest back azimuths (Figure 2).

The vertical and horizontal component P wave data ofFigure 2 were digitized at an equal time increment of 0.25 sfrom paper copies made from film chips. The top and bottomof each trace was digitized and then averaged together. Severalevents were chosen for each quadrant to serve as a consistencycheck on the digitizing and rotation procedure.

A few pertinent observations can be made about the Pwaves of Figure 2. The first obvious one is that each rotationyielded a significant tangential component approximately onethird to one half as large as the radial component. In all casesit is considerably above the seismic noise level. We must first

interface velocity contrasts may be obtained. See Burdick and determine if this is a real effect. Several possibilities concerningLangston [1977] for further details on this kind of time domain the origin of the tangential component come to mind otherinterpretation. Once it became apparent that vector rotation of than earth structure: source mislocation (wrong back azi-the horizontal P wave forms into the theoretical ray direction muth), digitizing error, and instrument calibration are some.yielded significant tangential motion for one back azimuth, Source mislocation can be ruled out, since the error requiredmore azimuthal coverage was needed to determine the nature would be thousands of kilometers. In any case, even if theof the probable nonhorizontal structure. Teleseisms were ob- wrong back azimuth was used, both horizontal componentstained to cover as many back azimuths as possible to apply should differ only by an amplitude factor and be identical indipping interface theory following a suggestion made by the wave shape, provided structure is horizontal. A simple com-

EVENT - 1D/716 EVENT - 1/lq/6qBAZ=2q3. 7 BAZ=3D6. 7DEL=76. 2 DEL=,62. 1

! ' I ' I ' I ' I ' I ' II ' I ' I ' I ' I ' I ' I

NS _ O. 2

I , I , I , I , I , I , IO 1O 20 SO 'I'D S;O 16O

SEC

EVENT - ;/7/66ABAZ=2q$:. 3DEL=32. 4

', ' , ' , ' , ' , ' ,

NS D. 31

EW o;3-

_R 1_

0.36

RAD o. 4;

I , , I , I , I , I , I I , I , I , I , I , I , ID D 20 30 d'O S;:O 6D rl 10 2D 30 'l,O S;:O 6D

SEC SEC

EVENT - 1/2q/71BAZ=3nq. 7DEL=;3. q

I ' I ' I ' I ' I ' I ' I

_TAN 2_I , I , I , I , I , I ,

O 10 2D 3D 'I'D FO 60SEC

Fig. 2c. Same scheme as Figure 2a for events on October 7, 1968; August 7, 1966 A; January 19, 1969; and January 29,1971.


4/14


TABLE 1. Event Parameters

Date Time, UT Position, deg M H, km

Back

Distance, Azimuth,Location deg deg

March 17, 1966 1550:32.3 21.08S, 179.15W 5.9 627Aug. 7, 1966 A 0213:04.3 50.57N, 171.22W 6.3 29

Aug. 7, 1966 B 1736:28.5 31.74N, 114.51W 5.7 32Oct. 15, 1967 0800:52.6 11.92N, 85.98N 6.2 181Dec. 21, 1967 0225:21.0 21.89S, 70.07W 6.0 20Dec. 27, 1967 0917:50.3 21.29S, 68.20W 6.3 91July 25, 1968 0723:02.0 30.97S, 178.13W 6.5 17Oct. 7, 1968 1920:20.8 26.29N, 140.70E 6.1 518Jan. 19, 1969 0702:07.9 44.89N, 143.21E 6.3 238Jan. 29, 1971 2158:03.2 51.69N, 150.97E 6.0 515Feb. 21, 1971 1035:19.7 23.81S, 67.20W 6.0 166March 6, 1972 1850:16.8 50.14N, 148.80E 5.4 569March 23, 1974 1428:35.4 23.9S, 179.8E 6.1 535

Tonga 85.0 232.1Aleutians 32.4 295.3Gulf of California 16.0 156.9Nicaragua 46.0 127.2Chile 82.7 132.7Chile 83.2 130.9Kermadec 92.5 225.7Bonin Islands 76.2 293.6Hokkaido 62.1 306.7Sea of Okhotsk 53.9 309.7Chile-Argentina 85.8 131.5Sea of Okhotsk 55.9 309.0

Fiji 87.9 231.3

parison of any of the NS-EW components in Figure 2 showsthat there are always major differencesin wave shapes. Majordigitizing error was ruled out from comparison of the digitized

traces with the original data. Some error undoubtedly doesoccur, which is why many events were examined. Consistencyof equalized wave shapes of events from the same azimuth isgood, indicating little numerical error. This will be presentedin a later section.

Instrument miscalibration is the most serious factor whichmust be considered. Mitchell and Landisman [1969] indicatethat calibration of the long-period WWSSNsystem may be offby as much as 20% in the free periods and damping of theseismometer and galvanometer. Experience with calibration ofa similar long-period system at Pasadena has also indicatedpossible calibration errors of up to 20% (F. Lehner, personalcommunication, 1977). A parameter study was made whereseismograph constants were varied by 20%.Resulting theoreti-cal instrument responses [Hagiwara, 1958]were convolvedwith a simple pulse similar, for example, to the duration of theDecember 12, 1967, event. Differences in wave shapes wereslight and did not approach any of the characteristics of thedata. Figure 3 shows calibration pulses for the long-periodsystem at LON taken from the records containing the October7, 1968, event. The calibration current and electromechanicalconstant for each component is given in Table 2. Aside from asmall amount of long-period drift on the EW component theshapes and amplitudes of the pulses are similar. A check on therelative amplitudes expected between components based onthe data of Table 2 yields only a 3% difference from observedamplitudes. Therefore the LON long-period system was rea-

sonably calibrated.One more powerful argument can be made against the in-strument system being grossly miscalibrated. Figure 4 shows acomparison of wave forms for two events with similar pulse-like wave forms on the vertical component. The difference inback azimuth is almost 180 . Now, if one horizontal corn-

ponent was grossly in error so that it, say, produced an obvi-ously longer pulse than the other component, then it shouldbehave in a similar fashion for all recorded events. For ex-

ample, let us say that the EW component seen for the Decem-ber 27, 1967, event of Figure 4 consistently produces a long-duration pulse as compared to the NS component. However,the January 29, 1971, event data look reversed. In this case theNS component is obviously of longer duration compared tothe EW component. Therefore the assumption that in-struments are consistently miscalibrated is erroneous, whichleads to the conclusion that earth structure is causing theeffect.

