Ž .Physics of the Earth and Planetary Interiors 114 1999 71–80

Evolution of oceanic upper mantle structure

Yu-Shen Zhang ), Thorne LayInstitute of Tectonics, Earth and Marine Sciences Building, UniÕersity of California, Santa Cruz, CA 95064, USA

Received 8 December 1997; accepted 6 November 1998

Abstract

Love and Rayleigh wave phase velocities increase systematically with increasing age of oceanic lithosphere up to 150Ma. The rates of increase differ between oceans and vary for different age intervals. Modeling of lithospheric age–phasevelocity relations indicates that the high velocity seismic lid thickens and the velocities in the sub-lithospheric low velocity

Ž .zone LVZ increase with age at different rates between oceans. The initial thickness of the lithosphere near mid-oceanridges, as averaged by long-period surface waves, varies between 10 and 45 km, and 100 Ma lithosphere has thickness from83 to 110 km, for models with isotropic shear velocities. Hotspots in oceanic areas appear to modify oceanic lithosphereevolution, producing deviations from average patterns with increasing age. q 1999 Elsevier Science B.V. All rights reserved.

Keywords: Surface wave; Ocean mantle structure; Evolution

1. Introduction

Oceanic lithosphere is the stiff, coherently trans-Ž .lating portion plate of the chemical and thermal

boundary layer at the top of the Earth’s dynamicconvection system. Understanding how oceaniclithosphere evolves from its creation at upwellingsbeneath mid-ocean ridges until it plunges into theinterior at subduction zones is a key geodynamicproblem which is not yet resolved. Seismologicalevidence generally indicates that the oceanic‘seismic’ lithosphere, as defined by the depth to thebase of the high velocity ‘lid’ overlying the sub-lith-

Ž .ospheric low velocity zone LVZ , increases in thick-Žness with age Forsyth, 1977; Yu and Mitchell,

1979; Anderson and Regan, 1983; Wiens and Stein,

) Corresponding author. Fax: q1-831-459-3074; e-mail:yushen@es.ucsc.edu

1983; Nishimura and Forsyth, 1989; Zhang and Tan-.imoto, 1991, 1992, 1993 , but there is significant

variation in seismic models for oceanic upper mantlestructure. For example, for 100 Ma oceanic plate,various models indicate that the thickness of the

Žseismic lithosphere is as little as 50 km Anderson. Žand Regan, 1983 or as much as 90 km Forsyth,

.1977; Zhang and Tanimoto, 1991, 1992 .Ž .Zhang and Lay 1996 determined global Love

and Rayleigh wave phase velocity variations forperiods from 85 to 250 s. Their data and inversionstability were carefully checked, and the results arecompatible with other recent global and regional

Žseismic studies Laske and Master, 1996; Ekstrom et¨.al., 1997 . Using the Love and Rayleigh wave phase

Ž .velocity dispersion data Zhang and Lay, 1996 andŽ .the recent seafloor age map Mueller et al., 1995 ,

we obtained lithospheric age–phase velocity curvesŽ . Ž . Ž .for Pacific PAC , Indian IND and Atlantic ATL

0031-9201r99r$ - see front matter q 1999 Elsevier Science B.V. All rights reserved.Ž .PII: S0031-9201 99 00047-3

( )Y.-S. Zhang, T. LayrPhysics of the Earth and Planetary Interiors 114 1999 71–8072

oceans. We found that surface wave phase velocitiesincrease systematically with increasing age of oceaniclithosphere up to 150 Ma, and that the rates ofincrease differ between oceans and vary for differentage intervals.

Assuming that the surface wave phase velocitiesare determined by the upper mantle structure, andthat the oceanic upper mantle structures vary withage and depth, we modeled the lithosphere age–surface wave phase velocity relations in three oceansand constructed oceanic upper mantle structure mod-els. The results indicate that the high velocity seis-mic lid thickens and the velocities in the sub-litho-spheric LVZ increase with age. The rates are differ-ent between oceans. The initial thickness of thelithosphere near mid-ocean ridges varies from 10 and45 km, and 100 Ma lithosphere has thickness be-tween 83 and 110 km, for models with isotropicshear velocities.

