JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 95, NO. BI1, PAGES 17,475-17,502,OCTOBER 10, 1990

Evolution of the Walvis Ridge-Rio Grande Rise Hot Spot System' Implicationsfor African and South American Plate Motions Over Plumes

JOHN M. O'CONNOR AND ROBERT A. DUNCAN

College of Oceanography,Oregon State University,Corvallis

Crystallizationages of volcanicrocks, dredged or drilled from the Walvis Ridge (ten sites)and the Rio GrandeRise (one site), have been determined bythe 40Ar/39Ar incremental heating technique. The fundamentallyage-progressive distribution of thesebasement ages suggests a commonhot spot sourcefor volcanismon the island of Tristan da Cunha, along the Walvis Ridge and Rio Grande Rise, and for the formation of the continental flood basalts located in Namibia (Africa) and Brazil (South America). The WalvisRidge-Rio Grande Rise volcanic system evolved along a sectionof the SouthAtlantic spreading-axis, as the African and South American plates migratedapart, astride,or in closeproximity to, an upwelling plume. Reconstructionsof the spatialrelationship between the spreading-axis,the Tristan hot spot, and the evolvingWalvis Ridge-Rio Grande Rise volcanic feature show that, at about70 Ma, thespreading-axis began to migratewestward, away from the hot spot. The resultingtransition to intraplatehot spotvolcanism along the Walvis Ridge (and associatedtermination of Rio GrandeRise formation)also involveda northward migrationof previouslyformed African seafloor over the hot spot. Rotationparameters for Africanmotion overfixed hot spots(i.e., absolutemotion) have been recalculated such that the predictedtrail of the Tristan hotspot agrees with the distribution of radiometricand fossil basement ages along the Walvis Ridge. African absolutemotion has been extended to the Southand North American plates, by the additionof relativemotion reconstructionpoles.

INTRODUCTION island [McDougall and Ollier, 1982], with the most recent volcaniceruption having occurred in 1962 [Barker,1964]. Apartfrom the mid-Atlanticspreading-axis, the Walvis Ridge This paperreports the resultsof sixteen40Ar-39Ar (Figurela) and Rio GrandeRise (Figure lb) are the two most incrementalheating studies of volcanicrocks recovered from ten prominentbathymetric features in the South Atlantic basin. locationsdistributed along the Walvis Ridge and a singlesample Broadlyspeaking, the Walvis Ridge and the Rio Grande Rise site located on the Rio Grande Rise (Figure 2). The changing form a V-shaped pair of volcanic lineaments, whose axis of spatialrelationship between the spreading-axis,the Tristan hot symmetryis the South Atlantic spreading-center(Figure 2). spot,and the evolvingstructure of the Walvis Ridge-RioGrande Morgan [1971, 1972], following Wilson [1963], related the Rise system,has been reconstructedfor selectedtimes. These formationof thesevolcanic trails to hot spotvolcanism, presently reconstructionsprovide new insights into the influenceof hotspot active beneath the ocean island of Tristan da Cunha. and spreading-axis interactions on the structural and The Walvis Ridge can be subdivided into three compositionalevolution of the Walvis Ridge-RioGrande Rise morphologicallydistinct regions, two of which appearto have a volcanicsystem. Previously estimated finite reconstructionpoles comparableor "mirror image" region on the Rio Grande Rise. for African plate motionover hot spots(i.e., absolutemotion) Thebroad eastern plateau of the Walvis Ridge projectsseaward havebeen adjusted such that the predicted track of theTristan hot from the African coast to about 5øE. This plateau initially spoton the Africanplate reconstructs the distributionof the new evolvedin conjunctionwith the TorresArch, later with the broad WalvisRidge basement ages. The absolutemotions of the South plateauof the Rio GrandeRise, andfinally with the northeastern and North Americanplates have, in turn, beencalculated by the shoulder of the Rio Grande Rise. The central section of the additionof appropriaterelative motion reconstruction poles. Walvis Ridge (about 5øE to 2øW) includes three distinct, subparallelNE-trending ridges, which formed in conjunctionwith GEOLOGICAL SETTING the eastern,N-S ridge of the Rio Grande Rise. Finally, the WalvisRidge mergesinto a line of large seamounts,which The Parana (southeastBrazil) and the Etendeka(southwest extends southwestward from the central section of the Walvis Africa) continentalflood basalts(Figure 2) representthe earliest Ridgetoward Tristan da Cunha.A correspondingseamount chain known manifestationof the hot spotresponsible for formingthe doesnot appearto existon the SouthAmerican plate; however, Walvis Ridge-Rio GrandeRise. Theseflood basaltfields are theZapiola Seamount Complex, located to the southof the Rio shownto have formedas adjoiningneighbors, on reconstruction GrandeRise may be relatedin some•vay to theTristan hot spot. of the SouthAtlantic for early Cretaceoustime [Rabinowitzand The island of Tristan da Cunha is located N550 km to the east LaBrecque,1979] and Figure 9a. Geochemicaland isotopic data of thepresent day spreading-axis,on C6 oceaniccrust (19 Ma supporta commonsource for the Parana-Etendekaand Walvis after the time scale of Kent and Gradstein [1986], used Ridgebasalts [Hawkesworth et al., 1986]. K-Ar agesranging throughoutthis discussion). Apparent K-Ar agesranging from from 114 to 196 Ma have beenreported for the Etendekalavas 0.01 + 0.02 Ma to 0.21 + 0.01 Ma have been determined for this [Siednerand Miller, 1968;Siedner and Mitchell, 1976]. Erlank et al. [1984] have recognizedthree petrographicunits within thesebasaltic lavas: the coastalAlbin type, the widespreadand Copyright1990 by the American Geophysical Union. voluminousTafelberg type, and in the faultedcoastal zone, late Papernumber 90JB00782. stagecross-cutting dolerite dikes which have been grouped as the 0148-0227/90/90JB-0078255.00 Horingbaaitype. K-Ar agesfor these three units range from 88 to