P WAVE EQUALIZATION PROCEDURE

Shallow teleseisms and teleseisms with complicated appar-ent source functions had to be used to provide good azimuthal

coverage. Deep events with simple pulselike wave forms on thevertical component are preferred because later P to S con-versions on the horizontal components may then be investi-gated directly [Burdick and Langston, 1977]. Even when suchdata are available, however, there still remain differences be-tween effective source time functions, which can cause signifi-cant differences in synthetic seismogram calculations [e.g.,Langston, 1977a]. Thus a way of equalizing the data to com-pensate for differing source time functions would be helpful. Amethod of doing this is suggested by the form of the data andtheoretical considerations.

In the time domainthe form of the theoreticaldisplacementresponse for a P plane wave impinging under a stack of hori-zontal or dipping interfaces can be given by

Dr(t) = l(t) * S(t) * Ev(t)

D(t) = I(t) * $(t) * E(t)

Dr(t) = I(t) * S(t) * Er(t)

(1)

Fig. 3. Long-period calibration pulses taken from the October 7, 1968, recordings. Refer to Table 2 for calibrationconstants.


5/14

LANGSTON:STRUCTUREUNDER MOUNT RAINIER 4753

TABLE 2. Calibration Constants for the October 7, 1968, CalibrationCurves

Calibration Current, ElectromechanicalComponent mA Constant, N/A

Z 0.2 0.106NS 0.2 0.099

EW 0.2 0.097

where $(t) is the effectivesource time function of the imping-ing wave, l(t) is the instrument impulseresponse,and theEv(t),Es(t), and Er(t) are the vertical,radial, and tangentialstructure impulse responses, respectively.$(t) may be quitecomplicatedand related to dislocation time historyand sourcearea reverberations, for example. The E(t) are of the form

E(t)= {a,b(t-r,)+ $,H[b(t-r,)]} (2)where a, and , are constants related to the product of reflec-tion-transmission coefficients, fi(t) is the Dirac delta function,

error in assuming (3) can be explicitly examined. From raycalculations in a layer over a half-space model, the largestarrival on the vertical component besides direct P is the first Preverberation. Consider therefore

Ev(t) = a,fi(t- r,) + a,b(t - r,) (5)

where the first term on the right is direct P and the other isreverberated P. It can be shown, to first order, that the inerseoperation of Ev(t), Ev-(t), is given by

1 1 1 /

where

=

For simplicity, let a = 1 and r = 0. In practice, a/a -0.1. Therefore there will be no deconvolution error in apply-ingEv-X(t)or 1/Ew) to (4) until relativetimesequal to ra.Since aa/aa is smaller than relative amplitudes of converted

r is the travel time of the ith ray, and H[Hilbert transform operator. The summation is done over nrays. A common way to calculate the E(t) for plane layeredmodels is to use a propagator matrix technique in the fre-quency domain [Haskell, 1960,1962].In this case, for phasevelocitygreater than any P or $ velocity in the earth model,is zero for all i, n is infinity, and Er(t) does not apply. Equa-tions (1) also do not apply, in general, to the case of anincident $ wave. Even in the horizontally plane layered case,S(t) for $V waves will often be differentfrom that of the $Hwave.

A common observation made from teleseismic data ofsimpledeep events is that the verticalcomponent of groundmotion behaves as a pulselike time function convolved withthe instrument response with only minor later arrivals [Bur-dick and Helmberger, 1974].Theoretical calculations for typi-cal crustal structures show that crustal reverberations andconvertedphaseson the verticalcomponentof steeply incidentP waves are minor even if a high-contrast interface is allowedto dip [Burdick and Langston, 1977;Langston, 1977b].How-ever, this starts to break down for extremely high velocitycontrasts (> 2 km/s). Thus wecan approximate $(t) by letting

l(t) * S(t)- Dr(t) (3)

] representsthe phases on the horizontalcomponent,the error is small evenfor later times.Unfortunately, the deconvolution procedure of (4) is nu-

merically unstable because signals are band-limited and con-tain random noise. In practice, a technique suggestedby Helm-berger and Wiggins [1971] and Dey-Sarkar and Wiggins[1976b]is usedto estimate the deconvolution.This estimateisthen multiplied by the transform of a Gaussian to limit thefinal frequency band by excluding high-frequencysignals notobviouslypresent in the original recordings. In the frequencydomain this process is given by

Ds(o)Dv(o).G(o) (7)En'(co)=where

ss(o)-- max IDv(o)v(o),c max [Dv(o)Dv(o)]tand

G(co)= e

Es'(co)is the deconvolved'radial earth response,' and the barover Dr(co) denotes the complex conjugate. The function ss(co)can be thought of as simplybeing the autocorrelation

The implicit assumption is that Dr(t) behaves mostly like oneDirac delta function. Assuming that instrument responses arematched between components, Es(t) and Er(t) may be found

by deconvolvingl(t) * $(t) from Ds(t) and Dr(t). In thefrequency domain this process is given byDs(w) Ds(co)

Es(co)= l(w)S(co) '" Dr(w)(4)

Er(co) = l(w)S(co) - Dr(w)Es(co)and Er(co)are then retransformed back into the timedomain. Note the similarity of this technique with the spectralratio method previously used in these kinds of studies [e.g.,Kurita, 1973].The important differenceis that phase informa-tion is conserved. Furthermore, the resulting time series maybe interpreted directly as a seismogram, allowing times andamplitudes of arrivals to be examined in a relatively unam-biguous manner.

Assuming, for the moment, no noise, an estimate of the

12/27/67 1/29/71BAZ -- 130.9 BAZ-- 309.7

t = 83.2 t - 53.9

NS --

EW

Fig. 4. Comparison of P wave form data from two events withopposingback azimuths.Note that pulsecharacteralternatesfor eachhorizontal component, indicating that the effect is from earth structureand not instrument miscalibration.


