Crustal thickness along the western Gala¤pagos Spreading ... · J. Pablo Canales a;, Garrett...

Crustal thickness along the western Gala¤pagos SpreadingCenter and the compensation of the Gala¤pagos hotspot swell

J. Pablo Canales a;�, Garrett Ito b;c, Robert S. Detrick a, John Sinton c

a Department of Geology and Geophysics, Woods Hole Oceanographic Institution, 360 Woods Hole Road, Woods Hole, MA 02543,USA

b Department of Geology, University of California, Davis, CA, USAc Department of Geology and Geophysics, School of Ocean and Earth Science Technology, University of Hawaii, Honolulu, HI, USA

Received 1 April 2001; received in revised form 8 July 2002; accepted 10 July 2002

Abstract

Wide-angle refraction and multichannel reflection seismic data show that oceanic crust along the Gala¤pagosSpreading Center (GSC) between 97‡W and 91‡25PW thickens by 2.3 km as the Gala¤pagos plume is approached fromthe west. This crustal thickening can account for V52% of the 700 m amplitude of the Gala¤pagos swell. Aftercorrecting for changes in crustal thickness, the residual mantle Bouguer gravity anomaly associated with theGala¤pagos swell shows a minimum of 325 mGal near 92‡15PW, the area where the GSC is intersected by the Wolf^Darwin volcanic lineament (WDL). The remaining depth and gravity anomalies indicate an eastward reduction ofmantle density, estimated to be most prominent above a compensation depth of 50^100 km. Melting calculationsassuming adiabatic, passive mantle upwelling predict the observed crustal thickening to arise from a small increase inmantle potential temperature of V30‡C. The associated thermal expansion and increase in melt depletion reducemantle densities, but to a degree that is insufficient to explain the geophysical observations. The largest densityanomalies appear at the intersection of the GSC and the WDL. Our results therefore require the existence ofcompositionally buoyant mantle beneath the GSC near the Gala¤pagos plume. Possible origins of this excess buoyancyinclude melt retained in the mantle as well as mantle depleted by melting in the upwelling plume beneath theGala¤pagos Islands that is later transported to the GSC. Our estimate for the buoyancy flux of the Gala¤pagos plume(700 kg s31) is lower than previous estimates, while the total crustal production rate of the Gala¤pagos plume (5.5m3s31) is comparable to that of the Icelandic and Hawaiian plumes.; 2002 Elsevier Science B.V. All rights reserved.

Keywords: Galapagos Rift; crust; hot spots; swells; plumes; mid-ocean ridges

1. Introduction

About 30% of the ocean £oor is occupied by

depth anomalies (swells) that correlate with grav-ity anomalies and hotspot volcanism [1] likely tooriginate from mantle plumes. Understandinghow hotspot swells are formed and sustainedover time will provide insight into the dynamicsof plume^lithosphere interaction and transfer ofenergy from the deep mantle to the Earth’s sur-face. Studies of mid-plate swells (e.g. [2,3^6]) sug-

0012-821X / 02 / $ ^ see front matter ; 2002 Elsevier Science B.V. All rights reserved.PII: S 0 0 1 2 - 8 2 1 X ( 0 2 ) 0 0 8 4 3 - 9

* Corresponding author. Tel. : +1-508-289-2893;Fax: +1-508-457-2150.E-mail address: [email protected] (J.P. Canales).

EPSL 6362 25-9-02 Cyaan Magenta Geel Zwart

Earth and Planetary Science Letters 203 (2002) 311^327

www.elsevier.com/locate/epsl

mailto:[email protected]

http://www.elsevier.com/locate/epsl

gest that these features are supported by a combi-nation of buoyancy by an ascending mantleplume and lithospheric density anomalies of ther-mal and/or chemical origin. The primary sourcescontributing to swell uplift are [7] : (1) thermalreheating of the lithosphere by magma injection;(2) thermal expansion of hot plume materialponding at the base of the lithosphere; (3) com-positional buoyancy of the residual swell root de-pleted by melt extraction; and (4) crustal thicken-ing by hotspot magmatism. However, the relativeimportance of each of them remains debatable.

The importance of buoyancy caused by anom-alously hot mantle has been tested using heat £owmeasurements, but the results have been inconclu-sive [3^5,8]. However, anomalously low seismicvelocities support the possibility of elevated man-tle temperatures beneath some hotspots [9,10].Another potentially important source of buoy-ancy may arise from depletion of melt from themantle. Partial melting of fertile mantle results ina lighter, depleted residue by preferential extrac-tion of heavier elements such as iron and con-sumption of denser phases such as garnet[11,12]. The correlation between swell topographyand volcano volume in Hawaii [7] argues for de-pletion as the primary source of support for theHawaiian swell (although numerical modelingsuggests that depletion plays a secondary role[13]). As depletion would increase the seismic ve-locity of the residue [12], the observed elevatedmantle seismic velocities beneath some hotspotswells [14] also support models with depleted swellroots. While the relative importance of thermalversus compositional mantle buoyancy remainscontroversial, there is strong evidence that bothe¡ects are important. Magmatic underplatingand crustal thickening [15,16] may play an impor-tant role in supporting some hotspot swells, as inthe case of the Marquesas hotspot swell [6].

The study of swells associated with mantleplumes proximal to mid-ocean ridges can provideadditional constraints on the structure of oceanicswells. Plume-related mantle thermal anomaliesacross the swell may alter crustal productionalong the nearby mid-ocean ridge. Higher mantletemperatures will result in larger melt productionby adiabatic mantle upwelling [17]. Therefore, the

contribution of the thermal anomaly to swell up-lift can be evaluated from melt production indi-cators such as crustal thickness along the spread-ing center. An ideal case of an oceanic swell andnear-ridge hotspot is the Gala¤pagos plume^ridgesystem (e.g. [18]).

The Gala¤pagos hotspot swell encompasses alarge portion of the Gala¤pagos Spreading Center(GSC) in the eastern equatorial Paci¢c (Fig. 1a).The GSC bisects the swell in an east^west direc-tion, separating the Cocos and Nazca plates at anintermediate full rate that increases from 47 mmyr31 near 97‡W to 63 mm yr31 at 86‡W [19]. Thelarge morphological variations observed along theGSC [20] suggest along-axis changes in magmasupply related to the nearby Gala¤pagos plume.Dominant volcanic features include the Cocosand Carnegie ridges, the Gala¤pagos Archipelagoand platform, and the Wolf^Darwin lineament(WDL), a volcanic chain extending between theGala¤pagos platform and the GSC. All these fea-tures re£ect the activity and evolution of the Ga-la¤pagos hotspot and its interaction with the GSC.

Previous studies of the mantle density structurebeneath the GSC supporting the Gala¤pagos swellwere based on bathymetry and gravity observa-tions [21]. In this paper we use wide-angle andmultichannel seismic (MCS) data to constraincrustal thickness variations along the westernGSC, from which we then infer mantle temper-ature variations beneath the GSC. Additionalmodeling of topography and gravity data pro-vides constraints on the anomalous density distri-bution across the Gala¤pagos swell. The integratedinterpretation of all the geophysical data allowsus to discriminate between thermal and chemicalanomalies, and quantify their importance in com-pensating the Gala¤pagos swell.

2. Seismic crustal thickness along the western GSC

As part of the G-PRIME project (Gala¤pagosPlume^Ridge Interaction Multidisciplinary Ex-periment), wide-angle and MCS data were col-lected along the GSC between 97‡W and91‡20PW [22] (Fig. 1b). In this paper we used asubset of these data to constrain the long-wave-


J.P. Canales et al. / Earth and Planetary Science Letters 203 (2002) 311^327312

length variation in crustal thickness along thewestern GSC.

