Seismic evidence for a tilted mantle plume and north^south mantle...

Seismic evidence for a tilted mantle plume and north^southmantle £ow beneath Iceland

Yang Shen a;�, Sean C. Solomon b, Ingi Th. Bjarnason c, Guust Nolet d,W. Jason Morgan d, Richard M. Allen d;1, Kristin Vogfjo«rd e,

Steinun Jakobsdo¤ttir f , Ragnar Stefa¤nsson f , B.R. Julian g, G.R. Foulger h

a Graduate School of Oceanography, University of Rhode Island, South Ferry Road, Narragansett, RI 02882, USAb Department of Terrestrial Magnetism, Carnegie Institution of Washington, 5241 Broad Branch Road N.W., Washington, DC 20015,

USAc Science Institute, University of Iceland, Reykjavik, Iceland

d Department of Geosciences, Guyot Hall, Princeton University, Princeton, NJ 08544, USAe National Energy Authority, Grensasvegi 9, Reykjavik, Iceland

f Meteorological O⁄ce of Iceland, Bustadavegi 9, Reykjavik, Icelandg US Geological Survey, 345 Middle¢eld Road, Menlo Park, CA 94025, USA

h Department of Geological Sciences, University of Durham, Durham DH1 3LE, UK

Received 7 August 2001; received in revised form 23 January 2002; accepted 25 January 2002

Abstract

Shear waves converted from compressional waves at mantle discontinuities near 410- and 660-km depth recordedby two broadband seismic experiments in Iceland reveal that the center of an area of anomalously thin mantletransition zone lies at least 100 km south of the upper-mantle low-velocity anomaly imaged tomographically beneaththe hotspot. This offset is evidence for a tilted plume conduit in the upper mantle, the result of either northward flowof the Icelandic asthenosphere or southward flow of the upper part of the lower mantle in a no-net-rotation referenceframe. ß 2002 Elsevier Science B.V. All rights reserved.

Keywords: mantle plumes; transition zones; discontinuities; convection; Iceland

1. Introduction

Linear chains of volcanic centers displaying reg-ular age progressions are thought to result fromzones of upper-mantle melt production that arenearly stationary with respect to the overlyingtectonic plates [1]. Morgan [2] proposed thatsuch stationary regions of melt production, orhotspots, are maintained by long-lived upwellingof warm material from the lower mantle throughnarrow conduits that he termed plumes. The hy-

0012-821X / 02 / $ ^ see front matter ß 2002 Elsevier Science B.V. All rights reserved.PII: S 0 0 1 2 - 8 2 1 X ( 0 2 ) 0 0 4 9 4 - 6

* Corresponding author. Tel. : +1-401-874-6848;Fax: +1-401-874-6811.

E-mail address: [email protected] (Y. Shen).

1 Present address: Department of Geology and Geophysics,University of Wisconconsin, 1215 W. Dayton St., Madison,WI 53706, USA.

EPSL 6140 17-4-02 Cyaan Magenta Geel Zwart

Earth and Planetary Science Letters 197 (2002) 261^272

www.elsevier.com/locate/epsl

pothesis of ¢xed hotspots has provided a usefulmeans of determining lithospheric plate motionsfrom the geographic orientations and age distri-butions of volcanic chains, but the validity of thishypothesis on a global scale remains controversial[3,4]. Observational evidence for or against ¢xedhotspots has been derived, to date, mostly fromhotspot manifestations at the Earth’s surface.

The concept of ¢xed hotspots implies verticalplume conduits, because de£ection of plume con-duits by the convecting mantle ^ as suggested bymantle £uid dynamic models [5] ^ should lead torelative motions among hotspots over time scalesof plate (and thus mantle-£ow) reconstructions.Low-seismic-velocity anomalies in the lower man-tle beneath hotspots (e.g. Iceland and Hawaii) insome global tomographic models [6,7], anomaliesthat usually di¡er in shape from vertical columns,have been suggested as evidence for tilted plumesin the lower mantle [7]. However, the ability ofglobal models to resolve narrow, low-velocitystructures in the lower mantle remains a topic ofdebate [8], and alternative global tomographicmodels fail to show a low-velocity anomaly be-neath Iceland in the lower mantle [9,10].

An alternative approach to address the tiltingof plume conduits in the upper mantle is to com-pare measures of plume in£uence in the shallowupper mantle with the depths to seismic disconti-nuities near 410- and 660-km depth, global fea-tures that have been identi¢ed with the temper-ature-dependent transitions of (Mg,Fe)2SiO4 fromK-olivine to L-spinel (wadsleyite) and from Q-

spinel (ringwoodite) to (Mg,Fe)SiO3-perovskiteplus (Mg,Fe)O-magnesiowu«stite, respectively[11,12]. In an earlier study of mantle discontinu-ities beneath Iceland [13], we used receiver func-tions derived from body-wave records of teleseis-mic earthquakes from the broadband ICEMELTseismic network [14] and the permanent GlobalSeismographic Network station BORG (Fig. 1)to demonstrate that the transition zone is thinnerthan the average Earth [15] beneath central andsouthern Iceland but is of normal thickness be-neath surrounding areas. This result is consistentwith a hot and narrow plume originating from thelower mantle [13].