6/14

4754 LANGSTON:STRUCTUREUNDERMOUNTRAINIER

LAYER OVER HALF SPACE COR MODEL

i

1.0A PpPmpPsPmp+PpSmp1.0VERTICAL ,,PP --_

0.45l\ Pis PpPmsI

RADIAL mp ,Doovo_,T,O_/RESULT V - --

Fig.5. Testofdeconvolutiontechniquewithsyntheticdatausingtwoplanelayeredearthmodels(Table3).AGaussianeffectivetimehistoryhasbeenconvolvedwiththeverticalandradialdisplacementforeachmodel.Thetwobottomcurvesshowtheresultofthedeconvolutionprocedureinwhichaninstrumentoperatorandasimplepulseliketimefunctionwereincludedinthesyntheticdata.PhasenomenclatureisafterBthandStefdnsson[1966].

of Dv(o)with any spectral troughs filledto a level dependingon the parameter c. In every deconvolutionto be presented, c= 0.01, and a, the parameter controlling the width of theGaussian, was set to 0.7. A Gaussian was chosen for itssmooth symmetric shape and pulselike character in the timedomain. The width of the Gaussian was chosen to approxi-mately match observed pulse widths of the simpler verticalwave forms. This helps to insure against getting too muchdetail which is not warranted by the observations.

Figure 5 shows synthetic examples of the deconvolutionprocess.A layer-over-half-spaceand the COR model (Table 3)were chosen to test the assumptions inherent in (4) and (7).Synthetics were calculated using Haskell's [1962]method. Ineach casethe verticalcomponentshowssignificantP multipleslater in the wave train. The deconvolution result was obtainedusing (7) for synthetic wave forms which included con-volutions of the instrument and an effective time functioncomparable to the December 27, 1967, event (Figure 2). Care-

ful comparison of the synthetic radial response and the decon-volution result shows that wave shape differences are smalland consist primarily of removing arrivals which end up as Pwaves at the receiver. Note, for example, the absence of thefirst P reverberation, PpPmp, in the layer-over-half-spacemodel. This will be true in general for any plane layeredstructure under which incideqt waves have small angles of

TABLE 3. Earth Models

Layer Vp, km/s Vs, km/s p, g/cm S Th, km

Layer Over Half SpaceI 6.0 3.5 2.6 40.02 8.1 4.7 3.2 ...

COR ModelI 5.5 3.0 2.6 10.02 6.7 3.9 2.8 6.03 8.0 4.6 3.2 5.04 7.9 4.5 3.15 1.55 7.8 4.2 3.1 2.06 7.7 3.7 3.0 1.57 7.5 3.5 2.95 1.58 7.2 3.3 2.9 2.09 6.9 3.3 2.85 2.5

10 6.6 3.3 2.85 12.911 8.0 4.6 3.2 ...

L ON ModelI 6.0 3.0 2.7 20.0

2 8.0 4.6 3.2 70.03 7.8 4.0 3.2 35.04 8.1 4.65 3.2

incidence. This occurs, since converted S waves are almostexclusivelyfound on the horizontal components;the ampli-tude ratiosof rays whichare P waves in their last ray segments,in relation to the direct wave, will be the same on both com-

ponents. Therefore this deconvolution technique (and anyother spectral ratio technique) essentially removes P waves,except for the direct arrival, and leaves conversions and rever-berations of the P to $ type for the plane layered problem.This willnot be strictlytrue for the dippinginterfaceproblem,since each ray can have a different angle of incidence and

RADIAL P TANGENTIAL P

10/15/67BAZ: 129.2A: 46 0 12./27/67

130 9

83 2.

2_./21/71131.5

85.8

12/2.1/67132:.7

82.7

7/25/68225.7

92:. 5

3123174

231 3

87.9

3/17/66232.0

85.0

J

o 18

0.17

I0.32

10.32

I 0.12

0.20

0.10

0.06

10/7/68

293.

76.

10.31 0.14

8 / 7/66

2:95.3

32:.4

1 / 19/69

306 7

62 1

1/29/71309.7 o

53 9

0.48

0.36

0.13

J 0.18

I 0.42 I0.26

I

A C AB C

k 30 sec I

Fig. 6. Deconvolution of effectivetime functionsand instrumentoperatorsof the radialand tangentialP wave formsin Figure2. Thedashedlinesindicateinterpretationsof arrivals.Amplitudes,relativeto the vertical component, are shown by each time series.


7/14

LANGSTON:STRUCTUREUNDERMOUNTRAINIER 4755

z

azimuth of arrival, ruining the relative amplitude similarity ofP waves between vertical and radial components.

The radial and tangential P wave forms of Figure 2 weredeconvolved using this technique and are shown in Figure 6. A5-s cosine taper was applied to the ends of each time seriesbefore deconvolution. Although the wave shape details ofwaves from similar back azimuths can vary, there is good

consistency in overall shape and timing. Spurious arrivalsbefore direct P are caused from a combination of originalseismic noise and the filling of troughs in the autocorrelationfunction of the vertical component. In other words, the origi-nal effective source time function coupled with noise can stillcause some effect through the deconvolution procedure. Thestability of tangential and radial wave shapes for events ofsimilar back azimuth over time, along with previous argu-ments, shows that particle motion effects are due to earthstructure and not instrument problems.

The peak times of three prominent arrivals are shown by thedashed lines of Figure 6. Arrival A is the peak of the direct Pwave. On the tangential components it is seen to be small andinterfering with arrival B. Arrival B is predominant in thesoutheast and northwest quadrants; note also the general waveform antisymmetry about northeast-southwest azimuths. Ar-rival C is represented primarily by the tangential P wave formsof southwest back azimuths. It also seems to occur in the

radial P wave forms as a positive peak at the top of Figure 6changing to a negative arrival near the bottom. Figure 7 showsnormalized radial-tangential particle motion for the three rep-resentative wave form groups of Figure 6. These plots demon-strate the off-azimuth arrival of direct P and the large tan-gential secondary arrivals.

INTERPRETATION OF THE P WAVE FORMS

We are now faced with the rather formidable task of ex-

plaining the wave forms of Figure 6. Ideally, the goal is tomake a physical model of earth structure under LON whosewave response matches the observed wave forms. However,the inverse problem is ill posed because there are few outsideconstraints on crustal properties, the crust is likely to be later-ally heterogeneous from geologicobservations,and the dataarearesultofnearlyverticalpropagationofbodywaveseventhough the events are azimuthally distributed. This kind ofdatacanonlyresolvevelocitycontrastsand averagecrustaltravel times.Becauseof these problemsand keepingOcham'sRazor in mind, a hypothesis-testing approach will be taken byexamining the effects of simple planar dipping earth models.