17,475 17,476 O'CONNORAND DUNCAN: WALVIS RIDGE AND RIO GRANDE RISE HOT SPOTSYSTEM

ETENDEKA

24'

34'

Fig.1. (a)Bathymetry forthe Walvis Ridge, contoured in1000 m intervals[after Needham etal., 1986]. (b) Bathymetry forthe Rio GrandeRise, contoured in 1000m intervals[after Cherkis et al., 1989]. Mercatorprojections. O'CONNORAND DUNCAN:WALVIS RIDGEAND RIO GRANDERISE HOT SPOT SYSTEM 17,477

I I 1 i i I I 10S Africa South America

Parana Spreading-Axis 13 f -' 34 i

/ s_•.o...u,o

'•! !..• 31 25 21 13 6

a 13

II, ,•, ß • 40S •1 13I

I I I I I ! I 50 s 6O W 50 w 40 w 30 w 20 w 10 w 0 10 E 20 E

Fig. 2. Solidcircles show the locationsof the samplesites from whichthe rocksanalyzed in thisstudy were recovered. Seafloor spreadinganomalies and tYacture zones are from Candeet al. [ 1989]. A fossilspreading center south of the Rio GrandeRise is shownby a "railroadtrack". The locationto whichthe spreading-axisjumped, between C32 andC30 times,is shownas a lineof opencircles on boththe Rio GrandeRise and the Walvis Ridge. Mercatorprojection.