2.1. Wide-angle seismic refraction

2.1.1. DataThree axis-parallel wide-angle seismic pro¢les

(Gala-1, Gala-2 and Gala-3; Fig. 1b) were shotalong the western GSC. The seismic source wasthe R/V Maurice Ewing’s 143-l air gun arraytowed at a depth of 8 m, with a shot interval of210 s at a nominal speed of 4.5 knots. Shot loca-tions were obtained from shipboard Global Posi-tioning System (GPS) positions corrected for the

Fig. 1. (a) Bathymetry map of the Gala¤pagos region [53]. GSC is the Gala¤pagos Spreading Center and WDL is the Wolf^Darwinlineament. Contours every 500 m. The Gala¤pagos swell is roughly marked by the shallow depths (6 3000 m) extending betweenV97‡W and 85‡W. The green circle marks the present location of the Gala¤pagos hotspot beneath Fernandina [54]. Black arrowsshow the sea£oor spreading direction and the gray arrow shows the migration of the ridge in a hotspot reference frame at a rateof 47 mm/yr [55]. (b) Bathymetry map of the study area (western section of the GSC). Bathymetry data include global bathyme-try derived from satellite altimetry data [53] and multibeam bathymetry (National Geophysical Data Center, Canales et al. [20]and Detrick et al. [22]). Dashed lines mark the location of the spreading axis, thin solid lines are wide-angle seismic refractionpro¢les (labeled Gala-1, 2 and 3), and gray lines are multichannel seismic pro¢les (labeled RF-2 and RF-6 in Fig. 4).


J.P. Canales et al. / Earth and Planetary Science Letters 203 (2002) 311^327 313

distance between the GPS antenna and the airgun array. The shots were recorded by ocean bot-tom hydrophones (OBHs). Locations of the in-struments on the sea£oor were determined fromthe shot position and water-wave travel times us-ing a Monte Carlo simulation [23]. The waterdepth at the relocated position of the instrumentswas taken from the Hydrosweep multibeam ba-thymetry.

The 100 km long pro¢le Gala-3 was centered at92‡W (Figs. 1b and 2a), where the GSC is char-acterized by a 30 km wide, East Paci¢c Rise(EPR)-like axial high elevated V700 m abovethe surrounding sea£oor. The pro¢le was located25 km to the north of the spreading axis to avoidthe in£uence of a possible axial magma chamberand low velocity zone in estimating crustal thick-ness. The pro¢le was recorded with ¢ve OBHs

Fig. 2. (a) Bathymetry along pro¢le Gala-3. (b) Wide-angle seismic record section from instrument OBH-27, located at the west-ern end of pro¢le Gala-3. Travel time is reduced to 7 km s31. Solid lines are predicted travel time curves by the preferred 1-Dvelocity model (Fig. 3b) for crustal arrivals (Pg), Moho re£ections (PmP) and upper mantle refractions (Pn). (c) Synthetic recordsection predicted by the preferred 1-D velocity model (Fig. 3b). (d^f) same as (a^c) for instrument OBH-26, located at the west-ern end of pro¢le Gala-2. (g^i) same as (a^c) for instrument ORB-5, located at the center of pro¢le Gala-1.



evenly spaced every 15 km. All of the instrumentsreturned data with high signal-to-noise ratio. Themaximum shot-receiver aperture was 80 km,which allowed recording crustal refractions (Pg),Moho re£ections (PmP) and upper mantle re£ec-tions (Pn). Fig. 2b shows an example of a seismicrecord section recorded at Gala-3. The Pg phaseis clearly observed as ¢rst arrival at ranges 6 55km, the PmP phase is observed as a high-ampli-tude, secondary arrival between 30 and 80 kmranges, and the Pn phase is observed as ¢rst ar-rival at ranges s 55 km.

The 90 km long pro¢le Gala-2 was centered at94‡15PW (Figs. 1b and 2d) 15 km to the north ofthe spreading axis, which is characterized by arough morphology transitional between axialhigh and valley [20]. The pro¢le was recordedwith ¢ve OBHs evenly spaced every 15 km. Allof them returned data with high signal-to-noiseratio, but a clock failure in the middle instrumentprevented us from using the data from this OBH.The maximum shot-receiver aperture was 75 km.Fig. 2e shows a record section representative ofthis pro¢le with the identi¢ed seismic phases.

The 95 km long pro¢le Gala-1 was centered at97‡W (Figs. 1b and 2g) along a deep axial valleysimilar to those found at slow-spreading ridgeslike the Mid-Atlantic Ridge (MAR). This sitewill serve as our reference, unperturbed crustalmodel since bathymetry and geochemical data in-dicate that this part of the GSC is not in£uencedby the Gala¤pagos plume [22]. The pro¢le was lo-cated in near-zero age crust because previousstudies (e.g. [24,25]) show that at ridge segmentswith axial-valley morphology, the oceanic crustacquires its full thickness at or very near thespreading axis. Five OBHs spaced every 19 kmwere deployed along Gala-1, but only the threeeasternmost instruments returned good-qualitydata. Fig. 2h shows a record section representa-tive of this pro¢le with the identi¢ed seismicphases.

2.1.2. Modeling and resultsTo obtain the crustal structure at the three sites

we modeled Pg, PmP and Pn travel times using aforward ray-tracing algorithm [26]. The traveltime picks of the Pg, PmP and Pn phases were

handpicked, with an estimated uncertainty of 20,30 and 25 ms, respectively. In addition to model-ing the travel times, we also computed syntheticseismograms [26] to ensure that the preferredmodels predict relative amplitudes consistentwith the observations.

Inspection of the Pg travel time against shot-receiver o¡set (Fig. 3a) shows that Pg picks alongpro¢les Gala-3 and Gala-2 have very little vari-ability (6 150 ms) for a given o¡set. This suggeststhat the crustal velocity structure along these pro-¢les is basically one-dimensional (1-D), with nosigni¢cant lateral variability along the pro¢les.We therefore modeled the wide-angle travel timesto obtain the best 1-D structure and averagedcrustal thickness at each of the pro¢les. Modelingthe two-dimensional structure along the pro¢leswould add very little information to our objectiveof ¢nding the long-wavelength variation in crustalthickness between the three sites. In contrast, Pgpicks from pro¢le Gala-1 show larger variability(Fig. 3a), suggesting larger lateral velocity varia-tions. To be consistent with the modeling ap-proach in the three pro¢les, we obtained the aver-aged 1-D structure in two distinct domains (eastand west of 96‡55PW) along pro¢le Gala-1.

The preferred 1-D structures and their esti-mated uncertainties (see Appendix) are shown inFig. 3b,c, respectively. Travel times predicted byour preferred models for each of the representa-tive instruments are shown in Fig. 2c,e,h and thesynthetic seismograms are shown in Fig. 2c,f,i.We ¢nd little di¡erence in crustal velocities be-tween Gala-2 and Gala-3, which are similar tothose of the EPR. In contrast, we ¢nd lower crus-tal velocities at Gala-1, similar to those found inthe slow-spreading MAR, probably re£ecting amore tectonized and fractured crust. Averagedcrustal thicknesses along each pro¢le (Fig. 3b)show a progressive increase from west to east:5.60 km at Gala-1, 5.90 km at Gala-2, and 7.45km at Gala-3. The increase in crustal thickness isconsistent with the observation that the crossoverdistance between Pg and Pn increases from 25 kmat Gala-1 (Fig. 2h) to 32 km at Gala-2 (Fig. 2e)and to 55 km at Gala-3 (Fig. 2b). Although thesmall di¡erences in crustal age among the threepro¢les may account for some of the di¡erences in



crustal velocity structure (e.g. [27]), they cannotexplain the variation in crustal thickness (e.g.[28]).