We start (and end) witha singledippinginterfacemodel

Calculations of responses for such models indicate that tan-gential P wave components are azimuthally antisymmetricabout an azimuth perpendicular to the strike of the interface[Langston, 1977b]. The polarities of tangential P and the Psconversion at the interface indicate the direction of dip. Figure8 shows a plot of arrival B polarity versus back azimuth. Fromits reverse polarity compared to direct tangential P, its relativearrival time of 3 s, and its azimuthal symmetry characteristics,arrival B is interpreted to be a Ps conversion occurring at aninterface from 15 to 20 km in depth, depending on crustalvelocities. Assuming, for the moment, that the velocity con-trast is from higher to lower velocity as depth decreases, theallowable dip azimuth lies within the northern sector boundedby solid lines on Figure 8. Note that polarity information wastaken from the August 7, 1966 B event (BAZ = 156.9). Thedeconvolution of this event was poor as determined frompresignal noise in the final result. This was probably partly due


8/14


o NEGATIVE

POSITIVE

Fig. 8. Polar plot of Ps tangential polarity versus back azimuth.There is no significance in the radius length to each point. Positivetangential polarity is taken to be clockwise, looking downward aroundthe source. Direct tangential P wave polarity is opposite for eachdatum. Acceptable interface dip directions are located in the northernsector bounded by the two solid arrows. The dotted arrow indicatesdip direction based on symmetry of NW and SE wave forms.

to digitizing problems, since the vertical component was highamplitude, but also partly due to the event's short range.Many different phases with differing phase velocities are un-doubtably present in these data, invalidating the assumptionof one incident angle for the incident wave. The dashed line(Figure 8) is dip azimuth for the interface inferred from sym-metry consideration. The SE and NW groups are more mirrorsymmetric than are the SE and SW wave groups; hence, andon the basis of model calculations, it is more reasonable with-out other constraining data to locate the $E and NW groups atsimilar azimuthal positions relative to dip azimuth. Note thatwave antisymmetry, off-azimuth P arrivals, and high tan-gential wave amplitude all argue against /nisotropic mediumeffects as discussed by Keith and Crampin [1977].

Although polarities can indicate interface dip and strike,wave amplitudes must be used to determine interface velocitycontrast and dip magnitude. The radial and tangential waveforms of Figure 6 exhibit large variation in amplitudes relativeto the vertical component. Part of this variation is due todiffering incident angles, but probably the greater part is dueto numerical digitization and rotation error. To get an esti-

P AND Ps MODEL SPACE

LAYER Of. , P Th O'k 1 6.0 3.0 2.7 20. 0.33

__ \

\\ 2 .... o.25 STRIKE=-55

$ oe o

- _ __oo e o o o

success - - --

0 I I I0 10 20 30

DIP ()(TOWARDS NNE)

Fig. 9. Inversion result to determine bounds of interface dip and Svelocitycontrast which explaintangential P and Ps relativeamplitudesassuming a 70% standard error. See text for details.

mate of interface contrast and dip and also to see the trade-offsinvolved, a numerical search and test inversion procedure wasset up to model inferred P and Ps tangential relative ampli-tudes. Amplitudes we, read from four distributed events(De-cember 27, 1967; July 25, 1968; October 7, 1968; and January19, 1969) and were assigned 70% standard errors. This largeerror conservatively incorporates the factor of 2 variation seen

in the relative amplitude data. Crustal velocities and densitiesfor the upper layer and interface strike were assumed withinterface dip and velocity contrast being incrementally varied.P and Ps amplitudes were calculated for each model andcompared to the observations. Those falling within the allow-able 70% were deemed successes, and those outside failures. Ifall four datums were satisfied by some structure model, thenthat model was considered viable and was output from theprogram.

Figure 9 shows the result from one such inversion. Thechoice of upper crustal velocities is not very crucial to thelocation of the boundaries of the model space for P and Psamplitudes, although a high Poisson ratio helps the P/Psratio. Lower P and S velocities would tend to raise the bound-aries, preserving the important general characteristics. Chang-ing interface strike within the bounds of Figure 8 also doeslittle to these results. First note that there is a pronouncedtrade-off between interface velocity contrast and dip; highcontrasts require small dips, and vice versa. Also note that anysuccessful model requires large contrasts and large dips toapproximate observed amplitudes. Thus the causative struc-ture can be considered a major tectonic feature of the area.

By symmetry in the problem and considering the behaviorof the transmission coetficientsat the interface, it is logical toassume that similar results could be obtained for an interface

RADIAL P TANGENTIAL P

12/27/67BA Z = 130 9 A = 8:5 2 OBS2 RAYS _

lORAYS3123174

231.5 o87.9

2 RAYS

10 RAYS

1 / 19/69:506 7

62 1

2 RAYS

0.2

0.07

0.18

lO RAYS

-30sec-- --Fig. 10. Comparison of deconvolved radial and tangential ? waves

with synthetics calculated for the single dipping interface model.


9/14

LANGSTON.'STRUCTUREUNDERMOUNTRAINIER 4757

EVENT - ;/6/72BAZ=50q. 0DEL=F9. q

I ' I ' I ' I ' I ' I ' I

z 1, oo

1.09

_E_1. O:

1. D;

_TA_I , I , I , I , I , I , I

O 10 20 50 40 90 60SEC

Fig. 11. Vector rotation of short-period P wave data. Note largeroverall amplitude of horizontal components relative to the vertical.

dipping in the opposite direction with the opposite velocitycontrast. However, when this kind of low-velocity zone modelwas tested, no model could be found to explain the observedamplitude. Angles of incidence of waves in low-velocity zonemodels were much less than those occurring with the previousnormal contrast models; i.e., incident horizontal ray parame-ter (phase velocity) was the same. Hence induced ray azimuthanomalies and surface incident angles were smaller, reducingtangential amplitudes.