377 Ma. The authorsattribute these anomalously high K-Ar ages continentalcrust, as shownby seismicreflection studies [Sibuet et (as indicated by stratigraphic relationships) to excess argon al., 1984]. This portionof the ridge is characterizedby internal "inherited"at the time of crystallization.Feldspar separates taken basementreflectors which dip to the west, while still further to from two fresh Horingbaai dolerites yielded fairly convincing the west (10øE to 8øE) the acousticbasement changes to that of 40Ar-39Arplateau ages of 125,and 125 to 130Ma [Erlanket al., typical oceanic crust [Sibuet et al., 1984]. On the South 1984]. These late stage intrusives may represent a final American margin, the Parana lavas stretch seaward toward the upwelling of continental hot spot volcanism immediately Vema Channel in the form of the Torres Arch •,Figure lb) precedingthe initiation of Walvis Ridge construction. In Brazil, [Kowsmann et al., 1977; Kumar, 1979]. Paranavolcanism also appearsto have peaked at between 120 and The presentday positionof the spreading-axisin the latitudes 130Ma [Amaralet al., 1966;McDougall and Ruegg, 1966; Melfi, of the Walvis Ridge and Rio Grande Rise is asymmetric with 1967; Siedner and Mitchell, 1976; Cordani et al., 1980]. respect to the coasts of Africa and South America; i.e., the Basedon theresults of seismicreflection profiling, Austin and spreading-axisis locatedcloser to the African than to the South Uchupi[1982] havesuggested that the SouthAtlantic began to American coast (Figure 2). The southern boundary of this openin the form of a northwardpropagating rift. The oldest spreading-axis asymmetry would appear to be the unnamed seafloorspreading anomaly identified in the SouthAtlantic at the fracturezone locatedto the north of Tristan da Cunha (Figure 2). latitudeof Capetownis M11 (133 Ma), while the oldestanomaly This asymmetryhas been attributedto a successionof easterly recognizedoff of the Namibian coastis M4 (126 Ma) [Cande et and southeasterlyspreading-axis migrations [Mascle and Phillips, al., 1989]. As the Etendekaand Paranaflood basaltsappear to 1972; Sclater and McKenzie. 1973; Ladd, 1974, 1976], which have erupted at about the same time that the northward- began prior to C34 (84 Ma) time. The first recordedeastward propagatingSouth Atlantic rift reachedthe latitudeof Namibia, a migration of the spreading-axisoccurred to the north of the closeassociation may haveexisted between hot spotvolcanism Walvis Ridge-TorresArch lineamentat early Albian-late Aptian andcontinental rifting. However,due to uncertaintyin locating time (97.5 to 115 Me) [Leyden et al., 1976; Ponte and Asmus, the continent-oceanboundary [e.g. Rabinowitz, 1976, 1978; 1976; Cande and Rabinowitz, 1978, 1979; Rabinowitz and Scrutton,1978], it is presentlydifficult to determinewhether or LaBrecque, 1979; Kumar and Gainboa, 1979]. The primary not,on reconstructionof the African and SouthAmerican plates evidence for the timing of this axis jump is the asymmetric to a preriftingconfiguration, overlapping or misfit existsat the distribution of continentalmargin salt deposits,which are more latitudeof theWalvis Ridge-Rio Grande Rise. Theexistence of extensive in the Brazilian compared with the Angolan basin. suchan overlap would suggest that crustal stretching occurred in African plate Mesozoicmagnetic anomalies located to the north response to doming of the continental lithosphere by the of the Rio Grande Fracture Zone were transferred westward to the upwellingplume, so supporting a relationshipbetween rifling and Sao Pauloplateau [Cande and Rabinowitz,1978, 1979];however, hotspot volcanism. Morgan [ 19721,Anderson [ 1982] and White to the south of this fracture zone Mesozoic anomalies remained andMcKenzie [1989] havesuggested such a causeand effect on the African plate [Rabinowitz,1976]. The timing and location relationshipbetween Atlantic basin hot spots and the initiation of of the beginning of this second phase of spreading-axis continentalrifting. migrationsare