2.2. MCS re£ection

While refraction experiments provide accuratecrustal velocity and thickness information at threelocalized regions along the GSC, MCS re£ectionsfrom the Moho provide crustal thickness con-straints between the refraction experiments alongV370 km of the GSC (Fig. 1b). The MCS surveyincluded six axis-parallel lines 15^30 km north ofthe ridge axis on mature crust. Energy from a 72-lair gun array was recorded with a 6.1 km long,480-channel, digital streamer every V15 s to ob-tain 80-fold common mid-point (CMP) gathers.In the CMP gathers, re£ections from the Mohoappeared as a nearly £at arrival, typically extend-ing from zero to 3.5^5 km shot-receiver o¡sets.Prior to stacking, traces with o¡sets less than0.5^1.5 km were muted to minimize coherent

noise with moveout larger than Moho. We thencorrected for normal moveout and stacked theCMPs.

We con¢dently image Moho along V80% ofthe stacked MCS sections. In the westernmostend of the survey Moho appears as a relativelyweak, £at lying re£ector at V2.0 s two-way traveltime (TWTT) below the sea£oor (Fig. 4a). Thesub-sea£oor TWTT to Moho increases to theeast, reaching V2.5 s near 92‡W (Fig. 4b), wherethe Moho re£ection is the strongest. To computecrustal thickness from TWTT to Moho we con-verted the velocity^depth pro¢les from each of thethree refraction experiments (Fig. 3b) to velocity^TWTT pro¢les and assigned each pro¢le to thecenter of the associated refraction line. We thenused a cubic spline to interpolate velocities hori-zontally between the refraction line centers ateach CMP location. The TWTT to Moho pickswere then converted to crustal thickness from theinterpolated velocity^TWTT pro¢les.

Fig. 5b shows crustal thicknesses along the

Fig. 3. (a) Pg travel time picks for the OBHs along pro¢les Gala-1 (squares), Gala-2 (triangles) and Gala-3 (circles). Note thatpicks for Gala-2 and Gala-3 show very little travel time variability for a given shot-receiver o¡set, suggesting that the crustal ve-locity structure at these sites is basically 1-D. Instead, picks for Gala-1 show larger variability due to a more heterogeneous crus-tal structure. (b) Preferred 1-D velocity models for the three wide-angle seismic pro¢les. Crustal thicknesses are indicated. Veloc-ity structures representative of fast-spreading (EPR) [34] and slow-spreading (MAR) [24] ridges are shown for comparison. (c)Estimated uncertainties of the velocity models. See Appendix for details.



GSC derived from both the MCS and refractiondata. Consistency between the two datasets isdemonstrated by the similarity in crustal thicknessmeasurements where the two datasets overlap (i.e.Gala-2 and Gala-3). The MCS picks show short-wavelength (6 50 km) undulations in crustalthickness typically S 0.5 km in amplitude. Someof these undulations may arise from sea£oorroughness that introduces some uncertainty inthe identi¢cation of the Moho re£ection. On theother hand, some of these undulations likely re-£ect true structural variations, in particular eastof 92‡30PW where the Moho re£ection is high inamplitude and sea£oor topography is smooth.For our purpose of determining the origin ofthe Gala¤pagos swell, however, we will ignorethese short-wavelength variations and focus onlyon the longest-wavelength regional variation. Toproduce a regional, smooth crustal thicknessmodel we use a third-order polynomial that ¢tsthe MCS and refraction results in a least-squaressense. This regional crustal model shows mini-mum crustal thicknesses of V5.6 km, closelymatching the refraction results at Gala-1. Thisreference value is consistent with mean values ofcrustal thickness in fast-spreading crust (e.g. [28]),and slow-spreading crust averaged along individ-ual ridge segments [25]. The maximum crustalthickness approaches 8.0 km at the eastern endof the survey. We thus ¢nd a total thickening of

the crust by V2.3 km, an increase by V40% ofthe minimum crustal thickness (Fig. 5b).

3. Compensation of the Gala¤pagos swell

3.1. Depth anomaly

Hydrosweep multibeam bathymetry acquiredduring the G-PRIME cruise has allowed us forthe ¢rst time to map continuously the ridge crestalong the GSC between 97‡50PW and 90‡50PW(Fig. 1b). Along the western section of the GSCthe axial depth shoals from west to east by 1800 m(Fig. 5a). While part of this change in axial depthcorresponds to the bathymetry anomaly of theGala¤pagos swell, a signi¢cant part of it is prob-ably related to changes in axial morphology. Lith-ospheric stresses, which are controlled by the ther-mal state of the lithosphere, are responsible forthe origin of mid-ocean ridge topography suchas axial valleys and highs (e.g. [29,30]).

We use a two-dimensional ¢ltering approach toremove from the axial depth pro¢le the topogra-phy variations that result from changes in axialmorphology. Low-pass ¢ltering the bathymetrywith cuto¡ wavelengths (V) smaller than 85 kmresults in swell amplitudes that decrease progres-sively as V increases. In contrast, for Vs 85 kmthe swell amplitude remains constant indepen-

Fig. 4. Stacked multichannel seismic section along pro¢les (a) RF-2 and (b) RF-6. See Fig. 1b for location. Prominent sub-sea-£oor re£ectors are interpreted as the base of layer 2A and the Moho. Note that both panels are roughly aligned with the sea£oorre£ection to enhance the di¡erences in crustal TWTT to the Moho re£ection between both pro¢les. The Moho in RF-2 is atV2 s TWTT below the sea£oor, while at RF-6 it is deeper (V2.3 s TWTT).



∆Fig. 5. (a) Axial depth along the GSC. Thick solid line is the axial depth taken from the low-pass ¢ltered bathymetry (cuto¡wavelength of 85 km), interpreted as the swell topography. Dashed line shows the predicted topography anomaly due to varia-tions in crustal thickness, calculated from Eq. 1. Insets show the cross-axis topography (¢ltered and un¢ltered) at three locationswith distinct axial morphology. In areas characterized by axial-valley morphology (west of 95‡30PW), ¢ltering removes the ele-vated shoulders and the deepening of the valley £oor. Where an axial high is present (east of 92‡40PW), ¢ltering attenuates theamplitude of the high. And where the morphology is a transition between axial high and valley, ¢ltering does not signi¢cantlychange the axial depth. (b) Crustal thickness along the GSC from the multichannel seismic data (solid circles). The open squaresare the crustal thickness measurements from the wide-angle pro¢les. The solid line is a smooth crustal thickness pro¢le derivedfrom both the MCS and wide-angle observations. (c) Thick solid line is the mantle Bouguer gravity anomaly (MBA) along theGSC computed assuming constant crustal thickness (see text for details). The gray line is the RMBA calculated using the smoothMoho pro¢le in (b) (see text for details). Thin solid lines are gravity anomalies produced by the mantle density variations shownin (d) for di¡erent compensation depths (labels in km). (d) Along-axis variations in mantle density inferred from the isostasymodel (Eq. 2, see text for details), for di¡erent values of the compensation depth (labels in km). Dashed portions of the curvesin (c) and (d) are extrapolations where there are no crustal thickness constraints.



dently of V. As a ¢rst-order approximation, weconsider that a low-pass ¢lter with V=85 km ef-fectively removes the contribution of the axial to-pography while preserving the longer-wavelengthanomaly associated with the swell. The insets inFig. 5a illustrate the e¡ect of the ¢ltering oncross-axis bathymetry. We ¢nd that V60% ofthe observed variations in axial depth is relatedto changes in axial morphology. The amplitude ofthe bathymetry swell along the GSC is thereforegiven by the ¢ltered axial depth (Fig. 5a), whichhas a maximum elevation of V700 m near 91‡W.