Figure 10 displays synthetic seismogram comparisons withthe observed for the three best deconvolved events. These were

chosen on the basis of low presignal noise and representationof their respective back azimuth groups. The crustal model ofFigure 9 was used with an S velocity contrast and dip of 1.08km/s and 12.5 , respectively. Two calculations are shown foreach observation. In the first, only the rays P and Ps wereincluded. The other calculation includes eight other importantmultiples and converted reverberations in the single dipping

interface model. These rays produce the secondary oscillationsseen after Ps in the 10-ray model. The purpose in showingthese two calculations is to demonstrate both the partial suc-cess and obvious limitations of naively assuming the singledipping interface model. First note how well polarities andwave shapes are explained with inclusion of only P and Ps inthe model. This is true even of the March 23, 1974, event if the

large observed downswing (arrival C) is ignored for the mo-ment. Inclusion of the other conversions and reverberations do

little to help matters, however. Although the size of the rever-berations approximate the observed for December 27, 1967,

,they do little to explain arrival C for March 23, 1974. Someinsight into the physics of these rays provides an excuse, if notan explanation, for why the model breaks down so quickly.Azimuth anomalies for P and Ps refracting through the inter-face are of the order of 10-20 . That is, their apparent azi-muths of arrival are off by this amount with respect to trueback azimuth. Rays which reverberate even once in the struc-ture may have very large azimuth anomalies approaching 90or more, depending on the phase conversion sequence. Thus

coherent arrival of these multiples depends on the interfacebeing planar for large horizontal distances, since typical raypaths follow a tortuous route through the structure to thereceiver. If we assume that arrival C is an off-azimuth rever-

beration by this argument, then the structure is obviously notplanar over the distances required. Arrival C could also repre-sent a diffraction effect from unmodeled topography or otherdiscontinuous structure. There have been previous observa-tions of large secondary arrivals associated with body wave-to-surface wave conversion at topographic discontinuitiesfrom analysis of array data [Key, 1967]. However, without dT/dA and azimuth information for these later arrivals, not muchmore can be said about them.

Further insight and anxiety can be obtained by examiningthe effect of LON structure on short-period P waves. Figure 11is an example of a P wave vector rotation of an event from theSea of Okhotsh recorded at Longmire. First note that both theradial and tangential components attain larger overall ampli-tudes than the vertical component. Assuming that the first 3 sof the wave forms represents the direct P wave (as inferredfrom the long-period observations), the radial-vertical ampli-

EVENT - 11'I/7/6;BAz=2q:. 6DEL=7,4. 2

I ] ' I ' I ' [S..L.', ' i ]

--

EW D.41

- Fl. 57^D=

D.116I , I , I , I , I , I ,

D 2L1 4D 6D :$13 1DD$EC

EVENT - 1D/2F/6FBAZ=:DF. DDEL :61. 5

I ' I ' I ' I ' I ' I ' I

Z 1. DO

--

_ _

I , I , I , I , I , I , I

O 20 4D 60 :D 1DD 1211SEC

EVENT - 1/2q/71BAZ=SDq. 7DEL:FS. q

i , m , I ' ) , I , I ' )

_Z ll__NSS__:.74

I , m , m , m , m , m , m

D 2D 4D 6D D 1DD 12DSEC

Fig. 12. Vector rotation of S wave data from three deepevents.Interpretationsof S and Sp phasesare shown.The EWcomponent of the October 7, 1968, event is truncated because the trace was off scale.


10/14


EVENT - 113/7/i)AZ=2q:.DEL=7,. 2

I ' I ' I ' I ' I ' I ' I

Z 1. DDm

NS i 13.1Fi

I u- 2iqI

:q

.TANGI , I , I , I , I , I , I

0 10 2aSEC

EVENT - 111/2F/6FBAZ=DF. DDEL =fi 1.

RAD D. 66--

m

p , I , m , I , m , I , m

D 1D 2D D D FD 6DSEC

EVENT - 1/2q/71BAZ=:Dq. 7DEL=F:L q

I ' I ' I ' I ' I ' I ' II

1. aa

_E. a.6F_RR,!D.F7D. 7DTANGI , I , i , i , I , I , IO 113 20 IO 40 FD 60

O. fid

Fig. 13. Seismogram detail of the portion before direct S for the three events of Figure 12.

tude ratio does seem normal for the angle of incidence in-volved. However, if the large-amplitude arrival 3-4 s afterdirect P on the radial component is the short-period equivalentof arrival B, the long-period Ps conversion, then it impliesextreme frequency dependence not supplied by the singleplanar dipping interface model. No such amplitude ratio isseen for long-period radial P waves from any back azimuth(Figure 6). A possible mechanism for this effect could befrequency-dependent geometric spreading effects caused byundulations on the interface, provided the wavelength of thesepossible undulations are of the order of the wavelength ofshort-period waves. Long-period waves could be expected to

average over these possible undulations. SH ray calculationsin severe basin type structures indicate that internal causticsand focusing can produce large secondary arrivals [Hong andHelmberger, 1978].

In any case it is plain that there remain many problems inexplaining even some of the first-order effects observed in thewave form data. What is important is the compelling evidencethat there exists a major laterally heterogeneous structure un-der LON capable of severely distorting long-period seismicwaves. To a fair approximation, part of this structure includes

a high-contrast dipping interface, approximately planar, atlower crustal depths.

LONG-PERIODS WAVE FORMSANDSp CONVERSIONS

Long-period S wave forms from several deep events wereinvestigated with the hope of finding Sp conversions to deter-mine more constraints on the dipping interface model sug-gested by the P wave form data. S wave form data are neces-sarily limited to those of simple deep events in the distancerange 450-85 . P to S and S to P near-source conversionsproduce complicated effective source functions for shallowsource SV waves masking the effects of near-receiver con-

versions. S-coupled PL wave distortion for distances less than45 and core phases for distances greater than about 85 further complicate interpretation. It is also desirable to obtainS data from several events to determine the stability of any Spprecursor observation [Burdick and Langston, 1977].

S data from three events which meet these criteria are dis-

played in Figure 12. Because of LON's particular position withrespect to zones of deep earthquakes, satisfactory data arelimited to events occurring in the northwest Pacific. Of partic-ular interest in Figure 12 is the prominent Sp precursor with an

VERTICAL S

s

BAZ = 309 7 oA:539

WCALONMODEL * /'

\F'--LONMODEL

WITH90MOHODIPANDE=66

RADIAL S

2.2

2

2

100 sec H

Fig. 14. Comparison of observed and synthetic S waves for the January 29, 1971, event. See text for details.


11/14

LANGSTON:STRUCTUREUNDERMOUNTRAINIER 4759

18-s lead time before direct S. This phase is of the samepolarity as direct S on the vertical component but is of oppo-site polarity on the radial component. It is also largest on thevertical component, which is consistent with its P wave inter-pretation. The relative polarity and timing of this Sp phasesuggest conversion at an interface at about 145-km depth witha velocity contrast from higher to lower velocities as depthdecreases.

Figure 13 shows rotations of the same data up to the arrivalof direct S. Although there is some noise at this level on thetangential component, Sp does seem to have tangential par-ticle motion. This observation is consistent in polarity with anSp to S conversion at the dipping structure seen by the Pwaves. It may also be due in part to dip on the 145-km-depthinterface.