Hotspots, the surface feature of mantle plumes,are generally accepted to be connected to the deep

Žconvective systems Morgan, 1972, 1984; Wilson,.1973 . Several recent seismic studies indicated that

some hotspots are associated with slow velocityŽanomalies in the mantle Zhang and Tanimoto, 1992,

.1993; Grand, 1994; Wolfe et al., 1997 , but theirnature remains enigmatic. Using obtained oceanicupper mantle models, we calculated surface wavephase velocity dispersions, and constructed surfacewave phase velocity residual maps for the PAC,IND, and ATL in the current study. We found thatmany hotspots are in or near to the slow velocityregions in the residual maps. Hotspots in oceanicareas appear to modify oceanic lithosphere evolution,producing deviations from average patterns with in-creasing age.

2. Phase velocity vs. lithosphere age

Using about 30,000 seismograms from earth-Ž .quakes with MG6.0, Zhang and Lay 1996 deter-

mined Love and Rayleigh wave phase velocity varia-tions from 85 to 250 s. All seismograms underwentcareful quality control in the time and frequencydomains. Fig. 1 shows maps of resulting Love andRayleigh wave phase velocity variations at a periodof 150.1 s. The seismic phases G1, R1, G2 and R2

are used. These phase velocity variation maps arewith a hybrid parameterization, in which an initialiteration retrieves the low order spherical harmoniccomponents that are used as an aspherical referencemodel for performing final block model inversions.These models are corrected for crustal thicknessvariations, which mainly affect the baseline betweenoceans and continents. Results for other periods can

Ž .be found in the work of Zhang and Lay 1996 , andthe patterns in the period range 85 to 150 s are quitesimilar. These results are used here to analyze first-order features of global oceanic upper mantle struc-ture, allowing for the spatial smoothing of the inver-sions and the intrinsic averaging properties of inter-mediate to long-period surface waves. Casual inspec-tion of the surface wave phase velocity maps indi-cates that velocities generally increase with litho-spheric age in oceanic regions, but there are some

Žstrong departures from this pattern note, for exam-.ple the relatively low velocities northwest of Hawaii .

ŽUsing the recent seafloor age map Mueller et al.,.1995 , we determine lithospheric age–phase velocity

relationships for both Love and Rayleigh waves withperiods from 85 to 250 s in the PAC, IND, and ATL,

Ž .along with an average AVE for all three oceans.Marginal basins and some boundary areas of oceansare not included because their ages are not wellcharacterized.

Fig. 2 shows lithospheric age–phase velocity rela-tions for PAC, IND, ATL and AVE for Love waveperiods of 85.5, 100.6, 150.1 and 200.8 s, andRayleigh wave periods of 85.5, 100.6, and 150.1 s.Phase velocities increase systematically with oceanfloor age in every region, but this age-dependence isstrongest for short-period surface waves and de-creases with period, indicating that plate effects areconfined to the uppermost mantle. The other interest-ing feature is the lithospheric age–phase velocityrelation differences between oceans. For young litho-sphere, the fast spreading PAC has the lowest phasevelocities, while the slow spreading ATL has thehighest phase velocities. This indicates a spreading

Žrate control on the seismic structure Zhang and.Tanimoto, 1991, 1993 . Ridge with a slow spreading

rate will have a slower upwelling, which producesmore cooling to the surface beneath the ridge, andthickens the lithosphere at the axis and terminatesmelting at a greater depth. Given the smoothing

( )Y.-S. Zhang, T. LayrPhysics of the Earth and Planetary Interiors 114 1999 71–80 73

Ž .Fig. 1. Surface wave phase velocity variations at a period of 150.1 s. These results are from Zhang and Lay 1996 . The plate boundary andŽ .coast lines are shown. The contour interval is at half of the shading scale. The solid circles indicate locations of surface hotspots. a

Ž .Rayleigh waves; b Love waves.

effects of the tomographic inversion, one must allowfor the difference in lateral averaging in each regionŽa larger range of lithospheric age is sampled by agiven long-period surface wave in the ATL than in

.the PAC , but the differences in Fig. 2 are notaccounted for by this effect. Variations are alsofound between plates with different spreading ratesfor heat flow measurements within regions less than

Ž80 Ma Parsons and Sclater, 1977; Sclater et al.,1980; Anderson and Skilbeck, 1981; Stein and Stein,

.1992 and in seismic attenuation structure beneaththe Mid-Atlantic Ridge and the East Pacific RiseŽ .Canas and Mitchell, 1981 .