Additional broadband seismic data were col-lected during the Iceland Hotspot Project (or theHOTSPOT experiment) [16] from the spring of1996 to the fall of 1998 (Fig. 1). The combinedICEMELT and HOTSPOT data set is approxi-mately four times as large as the ICEMELTdata used in the previous study [13]. By incorpo-rating a substantially greater number of receiverfunctions than in the previous study, we are nowable to identify features of the transition-zoneanomaly that were not resolved earlier.

2. Data processing and signal quality

The calculation and stacking of receiver func-tions follow procedures described previously[13,17]. To ensure that random noise is far belowthe signal (the conversion of a P-wave to an S-

Fig. 1. Seismic stations used in the ICEMELT experiment (A) and the Iceland Hotspot Project (B). The square denotes theGlobal Seismic Network station BORG. Dashed lines delineate the northern (NVZ), eastern (EVZ), and western (WVZ) volcaniczones.


Y. Shen et al. / Earth and Planetary Science Letters 197 (2002) 261^272262

wave at depth d, or Pds), we select seismogramswith noise levels (de¢ned as the standard devia-tion of the values on the radial component in an80-s window before the P arrival) less than 0.1times the amplitude of the direct P-wave on thevertical component. Because many seismogramshave signal-to-noise (S/N) ratios greater than 10and because we restrict considerations to stackswith more than 60 receiver functions (Fig. 2B),the noise levels in linearly stacked receiver func-tions are 0.03^1.0% of the amplitude of the P-wave (Fig. 3), much less than the amplitude oflinearly stacked P410s and P660s phases (2^6%of the amplitude of the P-wave). Improvementin S/N ratios is achieved with an nth-root (n = 2)stacking process [18], a non-linear method thatsuppresses random noise and enhances coherentsignals. The relative amplitudes of nth-root-stacked P410s and P660s phases are 10^80 timesgreater than the levels of similarly processed andstacked records of noise immediately before thecorresponding P arrivals (Fig. 2E,H,K).

Rather than using the peaks of P410s andP660s as in the previous study [13], the arrivaltimes of the converted phases are weighted line-arly by the value of the amplitude of the wave-form of the converted phases. For symmetric con-verted-phase waveforms, the expected picks of thenew method are the peaks of the waveforms, butthese picks are less susceptible to random noisethan in the previous study [13]. Bootstrap analysis[19] shows that the greater number of high-qualityreceiver functions than in the previous study andthe new arrival-time picking method yield 95%con¢dence limits on the arrival times that aremuch smaller than the lateral variation in the ar-rival times of the converted phases (Fig. 2G,J).

3. Transition-zone thickness

Images of P410s and P660s arrivals derived bynth-root stacking of receiver functions, andP660s^P410s di¡erential arrival times, are shownalong selected pro¢les in Fig. 2. To the north ofIceland, variations in P410s and P660s times arepositively correlated and comparable in magni-tude (Fig. 2C), re£ecting the dominant in£uence

of velocity heterogeneities shallower than the 410-km discontinuity because of the nearly identicalpaths of P660s and P410s over that depth interval.P660s^P410s di¡erential times, which provide in-formation on the thickness of the transition zone,are not sensitive to heterogeneities shallower thanthe 410-km discontinuity and are similar to thatpredicted by the iasp91 global seismic velocitymodel [15] (Fig. 2D). The positive correlation be-tween P410s and P660s times breaks down be-neath central and southern Iceland (Fig. 2F,I) ;however, where observed P660s^P410s di¡erentialtimes are less than predicted by iasp91 (23.9 s) byas much as 1.6 s (Figs. 2G,J and 3).

Excluding central and southern Iceland, theaverage di¡erential time beneath the remainingareas is 24.2 þ 0.1 s, comparable to, and withinthe uncertainty of, values for the mantle beneaththe southern East Paci¢c Rise (24.7 þ 0.6 s) [20] aswell as the global average (24.0 þ 0.6 s) [21]. Ob-servations of long-period SS precursors indicatethat the mantle transition zone in most oceanicareas is unlikely to be signi¢cantly thicker thanin the iasp91 model [12].

The mantle transition-zone thickness beneathcentral and southern Iceland implied by theP660s^P410s di¡erential times is less than beneathsurrounding areas by V19 km. Given Clapeyronslopes of 2.9 and ^2.1 MPa/K for the 410- and660-km discontinuities [22], respectively, the re-duction in the transition-zone thickness is equiv-alent to an excess temperature of 140 K. The ¢-nite sizes of both the Fresnel zone of convertedphases and the patches used for stacking [13] tendto smooth lateral variations in discontinuitydepths. For example, for a dome-shaped elevationof the 660-km discontinuity having a radius of150 km and a height of 15 km, we can retrievefrom similarly processed and stacked syntheticwaveforms about 65% of the maximum amplitudeof the topography over a 400-km-diameter region.Furthermore, lower velocities associated with ex-cess temperatures within the transition zone in-ferred from the transition-zone thickness anomalywould increase the P660s time (V0.2 s per 100 Kexcess temperature) and decrease the apparent re-duction in the transition-zone thickness. Our esti-mate of the excess temperature should therefore


Y. Shen et al. / Earth and Planetary Science Letters 197 (2002) 261^272 263

be regarded as only an apparent value and mostlikely a lower bound. With these factors takeninto consideration, an apparent excess tempera-ture of at least 140 K is in agreement with esti-

mates of the thermal anomaly (V200 K) at thedepth of melt generation (6 200 km) [23], valuesthat re£ect averages of the melt-generation andmelt-migration processes.