Interpretation of the relative amplitudes of Sp to S is ratherdifficult because there is poorer azimuthal and effective sourcetime function control. If the tangential SH wave is any in-dication of SV effective time function, then Figure 12 suggestsfairly simple pulselike incident waves. However, if the con-

version is dipping, there could be large azimuthal variation inthe Sp/S amplitude ratio, and hence data from other azimuthsare needed to exclude this possibility. Nevertheless, some fruit-ful insight may be gained by examining a few model calcu-lations.

Figure 14 displays four different model studies of the Janu-ary 29, 1971, event data. WCA is an upper mantle modelderived from short-period P wave form data, appropriate tothe LON area, by Dey-Sarkar and Wiggins [1976a]. Model T7was derived from short-period and long-period P wave formdata by Burdick and Helmberger [1978]and is appropriate forthe western United States. S wave velocities were derived from

these P wave models by assuming a Poisson solid (a = 0.25).The responses to WCA and T7 were calculated using Haskell's[1962] method for plane layered structures. Note from Figure14 that an Sp arrival is produced from conversion at the lowerboundary of the upper mantle low-velocity zone which hasapproximately the correct time advance, relative to S, butseverely underestimates the observed Sp/S ratio. The large Sparrival several seconds before S in both WCA and T7 calcu-lations is the conversion from the Moho.

In the absenceof dip, largerSp/S amplituderatiosmaybeobtained by increasing the S velocity contrast across the inter-face [Jordan and Frazer, 1975]. This is demonstrated in Figure14 by the LON model with no dipping interfaces (Table 3).This model is, of course, very idealized; it is only presented toshow the type of contrasts needed to explain the Sp/S ampli-

tude ratio. Crustal structure in the LON model is limited to asingle layer 20 km thick. Note that although the amplitude ofthe Sp conversion from the base of the LVZ is approximatelycorrect, the Sp conversion from the crust-mantle interface at20 km is apparently much too large. However, the effects ofthe dipping interface inferred from the P wave form data havenot been included in these models. If the same velocity con-trast at this crust-mantle interface is assumed, a dip of 9 maybe determined from Figure 9, which explains the P and Pstangential amplitudes. S wave polarization angle must also beestimated, since responses from dipping structures depend onlocal tangential and radial coordinates at each interface [Lang-ston, 1977b].The polarization angle change induced in S wavesfrom simple dipping interface models is of the order of themagnitude of the dip, depending on back azimuth. Thereforean estimate of polarization angle may be obtained directlyfrom the data. From model studies, variations of approxi-

mately 15 in polarization do not cause significant wave formvariations in calculated vertical and radial components unlessthe incident wave is nearly pure SH in the horizontal frame.

The bottom model in Figure 14 includes the same velocitystructure as the plane layered LON model except that theMoho is allowed to dip 9 toward the northeast (strike equals-55 ). A polarization angle of 66 , appropriate for the Janu-ary 29, 1971, event, was also assumed. Note that the combinedeffects of interface dip and S polarization significantly reducethe Sp(m) arrival.

Besides interface contrast, interface dip toward the north-west can increase the Sp/S ratio for S waves from northwestback azimuths. Thus there is some ambiguity in assigningphysical significance to the velocity contrast given in the LONmodel. The most important pieces of information are therelative timing and polarity of the Sp phase.

DISCUSSION

Like any other geophysical or geological study, this studysuffers from nonuniqueness problems because of theory limita-

tions and structure complexity. However, it is surprising inmany respects that so much general information can be de-duced from teleseismic data recorded at onl3/one receivingstation. Further constraints on details of the dipping structureunder LON could be obtained by a logical extension of thetechniques used here. A small broadband seismic array couldadd important dT/dA and azimuth of arrival information[e.g., Hayskor and Kanasewich, 1978]. These data, coupledwith ray arrival time and amplitude, would lend themselves togreater control over structure geometry, or at least provideexplanations for the behavior of the data.

The results of this study also suggest that useful reconnais-sance of structure at-specific areas could be obtained with thetemporary installation of only one well-calibrated, broadband,three-component seismic station. This would be important inaiding either earth structure studies or site selection for apermanent seismic station. In the later case the absence oflarge secondary and/or off-azimuth arrivals would form posi-tive criteria for choosing the site.

The deconvolution technique for removing the effects ofinstrument and time function seems to be a useful way ofmanipulating and equalizing wave form data for these kinds ofproblems. The effects of numerical error and time series trun-cation can be estimated more easily than in spectral inter-pretations, since phase information is implicitly included in thedeconvolved time series. For example, noncausal or presignalarrivals above the original seismic noise level are an immediate

indicator of the quality of the deconvolution. Error in theassumption that the vertical P component consists of only thedirect wave is small for most any horizontally layered structureand simple dipping structures. In these cases, if the verticalcomponent has sizable secondary P arrivals, then the horizon-tal components have even larger amplitude converted P to Sphases in relation to horizontal P. If, in a particular study, it isdeemed that more resolution is justified by the data than issupplied by this assumption, it becomes a simple matter tocalculate synthetic time series of transfer function ratios whichcan be compared with the data wave forms.

The tectonic and geologic implication of a high-contrastdipping interface under Mount Rainier is not clear. Becauselong-period waves are strongly affected by this structure it islikely that it has lateral dimensions exceeding typical long-period wavelengths of approximately 60 km or so. The 3-srelative arrival time between inferred Ps and P arrivals sug-


12/14

4760 LANGSTON:STRUCTUREUNDERMOUNTRAINIER

125 124 125o 122 121 /r

\

/

( ,I

49

48

47

46

0 20 40 60 80 100 KMI I I I I I

Fig. 15. Simplified Bouguer gravity anomaly map of western Washington showing tectonic trend of the OlympicMountains and Puget Sound horst structure.

gests an interface depth of 15-20 km. The interpretation fa-vored here is that this interface represents the Moho underMount Rainier. From a travel time study of P waves frommine blasts in the southern Puget Sound area, Zuercher [1975]

suggested that Moho depths just to the west of Mount Rainierand south of Puget Sound are only 18 or 19 km. These shallowdepths are similar to a 16-km M oho depth obtained in thecoast ranges of Oregon and Washington by Berg et al. [1966].Zuercher's study also suggested that the Puget Sound area andenvirons are regions of great crustal variability with majorundulations and dips on the Moho. Other geophysical studiessuggest that crustal structure for the area is discontinuous andcomplex [e.g., Dane et al., 1965]. Thus the laterally hetero-geneous structure inferred from this study is no real surpriseand is consistent in character with area structure.

