Mantle thermal structure and active upwelling during ...science.whoi.edu/labs/mclean210/kelemen/pages... · W.S. Holbrook et al./Earth and Planetary Science Letters 190 (2001) 251^266

Mantle thermal structure and active upwelling duringcontinental breakup in the North Atlantic

W. Steven Holbrook a;*, H.C. Larsen b, J. Korenaga c, T. Dahl-Jensen b,I.D. Reid b, P.B. Kelemen d, J.R. Hopper b, G.M. Kent e, D. Lizarralde f ,

S. Bernstein b, R.S. Detrick d

a Department of Geology and Geophysics, University of Wyoming, Laramie, WY 82071-3006, USAb Danish Lithosphere Centre, Òster VÖldgade 10, DK1350 Copenhagen, Denmark

c Department of Earth, Atmospheric and Planetary Sciences, Massachusetts Institute of Technology, Cambridge, MA 02139, USAd Department of Geology and Geophysics, Woods Hole Oceanographic Institution, Woods Hole, MA 02543-1542, USA

e Scripps Institution of Oceanography, University of California, San Diego, La Jolla, CA 92093-0225 USAf School of Earth and Atmospheric Sciences, Georgia Institute of Technology, 221 Bobby Dodd Way, Atlanta, GA 30332, USA

Received 11 June 2000; received in revised form 1 June 2001; accepted 2 June 2001

Abstract

Seismic reflection and refraction data acquired on four transects spanning the Southeast Greenland rifted margin andGreenlandÎceland Ridge (GIR) provide new constraints on mantle thermal structure and melting processes duringcontinental breakup in the North Atlantic. Maximum igneous crustal thickness varies along the margin from s 30 kmin the near-hotspot zone (6 500 km from the hotspot track) to V18 km in the distal zone (500^1100 km). Magmaticproductivity on summed conjugate margins of the North Atlantic decreases through time from 1800 þ 300 to 600 þ 50km3/km/Ma in the near-hotspot zone and from 700 þ 200 to 300 þ 50 km3/km/Ma in the distal zone. Comparison of ourdata with the British/Faeroe margins shows that both symmetric and asymmetric conjugate volcanic rifted marginsexist. Joint consideration of crustal thickness and mean crustal seismic velocity suggests that along-margin changes inmagmatism are principally controlled by variations in active upwelling rather than mantle temperature. The thermalanomaly (vT) at breakup was modest (V100^125³C), varied little along the margin, and transient. Data along the GIRindicate that the potential temperature anomaly (125 þ 50³C) and upwelling ratio (V4 times passive) of the Icelandhotspot have remained roughly constant since 56 Ma. Our results are consistent with a plumeîmpact model, in which(1) a plume of radius V300 km and vT of V125³C impacted the margin around 61 Ma and delivered warm material todistal portions of the margin; (2) at breakup (56 Ma), the lower half of the plume head continued to feed activelyupwelling mantle into the proximal portion of the margin; and (3) by 45 Ma, both the remaining plume head and thedistal warm layer were exhausted, with excess magmatism thereafter largely confined to a narrow (6 200 km radius)zone immediately above the Iceland plume stem. Alternatively, the warm upper mantle layer that fed excess magmatismin the distal portion of the margin may have been a pre-existing thermal anomaly unrelated to the plume. ß 2001Elsevier Science B.V. All rights reserved.

Keywords: volcanism; rift zones; refraction methods; large igneous provinces; North Atlantic

0012-821X / 01 / $ ^ see front matter ß 2001 Elsevier Science B.V. All rights reserved.PII: S 0 0 1 2 - 8 2 1 X ( 0 1 ) 0 0 3 9 2 - 2

* Corresponding author. Tel. : +1-307-766-2427; Fax: +1-307-766-6679. E-mail address: [email protected] (W.S. Holbrook).

EPSL 5898 23-7-01 Cyaan Magenta Geel Zwart

Earth and Planetary Science Letters 190 (2001) 251^266

www.elsevier.com/locate/epsl

1. Introduction

When a continent ruptures, prodigious volcan-ism can occur, producing enormous volumes ofigneous rocks in an elongate magmatic provinceextending along the entire rifted edge of the con-tinent (e.g., [1^3]). These volcanic rifted margins(VRM) are the dominant type of rifted margin inthe Atlantic (e.g., [4^6]), but the cause of theanomalous volcanism is poorly understood. Theigneous crust emplaced at VRM has three distin-guishing geophysical characteristics. First, its low-er crust has high P-wave velocities (s 7.2 km/s),suggesting compositions more ma¢c than normaloceanic lower crust [3,7]. Second, seismic re£ec-tion data over VRM show wedges of seaward-dipping re£ection sequences (SDRS, [1,8]) thatcorrespond to subaerially deposited basalt £ows[9^11]. Third, VRM crust often exceeds 20 kmin thickness ^ more than three times as thick asnormal oceanic crust, which is produced by pas-sively upwelling mantle [12] with a potential tem-perature of about 1280³C [13,14]. VRM produc-tion thus requires mantle that is anomalouslywarm [3], anomalously fertile, actively upwelling[15,16], or some combination thereof [7].

The margins of the North Atlantic have playeda key role in the development of ideas regardingVRM genesis. Breakup of Greenland from north-ern Europe about 56 Ma left behind massive se-quences of igneous rocks, including £ood basaltson Greenland, the Faeroe Islands, and Great Brit-ain (e.g., [17,18]) and SDRS that extend at least2600 km along strike (e.g., [2,4,8,19^21]). The ap-parent association of North Atlantic VRM andthe Iceland hotspot led to the hypothesis thatVRM are the expression of hot mantle temper-atures resulting from a mantle plume. Variousplume structures at VRM have been proposed,including steady-state incubating plume heads[22], large mushroom heads of impacting plumes[23,24], and rising, vertical sheets [25,26]. Alterna-tively, it has been proposed that (1) hot materialderived from plumes ponds beneath thick conti-nental lithosphere over a wide region prior to rift-ing [27], (2) a hot mantle reservoir develops belowcontinental lithosphere independently of plumes[28,29], or (3) rifting itself instigates active mantle

upwelling and a melting anomaly without a sig-ni¢cant temperature anomaly (e.g., [15,16]).