Conventional models for oceanic thermal evolu-Žtion Parsons and Sclater, 1977; Sclater et al., 1980;

.Stein and Stein, 1992 suggest that oceanic litho-sphere cools and subsides as the thermal boundarylayer thickens in the first 60–80 Ma, but then thethickness of the thermal lithosphere becomes con-

Žstant, possibly due to small-scale convection Par-.sons and McKenzie, 1978 , viscous shear stress heat-

Žing of the base of the lithosphere Schubert et al.,.1976 , andror thermal rejuvenation effects of super-

Žplumes and hotspots McNutt and Judge, 1990; Lar-.son, 1991; Larson and Olsen, 1991 . Fig. 2 indicates

that surface wave phase velocities continue to in-crease with lithospheric age up to 150 Ma in eachocean, in contrast to the behavior of heat flow andwater depth. The lithospheric age–phase velocitycurves can be roughly divided into three domains.

( )Y.-S. Zhang, T. LayrPhysics of the Earth and Planetary Interiors 114 1999 71–8074

The first is for ages less than 40 Ma, where surfacewave phase velocities in IND, PAC, and AVE in-crease with age rapidly and are correlated withspreading rate. The rate of increase for ATL islower. The second domain is from 40 to about 100Ma. The rate of increase of phase velocities reducesrelative to the first domain, and there are evendecreases in velocity at some frequencies. Love wavephase velocities for IND increase to exceed those ofATL, while Rayleigh wave phase velocities for INDremain lower than those for ATL. Observed heatflow measurements for ages from 40 to 100 Magenerally have higher values than in theoretical cal-

Žculations Parsons and Sclater, 1977; Sclater et al.,1980; Anderson and Skilbeck, 1981; Stein and Stein,

.1992 , suggesting that thermal perturbations affectthis age range. Many hotspots, such as Arnold,Crozet, Discovery, Hawaii, Marquesas, MacDonald,Reunion, St. Helena, Tahiti, and Trindade, are lo-

Ž .Fig. 2. a Ocean floor age vs. Love wave phase velocity fordifferent oceans. The phase velocities are averaged in 10 Masegments with 1 standard deviation being shown by the error bars.

Ž .The phase velocities are from Zhang and Lay 1996 , and the ageŽ .data are from Mueller et al. 1995 . ATL: Atlantic Ocean; IND:

Ž .Indian Ocean; PAC: Pacific Ocean, AVE: three ocean average. bŽ .Same as a , but for Rayleigh wave.

Ž .Fig. 2 continued .

Žcated in ocean regions in the 40–100 Ma range Fig..1 , and thermal anomalies associated with these up-

wellings may account for both the seismic and heatflow behavior. Love wave phase velocities in thethree oceans converge near ages of 100 Ma, anddiverge at larger ages. The third domain is from agesof 100 to 150 Ma. Phase velocities in this domainincrease, but at different rates between the oceans.Phase velocities tend to decrease beyond 150 Ma,but this may be an artifact of small ocean floor areasand seismic velocities that are biased by proximity tosubduction zones.

3. Modeling oceanic upper mantle structure

We modeled the upper mantle structure causingthe lithospheric age–phase velocity patterns usingsimply parameterized models, which fit the Love andRayleigh wave dispersion simultaneously. After in-specting the lithospheric age–phase velocity rela-tions, we excluded Love waves with periods longerthan 200 s and Rayleigh waves with periods longerthan 150 s, because the data and relationships appearunstable in those cases. Ocean bathymetry and crustalthickness have large effects on surface wave propa-

( )Y.-S. Zhang, T. LayrPhysics of the Earth and Planetary Interiors 114 1999 71–80 75

gation, and the observations are corrected to a uni-form crustal model given by the Preliminary Refer-

Ž . Žence Earth Model PREM Dziewonski and Ander-.son, 1981 . We also did tests to change the bottom

depth of the LVZ from 200 to 250 km, and foundthat the surface wave phase velocity, which are usedin this study, have variation less than 0.3%. Notethat previous seismological studies indicate that thevelocity variations deeper than 220 km are much

Žsmaller than in the uppermost mantle Zhang andTanimoto, 1991, 1992; Laske and Master, 1996;

.Zhang and Lay, 1996 , so we assumed that thevelocity structure below 220 km depth is the same asPREM and that the bottom depth of LVZ is at 220km. The eigenfunction kernels of our intermediateand long-period surface wave phase velocities aremainly affected by the velocity structure in the high

Ž .velocity uppermost mantle layer LID and in LVZ,so we attempt to fit the Love and Rayleigh waveregionalized dispersion observations using simplemodels with varying LID and LVZ structures.