not well constrained. The Walvis Ridge extends southwestwardlyfrom the LaBrecqueet el. [1984] mappedyounger anomalies extending Etendekaflood basalts to about 10øE in the form of thinned progressivelysouthward and terminating againstthe northern 17,478 O'CONNORAND DUNCAN: WALVIS RIDGE AND RIO GRANDE RISE HOT SPOT SYSTEM flank of the Wa!vis Ridge (Figure 2) and proposedthat a composedoflaths ofplagioclase, granular pyroxene, Fe-oxides, southwardlypropagating spreading-axis was involved in the andglass. Rock compositions contain between 1.0 and4.8% formationof the Walvis Ridge-Rio GrandeRise volcanic system. MgO (dredgedsamples) and 0.9 and 7.2% MgO (drilled Barker [1984] suggestedthat such easterly and southerly samples),representing liquids that are far evolved from primary migrationsof thespreading-axis represent a "recapturingprocess" melts. Brief sampledescriptions and sample site coordinates are of the westwarddrifting spreading-axis, by theTristan da Cunha givenin Table1, with samplesite locations shown in Figures2 hot spot. Suchspatial readjustments of the spreading-axismay and 5. thereforerepresent a processby which the axis maintainedits Mostof therocks show patchy alteration of thegroundmass to positionon, or closeto, the Tristanhot spotfor the durationof low-temperatureFe-oxides and clay minerals.This appearsto Rio Grande Rise formation. haveaffected mainly the groundmass,but in somecases even the Derrick and Watts [1979] and Humphris and Thompson pyroxenemicrophenocrysts arereplaced. Glass has devitrified, [1983] haveproposed that at about80 Ma themid-Atlantic axis andvery finely crystallized mesostasis (as muchas 40%by began a westwardmigration away from the hot spot. The volume of somesamples) has altered in placesto smectitesand resultingtransition to intraplatehot spotvolcanism along the zeolites.This is furtherdemonstrated by themoderately high Walvis Ridge alsoinvolved a pronouncedeast-southeast ridge totalvolatile contents (CO2 plus H20+), whichrange from 1.2 to crest jump, sometimebetween anomaly 32 (71.4 Ma) and 3.9%. Olivine is rarely presentas a phenocrystphase or in the anomaly30 (67 Ma) times,which has been mapped south of the groundmass. Rio GrandeRise (Figure 2) [Cande et al., 1988; after J. L. Much of the potassiumin theserocks is likely to residein the LaBrecqueand J. M. Brozena,manuscript in preparation].This variablyaltered groundmass. The radiogenic 40Ar (40Ar*) ridgejump left a fossilspreading center roughly 200 km long,the generatedbythe decay of 40K at these sites (after crystallization) locationof which is consistentwith a southwardpropagating rift can be lost bothby diffusionand the low-temperatureformation modelfor spreading-axisinvolvement in the formationof the of clay minerals,resulting in erroneouslyyoung conventional WalvisRidge-Rio Grande Rise [LaBrecque et al., 1984]. K-Ar ages [Schlangeret al., 1984]. Smectitesand zeolites crystallizefrom volcanic glass, infill vesiclesand replace primary minerals. SAMPLE SELECTION Other complicationsencountered in applying the K-^r Thepetrology and geochemistry ofthe Walvis Ridge and Rio techniqueto the dating of submarinevolcanic rocks may arise GrandeRise samples selected for agedeterminations have been from the fact that glassy submarinelavas may containexcess reportedby Humphrisand Thompson [1982, 1983], Thompson 40Ar,due to theretention of mantle-derivedAr in magmaerupted and Humphris[1984], and Richardson et al. [1982, 1984]. underhigh hydrostaticpressure [e.g., Dalrymple and Moore, Overall,the Walvis Ridge and Rio Grande Rise samples range in 1968;Dalrymple and Lanphere,1969; Dymond, 1970; Aumento, compositionfrom tholeiitic to alkalic basalts. These rocks 1971]. Metasomaticaddition of K, derived from seawater,to generallycontain phenocrysts of plagioclasefeldspar up to 8 mm grainsurfaces [Melson and Thompson,1973; Seidemann, 1977] long and, lesscommonly, titaniferous augite in a groundmass may leadto erroneouslyyoung K-Ar agedeterminations.