We note that the southern boundary of thetransition-zone anomaly beneath Iceland (Fig. 4)remains to be mapped, and we therefore cannotrule out the possibility that the transition-zoneanomaly beneath Iceland is part of a large region-al anomaly that extends south of Iceland. How-ever, the close proximity of the transition-zoneanomaly to the low-velocity body in the shallow

mantle [24^26] and the fact that the thinner tran-sition zone does not follow the geometry of theplate boundary support the interpretation that thetransition-zone anomaly beneath Iceland is asso-ciated with the Iceland mantle plume.

4. The 410- and 660-km discontinuities

Within the area of anomalously thin mantletransition zone, the reduction in P660s^P410s dif-ferential time correlates with a decrease in P660sarrival time (Figs. 2I and 5) and does not show acomparable correlation with P410s arrival time(Fig. 5). The individual and di¡erential Pds timesreported here have not been corrected for velocityheterogeneity or anisotropy in the upper mantle,but such corrections do not change the generalrelationships between di¡erential times and indi-vidual arrival times of the converted phases [13].Any correction for delays expected from greater-than-normal temperatures in the transition zoneinferred from the transition-zone thickness anom-aly would further reduce P660s times andstrengthen the correlation between di¡erentialand P660s times.

Several conditions pertinent to the Icelandplume may reduce the depth to the 410-km dis-continuity and thus at least partially balance thee¡ect of excess temperatures and a positive Cla-peyron slope for the 410-km discontinuity. Alongthe Reykjanes Ridge, basalt samples show in-creasing H2O concentrations toward Iceland, in-dicating that Icelandic basalts prior to degassing

Fig. 3. Waveforms of linearly stacked (A) and nth-root-stacked (B) receiver functions along the north^south pro¢lethrough central Iceland and their 95% con¢dence limits (thinlines). The traces are aligned on the P410s arrival time deter-mined by bootstrap [19] and weighted linearly by the valueof the amplitude of the waveform of the converted phase.They are normalized by a constant for each panel and plot-ted as a function of latitude in degrees. The vertical linesmark the values predicted for the iasp91 model [15].

Fig. 2. Images of P410s and P660s in stacked receiver functions and P660s^P410s di¡erential times indicate an anomalously thintransition zone beneath central and southern Iceland. (A) Locations of the pro¢les of receiver function stacks. (B) The numberof receiver functions in each stack at a depth of 660 km. (C) Relative amplitudes of receiver function nth-root stacks along aneast^west pro¢le north of Iceland. Red and yellow colors represent positive- and relatively higher-amplitude arrivals. Note thatthe nth-root stacking ampli¢es weak signals in noisy data but does not preserve amplitudes and waveforms. The vertical axis isthe time after the compressional-wave (P) arrival. (D) P660s^P410s di¡erential times along this northern pro¢le compared withthat predicted (horizontal line) for the iasp91 global model [15]. The di¡erential times and their 2c errors are estimated using abootstrap method. A 1-s change in di¡erential time is equivalent to an about 10-km change in transition-zone thickness. (E) TheS/N ratio of identi¢ed P410s (black line) and P660s (red line) phases along this northern pro¢le, obtained by dividing the ampli-tude of the converted phases by the levels of similarly processed and stacked records of noise immediately prior to the corre-sponding P arrivals. (F^H) Relative amplitude, di¡erential times, and S/N ratios ratios of converted phases along an east^westpro¢le through central Iceland. (I^K) Relative amplitude, di¡erential times, and S/N ratios of converted phases along a north^south pro¢le through central Iceland.6



have at least 0.35 wt% H2O [27], an amount great-er than in normal mid-ocean ridge basalt (MORB,0.1^0.2 wt%) [28]. Since water is preferentiallypartitioned into the melt, high H2O concentra-tions in Icelandic basalts and a higher-than-nor-mal extent of melting beneath Iceland requirehigh concentrations of H2O in the plume source.H2O in the (Mg, Fe)2SiO4 system stabilizes L-spinel, resulting in a broader and shallower K^Ltransition [29]. The onset of the K^L transforma-tion for a mantle source with 500 ppm H2O iscalculated to occur about 8 km shallower thanfor a normal MORB source with 200 ppm H2O[29]. In the postspinel transition, H2O shifts thephase boundary (the 660-km discontinuity) tohigher pressures than in the anhydrous case [30].

Other possible mechanisms for reducing thedepth of the 410-km discontinuity beneath Ice-land include a lower-than-normal Al content inthe plume source [11] and a non-equilibrium con-dition at the depth of phase transformation [31].