Three questions crucial to examining any of theabove hypotheses are: (1) What was the magni-tude of the mantle temperature anomaly, if any,during breakup? (2) Did active upwelling of themantle take place? And (3) how did the thermalanomaly and active upwelling vary in time andspace? (We de¢ne active upwelling in a purelykinematic sense, where the upwelling rate of man-tle through the melting zone exceeds the half-spreading rate of the rift. Passive upwelling, incontrast, is considered as simple corner £ow inwhich the upwelling rate equals the half-spreadingrate. This de¢nition of active upwelling di¡ersfrom the common usage in the mid-ocean ridgeliterature, in which retained melt is invoked toenhance buoyancy [30,31]. Because continentalrifting is a transient process, decoupling of mantle£ow from surface motion is viable; our active up-welling ratio is designed to quantify that decou-pling.) Understanding the spatial and temporalvariation of mantle temperature and active up-welling along the evolving margin would placestrong constraints on models of rift dynamics,plume^rift interaction, and rift-related convection.In particular, we need tests that might distinguishbetween models that invoke high mantle temper-atures with passive upwelling (e.g., 1550³C; [26]),no thermal anomalies but strongly active upwell-ing (e.g., [16]), and more recent models that com-bine modest mantle temperatures and active up-welling (e.g., [32,33]).

Data from the SIGMA study (SIGMA: SeismicInvestigation of the Greenland Margin) allow usto address the relative roles of temperature andactive upwelling in the North Atlantic on the ba-sis of: (1) precise crustal velocity control on tran-sects spanning an entire rifted margin, allowingmean crustal composition to be estimated;(2) systematic conjugate margin coverage at vari-able o¡set from the Iceland hotspot track; and(3) critical in situ geochemical and geochronolog-ical data, including samples from deep ocean drill-ing. The SIGMA data suggest a relatively modestthermal anomaly (100³C) and that active upwell-ing, probably associated with a rising plume head,strongly contributed to breakup magmatism near


W.S. Holbrook et al. / Earth and Planetary Science Letters 190 (2001) 251^266252

the plume track. Active upwelling continuedalong the Iceland hotspot track and contributesto melting beneath present-day Iceland. By as-suming passive upwelling, prior studies mayhave signi¢cantly overestimated mantle tempera-tures in the Iceland hotspot and along the marginduring continental breakup.

2. Experiment design and data modeling

In order to assess the temporal and spatial evo-

lution of the breakup melting anomaly, we ac-quired high-precision, crustal-scale, wide-angleseismic data along four pro¢les situated at variouso¡sets from the Iceland hotspot track (Fig. 1).Line 1 extended along the entire GreenlandÎce-land Ridge (GIR) to measure post-breakup mag-matism produced by the Iceland hotspot. Thesouthernmost pro¢le, Line 4, crossed the southerntip of Greenland, 1050 km from the hotspottrack. Pro¢les were located to provide in situ sam-ple control from drillholes into the upper crust(e.g., [9]) as well as to lie in approximately con-

Fig. 1. Location of SIGMA seismic pro¢les 1^4 and pro¢les on the conjugate Faeroe^Hatton^Rockall margins, on an anomaly18 (V40 Ma) reconstruction of the North Atlantic. Shading represents positive (light) and negative (dark) magnetic anomalies[59]; anomalies 21n and 24n are labeled. FIRE, FaeroeÎceland Ridge Experiment [34]; HB, Hatton Bank [2,20]; EB, EdorasBank [26]; Rapids, [46]. Inset shows pro¢le locations on present-day bathymetry. Circles represent ocean-bottom seismic instru-ments; short solid lines represent expanding spread pro¢les on the HB transect.


W.S. Holbrook et al. / Earth and Planetary Science Letters 190 (2001) 251^266 253

jugate positions to seismic pro¢les on the Faeroe^Rockall^Hatton margin (Fig. 1; [2,26,34]). Datawere acquired using the R/V Maurice Ewing, witha 20-element, 137-l airgun array as a source andthree sets of recording instruments: the Ewing's4.0-km-long hydrophone streamer, ocean-bottomseismic hydrophones and seismometers (64 total

stations), and Reftek portable seismometers de-ployed on Greenland and Iceland (Lines 1, 2,and 3; 20 stations total).

The resulting data set was processed intostacked multichannel seismic (MCS) sections(not reported on here) and wide-angle record sec-tions for all four pro¢les. Wide-angle data pro-

Fig. 2. P-wave velocity models for the four SIGMA transects. Horizontal and vertical scales in km. Arrows show coastlines (pro-jected on Line 4). Numbers along top of models (e.g., 24, 23, etc.) are magnetic anomalies; Bk, interpreted breakup position(100% igneous crust). Color bar shows P-velocity values in km/s. Transect 2 published as Plate 1 in Korenaga et al. [36].



cessing included instrument relocation, bandpass¢ltering (4^16 Hz), previous shot noise suppres-sion, and spiking deconvolution. P-wave velocitymodels (Fig. 2) were derived by inverting traveltimes of refractions and wide-angle re£ections re-corded on the ocean-bottom instruments and on-shore seismometers [35,36]. Tight constraints oncrustal velocities and thicknesses were providedby refractions and wide-angle re£ections observedon the ocean-bottom instruments. Vertical-inci-dence re£ections from the MCS stacks werejointly inverted with wide-angle traveltimes toconstrain sediment thickness. Data quality andseismic ray coverage of the wide-angle data areexcellent on all four pro¢les (Fig. 3; [36]). Datafrom Transect 2 are published [36] ; manuscriptsdetailing data from the other three transects are inpreparation.

3. Magmatism on the Southeast Greenland margin

3.1. Thickness of igneous crust

The wide occurrence of thick igneous crust, in-cluding high-velocity lower crust (HVLC) and athick extrusive layer with systematically seaward-dipping re£ectors (Fig. 2), con¢rms that theSoutheast Greenland margin is volcanic along itsentire length (e.g., [37]). Seaward-dipping re£ec-tions exist beneath much of Line 1, and wedgesof seaward-dipping re£ections are observed onLines 2, 3, and 4.

SIGMA-1 shows s 30-km-thick crust and low-er-crustal velocities as high as 7.4^7.5 km/s ex-tending continuously from Greenland to Iceland;a similar crustal structure is reported beneath Ice-land [34,38]. Data from the westernmost seismicreceivers indicate that the HVLC extends at leastto 33 km depth, and possibly to 40 km, beneaththe western end of the line (Fig. 2). Gravity mod-eling also shows landward crustal thickening nearthe landward end of SIGMA-1 (Reid et al., inpreparation), and a similarly large crustal thick-ness is reported from a conjugate pro¢le o¡ theFaeroes [34]. However, lack of seismic data fromonshore Greenland makes the magnitude andcomposition of such crustal thickening an unre-

solved question, and unless otherwise stated, weassume a maximum crustal thickness of 33 kmalong SIGMA-1. Maximum crustal thicknessalong SIGMA-2, -3, and -4 is less than on SIG-MA-1, though still close to 30 km at SIGMA-2.Unlike SIGMA-1, all three southern pro¢les showrapidly seaward decreasing crustal thickness.