We calculate the topography anomaly relatedto crustal thickness variations, vHc, by assuminglocal Airy isostasy (see Table 1 for de¢nitions):

vHc ¼vbmc

vbmwWvC ð1Þ

Fig. 5a shows that vHc underestimates the ob-served swell amplitude. On average, east of 94‡Wcrustal thickness variations support 52S 11% ofthe depth anomaly (although locally near 91‡30PWit can be as much as 80%). This indicates thatvariations in crustal thickness cannot be theonly source of support for the Gala¤pagos swell ;an eastward decrease in mantle density is alsorequired.

3.2. Gravity anomaly

Gravity data were acquired continuously dur-ing the G-PRIME cruise using the R/V Ewing’sBell Aerospace BGM-3 gravity meter. Free airanomalies (FAA) were obtained from the totalgravity ¢eld measurements after Eo«tvos and in-strumental drift corrections. FAA was reduced

Table 1Notation

Variable De¢nition [units] Value

Ac cross-sectional area of the excess crust [m2] 8.9U108a

As cross-sectional area of the swell (mantle contribution) [m2] 1.8U108a

C crustal thickness [m]Cp heat capacity [J kg31 K31] 1256vC crustal thickness variation [m]dF/dP melt productivity [% GPa31]dT/dPMadb slope of the adiabat [‡C GPa31] 10dT/dPMsol slope of the solidus [‡C GPa31] 130F melt fractionHf heat of fusion [J kg31] 754U103

vH swell topography [m]vHc swell topography due to crustal thickness variations [m]P pressure [GPa]T temperature [‡C]Tsol0 intercept of the solidus [‡C] 1100

U full spreading rate [mm yr31] 54Z sea£oor depth [m]Zp compensation depth [m]K coe⁄cient of thermal expansion [K31] 3.1U1035

L coe⁄cient of depletion density reduction 0.06bc crustal density [kg m33] 2800bm reference mantle density [kg m33] 3330vbcw crust^water density contrast [kg m33] 1770vbm mantle density anomaly [kg m33]vbmc mantle^crust density contrast [kg m33] 530vbmw mantle^water density contrast [kg m33] 2300a Ac is the cross-sectional area of the excess crust (in excess of 5.6 km), and As is the cross-sectional area of the swell topogra-phy after removing the crustal support (i.e. the area comprised between the thick solid and dashed lines in Fig. 5a, east of95‡30PW). Ac and As are measured along the ridge, i.e. perpendicular to the spreading direction, assuming symmetry about the90‡50PW FZ.



to mantle Bouguer anomaly (MBA) using stan-dard procedures (e.g. [31]) in order to isolate thegravity signature of the subsurface density struc-ture. From the FAA we subtracted the contribu-tions of the sea£oor topography and the Mohointerface [32] assuming a 6 km thick, constant-density crust (see Table 1 for density contrasts).The resulting MBA (Fig. 5c) shows an eastwarddecrease of V70 mGal along the GSC, arisingfrom variations in crustal thickness and/or mantledensity (e.g. [31]).

Since we have independent seismic constraintson crustal thickness variations, we can estimatetheir contribution to the MBA. We calculated aresidual MBA (RMBA) by subtracting from theFAA the gravity contributions of the sea£oor andMoho interfaces assuming a constant-densitycrust with thickness varying according to thesmooth crustal model of Fig. 5b. We did not cor-rect for the lithospheric cooling e¡ect (e.g. [33])because our analysis is restricted to the zero-age,along-axis gravity anomalies. If all of the gravityvariations were caused by variations in sea£oordepth and crustal thickness, then the RMBAwould be near zero. However, the RMBA (Fig.5c) becomes increasingly negative to the east,reaching a minimum of V325 mGal at92‡15PN, the intersection between the GSC andthe WDL. Therefore, on average, east of 94‡Wcrustal thickening accounts for 60S 10% of theobserved MBA. The negative RMBA is yet fur-ther evidence for substantial variation in mantledensity, with lower mantle densities beneath theridge near the hotspot.

The previous calculations are made with theassumption that no signi¢cant changes in crustaldensity occur along the GSC. The large variationsin morphology along the GSC suggest the pres-ence of signi¢cant variability in the thermal (i.e.density) structure of the axial crust. We calculatedthe gravitational e¡ect of an 8 km wide, low-den-sity, high-temperature axial crustal zone [34] witha progressive density decrease between 97‡W and91‡W corresponding to a temperature increase of500‡C (a maximum estimate for the di¡erence inaverage axial crustal temperature beneath a slow-spreading ridge without a steady-state crustalmagma chamber, and a fast-spreading ridge with

a steady-state crustal magma chamber [35]). We¢nd the e¡ects of this structure to be small, with amaximum contribution of 5^7 mGal near 91‡W.

3.3. Mantle density anomaly

The previous analysis of topography and grav-ity data shows that the variation in crustal thick-ness along the GSC is an important, but not theonly, source of support for the Gala¤pagos swell.Variations in mantle density are also required. Todetermine the mantle density structure beneaththe GSC we assume local isostatic equilibriumalong the ridge as a combination of crustal Airyisostasy and Pratt isostasy within the mantle. Thevariation in mantle density vbm along the GSCcan be inferred from a mass balance as (see Table1 for de¢nitions) :

vbm ¼ vbmcWvC3vbmwWvHZp3Z3C

ð2Þ

The only unconstrained parameter in Eq. 2 isZp, an e¡ective compensation depth above whichwe assume all the density anomalies are con¢ned.Fig. 5d shows the variations in mantle densityderived from Eq. 2 for several values of Zp. Ifthe density anomalies are shallow (Zp = 50 km),the magnitude of the density decrease east of95‡W is V10 kg m33, with a maximum amplitudeof V19 kg m33 at 92‡15PN, the intersection be-tween the GSC and the WDL. For Zp = 100 km,the density anomaly east of 95‡W is V5 kg m33,with a maximum amplitude of V8 kg m33 at92‡15PN. For deeper compensation depths themaximum density anomaly is less than 5 kg m33.

To constrain Zp we calculate the gravity con-tribution of each of the density structures dis-played in Fig. 5d and then compare them to theRMBA. The results (Fig. 5c) show that the grav-ity data are consistent with compensation depthsof 50^100 km, indicating signi¢cant mantle den-sity reduction near the hotspot, with values of 4^8kg m33 near 91‡25PW and maximum values of8^19 kg m33 near the intersection of the WDL.These calculations show that shallow mantle den-sity variations are su⁄cient to support the swell,



without the need to invoke dynamic uplift unre-lated to shallow density variations (e.g. [36]).