It has been suggested that the Iceland plumesource consists of an Al-poor, refractory matrixand veins and blobs containing enriched compo-nents [32]. The geochemical signature of depletionin Icelandic basalts, however, can also be matchedby a multicomponent mixing model involving twoincompatible, trace-element-enriched componentsand the usual normal MORB source [33]. A man-tle source with a lower-than-normal Al contentand a higher-than-normal percentage of olivinewould be predicted to display greater-than-normalvelocity jumps at the 410- and 660-km disconti-nuities and high amplitudes for phases convertedat the discontinuities. The amplitudes of the line-arly stacked P410s and P660s phases we observe,in fact, are correlated with the reduction inP660s^P410s di¡erential time. In particular, thelargest P410s amplitude beneath southern Iceland(near 63.5‡N, 17.5‡W, at 4^5% of the amplitudeof the P-wave on the vertical component) is nearlytwice as large as beneath surrounding areas. In

Fig. 4. Map view of di¡erences between the observed P660s^P410s di¡erential times and the value predicted for the iasp91 model[15]. Red and yellow colors indicate signi¢cantly smaller di¡erential times (thinner mantle transition zone) than in iasp91, whileblue colors denote normal or somewhat greater di¡erential times (normal or slightly thicker mantle transition zone). The crossmarks the center of the mapped area of thinned transition zone. The circle, square, and triangle represent the locus of thickestcrust and the centers of low S and P velocities in the uppermost mantle [24], respectively. The image has been smoothed by atwo-dimensional, ¢ve-point moving average. Dashed gray lines delineate Icelandic volcanic zones and the axis of the Mid-AtlanticRidge.



contrast to the situation at 410-km depth, varia-tions in the Al content of the mantle should havelittle e¡ect on the depth of the transformationbetween Q-spinel and perovskite plus magnesio-wu«stite [11].

If the upper-mantle phase transitions within theupwelling plume conduit are limited by nucleationand kinetics [31], then rather than thermodynamicequilibrium the 660-km discontinuity as well asthe 410-km discontinuity would be shoaled withinthe plume. Because the transformation from per-ovskite plus magnesiowu«stite to Q-spinel occursover a narrow pressure interval [11], any non-equilibrium e¡ect on the depth of the 660-km dis-continuity is unlikely to be substantially greaterthan that on the depth of the 410-km discontinu-ity and, in any case, is probably much less thanV20 km.

Of the possible contributors to reducing thedepth of the 410-km discontinuity within the Ice-land plume ^ high H2O content, low Al content,

or a non-equilibrium phase transition ^ none arelikely to reduce the thickness of the mantle tran-sition zone. The low values of P660s^P410s di¡er-ential time beneath central and southern Icelandare therefore predominantly the consequence ofthe excess temperature of the plume and the pos-itive and negative Clapeyron slopes of the phasetransformations near 410- and 660-km depth, re-spectively. This inference is consistent with theresults of recent mineral-physics experiments[34,35] that the 660-km discontinuity correspondsto the transformation between Q-spinel and perov-skite plus magnesiowu«stite, rather than the gar-net^perovskite transformation [11,36], which hasa positive Clapeyron slope.

5. Tilting of the Iceland plume and implications formantle £ow beneath Iceland

Most of the area of thinner transition zone (de-¢ned as the region where P660s^P410s di¡erentialtimes are less than predicted by iasp91 by at least0.5 s, or the transition-zone thickness is at least5 km thinner than in iasp91) resolved in our studyis south of central Iceland (Fig. 4). The center ofthe mapped area of thinned transition zone, de-¢ned as the average location of the thinner-than-normal transition zone weighted linearly by themagnitude of reduction in transition-zone thick-ness, is beneath southern Iceland near 63.8 þ0.1‡N, 17.9 þ 0.1‡W (Fig. 4). If the smallestP660s^P410s di¡erential time indicates the highestexcess temperature at the center of the plume,then the center of the transition-zone anomalymay be south of 63.8‡N by 1‡ of latitude ormore (Figs. 2J and 4).

The center of the area of thinned transitionzone lies south of the inferred location of theplume at the depth of primary melt generation(6 200 km) by at least 100 km. The location ofthe plume core in the shallow mantle is con-strained in several ways. First, crustal thicknesscan be regarded as a proxy for integrated meltproduction in the upper 200 km or so of mantle.Increased temperature and mass £ux within therising plume conduit are expected to result ingreater than average melt production and thus

Fig. 5. Correlations between the P660s^P410s di¡erentialtimes and individual arrival times of the P410s (A) andP660s (B) phases in the area of anomalously thin transitionzone. The straight lines represent two end-member models inwhich the reduction in P660s^P410s di¡erential times iscaused by (I) a decrease in P660s arrival time (relative to theP arrival time), or (II) an increase in P410s arrival time.



an anomalously thick crust above the plume core.Inversions of waveforms of surface waves fromlocal earthquakes [37], travel times of refractedphases [38], and receiver functions for local crus-tal structure [39,40] show that the center of thearea of the thickest crust (in excess of 40 km) isnear 64.7‡N, 17.3‡W, coinciding with a regionalBouguer gravity minimum [41]. Second, tomo-graphic inversions of body-wave travel-time de-lays show a low-velocity anomaly in the shallowupper mantle (6 250 km) beneath central Icelandnear 64.7‡N, 17.5‡W [24^26]. Third, olivine tho-leiite and picrite samples from the Icelandic neo-volcanic zones reveal a ‘plateau’ of high 3He/4He(V20 R/Ra, sample values normalized by the at-mospheric value) that is approximately 100 km indiameter and coincides with the Bouguer gravityminimum, the maximum crustal thickness, andthe upper-mantle low-velocity anomaly [42]. Bred-dam et al. [42] suggest that this zone of elevated3He/4He ratios outlines the plume conduit at thedepth of melting of the Iceland mantle plume.