The velocity structures (Fig. 2) of SIGMA-2,-3, and -4 de¢ne a compositional change fromcontinental to oceanic crust over a 50- to70-km-wide transition zone. In this transitionzone, lower-crustal P-wave velocities increasefrom 6.5 to 6.6 km/s below the continent to wellover 7 km/s within the HVLC, and upper-crustalvelocities decrease from 6.0 to 6.3 in the continen-tal crust to 4^5 km/s o¡shore. The transition zoneand its seaward end (the continentôcean bound-ary, COB) are independently de¢ned by drillingand deep seismic re£ection data along SIGMA-3[9,39] to lie about 20 km landward of the shelfbreak [39]. On SIGMA-1 the lack of an upper-crustal lateral velocity gradient indicates that theCOB is located landward of the western end ofSIGMA-1, i.e., close to the coast, consistent withcoastal exposures of sheeted dyke complexes andlayered gabbros within a deformed, gneissic hostrock [40,41]. Except for SIGMA-3, the location ofthe COB (Fig. 2) is 180 þ 10 km landward of thewell-de¢ned magnetic anomaly 21n (Fig. 1) on allSIGMA pro¢les. The age of the COB at SIGMA-1 and 3 is determined through combined 40Ar/39Ar radiometric age determinations and magne-tostratigraphy to follow very shortly after mag-netic Chron 25n between 56 and 55.5 Ma[40,42]. Ages seaward of the COB are obtainedfrom sea£oor spreading anomalies (Fig. 1).

The variation of magmatism in time and spaceshows a progressive concentration of magmatismtoward the Iceland hotspot track (Fig. 4). On allfour transects, peak magmatism occurred duringthe earliest part of breakup, very close to theCOB; on the three southern transects, magmatismdeclined signi¢cantly with time. Two zones arede¢ned: a hotspot-proximal zone (SIGMA-1)with consistently thick (s 30 km) crust, and adistal zone (SIGMA-3 and -4) with more modestpeak volcanism (18 km) that thins rapidly to 9 km.SIGMA-2 occupies a transition between these



Fig. 3. Ray coverage in the igneous crust on all four pro¢les. Rays drawn by two-point ray tracing to observed travel time picks.Transect 1 (top) shown as ray density, with darker shades denoting greater ray density. Transect 2 shown as derivative weightsum [36], with lighter shades denoting greater ray density. Transects 3 and 4 show individual rays (solid lines), bottoming pointsof diving rays (dots), bottoming points of re£ections (dots or lines).



two zones: its maximum crustal thickness (V30km) is initially close to that of SIGMA-1 butdecreases over 10^12 Myr to only 8^9 km, similarto SIGMA-3 and -4 (Fig. 4). In the distal zone,crustal thickness decays to about 9 km, ¢rst (51Ma) on SIGMA-4, and slightly later (49 Ma) onthe more proximal SIGMA-3. Thus, the initiallywide melting anomaly (s 1200 km) became re-stricted to the GIR in only 6 Myr [43].

3.2. Conjugate margin structure

Comparison of three of the SIGMA pro¢leswith pro¢les in approximately conjugate positionson the British/Faeroe margin shows evidence ofboth symmetric and asymmetric volcanic marginsegments. Crustal thickness variations are sym-metric, within error, on two of the three conjugatemargin^pro¢le pairs: SIGMA-4/Edoras Bank [26]and SIGMA-1/FIRE [34,44] (Fig. 4). In contrast,the SIGMA-3 pro¢le shows a 150-km-wide zoneof thick igneous crust, in contrast to a 50-km-wide zone on the conjugate Hatton Bank pro¢le[19,20] (Fig. 4). Indeed, Larsen and Saunders [39]previously noted an unusually high half-spreadingrate near SIGMA-3, and regional studies [45] alsosuggest asymmetry along this part of the margin.An apparent asymmetry in the timing of peakmagmatism (Fig. 4) may be illusory, because mag-netic anomaly identi¢cation is uncertain at thatlatitude seaward of Hatton Bank.

This contrast in conjugate structure impliesasymmetric spreading at the latitude of SIGMA-3; major ridge jumps after spreading are ruled outby the lack of dip reversals in SDRS on the pro-¢le. Two possible explanations would circumventthe conclusion of asymmetric conjugate marginstructure. First, the SIGMA-3 and Hatton Bankpro¢les may not be perfectly conjugate, and theirtrue conjugates might show symmetric structures.This would imply a major transform structureand strong magmatic segmentation along themargin, which is not evident in bathymetric orpotential ¢eld data. Second, if existing interpreta-tions of the Hatton/Edoras Bank as continentalcrust are incorrect, the missing magmatic materialmight be located there. Existing seismic modelshave limited resolution beneath the axis of the

Hatton/Edoras Bank; wide-angle lines shot thereto date end on the axis and thus do not under-shoot the Bank [20,26,46]. Nevertheless, the lim-ited existing seismic data, including an axial ex-panding spread pro¢le [19], suggest continentalcrust, and interpreting the Hatton/Edoras Bankas Cenozoic igneous crust would create an asym-metry at the latitude of SIGMA-4/Edoras Bank.

3.3. Magmatic productivity through time

The SIGMA data allow accurate estimates of

Fig. 4. (Top) Igneous crustal thickness on SIGMA lines plot-ted as a function of distance from the hotspot track (Line1). Labeled curves show igneous thickness at di¡erent times(56 Ma, 53 Ma, etc). (Bottom) Igneous crustal thickness onSIGMA transects (solid lines) and pro¢les on conjugate mar-gins (dashed and dotted lines), plotted as a function of time.Breakup is assumed to occur at 56 Ma. Dashed lines FIR,FaeroeÎceland Ridge [34]; EB, Edoras Bank [26]; HB, Hat-ton Bank [2,20]. Dotted line is Rapids pro¢le [46].



magmatic productivity through time in the NorthAtlantic (Fig. 5). Although the SIGMA-3/Hattonasymmetry is clear (Fig. 5), when conjugate mar-gin pairs are summed, magmatic productivity onthe distal transects (SIGMA-3/Hatton and SIG-MA-4/Edoras) is identical. Magmatic productivityis two to three times greater in the proximal zone(V1800 þ 300 km3/km/Ma) than in the distal zone(V700 þ 200 km3/km/Ma) during breakup andtwice as great after 42 Ma (600 þ 50 km3/km/Mavs. 300 þ 50 km3/km/Ma). Along the distal half ofthe margin, magmatic productivity is virtually in-dependent of distance from the hotspot track. De-creasing crustal thickness (except on the SIGMA-1/FIRE pair) as well as decreasing spreading ratescontribute to the marked decrease in productivitywith time (Fig. 5). Errors in productivity comeprincipally from uncertainties in magnetic anom-aly correlation and are greatest on Line 1 andduring breakup on Line 2. The major features

of Fig. 5, namely, the strong decrease in produc-tivity through time on all transects and the signif-icantly higher productivity in the near-hotspotzone, are robust.