3.4. Melting models and inferred excesstemperature

The mantle density anomalies may arise fromthermal expansion and/or from compositionalvariations. To estimate the thermal contributionto the mantle density anomaly we assume that theeastward increase in crustal thickness along theGSC originates from temperature variations inthe underlying mantle. To calculate the temper-ature increase required to produce the crustalthickness variations, we assume passive mantleupwelling and we integrate the melt producedalong adiabatic melting paths for di¡erent mantlepotential temperatures (assuming that all the meltis extracted) (e.g. [17]). We use the melting func-tion [37] (see Table 1 for de¢nitions):

dFdP

¼ dT=dPMadb3dT=dPMsolHf =Cp þ dT=dF

ð3Þ

with

dT=dF ¼ 350ð13P=8:8Þ for F622%680ð13P=8:8Þ for Fv22%

ð4Þ

Fig. 6a shows the relationship between incre-ment of mantle temperature and the increase inmelt thickness. The maximum observed increase

6

Fig. 6. (a) The thin solid line is the increase in mantle poten-tial temperature versus increase in melt thickness derivedfrom the melting model of Langmuir et al. [37]. For a givenmantle potential temperature, Eqs. 3 and 4 were solved nu-merically for P, T and F using an explicit Runge^Kuttascheme. The melt thickness was then estimated by integratingthe volumetric melt fraction along the melting path abovethe solidus [17]. Eq. 3 yields melt productivities of dF/dP=12 and 18% GPa31 for F6 and v 22%, respectively(the reduction in productivity at 22% simulates the e¡ects ofexhausting clinopyroxene as a melting phase). We also ex-plored the dependence of our results on the melt productivityfunction. Allowing dF/dP to increase with F [56], the increasein mantle potential temperature required to match the crustalthickness variations is a few ‡C smaller than using Eqs. 3and 4, indicating that our results are not signi¢cantly sensi-tive to the melt productivity function adopted. Because latentheat removal during melting decreases mantle temperaturesabove the solidus [38], the mean temperature variationpresent beneath the GSC should be smaller than the varia-tion in potential temperature. We included this e¡ect by cal-culating, for a given potential temperature, the temperaturedi¡erence between a melting path and the adiabat withoutmelting, averaged above the compensation depth. At thesame potential temperature, shallower compensation depthstherefore yield smaller mean temperature anomalies thangreater compensation depths because a larger portion of themantle contributing to the swell bathymetry is cooled by la-tent heat loss. This e¡ect is illustrated with bold lines forour preferred compensation depths (labeled in km). Graybands mark the excess crustal thickness at 91‡25PW, 92‡15PW(the intersection of the WLD) and 94‡10PW. (b) Reductionin mantle density against increase in melt thickness derivedfrom the melting calculations, for our preferred compensa-tion depths (labeled in km). Thin solid lines correspond tothe decrease in density due to thermal expansion. Thick solidlines correspond to the decrease in density due to the com-bined e¡ects of thermal expansion and melt depletion forL=0.060. Also shown by dashed and dash-dotted gray linesare the results for minimum (0.024) and maximum (0.077) es-timates of L (see text for details.) Gray bands mark the ex-cess crustal thickness at three locations along the axis, as in(a), and the range of mantle density anomaly inferred fromthe isostasy model for our preferred compensation depths asillustrated in Fig. 5d.



in crustal thickness derived (2.3 km) requires anincrease in mantle potential temperature of 30‡C.However, the actual increase in mean mantle tem-perature above the compensation depth is some-what smaller (18^24‡C for Zp = 50^100 km) dueto the loss of latent heat of fusion (e.g. [38]). Thisresult indicates that a moderate mantle thermalanomaly along the GSC is su⁄cient to explainthe observed variation in crustal thickness.

The mantle density reduction due to the ele-vated temperatures and enhanced melt productioncan be expressed as the sum of two components:one purely thermal and a second one related tothe loss of dense phases (spinel) and elements (Fewith respect to Mg) during melting (mantle deple-tion) [11,12] (see Table 1 for de¢nitions):

vb ¼ 3bmðK WvT þ L WvFÞ ð5Þ

Here vF is the excess in average melt depletionbeneath the GSC produced by the mantle temper-ature anomaly and L is the coe⁄cient of depletionbuoyancy. The increase in MgO/FeO ratio in themantle residue for 25% melting reduces the den-sity of the depleted mantle by 0.6% with respectto the fertile mantle [11], yielding L=0.024. Ifmelting occurs primarily within the spinel stabilityzone, the consumption of spinel and clinopyrox-ene and the increasing MgO/FeO leaves a residue1.5% less dense than a fertile spinel lherzolite [11],yielding L=0.060. If signi¢cant melting occurs inthe garnet stability ¢eld, the total density reduc-tion is 2% with respect to a fertile garnet lherzolite[11], yielding a maximum estimate of L=0.077.We adopt L=0.060 as our preferred value sincemost of the melting beneath the GSC likely occursabove 70^90 km [39], values consistent with ourpreferred range of compensation depths of 50^100km.

Fig. 6b shows that the density reduction due tothermal expansion for the temperature incrementof 18^24‡C associated with the 2.3 km of crustalthickening is very small, less than 33 kg m33.By including the e¡ects of mantle depletion(L=0.06), the predicted maximum density anom-aly near 91‡W is V7^10 kg m33, while near theWDL it is V4^7 kg m33. As reference we alsoshow in Fig. 6b the density anomalies predicted

for the end-member cases of L=0.024 and L=0.077. By comparing the mantle density anom-alies predicted by the melting model with thoserequired by the geophysical data (shaded boxesin Fig. 6b), we ¢nd that depletion and thermalexpansion are su⁄cient to explain the densityanomaly near 91‡W, but not beneath other sec-tions of the GSC. This is true especially at92‡15PW, where the WDL intersects the GSC.Here the melting model predicts only V4^7 kgm33 of the V9^20 kg m33 density reduction esti-mated from the isostasy model for our preferredcompensation depths of 50^100 km. Thus a largeportion, but not all, of the GSC swell and gravityanomaly can be explained by the combined e¡ectsof an eastward thickening of the crust and asso-ciated reduction in mantle density caused by in-creasing temperatures and degree of melt deple-tion.

4. Discussion

4.1. Crustal thickness variations

The results from our seismic experiment repre-sent the ¢rst direct measure of crustal thicknessalong the GSC. Taking crustal thickness as aproxy for magma supply, the seismic measure-ments provide important constraints on the lateralextent of the plume-related thermal anomalyalong the GSC. While crustal thickening alongthe western GSC is very gradual, there are twodistinct gradients in the pattern of crustal thicken-ing (Fig. 5b). West of 94‡W the crust thickens tothe east by only a few hundred meters at a rate ofV130 m per 100 km, while most of the thickening(V2 km) occurs east of 94‡W at a more rapidrate (V620 m per 100 km). This pattern suggeststhat the plume in£uence on magma supply at theGSC is con¢ned primarily to within V350 km ofwhere the western GSC is closest to the hot spot,consistent with the steeper geochemical gradientsobserved east of 94^93‡W [22]. The relative en-richment in K/Ti and Nb/Zr in basalts collectedeast of the 95‡30PW propagator with respect tosamples to the west of it suggests that the propa-gator might be the western limit of plume-a¡ected



mantle [22]. Therefore, while the compositionalplume anomaly may extend for s 500 km alongthe ridge, our constraints on crustal thickness sug-gest that the thermal anomaly is more con¢ned,probably due to conductive cooling. These obser-vations provide important constraints on the re-gional thermal structure and along-axis dispersalof plume material in the Gala¤pagos plume^ridgesystem.

4.2. Excess buoyancy

Our results agree with the main conclusion ofthe previous study of the Gala¤pagos swell by Itoand Lin [21], that is, both crustal thickening andhigher temperatures near the hotspot within theshallow mantle contribute to support the swell.We ¢nd that crustal thickness variations support,on average, 52% of the depth anomaly and create60% of the MBA. These are maximum estimatesof the crustal contribution to swell topographyand gravity anomaly due to our assumption ofconstant crustal density. Since most of the crustalthickening occurs in the lower crust (Fig. 3b), weare overestimating vbmc in the plume-a¡ected sec-tion of the ridge and therefore underestimatingthe amplitude of the decrease in RMBA and man-tle density along the GSC by V10%.

Quantitatively, our measured crustal thicknessvariation (2.3 km) and inferred excess temperature(30‡C) are at the lower bounds of Ito and Lin’spredictions of 3 S 1 km and 50S 25‡C, respec-tively. Since Ito and Lin had no seismic con-straints on crustal thickness, they explained thetopography and gravity anomalies with a slightlylarger crustal thickness and mantle temperatureanomalies. The new constraints on crustal thick-ness from this study show that an additionalsource of buoyancy in the mantle is required tosupport the swell.