Figure 15 shows a simplified Bouguer gravity anomaly mapfor western Washington [Bonini et al., 1974]. A line has beendrawn to indicate the transverse tectonic trend of the OlympicMountains and Puget Sound basement horst structure [Daneet al., 1965]. Most physiographic and tectonic trends in thePacific Northwest are usually aligned in a north-south fashion

(e.g., see Figure 1). It is interesting to note that Mount Rainierlies on the eastward extension of this transverse trend and thatthe trend azimuth is consistent with interface strike inferredfrom the P wave form data. Thus the Mount Rainier area may

be genetically linked with structure to the west but with acomplicated volcanic overprint.Alternatively, the anomalous crustal structure may be asso-

ciated with processes involving vulcanism or pluton emplace-ment. LaFehr [1965] interpreted gravity anomalies over severalSouth Cascade volcanoes and inferred relatively large low-density aureoles under them. Although apparently masked byan upper crustal pluton under Mount Rainier (Z. F. DaneS,personal communication, 1978), this kind of lenslike structuremay exist there at midcrustal depth and be responsible for theobserved dipping structure. In any case, since some long-period effects were not explained by the single dipping inter-face model, it is very likely that the presence of Mount Rainierand its associated structures has an important effect on localwave propagation.

The most straightforward interpretation of the long-periodSp observations is that they are produced by conversion at the


13/14

LANGSTON: STRUCTURE UNDER MOUNT RAINIER 4761

base of the upper mantle LVZ. It is also apparent that radiallysymmetric earth models appropriate for the area do not ex-plain the magnitude of the effect. This seems to imply thateither structure is not radially symmetric or that there aresome resolution problems with the body wave techniques em-ployed. This assumes, of course, that the observed Sp ampli-tudes are not a fortuitous result of effective source time inter-

ference for all three events.One purpose in looking for Sp arrivals was to determine if

there is any evidence for a dipping slab under the Cascades[McKenzie and Julian, 1971; Lin, 1974] following suggestionsmade by Sacks and Snoke [1977] and Snoke et al. [1977].However, other than the Sp arrival from 145-km depth, theredoes not seem to be any other evidence for interfaces between145-km and 20-km depth. The relatively shallow LVZ struc-tures inferred from COR receiver data and teleseismic wave

forms from the 1965 Puget Sound earthquake do not seem tohave an analogue beneath LON.

SUMMARY

Wave forms from a suite of azimuthally distributed tele-seismic earthquakes were vectorially rotated into theoreticalray directions based on published epicentral parameters.cause a mixture of deep and shallow events was used, anequalization procedure was developed to remove the effect ofdiffering source time functions and the instrument response.This method consisted of deconvolving the vertical P wavecomponent from the radial and applying a Gaussian filter. Itwas shown that errors from including possible P wave multi-ples on the vertical component in the deconvolution algorithmwere small. Large P multiples generally imply larger P to Sconversions and reverberations in planar structures.

Equalized P wave forms showed good consistency for eventsfrom the same back azimuth. In particular, the tangential P

wave forms exhibited relatively large amplitudes of one-thirdto one-half the radial component. These tangential wave formswere also antisymmetric about a NNE-NE azimuth. It hasbeen shown through a series of arguments that the anomaloustangential P wave particle motion observed from the tele-seismic observations cannot be attributed to instrument mal-function or miscalibration.

Most wave forms were suciently modeled by the inter-ference of direct P and a Ps conversion from an interface at

to 20-km depth dipping toward the NNE. An inversion studyof the polarity and amplitude of tangential P and Ps indicatedthat major trade-offs occur between interface velocity contrastand dip. The single dipping interface model broke down whenreverberations in the structure were considered. Teleseismic

short-period P observations also suggested that the structurewave response is frequency dependent. Thus the single dippinginterface model can be considered a first approximation to theshallower structure under LON.

On the basis of nearby refraction profiles and tectonic trendsit is suggested that the interface is the Moho under MountRainier and that the area is intimately connected with PugetSound and Olympic Mountains structure.

S wave form data from three teleseisms were used in an

attempt to constrain interface parameters inferred from the Pdata. No constraints were found on shallow structure, but anSp precursor was observed with an 18-s lead time before directS. Using average upper mantle and crustal velocities, the tim-

ing suggested conversion at about 145-km depth. The polaritywas consistent with a normal velocity contrast from higher tolower velocities as depth decreases. Comparison of synthetic

seismograms calculated from earth models appropriate for thearea showed that the arrival is probably due to conversion atthe base of the upper mantle LVZ. However, higher S velocitycontrasts or dip on this interface is needed to explain therelatively large observed amplitude of Sp.

Acknowledgments. I would like to thank Z. F. Dane of the Uni-versity of Puget Sound for sending me a preprint of his gravity resultsfor Mount Rainier and Doug Baumgardt for reviewing the manu-script. This research was supported by the Advanced Research Proj-ects Agency of the Department of Defense and was monitored by theAir Force Once of Scientific Research under grant AFOSR-78-3528.

REFERENCES

Aki, K., A. Christoffersson, and E. S. Husebye, Determination of thethree-dimensional seismic structure of the lithosphere, J. Geophys.Res., 82, 277-296, 1977.

Atwater, T., Implications of plate tectonics for the Cenzoic tectonicevolution of western North America, Geol. $oc. Amer. Bull., 81,3513-3536, 1970.

Bth, M., and R. Stefnsson, $-P conversion at the base of the crust,Ann. Geofis., 19, 119-130, 1966.

Berg, J. W., Jr., L. Trembly, P. A. Emilia, J. R. Hutt, J. M. Kling,L. T. Long, W. R. McKinight, S. K. Sarmah, R. Soutiers, J. V.Thiruvathukal, and D. A. Vossler, Crustal refraction profile, Ore-gon coast range, Bull. Seismol. Soc. Amer., 56, 1357-1362, 1966.

Bonini, W. E., D. W. Hughes, and Z. F. Dane{, Complete Bouguergravity anomaly map of Washington, Geol. Map GM-I 1, State ofWash. Dep. of Natur. Resour., Seattle, 1974.

Burdick, L. J., and D. V. Helmberger, Time functions appropriate fordeep earthquakes, Bull. Seismol. Soc. Amer., 64, 1419-1428, 1974.