The total volume of thick igneous crust pro-duced along the conjugate margins from (and in-cluding) the Iceland hotspot track to south ofGreenland between breakup (56 Ma) and anom-aly 24n (53 Ma) is 3.7U106 km3 ; 5.1U106 km3

was produced between breakup and anomaly 23n(51 Ma). Including the margins north of Iceland(e.g., [4]), 9.6U106 km3 of Northeast AtlanticVRM igneous crust formed between breakupand anomaly 23n. This estimate is about 50%greater than reported by Eldholm and Grue [4].SIGMA pro¢les and their conjugate counterpartsjointly suggest that as much as 70% of this vol-ume lies within the hotspot-proximal zone extend-ing (south of Iceland) to a maximum of 500 kmfrom the center of the hotspot track.

Fig. 5. Magmatic productivity (km3/km/Ma) through time on individual pro¢les (top) and summed conjugate margins (bottom);SIGMA-2 is assumed to be symmetric with its unsurveyed conjugate. Conjugate Transects III-HB show strong asymmetry in pro-ductivity values, but when summed together, show virtually identical productivity to the more symmetric IV-EB conjugate pair.Conjugate margins are assumed to be symmetric at SIGMA Transect II, which lacks a conjugate pro¢le.



4. Melting anomaly: hot or actively upwellingmantle?

Excess magmatism may result from enhancedmantle temperature or active mantle upwelling:higher temperatures increase the fraction of par-tial melting in the mantle, while active upwellingprovides an enhanced £ux of mantle through themelting zone. Simple translation of crustal thick-ness into mantle temperature anomaly may there-fore overestimate the temperature by not account-ing for active upwelling [7]. However, for a givencrustal thickness these two mechanisms producecrust of di¡erent major element composition, be-cause the average pressure (P) and fraction (F) ofmelting di¡er [13]. Higher potential temperaturedeepens the solidus and thus increases the heightof the melting column, resulting in higher mean Pand F and, therefore, melts enriched in MgO andpoorer in SiO2 (e.g., [13]). Active upwelling, incontrast, achieves a given crustal thickness withrelatively lower P and F, thus producing meltswith lower MgO and higher SiO2. Such crustal-scale compositional di¡erences would a¡ect themean seismic velocity of the crust [7] and mighttherefore be independently constrained with pre-cise crustal velocity control.

We aim to achieve this by comparing observedcrustal thickness and mean crustal velocity on theSIGMA transects with the expected velocity^thickness systematics for a common mantle melt-ing model. We calculate the expected crustalthickness (Hc) and mean crustal velocity for agiven mantle potential temperature and upwellingratio (ratio of upwelling £ux to spreading half-rate) in the following way. First, we calculatemelt fraction at a given pressure along a decom-pression melting trend at a given potential tem-perature (Fig. 6) using the adiabatic, equilibriummelting model of McKenzie and Bickle [14]. Thismodel assumes a constant entropy change onmelting of 250 J/kg/K and neglects the e¡ect ofconductive cooling during rifting [47]. We thenuse equation 2 of Kelemen and Holbrook [7] todetermine mean crustal Vp given the mean pres-sure and fraction of melting. We assume that allmelt is emplaced into the crust and that this is,apart from volatiles, a closed system, with mean

crustal composition providing a true representa-tion of the original melt.

Observed mean crustal velocity is calculatedfrom the velocity models (Fig. 2) and correctedto a temperature of 400³C and con¢ning pressureof 600 MPa by applying a pressure correction of0.00022 km/s/MPa [48] and a temperature correc-tion of 30.0005 km/s/³C. Temperatures are calcu-lated for a conductive geotherm from 10³C at thesea£oor to 750³C at 40 km depth. All crustal ve-locities less than 6.85 km/s are assumed to bea¡ected by alteration or porosity and thereforeunrepresentative of bulk composition; these ve-locities are replaced by 6.85 km/s before applyingtemperature and pressure corrections. This re-placement velocity corresponds, using equation 1of Kelemen and Holbrook [7], to the representa-tive composition of seaward-dipping wedge ba-salts sampled on ODP Leg 152 (MgO = 7.5%;SiO2 = 51%).

These calculations were done for passive up-welling and various extents of active upwelling;the results are referred to below as MB/KH. For

Fig. 6. Predicted mean crustal velocity vs. crustal thicknesssystematics for two melting models: the Langmuir et al. pas-sive upwelling model (thin line) and the McKenzie andBickle model, assuming passive upwelling (bold line) and ac-tive upwelling ratios (X) of 2, 4, and 8. Dashed lines repre-sent mantle potential temperatures for the McKenzie andBickle model. OC represents the calculated mean crustal ve-locity of oceanic crust near the East Paci¢c Rise [49]. Largearrows represent the isolated e¡ects of increased temperature,which causes increases in both mean crustal velocity andcrustal thickness, and increased active upwelling, whichcauses increasing crustal thickness without a signi¢cantchange in mean crustal velocity.



comparison, we also used melt compositions forpassive mantle upwelling calculated from theLangmuir et al. fractional melting model [30] tocalculate mean crustal Vp. Both the Langmuirand MB/KH models show the same fundamentaltrend: at a given upwelling ratio, increasing man-tle potential temperature causes strong increasesin both crustal thickness and velocity (e.g., [3]).However, increasing upwelling ratio at constantpotential temperature produces thicker crustwith only a slight increase in mean velocity;thus the two variables, temperature and upwellingratio, produce signi¢cantly di¡erent trends in theVp^Hc diagram (Fig. 6). The choice of meltingmodel does not a¡ect these trends but does a¡ectthe estimate of absolute temperature: for a givencrustal thickness, the fractional melting model ofLangmuir et al. [30] predicts mantle potential tem-peratures about 50³C warmer than the MB/KHpassive upwelling model. However, temperatureanomalies remain essentially the same. The in-crease in melt volume is not linearly proportionalto upwelling ratio [7], because the ¢nal depth ofmelting is limited by the resultant crustal thick-ness. The slight increase in average crustal veloc-ity re£ects higher mean P induced by a deeper¢nal depth of melt segregation ^ the `lid' e¡ectof thicker crust.