The o¡set between the center of the transition-zone thickness anomaly and the center of shallowmeasures of plume in£uence suggests that theplume is tilted in the upper mantle, with an angleof tilt of at least 9‡ from the vertical (Fig. 6). Forcomparison, the sharpness of the bend in the Ha-waiian^Emperor volcanic chain implies that thede£ection of the underlying plume due to achange in plate motion is less than 200 km [43].

Current tomographic models of the Icelandicupper mantle provide no strong evidence eitherfor or against a tilted Iceland mantle plume. In-versions of the ICEMELT data by Wolfe et al.[24] show that the center of the low P-wave veloc-ity at 300-km depth is noticeably south of thecenter of the velocity anomaly at 125-km depth,but there is no lateral o¡set of the center of theanomaly for S-wave velocity. Foulger et al. [25]found from inversions of HOTSPOT data that thelow-velocity anomaly appears to be elongatednorth^south at depths of 250^400 km and inter-preted this as evidence that the low-velocityanomaly extends no deeper than the mantle tran-sition zone. While Allen et al. [26], who inverteddata from the same experiment, found a similarelongation of the low-velocity anomaly for high-

frequency P-waves, they report a more circularlow-velocity anomaly for low-frequency P- andS-waves. The di¡erences between P- and S-veloc-ity models and the presence or absence of a lateralo¡set of the center of the low-velocity anomaly inthe uppermost 400 km may result from limits toresolution at depths near 400 km beneath Icelandfrom the limited aperture of the seismic networks(Fig. 1) and the assumption of mantle isotropy.Furthermore, a lower viscosity in the astheno-sphere than in the deeper mantle [44] could resultin a conduit more vertical in the asthenospherethan at greater depths. In the following discus-sion, a straight plume conduit is considered as a¢rst-order approximation.

Tilting of the plume may account for some ofthe apparent north^south elongation of both thetransition-zone thickness anomaly (Fig. 4) and thelow-velocity anomaly beneath southern Iceland

Fig. 6. Schematic north^south cross-section through the cen-ter of the Iceland mantle plume. The southward shift of thecenter of the transition-zone thickness anomaly relative tocentral Iceland, where the thickest crust, minimum Bouguergravity, and highest 3He/4He are found, indicates a tiltedplume conduit. Incipient hydrous melts migrate vertically tothe surface and erupt as high-3He/4He alkaline basalts in thesouthern portions of the eastern volcanic zone. The rising ve-locity of the plume conduit (Vc) can be decomposed into acomponent parallel to the plume conduit (V//) and a horizon-tal component (Vh) at least partly balancing a generallynorthward £ow of the upper mantle beneath the North At-lantic region near Iceland (U).



[25,26]. The tilt could also have important impli-cations for the distribution of geochemicalanomalies in Icelandic basalts. Alkaline basaltsfrom southern Iceland high in 3He/4He (13.9^26R/Ra) have been interpreted as products of low-degree, volatile-rich hydrous melting from theplume mantle de£ected horizontally below ahigh-viscosity lid having a base at the dry solidus[42]. Such a mechanism, however, should lead to asymmetric distribution of high-3He/4He alkalinebasalts about the center of the plume, but high-3He/4He alkaline basalts are not observed in thenorthern volcanic zone [42]. The tilted plumemodel provides a simple and self-consistent alter-native mechanism: Incipient hydrous melts gener-ated within the plume conduit migrate verticallyto the surface and erupt as high-3He/4He alkalinebasalts in southern Iceland (Fig. 6).

A tilt in the Iceland plume indicates relativeshear between horizontal £ows at asthenosphericlevels and those at greater depth. The relativeshear can result from southward £ow of the upperpart of the lower mantle, northward £ow of theIcelandic upper mantle, or a combination of thetwo in a reference frame having no net rotation ofthe mantle. Recent mantle circulation models, inwhich driving forces are density heterogeneitiesconstrained by global seismic tomographic mod-els, predict southward £ow of the lower mantlebeneath Iceland [45]. If the Iceland plume origi-nates at or near the core^mantle boundary, as hasbeen inferred from the detection of an ultralow-velocity zone above the core^mantl boundary be-neath Iceland [46], then deformation of the plumeconduit by such lower-mantle £ow would yield aplume source located near 60^61‡N, 19^22‡W[45], and a tilted plume conduit with a directionand magnitude of tilt in the upper mantle consis-tent with our observations.