Burdick, L. J., and D. V. Helmberger, The upper mantle P velocitystructure of the western United States, J. Geophys. Res., 83, 1689-1712, 1978.

Burdick, L. J., and C. A. Langston, Modeling crustal structurethrough the use of converted phases in teleseismic body-wave forms,Bull. Seismol. Soc. Amer., 67, 677-691, 1977.

DaneL Z. F., M. M. Bonno, E. Brau, W. D. Gilham, T. F. Hoffman,D. Johansen, M. H. Jones, B. Malfait, J. Masten, and G. O. Teague,Geophysical investigation of the southern Puget Sound area, Wash-ington, J. Geophys. Res., 70, 5573-5580, 1965.

Dehlinger, P., E. F. Chiburis, and M. M. Colver, Local travel-timecurves and their geologic implications for the Pacific Northweststates, Bull. Seismol. Soc. Amer., 55, 587-607, 1965.

Dey-Sarkar, S. K., and R. A. Wiggins, Upper mantle structure inwestern Canada, J. Geophys. Res., 81, 3619-3632, 1976a.

Dey-Sarkar, S. K., and R. A. Wiggins, Source deconvolution of tele-seismic P wave arrivals between 14 and 40 , J. Geophys. Res., 81,3633-3641, 1976b.

Fiske, R. S., C. A. Hopson, and A. C. Waters, Geology of MountRainier National Park, Washington, U.S. Geol. Surv. Prof Pap.,444, 93 pp., 1963.

Hagiwara, T., A note on the theory of the electromagnetic seismo-graph, Bull. Earthquake Res. Inst. Tokyo Univ., 36, 139-164, 1958.

Haskell, N. A., Crustal reflection of plane SH waves, J. Geophys. Res.,65, 4147-4150, 1960.

Haskell, N. A., Crustal reflection of plane P and SV waves, J.Geophys. ties., 67, 4751-4767, 1962.

Havskov, J., and E. R. Kanasewich, Determination of the dip andstrike of the Moho from array analysis, Bull. Seismol. Soc. Amer.,68, 1415-1419, 1978.

Helmberger, D., and R. A. Wiggins, Upper mantle structure of mid-western United States, J. Geophys. Res., 76, 3229-3245, 1971.

Hong, T.-L., and D. V. Helmberger, Glorified optics and wave propa-gation in nonplanar structure, Bull. Seismol. Soc. Amer., 68, 1313-1330, 1978. '

Johnson, S. H., and R. W. Couch, Crustal structure in the NorthCascade Mountains of Washington and British Columbia fromseismic refraction measurements, Bull. Seismol. Soc. Amer., 60,1259-1269, 1970. ,

Jordan, T. H., and L. N. Frazer, Crustal and upper mantle structurefrom Sp phases, J. Geophys. Res., 80, 1504-1518, 1975.

Keith, C. M., and S. Crampin, Seismic body waves in anisotropicmedia: Synthetic seismograms, Geophys. J. Roy. Astron. So.c., 49,225-243, 1977.


14/14

4762 LANGSTON: STRUCTUREUNDER MOUNT RAINIER

Key, F. A., Signal-generated noise recorded at the Eskdalemuir seis-mometer array station, Bull. Seismol. Soc. Amer., 57, 27-37, 1967.

Kurita, T., A procedure for elucidating fine structure of the crust andupper mantle from seismological data, Bull. Seismol. Soc. Amer., 63,189-209, 1973.

LaFehr, T. R., Gravity, isostasy, and crustal structure in the SouthernCascade Range, J. Geophys. Res., 70, 5581-5597, 1965.

Langston, C. A., Corvallis, Oregon, crustal and upper mantle receiverstructure from teleseismic P and S waves, Bull. Seismol. Soc. Amer.,67, 713-724, 1977a.

Langston, C. A., The effect of planar dipping structure on source andreceiver responses for constant ray parameter, Bull. Seismol. Soc.Amer., 67, 1029-1050, 1977b.

Langston, C. A., and D. E. Blum, The April 29, 1965, Puget Soundearthquake and the crustal and upper mantle structure of westernWashington, Bull. Seismol. Soc. Amer., 67, 693-711, 1977.

Lin, J. W., A study of upper mantle structure in the Pacific Northwestusing P waves from teleseisms, Ph.D. thesis, 98 pp., Univ. of Wash.,Seattle, 1974.

McKenzie, D., and B. Julian, Puget Sound, Washington, earthquakeand the mantle structure beneath the northwestern United States,Geol. Soc. Amer. Bull., 82, 3519-3524, 1971.

Mitchell, B. J., and M. Landisman, Electromagnetic seismographconstants by least-squares inversion, Bull. Seismol. Soc. Amer., 59,1335-1348, 1969.

Riddihough, R. P., The Juan De Fuca plate, Eos Trans. AGU, 59, 836-842, 1978.

Sacks, I. S., and J. A. Snoke, The use of converted phases to infer thedepth of the lithosphere-asthenosphere boundary beneath SouthAmerica, J. Geophys. Res., 82, 2011-2017, 1977.

Sengupta, M. K., and B. R. Julian, P-wave travel times from deepearthquakes, Bull. Seismol. Soc. Amer., 66, 1555-1579, 1976.

Smithson, S. B., Modeling continental crust: Structural and chemicalconstraints, Geophys. Res., Lett., 5, 749-752, 1978.

Snavely, P. D., and H. C. Wagner, Tertiary geologic history of westernOregon and Washington, Invest. 22, 25 pp., Div. of Mines andGeol., Olympia, Wash., 1963.

Snoke, J. A., I. S. Sacks, and H. Okada, Determination of the sub-ducting lithosphere boundary by use of converted phases, Bull.Seismol. Soc. Amer., 67, 1051-1060, 1977.

Unger, J. D., and R. W. Decker, The microearthquake activity of Mt.Rainier, Washington, Bull. Seismol. Soc. Amer., 60, 2023-2035,1970.

Unger, J. D., and K. F. Mills, Microearthquakes at Mt. Rainier--1969, Bull. Seismol. Soc. Amer., 62, 1079-1081, 1972.

Zuercher, H., A study of the crust in Puget Sound using a fixed seismicsource, M.S. thesis, 62 pp., Univ. of Wash., Seattle, 1975.

(Received December 27, 1978;

revised February 23, 1979;accepted February 26, 1979.)

langston 79

Documents