To test our approach and establish a referencemantle temperature, we applied our method todata from 1-Myr-old, 5.5-km-thick crust at theEast Paci¢c Rise [49]. We applied the same pres-sure and temperature corrections to the EPR dataas to the SIGMA data, except that we assumed anin situ temperature of 500³C at the Moho (6 kmsubsea£oor depth [50]). The calculated mean crus-tal velocity of 6.96 km/s at the East Paci¢c Rise isconsistent, within error, with the predicted valuefor passively upwelling mantle with a potentialtemperature of 1300³C (OC, Fig. 6). Accordingly,we report deviations in mantle temperatures (vT)relative to a `normal' temperature of 1300³C.

Data from all SIGMA pro¢les at breakup plotin the area of above-normal mantle temperature(Fig. 7). On the three southern transects, the ve-locity/thickness data evolve with time toward low-er mantle temperature. Data from SIGMA-3 andSIGMA-4 lie close to the MB/KH passive upwell-

ing curve, whereas SIGMA-1 and SIGMA-2 atbreakup lie well in the ¢eld of active upwelling(Fig. 7). We conclude that: (1) a margin-widemantle thermal anomaly was present duringbreakup but was exhausted beneath most of themargin within 6^12 Myr; (2) active upwelling atV4^6 times the passive rate (or a process withsimilar e¡ect) is required near the hotspot trackbut not in the distal zone, which can be modeledby purely passive upwelling (at vT = 100 þ 50³C)or with only modestly active upwelling (two timespassive); and (3) after breakup, temperatures re-main elevated (vT V100 þ 50³C) on the hotspottrack compared to elsewhere (vT = 25 þ 50³C). Apossible along-margin variation in mantle temper-ature during initial breakup cannot be resolved byour data (although temperatures appear some-what low on Transect 2), and a uniformvT = 100 þ 50³C along the entire margin is permis-sible. However, if 40-km-thick igneous crust ispresent at SIGMA-1 (Fig. 2), higher, but rapidlydecreasing, temperature (max. vT = 200 þ 50³C) in

Fig. 7. Mean crustal velocity and crustal thickness data forthe four SIGMA pro¢les (I, squares; II, circles; III, dia-monds; IV, triangles) and for oceanic crust near the East Pa-ci¢c Rise (EPR; [49]). Representative error bars are shownfor thick- and thin-crust portions of each line. Melting modelsystematics as in Fig. 6. Color coding of symbols representscrustal age: yellow = breakup (56 Ma) to anomaly 24 (52.9Ma); red = anomaly 24 to anomaly 23 (51.2 Ma); darkblue = anomaly 23 to anomaly 22 (49.3 Ma); green = anomaly22 to anomaly 21 (47.1 Ma); light blue = anomaly 21 toanomaly 20 (43.1 Ma); dark gray = post-anomaly 20. Lightgray boxes correspond to alternative Line I model with thickunderplate (gray in Fig. 2).



the proximal zone during earliest breakup is indi-cated (Fig. 7). If a mantle component more fertilethan periodititic mantle, e.g., recycled oceaniccrust, is present in the Iceland mantle plume, atconstant vT it would melt to higher degrees thannormal mantle, and therefore increase the appar-ent ratio of active upwelling, though this e¡ect islikely small.

To test the robustness of these conclusions, weconducted extensive error testing of all four pro-¢les to determine ranges of acceptable velocitiesand crustal thicknesses. Models were tested bysystematically varying crustal velocity in startingmodels and ¢nding best-¢t models by invertingfor depths to the top of the lower crust and theMoho. Traveltime mis¢t statistics indicate an un-certainty in mean crustal velocity of þ 0.08 km/sin thick crust and þ 0.05 km/s in thin crust. (Eventhough individual model parameters may havelarger errors, the error in mean crustal velocityis lower, since errors in model parameters areoften negatively correlated. That is, increasingthe velocity at a lower-crustal node will requirea corresponding decrease in an upper-crustalnode in order to conserve PmP travel times;mean crustal velocity is thus little a¡ected.) Kore-naga et al. [36] conducted an extensive MonteCarlo error analysis for Line 2 and determined asimilar uncertainty. Acceptable models have re-stricted slopes in Vp^Hc space: all acceptablemodels for Line 2, for example, have a £atterVp^Hc slope than the passive upwelling curves,implying a decreasing upwelling ratio with time.Forcing SIGMA-1 Vp^Hc data onto the MB/KHpassive upwelling curve would require a mean Vp

greater than 7.3 km/s, well outside the permissiblevelocity ¢eld. Similarly, SIGMA-3 and -4 cannotbe modeled with constant potential temperature.(The only conceivable way of raising the meancrustal velocity on SIGMA-1 to values consistentwith passive upwelling would be to postulate a 15-km-thick layer of cumulate olivine of velocity 8.0km/s; the implied crustal thickness would be 45km, much greater than any published estimate,and the implied vT would be s 350³C, which isinconsistent with the chemistry of Iceland lavas[13].)

5. Implications for volcanic margin genesis

The SIGMA data show that the thermal anom-aly associated with the breakup of Greenlandfrom Europe was modest (V100 þ 50³C), rela-tively uniform along the margin, and transient.The thick crust (s 30 km) and relatively lowmean crustal P-velocity (V7.1 km/s) along theentire Iceland plume track suggests that active up-welling occurs above the present-day Icelandplume, and that a similar process has occurredduring the entire history of the plume, at leastsince breakup at V56 Ma. The active upwellingassociated with plume passage through the deepermantle thus continues into the shallow levels ofdecompression melting within extensional rifts.

The requirement for an Iceland melting anom-aly at early breakup implies that acceptable mod-els for the formation of the Northeast AtlanticVRM must include the proto-Icelandic plume.Two types of pure plume models are consistentwith our ¢ndings: the low-viscosity, thin plumehead of Sleep [51] and Larsen and Saunders [39]and the plume incubation models, in which theIceland plume ponded beneath the Greenlandcontinental lithosphere long before rifting. In ad-dition, models in which the entire margin wasunderlain by a thermal anomaly due to continen-tal insulation (e.g., [52]), with a additional activeupwelling provided by the Iceland hotspot, areequally consistent with our data. However, thecontrast between largely amagmatic Cretaceousrifting (e.g., [53]) and widespread, virtually iso-chronous basaltic magmatism at 61 þ 1.5 Ma(e.g., [42,54]) strongly suggests that the thermalconditions of the asthenospheric mantle changedas a result of a sudden plume impact rather thanlong-standing plume incubation.