Splitting of shear waves recorded at ICEMELTseismic stations indicates that the direction of fastpolarization of steeply incident shear waves in theIcelandic upper mantle is approximately north^south [47]. This pattern of shear-wave splittingcannot be explained by simple models of horizon-tally diverging £ow driven either by plate spread-ing or by radial horizontal £ow from the center ofthe hotspot. Given that shear-wave splitting arises

primarily in the upper 300 km of the mantle [48],the observed pattern of shear-wave splitting hasbeen interpreted as the result of a combination ofplate-induced £ow and a generally northward£ow of the upper mantle beneath the North At-lantic region near Iceland [47]. This interpretationis consistent with the direction of tilt of theplume. Entrainment of plume material by north-ward asthenospheric £ow beneath Iceland alsoprovides a possible reconciliation of northwardextension of the upper-mantle low-velocity anom-aly into northern Iceland [25,26] and the normaltransition zone beneath this region. Generallynorthward upper-mantle £ow is not only pre-dicted from some simple kinematic models inwhich £ow is determined by the mass £ux im-posed by plate motion [49] but also from dynamicmodels that include £ows due to internal mantledensity heterogeneities inferred from seismic to-mography (B. Steinberger, personal communica-tion, 2001). It contradicts, however, an interpre-tation [50] of the apparently more pronouncedgeochemical and geophysical anomalies alongthe Mid-Atlantic Ridge south of Iceland thanthose to the north as the result of shallow south-ward mantle £ow.

The two kinematic explanations for the tilt ofthe plume (northward asthenospheric £ow andsouthward £ow of the upper part of the lowermantle) are not mutually exclusive, but they pre-dict opposite e¡ects on the surface motion of theIceland hotspot. Southward £ow of the lowermantle predicts a southward component of mo-tion of the surface hotspot, as the tilted conduitrises in the mantle [5,45]. Northward £ow of theupper mantle beneath Iceland, in contrast, pre-dicts a northward component of motion of thesurface hotspot, but such motion is at least inpart balanced by southward motion of the hot-spot from the rise of the tilted plume conduit.For a relatively slow-moving lower mantle, thetilt of the plume conduit and hotspot locationadjust quickly to quasi steady-state after changesin upper-mantle £ow [45]. The north^south direc-tions of fast polarization of shear waves [47] sug-gest that the rates of north^south asthenospheric£ow are greater than the plate velocity (18 km/Mafull spreading rate) so that the net £ow direction



and thus the strain-induced, lattice-preferred ori-entation in the asthenosphere beneath Iceland areapproximately north^south. For a plume having aradius of 100 km [24^26], an excess temperatureof 200 K [23], a tilt of 10‡, and an average vis-cosity of the upper mantle of 1020^1021 Pa s [44],the predicted southward horizontal motion of thehotspot due to the rise of the tilted conduit [5,45]is 5^50 km/Ma. Signi¢cant southward astheno-spheric £ow beneath Iceland can be ruled out,because such a £ow together with the e¡ect ofthe rise of the tilted conduit would result in asouthward motion of the surface hotspot at arate greater than the spreading rate. Such a mo-tion, if coherent over V20 Ma, would be resolv-able but is not observed. To maintain a slow-moving or stationary Iceland hotspot, £ow ofthe Icelandic upper mantle must be northwardand comparable to the southward horizontal mo-tion of the hotspot due to the rise of the tiltedplume (Fig. 6).

Geodynamic interpretations of the tilt of theplume conduit are clearly model dependent andneed to be sharpened by a fuller understandingof the history of hotspots and plate motions.Nevertheless, this study provides strong seismo-logical evidence for a tilted plume conduit in theupper mantle, a consequence of large-scale, hori-zontal mantle £ow beneath the surface hotspot.

Acknowledgements

We thank Bergur H. Bergsson, Bjorn Bjarna-son, Birgir Bjarnason, Bryndis Brandsdo¤ttir, Hau-kur Brynjo¤ lfo¤sson, Kristinn Egilsson, Pa¤lmi Er-lendsson, Gunnar Gudmundson, Eytho¤rHannesson, Tryggvi Hardarson, La¤rus Helgason,Bogi Ingimundarson, Haraldur Jo¤nsson, EinarKjartansson, A. Kuehnel, R. Kuehnel, SturlaRagnarsson, Pa¤lmi Sigurdsson, Ragnar Thrud-marsson, and the sta¡ of the National ElectricCompany of Iceland (Landsvirkjun) for assistancewith ICEMELT and HOTSPOT ¢eld operations.We also thank Scott King, Jeroen Ritsema, Bern-hard Steinberger, Peter van Keken, and an anon-ymous reviewer for constructive comments. Thisresearch was supported by the National Science

Foundation under Grants EAR-9316137, OCE-9402991, EAR-9417918, and OCE-9906902.[SK]

References

[1] J.T. Wilson, A possible origin of the Hawaiian islands,Can. J. Phys. 41 (1963) 863^870.

[2] W.J. Morgan, Convection plumes in the lower mantle,Nature 230 (1971) 42^43.

[3] P. Molnar, J. Stock, Relative motions of hotspots in thePaci¢c, Atlantic and Indian Ocean since late Cretaceoustime, Nature 327 (1987) 587^591.

[4] J.A. Tarduno, R.D. Cottrell, Paleomagnetic evidence formotion of the Hawaiian hotspot during formation of theEmperor seamounts, Earth Planet. Sci. Lett. 153 (1997)171^180.

[5] M.A. Richards, R.W. Gri⁄ths, De£ection of plumes bymantle shear £ow: experimental results and a simpletheory, Geophys. J. 94 (1988) 367^376.