A plausible origin for the additional buoyancyis melt retention in the mantle [40]. If the £ux ofmelt rising through the mantle increases withmantle porosity, then it seems logical that varia-tions in melt retention would be positively corre-lated with variations in magma supply. However,we ¢nd the largest crustal thickness and inferredmelt £ux near 91‡30PW (Fig. 5b), o¡set from the

calculated maximum mantle density reductionnear 92‡15PW (Fig. 5d). This could re£ect a de-coupling between crustal thickness and melt £ux(e.g. by ductile crustal £ow [41] or along-axis meltmigration at crustal and/or sub-crustal levels), orargues against melt retention as the origin of theexcess buoyancy.

Another possible explanation for the additionalmantle buoyancy is related to the concept thatplume material from the Gala¤pagos hotspot feedsthe GSC (e.g. [22,42]). As plume material melts atthe upwelling plume beneath the Gala¤pagos Is-lands its density decreases due to depletion [7].Some of this material may £ow northward tothe GSC and introduce buoyancy unrelated tomelting beneath the GSC [18]. This hypothesis isconsistent with the low 3He/4He ratio observedalong the GSC that suggests that plume-a¡ectedmantle beneath the GSC has been degassed byprior melting in the upwelling plume [22], and itis also consistent with the signi¢cant depleted sig-nature exhibited by lavas erupted in the northernGala¤pagos Islands and along the WDL [43]. Onecaveat for this scenario is that mantle depletionincreases the temperature of the solidus at a givendepth (e.g. [37]); therefore the variation in mantlepotential temperature required to explain thealong-axis crustal thickness variations would belarger than our estimate, increasing the contribu-tion of thermal buoyancy.

A depleted mantle source for the excess buoy-ancy seems to contradict the observed enrichmentin incompatible elements along the GSC near thehotspot [22,42]. While modeling the dynamics ofthe Gala¤pagos plume^ridge system and its geo-chemical implications are beyond the scope ofthis paper, this apparent discrepancy could be ex-plained by a scenario in which £ow from theplume to the ridge occurs at two distinct levels.Depleted residue from the center of the plumechannel would £ow at shallow levels to the ridge,contributing to the excess buoyancy, while en-riched material without signi¢cant prior meltingfrom the outer zones of the plume might £ow atdeeper levels, contributing to the enriched, low-melt-fraction lavas found at the GSC [22].

The maximum excess buoyancy occurs at theintersection of the GSC with the WDL, which



was proposed to be the surface expression ofchannelized £ow from the plume to the ridge[44]. Narrow sublithospheric channels connectingplumes and mid-ocean ridges may develop inplume^ridge systems where the ridge migratesaway from the plume [45^47]. However, the in-ferred moderate temperature anomaly and viscos-ity of the Gala¤pagos plume argues for preferen-tially radial, rather than focused, lateral £ow ofplume material [18]. In this scenario, melt associ-ated with secondary, small-scale convection [48]or decompression during upslope £ow [45] maybe focused along pre-existing lithospheric frac-tures, enhancing locally the buoyancy by melt re-tention. Our data cannot discriminate betweenthis hypothesis and the hypothesis in which hotand compositionally buoyant plume material isdelivered to the GSC along the WDL. However,the striking correlation between the WDL and theexcess buoyancy suggests that the WDL plays animportant role in the Gala¤pagos plume^ridge sys-tem that should be further investigated in futureresearch.

4.3. Buoyancy £ux and crustal productivity

Our new results on swell topography and on itscrustal support allow us to place constraints onbuoyancy £uxes and crustal production rates atthe Gala¤pagos plume^ridge system. The buoyancy£ux of excess mass £owing away from the ridgeaxis is given by (see Table 1 for de¢nitions) :

B ¼ AsWvbmwWU ð6Þ

and it is the sum of thermal buoyancy BT and thebuoyancy £ux caused by melt depletion BF. Wecompute BF from (e.g. [13]):

BF ¼ QcaWb cWL ð7Þ

where Qca is the volume £ux of excess crustalproduction at the ridge axis de¢ned as:

Qca ¼ AcWU ð8Þ

From Eqs. 6^8 we ¢nd that B=700 kg s31,BF = 255 kg s31 and BT = 445 kg s31. These valuesindicate that melt depletion is important, contrib-uting V35% of the total buoyancy B at the GSC.

Our estimate of the thermal buoyancy BT is smallcompared to prior estimates of 103 kg s31 withoutcrustal thickness constraints [47,49]. The presenceof the excess mantle buoyancy requires thermalbuoyancy to be less than the above estimate.Low thermal buoyancy is consistent with ourlow excess temperatures along the GSC and sug-gests low excess temperatures and/or low volume£ux arising from the Gala¤pagos plume. Indeed, aplume temperature excess of considerably lessthan 100‡C can be inferred if the lateral thermalgradient in the mantle between the GSC and theGala¤pagos Islands is comparable to that esti-mated along the GSC.

Ito et al. [18] estimated the crustal volume £uxforming the Gala¤pagos platform to be V4 m3

s31. Therefore, considering the excess volume£ux at the GSC given by Eq. 8 (Qca = 1.5 m3

s31), the total crustal volume £ux of the Gala¤pa-gos plume is V5.5 m3 s31. This value is surpris-ingly high given the moderate buoyancy £ux andexcess temperature. While the Gala¤pagos crustalvolume £ux is V20% smaller than that of theIcelandic plume (7 m3 s31) [50], it is comparableto that of the Hawaiian plume (5 m3 s31) [51].Despite their di¡erence in excess plume temper-ature, the similar crustal volume £uxes of the Ga-la¤pagos and Hawaiian plumes are likely due tothe di¡erences in lithospheric thickness. By con-sidering the combined crustal production of theGala¤pagos plume and sea£oor spreading at theGSC, we ¢nd that the crustal productivity of theGala¤pagos plume^ridge system is V15.7 m3 s31,very similar to the V16 m3 s31 of the Iceland^MAR system [50]. This result indicates that thethree-fold increase in plate velocities at the Gala¤-pagos plume^ridge system o¡sets the larger islandvolume and wider along-axis plume in£uence ofthe Iceland^MAR system.

5. Conclusions

The integrated modeling and interpretation ofbathymetry, gravity and seismic data presented inthis study places constraints on the compensationmechanism of the Gala¤pagos swell, allowing us toquantify the roles of crustal thickness, mantle



temperature and mantle depletion in supportingthe Gala¤pagos hotspot swell. The main conclu-sions of our study are as follows:

1. Modeling of wide-angle refraction and multi-channel seismic data along the western GSCbetween 97‡W and 91‡25PW indicates that thecrust thickens by V2.3 km as the Gala¤pagoshotspot is approached from the west.

2. The eastward increase in crustal thickness isaccompanied by a decrease in axial depth of1800 m (60% of which is due to changes inaxial morphology) and a decrease in MBA of70 mGal. Crustal thickness variations explain52% of the depth anomaly and 60% of theMBA. The remaining anomalies require thepresence of low mantle densities beneath theGSC near the hotspot, constrained to be with-in the uppermost mantle (above 50^100 kmdepth).

3. Assuming passive upwelling beneath the GSC,we ¢nd that the crustal thickness variations areconsistent with an eastward increase in mantlepotential temperature of 30‡C.

4. The density anomaly produced by thermal ex-pansion and mantle depletion beneath theridge is smaller than that estimated from thegeophysical data, with the largest discrepanciesfound at the intersection of the GSC and theWDL. These discrepancies imply the existenceof an additional compositionally buoyantsource supporting the Gala¤pagos swell. Possi-ble origins of this anomaly are melt retained inthe mantle as well as mantle depleted by melt-ing in the upwelling plume beneath the Gala¤-pagos Islands that is then transported to theGSC.