Regardless of the origin of the thermal anom-aly, the pattern of active upwelling provided byour data yields insight into the evolution of melt-ing dynamics during continental breakup in theNorth Atlantic. If we adopt the plume impactmodel, then the SIGMA data provide a map ofthe evolution of the impacting plume head. Wesuggest that the region of strongly active upwell-ing marks the extent of the rising plume head,where a vertical component of motion in upwell-





ing plume material increases £ux through themelting zone. At breakup, the actively upwellingpart of the plume extended at least to SIGMA-2but not as far south as SIGMA-3. The risingplume head at breakup is thus constrained to bebetween 200 and 500 km in radius.

The distal parts of the margin require little orno active upwelling. In these regions, warm mate-rial, possibly originating from the plume head,was laterally emplaced (e.g., [27]) but lacked avertical component of motion in excess of thepassive upwelling velocity. Distal-zone magma-tism was enhanced by excess temperature(V100³C) but still fell short of magmatismin the proximal zone, where strongly activeupwelling occurred. The modest temperatureanomaly appears to con£ict with picritic lavas(up to s 20% MgO) on the distal margins,which have been interpreted to represent highdegree melts from a high temperature source(vT = 250³C; e.g., [54]). However, the downwardincrease in crustal velocities on SIGMA-3 from7.1 km/s to 7.4 km/s indicates that the high-MgO picrites most likely represent lower-crustalcumulate material rather than primary melts.Low-MgO picrites and olivine basalts (12^14%MgO) sample near SIGMA-3 ¢t the mean Vp

data and therefore are likely to represent the pri-mary, unfractionated melt compositions.

A scenario consistent with our results is as fol-lows (Fig. 8): A plume of radius V300 km andvT of V125³C impacted the margin around 61Ma and laterally delivered warm material from itsupper portions to more distal portions of the mar-gin (e.g., [27]). At breakup (56 Ma), the lower halfof the plume head continued to feed active up-welling into the proximal portion of the margin.By 45 Ma, both the remaining plume head andthe distal warm layer were exhausted, and excessmagmatism was largely con¢ned to a narrow

(6 200 km radius) zone immediately above theIceland plume stem.

This model of VRM genesis di¡ers from pre-vious models in several important respects. First,the inclusion of active upwelling provides a mech-anism for enhancing volcanism with only modestthermal anomalies. Barton and White [26], lack-ing constraints on active upwelling, interpreted amantle potential temperature of 1550³C(vT = 250³C) for the 30-km-thick crust on theFaeroeÎceland Ridge and about 100³C coolerfor the thinner crust far to the south. In contrast,our inference of an active upwelling ratio of V4yields a temperature anomaly of only 125 þ 50³Cfor the 30^33-km-thick crust beneath the conju-gate GIR and little or no along-margin temper-ature variation (Fig. 7). This inferred cooler tem-perature is consistent with inversion of traceelement concentrations from drill cores alongSIGMA-3, which suggest a mantle thermal anom-aly of about 100³C [55]. Active upwelling andenhanced temperature were transient along SIG-MA-2 but constant along the entire GIR. Wetherefore reject the notion of a margin-wide melt-ing anomaly caused solely by active mantle up-welling in response to rifting, without a temper-ature anomaly (e.g., [16]). On the contrary,strongly active upwelling, where indicated byour data, is hotspot-related. Our results agreewith the recent geodynamical models of Keenand Boutillier [33], who inferred a thin, moder-ately hot (1450³C) layer beneath the margin andtransient active upwelling during breakup.Although Keen and Boutillier infer active upwell-ing in the distal region, where we infer nearlypassive upwelling, our data are compatible withmodestly active upwelling there (Fig. 7). Modelsof sheet-like upwellings that impact the margin[26] are inconsistent with our data, since suchmodels would predict strongly active upwelling

6Fig. 8. Conceptual cartoon of a plumeîmpact scenario consistent with our results. Prior to breakup, a rising plume head(vT = 125³C) impacts the continental lithosphere and spreads laterally to emplace warm material beneath the continent. At break-up (55 Ma), strongly active upwelling (white arrows) occurs above remnant plume head and plume stem, building 30-km-thickigneous crust in the hotspot-proximal region, while passively upwelling mantle yields 18-km-thick igneous crust in the distal re-gion. Finally, at 45 Ma, the warm mantle reservoir has been exhausted, and excess magmatism is strongly focused above theplume stem. An alternative, steady-state hotspot scenario, which is allowable by our data, would require a pre-existing warmupper mantle unrelated to the plume.



along whole margin. Finally, numerical models ofthe present-day Iceland plume range from thosethat include active upwelling but overestimatecrustal thickness [56,57] to others [58] that matchcrustal thickness data with a hot (1530³C), narrowplume (radius 100 km). Our results suggest thatfuture numerical models must incorporate meancrustal velocity as a constraint on the relativeroles of active upwelling and enhanced tempera-tures.

An important result of our work is the demon-stration that active upwelling in the melting zoneoccurred both during continental breakup and,later, above the steady-state Iceland plume. Pre-vious estimates of mantle potential temperatureanomalies from crustal thickness have neglectedthis. Thus, incorporation of active upwelling ex-plains the apparently paradoxical result that thebreakup-related, thick igneous crust in the NorthAtlantic has a 50% greater volume than previ-ously thought, despite a mantle thermal anomalythat is only half as large as in most previous mod-els.

Acknowledgements

We thank the captain and crew of the R/V Ew-ing for their hard work during a long and di¤cultcruise. We thank B. Brandsdottir, R.S. White andthe Cambridge team for deployment of seismom-eters on Iceland. The data quality owes much tothe ¢eld e¡orts of Jim Dolan, David DuBois, RobHandy, Bob Busby, Paul Henkart, Anders Brun,Thomas Nielsen, and Bill Koperwhats. This workwas supported by National Science FoundationGrant OCE-9416631.[RV]

References

[1] J.C. Mutter, M. Talwani, P.L. Sto¡a, Origin of seaward-dipping re£ectors in oceanic crust o¡ the Norwegian mar-gin by `subaerial sea-£oor spreading', Geology 10 (1982)353^357.