Seismic velocity structure is a manifestation oftemperature, composition, partial melting and dy-

Ž .namic state via anisotropy of the mantle. However,there are few clear relationships between these pa-rameters other than a handful of laboratory experi-ments. Simple parametric forms of the velocity varia-tions are used given that we have little a prioriconstraint on the structure. We assume that the ve-locity in the LVZ is a function of depth and litho-spheric age given by:

EV EVV t , z sV q D tq D z , 1Ž . Ž .o

Et Ez

where V is the shear wave velocity beneath theo

mid-ocean ridge at the top of LVZ, t is the oceanlithosphere age in Ma, and z is the depth from areference position at the base of the region beingperturbed to the bottom of the lid. Two basic typesof thermal models, the half-space cooling modelŽTurcotte and Oxburgh, 1967; Parker and Oldenburg,

.1973; Crough, 1975; Yoshii, 1975 and the plateŽmodel McKenzie, 1967; Sleep, 1969; Parsons and

Sclater, 1977; Sclater et al., 1980; Stein and Stein,.1992 have been used to model young ocean heat

flow and water depth successfully. These modelspredict that heat flow and water depth vary withagey1r2 and age1r2, respectively. In this study, we

adopt an age1r2 relationship for thickness of theoceanic lithosphere as a function of age,

'HsAqB t , 2Ž .where A and B are constants, and t is the oceaniclithospheric age in Ma. We step through models withA, B, V , EVrEt, and EVrEz varying with incre-o

ments of 0.2 km, 0.1 km May1r2, 5=10y3 km sy1,10y5 km sy1 May1, and 10y3 sy1, respectively,calculating the misfit of observed phase velocitiesfor Love wave for periods of 85.5, 100.6, 150.1,200.1 s and Rayleigh wave for periods of 85.5,100.6, and 150.1 s. The misfit that we seek tominimize is:

1r2n1ss V yV , 3Ž .Ž .Ý i iny1 is1

where V is the observed average phase velocity ini

every 10 Ma increment for each frequency and wavetype, V is the calculated phase velocity, and n is thei

total number of data. The preferred model is thatwith the smallest misfit, although this assumes valid-ity of the basic model parameterization.

The lithosphere age–phase velocity relationchanges between oceans, and a single model cannotfit the observations for all three oceans to within theprecision of the measurements. Therefore, we deter-mine models for ATL, IND, and PAC independently,as well as for AVE. As noted above, the lithosphere

Ž .age–phase velocity curves Fig. 2 show some varia-tions as a function of age, so we determined threesets of models; the first uses data up to 40 Ma, thesecond uses data up to 110 Ma, and the third fits thedata up to 150 Ma. The results for data less than 40Ma can be compared to heat flow and ocean depthobservations in young oceanic regions.

Not all of the parameters in our simple modelsproved to be resolvable. Parameters A, B, and EVrEtaffect the misfit strongly, while parameters V ando

EVrEz are not resolvable. Table 1 gives the optimalparameters and standard deviations obtained for thethree cases. It appears that mid-ocean ridge spread-ing rate, which is related to both slab pull and ridgepush forces, plays an important role in determiningthe oceanic upper mantle structure. For example, theparameter B, which is associated with thermal evolu-tion of the boundary layer, correlates with spreading

( )Y.-S. Zhang, T. LayrPhysics of the Earth and Planetary Interiors 114 1999 71–8076

Table 1The best fitting models

Ocean A B V EVrEt EVrEz so

( )a Age less than 40 MaPAC 12.6 6.300 4.345 0.00098 0.000 0.0043IND 40.4 4.200 4.345 0.00128 0.000 0.0085ATL 44.6 0.800 4.360 0.00013 0.001 0.0110AVE 16.6 6.300 4.370 0.00078 0.000 0.0058

( )b Age less than 110 MaPAC 6.0 10.700 4.280 0.00045 0.001 0.0093IND 34.4 6.500 4.350 0.00074 0.000 0.0101ATL 20.4 6.300 4.360 0.00017 0.001 0.0107AVE 10.0 8.600 4.375 0.00023 0.00 0.0076

( )c Age less than 150 MaPAC 17.2 9.100 4.260 0.00097 0.001 0.0124IND 39.4 4.900 4.355 0.00073 0.000 0.0122ATL 15.8 7.300 4.360 0.00015 0.001 0.0110AVE 16.2 7.500 4.365 0.00052 0.000 0.0097

The units are, A: km; B: km May1r2 ; V : km sy1 ; t: Ma; z: km;o

s : km sy1.

rate. PAC has the largest value and ATL has thesmallest value. Parameter A, the initial lithospherethickness beneath the mid-ocean ridges, as averagedover by our surface wave data, has large variationsbetween oceans. It is 12"6 km in PAC, 36"4 kmin IND, and 30"15 in ATL.