[6] H. Bijwaard, W. Spakman, Tomographic evidence for anarrow whole mantle plume below Iceland, Earth Planet.Sci. Lett. 166 (1999) 121^126.

[7] D. Zhao, Seismic structure and origin of hotspots andmantle plumes, Earth Planet. Sci. Lett. 192 (2001) 251^265.

[8] G.R. Foulger, D.G. Pearson, Is Iceland underlain by aplume in the lower mantle? Seismology and helium iso-topes, Geophys. J. Int. 145 (2001) F1^F5.

[9] J. Ritsema, H.J. van Heijst, J.H. Woodhouse, Complexshear wave velocity structure imaged beneath Africa andIceland, Science 286 (1999) 1925^1928.

[10] C. Me¤gnin, B. Romanowicz, The three-dimensional shearvelocity structure of the mantle from the inversion ofbody, surface and higher-mode waveforms, Geophys. J.Int. 143 (2000) 709^728.

[11] D.J. Weidner, Y. Wang, Phase transformations: implica-tions for mantle structure, in: S. Karato, A. Forte, R.Liebermann, G. Masters, L. Stixrude (Eds.), Earth’sDeep Interior: Mineral Physics and Tomography Fromthe Atomic to the Global Scale, Am. Geophys. Union,Washington, DC, 2000, pp. 215^235.

[12] P.M. Shearer, Transition zone velocity gradients and the520-km discontinuity, J. Geophys. Res. 101 (1996) 3053^3066.

[13] Y. Shen, S.C. Solomon, I.Th. Bjarnason, C.J. Wolfe, Seis-mic evidence for a lower-mantle origin of the Icelandplume, Nature 395 (1998) 62^65.

[14] I.Th. Bjarnason, C.J. Wolfe, S.C. Solomon, G. Gud-mundson, Initial results from the ICEMELT experiment:body-wave delay times and shear-wave splitting acrossIceland, Geophys. Res. Lett. 23 (1996) 459^462.

[15] B.L.N. Kennett, E.R. Engdahl, Traveltimes for globalearthquake location and phase identi¢cation, Geophys.J. Int. 105 (1991) 429^465.

[16] R.M. Allen, G. Nolet, W.J. Morgan, K. Vogfjo«rd, B.H.



Bergsson, P. Erlendsson, G.R. Foulger, S. Jakobsdo¤ttir,B.R. Julian, M. Pritchard, S. Ragnarsson, R. Stefa¤nsson,The thin hot plume beneath Iceland, Geophys. J. Int. 137(1999) 51^63.

[17] K.G. Dueker, A.F. Sheehan, Mantle discontinuity struc-ture from midpoint stacks of converted P to S wavesacross the Yellowstone hotspot track, J. Geophys. Res.102 (1997) 8313^8327.

[18] E.R. Kanasewich, C.D. Hemmings, T. Alpaslan, Nth-rootstack nonlinear multichannel ¢lter, Geophysics 38 (1973)327^338.

[19] B. Efron, G. Gong, A leisurely look at the bootstrap, thejackknife, and cross-validation, Am. Stat. 37 (1983) 36^48.

[20] Y. Shen, A.F. Sheehan, K.G. Dueker, C. de Groot-Hed-lin, H. Gilbert, Mantle discontinuity structure beneath thesouthern East Paci¢c Rise from P-to-S converted phases,Science 280 (1998) 1232^1235.

[21] S. Chevrot, L. Vinnik, J.-P. Montagner, Global-scaleanalysis of the mantle Pds phases, J. Geophys. Res. 104(1999) 20203^20219.

[22] C.R. Bina, G. Hel¡rich, Phase transition Clapeyronslopes and transition zone seismic discontinuity topogra-phy, J. Geophys. Res. 99 (1994) 15853^15860.

[23] R.S. White, J.W. Bown, J.R. Smallwood, The tempera-ture of the Iceland plume and origin of outward-propa-gating V-shaped ridges, J. Geol. Soc. Lond. 152 (1995)1039^1045.

[24] C.J. Wolfe, I.Th. Bjarnason, J.C. VanDecar, S.C. Solo-mon, Seismic structure of the Iceland mantle plume, Na-ture 385 (1997) 245^247.

[25] G.R. Foulger, M.J. Pritchard, B.R. Julian, J.R. Evans,R.M. Allen, G. Nolet, W.J. Morgan, B.H. Bergsson, P.Erlendsson, S. Jakobsdo¤ttir, S. Ragnarsson, R. Stefa¤ns-son, K. Vogfjo«rd, The seismic anomaly beneath Icelandextends down to the mantle transition zone and no deep-er, Geophys. J. Int. 142 (2000) F1^F5.

[25] R.M. Allen, G. Nolet, W.J. Morgan, K. Vogfjo«rd, B.H.Bergsson, P. Erlendsson, G.R. Foulger, S. Jakobsdo¤ttir,B.R. Julian, M. Pritchard, S. Ragnarsson, R. Stefa¤nsson,Imaging plume-ridge interaction in the mantle beneathIceland, J. Geophys. Res., in press.

[27] P. Poreda, J.-G. Schilling, H. Craig, Helium and hydro-gen isotopes in ocean-ridge basalts north and south ofIceland, Earth Planet. Sci. Lett. 78 (1986) 1^17.