[2] R.S. White, G.D. Spence, S.R. Fowler, D.P. McKenzie,G.K. Westbrook, A.N. Bowen, Magmatism at rifted con-tinental margins, Nature 330 (1987) 439^444.

[3] R. White, D. McKenzie, Magmatism at rift zones: Thegeneration of volcanic continental margins and £ood ba-salts, J. Geophys. Res. 94 (1989) 7685^7729.

[4] O. Eldholm, K. Grue, North Atlantic volcanic margins:Dimensions and production rates, J. Geophys. Res. 99(1994) 2955^2968.

[5] T.P. Gladczenko, K. Hinz, O. Eldholm, H. Meyer, S.Neben, J. Skogseid, South Atlantic volcanic margins,J. Geol. Soc. Lond. 154 (1997) 465^470.

[6] W.S. Holbrook, P.B. Kelemen, Large igneous province onthe U.S. Atlantic margin and implications for magmatismduring continental breakup, Nature 364 (1993) 433^436.

[7] P.B. Kelemen, W.S. Holbrook, Origin of thick, high-ve-locity crust along the US East Coast margin, J. Geophys.Res. 100 (1995) 10077^10094.

[8] K. Hinz, A hypothesis on terrestrial catastrophes wedgesof very thick oceanward dipping layers beneath passivemargins, Geol. Jahrb. 22 (1981) 5^28.

[9] H.C. Larsen, A.D. Saunders, P. Clift (Eds.), Proc. ODP,Init. Rep., 152 (Pt. 1), Ocean Drilling Program, CollegeStation, TX, 1994.

[10] D.G. Roberts, J. Backman, A.C. Moarton, J.W. Murray,J.B. Keene, Evolution of volcanic rifted margins Synthesisof Leg 81 results on the west margin of Rockall Plateau,Init. Rep. Deep Sea Dril. Proj. 81 (1984) 883^911.

[11] O. Eldholm, J. Thiede, O.L.S. Party, Formation of theNorwegian Sea, Nature 319 (1986) 360^361.

[12] The MELT Seismic Team,, Imaging the deep seismicstructure beneath a mid-ocean ridge: The MELT experi-ment, Science 280 (1998) 1215^1218.

[13] E.M. Klein, C.H. Langmuir, Global correlations of oceanridge basalt chemistry with axial depth and crustal thick-ness, J. Geophys. Res. 92 (1987) 8089^8115.

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

[15] C.E. Keen, C. Beaumont, Geodynamics of rifted conti-nental margins, in: M.J. Keen, G.L. Williams (Eds.),Geology of the Continental Margin of Eastern Canada,The Geology of North America, v. I-1, Geological Societyof America, 1990, pp. 391^472.

[16] J.C. Mutter, W.R. Buck, C.M. Zehnder, Convective par-tial melting 1. A model for the formation of thick basalticsequences during the initiation of spreading, J. Geophys.Res. 93 (1988) 1031^1048.

[17] W.A. Deer, Tertiary igneous rocks between ScoresbySund and Kap Gustav Holm, East Greenland, in: A.Escher, W.S. Watt (Eds.), Geology of Greenland,GrÖnlands Geologiske UndersÖgelse, Copenhagen, 1976,pp. 404^429.

[18] R. Waagstein, Structure, composition and age of the Fae-roe basalt plateau, in: A.C. Morton, L.M. Parson, (Eds.),Early Tertiary Volcanism and the Opening of the NEAtlantic, Geol. Soc. Spec. Publ. 39, Blackwell Scienti¢c,Oxford, 1988, pp. 225^238.

[19] S.R. Fowler, R.S. White, G.D. Spence, G.K. Westbrook,



The Hatton Bank continental margin ^ II. Deep structurefrom two-ship expanding spread seismic pro¢les, Geo-phys. J. Int. 96 (1989) 295^309.

[20] J.V. Morgan, P.J. Barton, R.S. White, The Hatton Bankcontinental margin ^ III. Structure from wide-angle OBSand multichannel seismic refraction pro¢les, Geophys. J.Int. 98 (1989) 367^384.

[21] J.C. Mutter, M. Talwani, P.L. Sto¡a, Evidence for a thickoceanic crust adjacent to the Norwegian margin, J. Geo-phys. Res. 89 (1984) 483^502.

[22] 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.

[23] R.S. White, Initiation of the Iceland plume and openingof the North Atlantic, in: A.J. Tankard, H.R. Balkwill(Eds.), Extension Tectonics and Stratigraphy of the NorthAtlantic Margins, AAPG Memoir 46, Tulsa, OK, 1989,pp. 149^154.

[24] I.H. Campbell, R.W. Gri¤ths, Implications of mantleplume structure for the evolution of £ood basalts, EarthPlanet. Sci. Lett. 99 (1990) 79^93.

[25] R.S. White, D. McKenzie, Mantle plumes and £ood ba-salts, J. Geophys. Res. 100 (1995) 17543^17585.

[26] A.J. Barton, R.S. White, Crustal structure of the EdorasBank margin and mantle thermal anomalies beneath theNorth Atlantic, J. Geophys. Res. 102 (1997) 3109^3129.

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

[28] D.L. Anderson, Superplumes or supercontinents?, Geol-ogy 22 (1994) 39^42.

[29] M. Gurnis, Large-scale mantle convection and the aggre-gation and dispersal of supercontinents, Nature 332(1988) 695^699.

[30] C.H. Langmuir, E.M. Klein, T. Plank, Petrological con-straints on melt formation and migration beneath mid-ocean ridges, in: J. Phipps Morgan, D. Blackman, J.L.Sinton (Eds.), Mantle Flow and Melt Generation at Mid-Ocean Ridges (Geophys. Monogr.h 71), American Geo-physical Union, Washington, DC, 1992, pp. 183^280.

[31] D.L. Turcotte, J. Phipps Morgan, The physics of magmamigration and mantle £ow beneath a mid-ocean ridge, in:J. Phipps Morgan, D. Blackman, J.L. Sinton, (Eds.),Mantle Flow and Melt Generation at Mid-Ocean Ridges(Geophys. Monogr. 71), American Geophysical Union,Washington, DC, 1992, pp. 155^182.

[32] R.R. Boutilier, C.E. Keen, Small-scale convection anddivergent plate boundaries, J. Geophys. Res. 104 (1999)7389^7403.

[33] C.E. Keen, R.R. Boutilier, The interaction of rifting andhot horizontal plume sheets at volcanic margins, J. Geo-phys. Res. 105 (2000) 13375^13387.