Parameter B for ATL in the first set of models is0.8, much smaller than for the other two groups; theinitial lithosphere thickness, A, for ATL changesfrom 15 to 45 km for the three cases, and parameterB for ATL and IND swap in relative size in fittingthe data out to 150 Ma. Thus, the specific modelparameters are not uniquely defined in the study.One reason is that the horizontal resolution of thephase velocity maps used in this investigation isabout 1600 km, the ATL is associated with a halfspreading rate about 40 km May1, then, the usedsurface wave phase velocity data cannot resolve thedetail variation near the Mid-Atlantic Ocean Ridge,or the current results have large uncertainty. Notethat there are relatively few data points for areasolder than 100 Ma as well, so the results of the thirdgroup of models are less stable.

Even if some model parameters are not stable, thebasic features are robust in this study. V , the veloc-o

ity at the top of LVZ, increases from PAC to IND to

ATL in all three groups of models, and does not varymuch between sets of models. We found that EVrEtis resolvably non-zero, while EVrEz is essentiallyzero in all cases. The shear velocity in the LVZ

Ž .increases with age Yoshii, 1975; Forsyth, 1977 . Itis reasonable to suggest that cooling occurs withinthe LVZ as the age of the overlying plate increases.This feature is in conflict with the plate modelŽParsons and Sclater, 1977; Sclater et al., 1980; Stein

.and Stein, 1992 .Considering the data quality and stability of the

results, we prefer the models obtained by fitting data

Fig. 3. The top panel shows preferred seismic models for PAC,IND, ATL and AVE oceanic structure. The top three layers arethe ocean water, upper crust and lower crust, all of which arelaterally uniform. The age-dependent layers involve the mantle lidand the underlying LVZ. The velocity structure in the crust andbelow 220 km are the same as the PREM structure for all models.The lower panels are shear wave velocity structures for PAC,IND, ATL and AVE, respectively. The thin solid lines indicate thereference PREM structure. The thick solid lines indicate shearwave velocity for the region for 5 Ma lithosphere, and the thickdotted lines are for 105 Ma lithosphere.

( )Y.-S. Zhang, T. LayrPhysics of the Earth and Planetary Interiors 114 1999 71–80 77

out to 110 Ma. Fig. 3 shows the upper mantleseismic models for this case. The thickness of theoceanic lithosphere increases with age, consistent

Žwith previous seismological studies Forsyth, 1977;Yu and Mitchell, 1979; Anderson and Regan, 1983;Wiens and Stein, 1983; Nishimura and Forsyth, 1989;

.Zhang and Tanimoto, 1991, 1992, 1993 . However,the rate at which the lithosphere thickens varies fromplate to plate. The base of the lid for 100 Ma plate isat 113.0, 99.4, 83.4 and 96.0 km for PAC, IND,ATL and AVE, respectively. These values are con-sistent with the lithosphere thickness of the recent

Ž .plate model GDH1 Stein and Stein, 1992 , 95"15km, obtained using heat flow and sea floor depth.The velocity, V , at the top of LVZ below mid-oceano

ridges is 4.280, 4.350, and 4.360 km sy1 for PAC,

Ž .Fig. 4. a The observed surface wave phase velocities and onestandard deviation vs. the ocean floor age in PAC in every 10 Ma.The solid lines are calculated surface wave phase velocity using

Ž . Ž .the PAC upper mantle model in Fig. 3. b Same as a , except forATL.

Ž .Fig. 4 continued .

IND, and ATL, respectively. These subtle variationsmay be related to the composition, temperature, anddegree of partial melting under each mid-ocean ridge,which is expected to vary due to the plate spreadingrate and relative importance of passive vs. activeupwelling. The velocity structures are very similar inthe youngest oceans, and the differences increase outto an age of 110 Ma.