5. Depletion and thermal buoyancy accounts forV35% and V65%, respectively, of our esti-mate for the total buoyancy £ux of 700 kgs31. While buoyancy £ux is moderate com-pared to prior estimates of other hotspots,the rate of crustal production of the Gala¤pagosplume is comparable to that of the Iceland andHawaii mantle plumes, and the crustal produc-tivity of the Gala¤pagos plume^ridge system isalso comparable to that of the Iceland^MARsystem.

Acknowledgements

This study was supported by the National Sci-ence Foundation Grant OCE-9819117 to theWoods Hole Oceanographic Institution andOCE-9818632 to the University of Hawaii. Wethank Captain Mark Landow, Science O⁄cerChris Leidhold, and the crew of the R/V MauriceEwing Leg 00^04 for their help in the data acqui-sition. We also thank Ecuadorian authorities forgranting us permit to work within Ecuador’s ex-clusive economic zone, and the o⁄cers of the Ga-la¤pagos National Park and Charles Darwin Re-search Station for facilitating our work within thelimits of the Gala¤pagos National Park. We aregrateful to the WHOI OBS Group, namely,J. Bailey, T. Bolmer, D. DuBois and R. Handyfor their technical support ; and to all the partic-ipants of the G-PRIME cruise for their assistancein the ¢eldwork. M. Behn helped with the gravitydata reduction and calculations. G. Hirth, P. Ke-lemen and J. Lin are acknowledged for sugges-tions and comments on various aspects of thisstudy. The paper bene¢ted from reviews byI. Grevemeyer, Y. Shen and two anonymousreviewers. The GMT software [52] was used inthe preparation of this manuscript. Contributionnumbers WHOI-10758 and SOEST-5999.[SK]

Appendix

To estimate the uncertainty of the velocitymodels (Fig. 3c), we generated a large numberof crustal models by adding random velocity per-turbations at each depth node to our preferredmodels (imposing positive velocity gradients withdepth, and that the randomized velocity valueswere within reasonable bounds according to thedepth). We then calculated the travel times pre-dicted by the random models. The velocity uncer-tainties were then estimated as the standard devi-ation of all of the random models that providedreasonable ¢ts to the data. Our velocity error es-timates indicate that the MCS TWTT Moho pickscan be converted to depth with an accuracy ofS 150 m.

The uncertainty of the crustal thickness esti-



mates is mostly controlled by the tradeo¡ betweencrustal thickness and lower crustal velocity. Weexplored this ambiguity in our solutions by mod-eling the travel times with velocity models thatincluded all possible combinations of lower crus-tal velocities ranging between 6.9 and 7.4 km s31,and crustal thicknesses within S 1 km of our pre-ferred solution (the velocity structure of the upperand middle crust was our preferred solution in allcases). The models that yield reasonable ¢ts indi-cate that the velocity^depth ambiguity introducesan uncertainty of 0.2^0.3 km to the wide-anglecrustal thickness estimates (Fig. 3b).

References

[1] S.T. Crough, Hotspot swells, Annu. Rev. Earth Planet.Sci. 11 (1983) 165^193.

[2] M.K. McNutt, L. Shure, Estimating the compensationdepth of the Hawaiian swell with linear ¢lters, J. Geo-phys. Res. 91 (1986) 13915^13923.

[3] R.P. Von Herzen, M.J. Cordery, R.S. Detrick, C. Fang,Heat £ow and the thermal origin of hot spot swells: theHawaiian swell revisited, J. Geophys. Res. 94 (1989)13783^13799.

[4] R.C. Courtney, R.S. White, Anomalous heat £ow andgeoid across the Cape Verde Rise: evidence for dynamicsupport from a thermal plume in the mantle, Geophys. J.R. Astron. Soc. 87 (1986) 815^867.

[5] R.S. Detrick, R.P. Von Herzen, B. Parsons, D.T. Sand-well, M.E. Dougherty, Heat £ow observations on the Ber-muda Rise and thermal models of midplate swells, J. Geo-phys. Res. 91 (1986) 3701^3723.

[6] M.K. McNutt, A. Bonneville, A shallow, chemical originfor the Marquesas swell, Geochem. Geophys. Geosyst. 1,

[7] J. Phipps Morgan, W.J. Morgan, E. Price, Hotspot melt-ing generates both hotspot volcanism and a hotspotswell?, J. Geophys. Res. 100 (1995) 8045^8062.

[8] R.N. Harris, R.P. Von Herzen, M.K. McNutt, G.Garven, K.A. Jordahl, Submarine hydrogeology of theHawaiian archipelagic apron 1. Heat £ow patterns northof Oahu and Maro reef, J. Geophys. Res. 105 (2000)21353^21369.

[9] G. Laske, J. Phipps Morgan, J.A. Orcutt, First resultsfrom the Hawaiian SWELL Pilot Experiment, Geophys.Res. Lett. 26 (1999) 3397^3400.

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

[11] E.R. Oxburgh, E.M. Parmentier, Compositional and den-sity strati¢cation in oceanic lithosphere ^ causes and con-sequences, J. Geol. Soc. Lond. 133 (1977) 343^355.

[12] T.H. Jordan, Mineralogies, densities and seismic velocities

of garnet lherzolites and their geophysical implications,in: F.R. Boyd, H.O.A. Meyer (Eds.), The Mantle Sample:Inclusions in Kimberlites and Other Volcanics 2, AGU,Washington, DC, 1979, pp. 1^14.

[13] N.M. Ribe, U.R. Christensen, The dynamical origin ofHawaiian volcanism, Earth Planet. Sci. Lett. 171 (1999)517^531.

[14] R. Katzman, L. Zhao, T.H. Jordan, High-resolution, two-dimensional vertical tomography of the central Paci¢cmantle using ScS reverberations and frequency-dependenttravel times, J. Geophys. Res. 103 (1998) 17933^17971.

[15] D.W. Caress, M.K. McNutt, R.S. Detrick, J.C. Mutter,Seismic imaging of hotspot-related crustal underplatingbeneath the Marquesas Islands, Nature 373 (1995) 600^603.

[16] A.B. Watts, U.S. ten Brink, P. Buhl, T.M. Brocher, Amultichannel seismic study of lithospheric £exure acrossthe Hawaiian-Emperor seamount chain, Nature 315(1985) 105^111.

[17] D. McKenzie, The generation and compaction of partiallymolten rock, J. Petrol. 25 (1984) 713^765.

[18] G. Ito, J. Lin, C.W. Gable, Interaction of mantle plumesand migrating mid-ocean ridges; implications for the Ga-lapagos plume-ridge system, J. Geophys. Res. 102 (1997)15403^15417.

[19] C. DeMets, R.G. Gordon, D.F. Argus, S. Stein, E¡ect ofrecent revisions to the geomagnetic reversal time scale onestimates of current plate motions, Geophys. Res. Lett. 21(1994) 2191^2194.

[20] J.P. Canales, J.J. Dan‹obeitia, R.S. Detrick, E.E.E. Hooft,R. Bartolome¤, D.F. Naar, Variations in axial morphologyalong the Gala¤pagos spreading center and the in£uence ofthe Gala¤pagos hotspot, J. Geophys. Res. 102 (1997)27341^27354.

[21] G. Ito, J. Lin, Mantle temperature anomalies along thepresent and paleoaxes of the Gala¤pagos Spreading Centeras inferred from gravity analyses, J. Geophys. Res. 100(1995) 3733^3745.