[28] P. Michael, Regionally distinctive sources of depletedMORB: evidence from trace elements and H2O, EarthPlanet. Sci. Lett. 131 (1995) 301^320.

[29] B.J. Wood, The e¡ect of H2O on the 410-kilometer seis-mic discontinuity, Science 268 (1995) 74^76.

[30] T. Inoue, Y. Higo, T. Ueda, Y. Tanimoto, T. Futagami,T. Irifune, The e¡ect of H2O and CO2 on K-L-Q andpostspinel phase transformation of olivine, EOS Trans.Am. Geophys. Union 82 (2001) F1129^F1130.

[31] V.S. Solomatov, D.J. Stevenson, Can sharp seismic dis-continuities be caused by non-equilibrium phase transfor-mations?, Earth Planet. Sci. Lett. 125 (1994) 267^279.

[32] A.C. Kerr, A.D. Saunders, J. Tarney, N.H. Berry, V.L.Hards, Depleted mantle-plume geochemical signature noparadox for plume theories, Geology 23 (1995) 843^846.

[33] B.B. Hanan, J. Blichert-Toft, R. Kingsley, J.-G. Schilling,Depleted Iceland mantle plume geochemical signature:artifact of multicomponent mixing? Geochem. Geophys.Geosyst. 1 (2000).

[34] S.-H. Shim, T.S. Du¡y, G. Shen, The post-spinel trans-formation in Mg2SiO4 and its relation to the 660-km seis-mic discontinuity, Nature 411 (2001) 571^574.

[35] L. Chudinovskikh, R. Boehler, High pressure polymorphsof olivine and the 660-km seismic discontinuity, Nature411 (2001) 574^577.

[36] K. Hirose, Y. Fei, S. Ono, T. Yagi, K. Funakoshi, In situmeasurements of the phase transition boundary inMg3Al2Si3O12 : implications for the nature of the seismicdiscontinuities in the Earth’s mantle, Earth Planet. Sci.Lett. 184 (2001) 567^573.

[37] R.M. Allen, G. Nolet, W.J. Morgan, K. Vogfjo«rd, M.Nettles, G. Ekstro«m, B.H. Bergsson, P. Erlendsson,G.R. Foulger, S. Jakobsdo¤ttir, B.R. Julian, M. Pritchard,S. Ragnarsson, R. Stefa¤nsson, Plume driven plumbingand crustal formation in Iceland, J. Geophys. Res., inpress.

[38] F.A. Darbyshire, I.Th. Bjarnason, R.S. White, OŁ .G. Flo¤-venz, Crustal structure above the Iceland mantle plumeimaged by the ICEMELT refraction pro¢le, Geophys. J.Int. 135 (1998) 1131^1149.

[39] F.A. Darbyshire, K.F. Priestley, R.S. White, R. Stefa¤ns-son, G.B. Gudmundsson, S.S. Jakobsdo¤ttir, Crustal struc-ture of central Iceland from analysis of teleseismic receiverfunctions, Geophys. J. Int. 143 (2000) 163^184.

[40] Z. Du, G.R. Foulger, Variation in the crustal structureacross central Iceland, Geophys. J. Int. 145 (2001) 246^264.

[41] G. Thorbergsson, I.Th. Magnu¤sson, G. Pa¤lmason, Grav-ity data and a gravity map of Iceland, Report OS-93027/JHD-07, Orkustofnun IŁslands, 1993.

[42] K. Breddam, M.D. Kurz, M. Storey, Mapping out theconduit of the Iceland mantle plume with helium isotopes,Earth Planet. Sci. Lett. 176 (2000) 45^55.

[43] R.W. Gri⁄ths, M.A. Richards, The adjustment of mantleplumes to changes in plate motion, Geophys. Res. Lett.16 (1989) 437^440.

[44] K. Lambeck, P. Johnston, The viscosity of the mantle:evidence from analyses of glacial-rebound phenomena, in:I. Jackson (Ed.), The Earth’s Mantle: Composition,Structure, and Evolution, Cambridge University Press,New York, 1998, pp. 461^502.

[45] B. Steinberger, Plumes in a convecting mantle: modelsand observations for individual hotspots, J. Geophys.Res. 105 (2000) 11127^11152.

[46] D.V. Helmberger, L. Wen, X. Ding, Seismic evidence thatthe source of the Iceland hotspot lies at the core^mantleboundary, Nature 396 (1998) 251^255.

[47] I.Th. Bjarnason, P.G. Silver, G. Ru«mpker, S.C. Solomon,



Shear wave splitting across the Iceland hotspot: resultsfrom the ICEMELT experiment, J. Geophys. Res., inpress.

[48] S. Karato, P. Wu, Rheology of the upper mantle A syn-thesis, Science 260 (1993) 771^778.

[49] B.H. Hager, R.J. O’Connell, Kinematic models of large-scale £ow in the Earth’s mantle, J. Geophys. Res. 84(1979) 1031^1048.

[50] C.G. Chase, Asthenospheric counter£ow: a kinematicmodel, Geophys. J. R. Astron. Soc. 56 (1979) 1^18.



Seismic evidence for a tilted mantle plume and north^south mantle...

Documents

Transcript of Seismic evidence for a tilted mantle plume and north^south mantle...