Fig. 4a and b indicate the observed and calculatedLove and Rayleigh wave phase velocity using mod-els in Fig. 3 for PAC and ATL. Most of the calcula-tions are within one standard deviation, and Lovewave phase velocity has better fitting than Rayleighwave. The current study indicates that a simple

Žboundary layer model Turcotte and Oxburgh, 1967;.Parker and Oldenburg, 1973 , which has only two

parameters, the initial temperature and the age, can-not explain these differences between oceans, andthe seismic velocity structures motivate a more com-plex model for ocean lithosphere evolution.

( )Y.-S. Zhang, T. LayrPhysics of the Earth and Planetary Interiors 114 1999 71–8078

4. Phase velocity residual maps

Our preferred model has lateral variations anddifferences from plate to plate, but it is notable thatsatisfactory fits to the long-period dispersion datawere obtained with purely isotropic shear velocitymodels. This may reflect the averaging involved inthe construction of the age–phase velocity curves foreach plate, with azimuthal anisotropy effects averag-ing out. While dispersion curves for anisotropicstructures can be computed and compared with thedata quite readily, the number of additional parame-ters involved causes this to be an underdeterminedproblem. Our general sense is that the basic phasevelocity differences between oceanic regions arerather robust, and anisotropic modeling may changesome aspects of the isotropic models determined for

each region in this study, but the basic structuraldifferences will persist.

Differences between oceanic regions younger than40 Ma and regions from 40 to 110 Ma are found in

Žheat flow Parsons and Sclater, 1977; Sclater et al.,1980; Anderson and Skilbeck, 1981; Stein and Stein,

. Ž1992 , ocean water depth Parsons and Sclater, 1977;.Sclater et al., 1980; Stein and Stein, 1992 , and our

lithosphere age–phase velocity relations. To probethis issue further, we subtract parametric phase ve-locity predictions, calculated using the models foreach ocean in Fig. 3, from the observed phase veloc-

Ž .ity maps Fig. 1 . This results in surface wave ‘resid-ual’ phase velocity maps, which indicate the spatialpatterns in deviations from the simple velocity mod-els that fit each plate on average. Fig. 5 shows theresidual maps at a period of 150.1 s. At this period,

Fig. 5. ‘Residual’ maps of Love and Rayleigh waves at a period of 150.1 s in oceanic areas. The maps are obtained by subtraction ofŽ . Ž . Ž . Ž .lithospheric model phase velocities Fig. 3 from observed phase velocities Fig. 1 . a Rayleigh wave residual anomalies; b Love wave

residual anomalies.

( )Y.-S. Zhang, T. LayrPhysics of the Earth and Planetary Interiors 114 1999 71–80 79

the surface waves are quite sensitive to shear wavestructure in the upper 300 km of the mantle. Toallow application to the entire ocean, we extrapolatedthe ocean models up to 160 Ma old lithosphere.

Comparing Fig. 5 with Fig. 1, the mid-oceanridge features have disappeared, and there is noobvious age-dependence in the residual phase veloci-ties. This indicates that our models provide reason-able approximations to the average plate structure ineach ocean basin. There are relatively fast and rela-tively slow velocity regions in Fig. 5. There are 35hotspots with locations in the ocean basins, many ofthem are located in slow velocity areas in the resid-ual maps. Caroline, Easter, Galapagos, Hawaii, JuanFernandez, Marquesas and Tahiti in the PAC, Ascen-sion, Azores, Bermuda, Bouvet, Cape Verde, Dis-covery, Fernando, New England, St. Helena, Tristande Cunha and Vema in the ATL, Amsterdam andCrozet in the IND are located close to slow velocitypeaks. Given the general notion that many hotspots

Žare associated with deep mantle plumes Morgan,1972, 1984; Wilson, 1973; Yoshii, 1975; Vogt, 1981;Vink et al., 1985; White and McKenzie, 1989;Campbell and Griffiths, 1990; Griffiths and Camp-

.bell, 1990; Sleep, 1990; Duncan and Richards, 1991and have high temperature and high degree of partialmelting, it appears that hotspot upwellings are re-sponsible for perturbations about the mean platetrends solved for in Fig. 3. The local perturbations inlithospheric temperature structure associated with thehotspots may also be manifested in the heat flow andwater depth anomalies in oceanic lithosphere of in-termediate age. This does not preclude other factorssuch as basal heating or small scale convectionbeneath the plate from playing a role, but it appearsthat oceanic lithosphere should be modeled withdynamic parameters such as spreading rate andhotspot thermal resetting being included. Note somehotspots are associated with slow velocity peaks inone map, but not in the other map in Fig. 5, this mayindicate the depth, anisotropy, or other effects.Clearly, more detailed investigation is needed.