[22] R.S. Detrick, J.M. Sinton, G. Ito, J.P. Canales, M. Behn,T. Blacic, B. Cushman, J.E. Dixon, D.W. Graham, J.Mahoney, Correlated geophysical, geochemical and volca-nological manifestations of plume^ridge interaction alongthe Galapagos Spreading Center, Geochem. Geophys.Geosyst. (2002) in press.

[23] J.M. Sinton, R.S. Detrick, Seismic and Petrologic Inves-tigation of the E¡ects of Plume^Ridge Interaction at theGalapagos Spreading Center, Woods Hole Oceanogr.Inst., Woods Hole, MA, 2000, 136 pp.

[24] E.E.E. Hooft, R.S. Detrick, D.R. Toomey, J.A. Collins, J.Lin, Crustal and upper mantle structure along three con-trasting spreading segments of the Mid-Atlantic Ridge,33.5‡^35‡N, J. Geophys. Res. 105 (2000) 8205^8226.

[25] J.P. Canales, R.S. Detrick, J. Lin, J.A. Collins, D.R. Too-mey, Crustal and upper mantle seismic structure beneaththe rift mountains and across a nontransform o¡set at theMid-Atlantic Ridge (35‡N), J. Geophys. Res. 105 (2000)2699^2719.



[26] C.A. Zelt, R.B. Smith, Seismic traveltime inversion for2-D crustal velocity structure, Geophys. J. Int. 108(1992) 16^34.

[27] I. Grevemeyer, W. Weigel, C. Jennrich, Structure andageing of oceanic crust at 14‡S on the East Paci¢c Rise,Geophys. J. Int. 135 (1998) 573^584.

[28] J.S. McClain, C.A. Atallah, Thickening of the oceaniccrust with age, Geology 14 (1986) 574^576.

[29] J. Phipps Morgan, E.M. Parmentier, J. Lin, Mechanismsfor the origin of mid-ocean ridge axial topography: im-plications for the thermal and mechanical structure ofaccreting plate boundaries, J. Geophys. Res. 92 (1987)12823^12836.

[30] M.A. Eberle, D.W. Forsyth, An alternative, dynamicmodel of the axial topographic high at fast spreadingridges, J. Geophys. Res. 103 (1998) 12309^12320.

[31] B.-Y. Kuo, D.W. Forsyth, Gravity anomalies of theridge-transform system in the South Atlantic between31‡ and 34.5‡S: Upwelling centers and variations in crus-tal thickness, Mar. Geophys. Res. 10 (1988) 205^232.

[32] R.L. Parker, The rapid calculation of potential anomalies,Geophys. J. R. Astron. Soc. 31 (1972) 447^455.

[33] J. Phipps Morgan, D.W. Forsyth, Three-dimensional £owand temperature perturbations due to a transform o¡set:E¡ects on oceanic crustal and upper mantle structure,J. Geophys. Res. 93 (1988) 2955^2966.

[34] R.A. Dunn, D.R. Toomey, S.C. Solomon, Three-dimen-sional seismic structure and physical properties of thecrust and shallow mantle beneath the East Paci¢c Riseat 9‡ 30PPN, J. Geophys. Res. 105 (2000) 23463^23478.

[35] T.J. Henstock, A.W. Woods, R.S. White, The accretion ofoceanic crust by episodic sill intrusion, J. Geophys. Res.98 (1993) 4143^4161.

[36] M.M. Yale, J. Phipps Morgan, Asthenosphere £ow modelof hotspot^ridge interactions: a comparison of Icelandand Kerguelen, Earth Planet. Sci. Lett. 161 (1998) 45^56.

[37] C.H. Langmuir, E.M. Klein, T. Plank, Petrological sys-tematics of mid-ocean ridge basalts: Constraints on meltgeneration beneath ocean ridges, in: J. Phipps Morgan,D.K. Blackman, J.M. Sinton (Eds.), Mantle Flow andMelt Generation at Mid-Ocean Ridges, Geophys.Monogr. Ser. 71, AGU, Washington, DC, 1992, pp.183^280.

[38] D. McKenzie, M.J. Bickle, The volume and compositionof melt generated by extension of the lithosphere, J. Pet-rol. 29 (1988) 625^679.

[39] Y. Shen, D.W. Forsyth, Geochemical constraints on ini-tial and ¢nal depths of melting beneath mid-ocean ridges,J. Geophys. Res. 100 (1995) 2211^2237.

[40] K. Jha, E.M. Parmentier, J. Phipps Morgan, The role ofmantle-depletion and melt retention buoyancy in spread-

ing-center segmentation, Earth Planet. Sci. Lett. 125(1994) 221^234.

[41] R.E. Bell, W.R. Buck, Crustal control of ridge segmenta-tion inferred from observations of the Reykjanes Ridge,Nature 357 (1992) 583^586.

[42] J.-G. Schilling, R.H. Kingsley, J.D. Devine, Galapagoshot spot-spreading center system 1. Spatial petrologicaland geochemical variations (83‡W-101‡W), J. Geophys.Res. 87 (1982) 5593^5610.

[43] K.S. Harpp, W.M. White, Tracing a mantle plume: iso-topic and trace element variations of Gala¤pagos sea-mounts, Geochem. Geophys. Geosyst. 2,

[44] W.J. Morgan, Rodriguez, Darwin, Amsterdam,.., a sec-ond type of hotspot island, J. Geophys. Res. 83 (1978)5355^5360.

[45] N.H. Sleep, Lateral £ow of hot plume material ponded atsublithospheric depths, J. Geophys. Res. 101 (1996)28065^28083.

[46] N.M. Ribe, The dynamics of plume^ridge interaction 2.O¡-ridge plumes, J. Geophys. Res. 101 (1996) 16195^16204.

[47] J.-G. Schilling, Fluxes and excess temperatures of mantleplumes inferred from their interaction with migratingmantle plumes, Nature 352 (1991) 397^403.

[48] W.B. Moore, G. Schubert, P. Tackley, Three-dimensionalsimulations of plume^lithosphere interaction at the Ha-waiian swell, Science 279 (1998) 1008^1011.

[49] N.H. Sleep, Hotspots and mantle plumes: some phenom-enology, J. Geophys. Res. 95 (1990) 6715^6736.

[50] G. Ito, J. Lin, C.W. Gable, Dynamics of mantle £ow andmelting at a ridge-centered hotspot: Iceland and the Mid-Atlantic Ridge, Earth Planet. Sci. Lett. 144 (1996) 53^74.

[51] S. Watson, D. McKenzie, Melt generation by plumes: astudy of Hawaiian volcanism, J. Petrol. 32 (1991) 501^537.

[52] P. Wessel, W.H.F. Smith, New version of the genericmapping tools released, EOS Trans. AGU 76 (1995) 329.

[53] W.H.F. Smith, D.T. Sandwell, Global sea£oor topogra-phy from satellite altimetry and ship depth soundings,Science 277 (1997) 1957^1962.

[54] M.D. Kurz, D.J. Geist, Dynamics of the Galapagos hot-spot from helium isotope geochemistry, Geochim. Cosmo-chim. Acta 63 (1999) 4139^4156.

[55] A.E. Gripp, R.G. Gordon, Current plate velocities rela-tive to the hotspots incorporating the NUVEL-1 globalplate motion model, Geophys. Res. Lett. 17 (1990) 1109^1112.

[56] P.D. Asimow, M.M. Hirschmann, E.M. Stolper, An anal-ysis of variations in isentropic melt productivity, Phil.Trans. R. Soc. Lond. A355 (1997) 255^281.



Crustal thickness along the western Gala¤pagos Spreading ... · J. Pablo Canales a;, Garrett...

Documents

Transcript of Crustal thickness along the western Gala¤pagos Spreading ... · J. Pablo Canales a;, Garrett...