[34] J.R. Smallwood, R.K. Staples, K.R. Richardson, R.S.White FIRE Working Group,, Structure and developmentof the crust above the Iceland mantle plume: from con-

tinental rift to oceanic spreading center, J. Geophys. Res.104 (1999) 22885^22902.

[35] C.A. Zelt, R.B. Smith, Seismic traveltime inversion for 2-D crustal velocity structure, Geophys. J. Int. 108 (1992)16^34.

[36] J. Korenaga, W.S. Holbrook, G.M. Kent, P.B. Kelemen,R.S. Detrick, H.-C. Larsen, J.R. Hopper, T. Dahl-Jensen,Crustal structure of the Southeast Greenland margin fromjoint refraction and re£ection seismic tomography,J. Geophys. Res. 105 (2000) 21591^21614.

[37] H.C. Larsen, S. Jakobsdottir, Distribution, crustal prop-erties and signi¢cance of seawards-dipping sub-basementre£ectors o¡ E Greenland, in: A.C. Morton, L.M. Parson(Eds.), Early Tertiary Volcanism and the Opening of theNE Atlantic, Geological Society Special Publication no.39, Blackwell Scienti¢c, Oxford, 1988, pp. 95^114.

[38] R.K. Staples, R.S. White, B. Brandsdottir, W.H. Menke,P.K.H. Maguire, J. McBride, FaroeÎceland Ridge experi-ment ^ I. The crustal structure of north-eastern Iceland,J. Geophys. Res. 102 (1997) 7867^7886.

[39] H.C. Larsen, A.D. Saunders, Tectonism and volcanism atthe SE Greenland rifted margin: A record of plume im-pact and later continental rupture, in: A.D. Saunders,H.C. Larsen, S. Wise (Eds.), Proc. ODP, Sci. Res. 152,Ocean Drilling Program, College Station, TX, 1998, inpress.

[40] C. Tegner, R.A. Duncan, S.B. Bernstein, C.K. Brooks,D.K. Bird, M. Storey, 40Ar/39Ar geochronology of Ter-tiary ma¢c intrusions along the East Greenland riftedmargin: Relation to £ood basalts and the Iceland hotspottrack, Earth Planet. Sci. Lett. 156 (1998) 75^88.

[41] J.A. Karson, C.K. Brooks, Structural and magmatic seg-mentation of the Tertiary East Greenland volcanic riftedmargin, in: P. Ryan, C. MacNiocaill (Eds.), ContinentalTectonics, Geological Society of London, Special Publica-tion, London, 1999.

[42] M. Storey, R.A. Duncan, A.K. Pedersen, L.M. Larsen,H.C. Larsen, 40Ar/39Ar geochronology of the WestGreenland Tertiary volcanic province, Earth Planet. Sci.Lett. 160 (1998) 569^586.

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

[44] K.R. Richardson, J.R. Smallwood, R.S. White, D.Snyder, P.K.H. Maguire, Crustal structure beneath theFaroe Islands and the FaroeÎceland Ridge, Tectono-physics 300 (1998) 159^180.

[45] R.D. Mu«ller, W.R. Roest, J.-Y. Royer, Asymmetric sea-£oor spreading caused by ridge^plume interactions, Na-ture 396 (1998) 455^459.

[46] U. Vogt, J. Makris, B.M. O'Reilly, F. Hauser, P.W.Readman, A.W.B. Jacob, P.M. Shannon, The Hatton Ba-sin and continental margin: Crustal structure from wide-angle seismic and gravity data, J. Geophys. Res. 103(1998) 12545^12566.

[47] J.W. Bown, R.S. White, The e¡ect of ¢nite duration ex-



tension on melt generation at rifted continental margins,J. Geophys. Res. 100 (1995) 18011^18029.

[48] N.I. Christensen, W.D. Mooney, Seismic velocity struc-ture and composition of the continental crust: A globalview, J. Geophys. Res. 100 (1995) 9761^9788.

[49] J.P. Canales, R.S. Detrick, S. Bazin, A.J. Harding, J.Orcutt, O¡-axis crustal thickness across and along theEast Paci¢c Rise within the MELT area, Science 280(1998) 1218^1221.

[50] C.A. Stein, S. Stein, A model for the global variation inoceanic depth and heat £ow with lithospheric age, Nature359 (1992) 123^129.

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

[52] D.L. Anderson, The sublithospheric mantle as the sourceof continental £ood basalts; the case against the continen-tal lithosphere and plume head reservoirs, Earth Planet.Sci. Lett. 123 (1994) 269^280.

[53] J.A. Chalmers, T.C.R. Pulvertaft, F.G. Christiansen, H.C.Larsen, K.H. Laursen, T.G. Ottesen, The southern WestGreenland continental margin; rifting history, basin de-velopment, and petroleum potential, in: J.R. Parker (Ed.),Petroleum geology of Northwest Europe, Proc. 4th Conf.,Geological Society of London, London, 1993, pp. 915^931.

[54] A.D. Saunders, J.G. Fitton, A.C. Kerr, M.J. Norry, R.W.Kent, The North Atlantic igneous province, in: J.J. Ma-honey, M. Co¤n (Eds.), Large igneous provinces, Amer-ican Geophysical Union, Washington, DC, 1997, pp. 45^94.

[55] M.S. Fram, C.E. Lesher, A.M. Volpe, Mantle meltingsystematics: The transition from continental to oceanicvolcanism on the Southeast Greenland margin, in: H.C.Larsen, A. Saunders, P. Clift (Eds.), Proc. Ocean Dril.Program Sci. Results, v. 152, Ocean Drilling Program,College Station, TX, 1998.

[56] N. Ribe, U.R. Christensen, J. Theissing, The dynamics ofplume^ridge interaction, 1: Ridge-centered plumes, EarthPlanet. Sci. Lett. 134 (1995) 155^168.

[57] 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.

[58] G. Ito, Y. Shen, G. Hirth, C.J. Wolfe, Mantle £ow, melt-ing, and dehydration of the Iceland mantle plume, EarthPlanet. Sci. Lett. 165 (1999) 81^96.

[59] R. Macnab, J. Verhoef, W. Roest, J. Arkani-Hamed, Newdatabase documents the magnetic character of the Arcticand North Atlantic, Eos Trans. Am. Geophys. Union 76(1995) 449^458.



Documents

Mantle thermal structure and active upwelling during ...science.whoi.edu/labs/mclean210/kelemen/pages... · W.S. Holbrook et al./Earth and Planetary Science Letters 190 (2001) 251^266