5. Conclusions

Rayleigh and Love phase velocities in PAC, IND,and ATL are used to probe the oceanic upper mantle

structure simultaneously. The Rayleigh and Lovewave phase velocities in period range 85 to 150 sincrease systematically with increasing age of oceaniclithosphere up to 150 Ma. Using the recent seafloor

Ž .age map Mueller et al., 1995 , we found that theage-dependence is strongest for short-period surfacewaves and decreases with period, indicating thatplate effects are confined to the uppermost mantle,and that the fast spreading PAC has the slowestphase velocities, while the slow spreading ATL hasthe highest phase velocities in the young ages, indi-cating that the spreading rate controls the oceanicupper mantle evolution.

Assuming that the lithosphere structure and seis-mic velocity in the upper mantle varies with depthand age, we modeled the oceanic upper mantle struc-ture in PAC, IND, ATL and AVE, that the velocitychange with age affects oceanic upper mantle struc-ture strongly, while the depth effect is small. Theinitial thickness of the lithosphere near mid-oceanridges, as averaged by long-period surface waves,varies between 10 and 45 km, and 100 Ma litho-sphere has thickness from 83 to 110 km.

To further probe oceanic upper mantle structure,we subtracted parametric phase velocity predictions,calculated using the obtained models for each ocean,from the observed phase velocity maps. The age-de-pendence and mid-ocean ridges disappeared in theresidual maps. Many hotspots are located in slowvelocity areas, and many of them are associated withor near to slow velocity peaks in the residual maps,suggesting that hotspot in oceanic areas modifyoceanic lithosphere evolution and produce deviationsfrom average patterns with increasing age.

Acknowledgements

This research was supported by National ScienceFoundation grant No. EAR-9219607 and by IGPP-LANL UCRP grants No. 354 and 354R. Y.-S. Zhangwas supported by the distinguished scholarship ofthe Chinese National Science and Technology Foun-dation. Two anonymous reviewers provided con-structive comments and suggestions. ContributionNo. 389, Institute of Tectonics and W.M. KeckSeismological Laboratory, University of California,Santa Cruz.

( )Y.-S. Zhang, T. LayrPhysics of the Earth and Planetary Interiors 114 1999 71–8080

References

Anderson, D.L., Regan, J., 1983. Uppermantle anisotropy and theoceanic lithosphere. Geophys. Res. Lett. 10, 841–844.

Anderson, R.N., Skilbeck, J.N., 1981. Ocean heat flow. In: TheSea, Vol. 7. Wiley-Interscience, New York, pp. 489–523.

Campbell, I.H., Griffiths, R.W., 1990. Implications of mantleplume structure for the evolution of flood basalts. Earth Planet.Sci. Lett. 99, 79–93.

Canas, J.A., Mitchell, B.J., 1981. Rayleigh wave attenuation andits variation across the Atlantic Ocean. Geophys. J. R. Astron.Soc. 67, 159–176.

Crough, T.S., 1975. Thermal model of oceanic lithosphere. Nature256, 388–390.

Duncan, R.A., Richards, M.A., 1991. Hotspots, mantle plumes,flood basalts, and true polar wander. Rev. Geophys. 29,31–50.

Dziewonski, A.M., Anderson, D.L., 1981. Preliminary referenceEarth model. Phys. Earth Planet. Int. 25, 297–356.

Ekstrom, G., Tromp, J., Larson, E.W.F., 1997. Measurements and¨global models of surface wave propagation. J. Geophys. Res.102, 8137–8157.

Forsyth, D.W., 1977. The evolution of the upper mantle beneathmid-ocean ridges. Tectonophysics 38, 89–118.

Grand, S.P., 1994. Mantle shear structure beneath the Americasand surrounding oceans. J. Geophys. Res. 99, 11591–11621.

Griffiths, R.W., Campbell, I.H., 1990. Stirring and structure inmantle starting plumes. Earth Planet. Sci. Lett. 99, 66–78.

Larson, R.L., 1991. Latest pulse of Earth: evidence for a mid-Cretaceous superplume. Geology 19, 547–550.

Larson, R.L., Olsen, P., 1991. Mantle plumes control magneticreversal frequency. Earth Planet. Sci. Lett. 109, 437–447.

Laske, G., Master, G., 1996. Constraints on global phase velocitymaps from long-period polarization data. J. Geophys. Res.101, 16059–16075.

McKenzie, D.P., 1967. Some remarks on heat flow and gravityanomalies. J. Geophys. Res. 72, 6261–6273.

McNutt, M.K., Judge, A.V., 1990. The superswell and mantledynamics beneath the south Pacific. Science 248, 969–975.

Morgan, W.J., 1972. Plate motions and deep mantle convection.Geol. Soc. Am. Mem. 132, 7–22.

Morgan, W.J., 1984. Hotspot tracks and the opening of theAtlantic and Indian Oceans. The Sea, Vol. 7. Wiley, NewYork, pp. 443–487.

Mueller, R.D., Roest, W.R., Royer, J.-Y., Gahagan, L.M., Sclater,J.G., 1995. A digital age map of the ocean floor. SIO Refer-ence Series, 93–30.

Nishimura, C., Forsyth, D.W., 1989. The anisotropic structure ofthe upper mantle in the Pacific. Geophys. J. 96, 203–229.

Parker, R.L., Oldenburg, D.W., 1973. Thermal model of oceanridge. Nature 242, 137–139.

Parsons, B., McKenzie, D.P., 1978. Mantle convection and thethermal structure of the plates. J. Geophys. Res. 83, 4485–4496.

Parsons, B., Sclater, J.G., 1977. An analysis of the variation ofocean floor bathymetry and heat flow with age. Geophys. J.82, 803–827.

Schubert, G., Froidevaux, C., Yuen, Y.A., 1976. Oceanic litho-sphere and asthenosphere: thermal and mechanical structure. J.Geophys. Res. 81, 3525–3540.

Sclater, J.G., Jaupart, C., Galson, D., 1980. The heat flow throughoceanic and continental crust and the heat loss of the earth.Rev. Geophys. 18, 269–311.

Sleep, N.H., 1969. Sensitivity of heat flow and gravity to themechanism os sea-floor spreading. J. Geophys. Res. 74, 542–549.

Sleep, N.H., 1990. Hotspots and mantle plumes: some phe-nomenology. J. Geophys. Res. 95, 6715–6736.

Stein, C.A., Stein, S., 1992. A model for the global variation inoceanic depth and heat flow with lithospheric age. Nature 359,123–129.

Turcotte, D.L., Oxburgh, E.R., 1967. Finite amplitude convectioncells and continental drift. J. Fluid Mech. 28, 29–42.

Vink, G.E., Morgan, W.J., Vogt, P.R., 1985. The Earth’s hotŽ .spots. Scientific American April 50–57.

Vogt, P.R., 1981. On the applicability of thermal conductionmodels to mid-plate volcanism: comments on a paper by Gasset al. J. Geophys. Res. 86, 950–960.

White, R.S., McKenzie, D.J., 1989. Magmatism at rift zones: thegeneration of volcanic continental margins and flood basalts.J. Geophys. Res. 94, 7685–7729.

Wiens, D.A., Stein, S., 1983. Age dependence of oceanic in-traplate seismicity and implications for lithospheric evolution.J. Geophys. Res. 88, 6455–6468.

Wilson, J.T., 1973. Mantle plumes and plate motions. Tectono-physics 19, 149–164.

Wolfe, C.J., Biarnason, I.T., VanDecar, J.C., Solomon, S.C.,1997. Seismic structure of the Iceland mantle plume. Nature385, 245–247.

Yoshii, T., 1975. Regionality of group velocities of Rayleighwaves in the Pacific and thickening of the Plate. Earth Planet.Sci. Lett. 25, 305–312.

Yu, G.-K., Mitchell, B., 1979. Regionalized shear velocity modelsof the Pacific upper mantle from observed Love and Rayleighwave dispersion. Geophys. J. R. Astron. Soc. 57, 311–341.

Zhang, Y.-S., Lay, T., 1996. Global surface wave phase velocityvariations. J. Geophys. Res. 101, 8415–8436.

Zhang, Y.-S., Tanimoto, T., 1991. Global Love wave phasevelocity variation and its significance to plate tectonics. Phys.Earth Planet. Int. 66, 160–202.

Zhang, Y.-S., Tanimoto, T., 1992. Ridges, hotspots and theirinteraction as observed in seismic velocity maps. Nature 355,45–49.

Zhang, Y.-S., Tanimoto, T., 1993. High resolution global uppermantle structure: plate tectonics. J. Geophys. Res. 98, 9793–9823.