TABLE1. Walvis Ridge and Rio Grande Rise Basalts Used for 40Ar-39Ar Age Determinations

,,

Sample Location(øS, øE) Depth,m Description

WatvisRidge AII-93-3-1 37ø 05.7',-7' 46.7' 2600-2000 vesicularpl-basalt AII-93-3-25 37ø 05.7',-7 ø 46.7' 2600-2000 vesicularpl-basalt AII-93-5-3 34ø 17.3', -5 ø 01.5' 3100-3000 vesicularpl-basalt AII-93-7-1 34ø 30.1',-3 ø 37.6' 2200-2100 alkali pl-basalt AII-93-8-11 34ø 29.9',-3 ø 28.7' 2000-1500 alkali basalt AII-93-10-11 34ø 20.1',-1 ø 34.4' 2300-2000 basalt AII-93-11-8 32ø 58.2', -0 ø 01.1' 3100-1600 vesicularbasalt V29-9-1 32ø 38.0', 1ø07.0' 3500 pl-basalt AII-93-14-1 31ø 59.6', 2 ø 23.6' 2300-1600 vesicularbasalt AII-93-14-19 31o 59.6', 2ø 23.6' 2300-1600 aphyricbasalt AII-93-21-1 25ø 26. I', 6ø 42.2' 3200-2600 pl-basalt DSDP 74-525-57-5,104--106 29ø 04.2', 2ø 59.1' 3087 aphyrictholeiitic basalt -6, 41-43 DSDP 74-528-40-5,70-75 28ø 31.5', 2ø 19.4' 4280 phyricpl-basalt DSDP 74-528-41-2,40-42 28ø 31.5', 2ø 19.4' 4291 aphyricbasalt

Rio Grande Rise

DSDP 72-516F-128-1, 22-24 30ø 16.6',-35ø17.1' 1268 phyric basalt DSDP 72-516F-128-1,104-107 30 ø 16.6',-35ø17.1' 1268 phyricbasalt O'CONNORAND DUNCAN: WALVIS RIDGE AND RIO GRANDE RISE HOT SPOTSYSTEM 17,479

The available Walvis Ridge and Rio Grande Rise samples from slightlyaltered volcanic rocks during the irradiationprocess. wereexamined in thinsection in orderassess the original igneous Thus,in conjunctionwith the loss of 40Ar*discussed earlier, texture,crystallinity, and subsequent degree of alterationof the 39Armay also be lost from, or relocatedwithin, a multiphase potassium-bearingphases. While no sample was regarded as sample. beingideal for radiometricdating purposes, well-crystallized Examination of apparent K/Ca ratios can be helpful in samples,displaying the lowest degrees of groundmassalteration, identifyingthe mineral phasesthat are contributingto different wereselected for40Ar-39Ar age incremental heating studies. regionsof the age spectrum[Walker and McDougall, 1982] (e.g., Figure 3f). With respect to the study reported here, low AGE MEASUREMENT METHODS temperaturesteps generally correspondto high K/Ca phases, characteristicof K-rich alterationproducts such as smectite. The The40Ar-39Ar method is basedon the generation of 39Ar high apparentages of theseinitial heatingsteps may reflect the from39K by the irradiation of K-bearingsamples with neutrons loss or redistributionof 39At due to neutron-inducedrecoil in a nuclearreactor. The principlesof thismethod of datinghave effects. The lower K/Ca valuesof the middle to high temperature beenpresented by Dairytopicand Lanphere [ 1971 ], Dairytopicet steps are characteristicof unaltered, low K phases such as al. [1981a],and McDougall and Harrison [1988]. Samples plagioclase. The plateauages usually evident at theseheating selectedfor age studieswere crushedafter the removalof any stepsmay be attributed tothe diffusion of 40Arand 39Ar from obvioussurface weathering and vesiclefillings. The 0.5 to 1.0 suchunaltered phases. The highesttemperature spectrum steps mm rock fragmentswere retained for dating experiments, are commonly lower in age than that of the mid-temperature ultrasonicallywashed in distilledH20, and dried. About one plateau and correspondto very low K/Ca ratios, which are gramof eachof thesamples was encapsulated in an evacuated characteristic of low K-phases,such as pyroxene.Recoil-induced quartzvial and irradiated in thecentral thimble position of the additionof39Ar from low temperature high K-phases toK-poor U.S.Geological Survey TRIGA reactorfor 15 hoursat 1 MW high temperaturephases, might account for such erroneously powerlevel, using the procedures described byDalrymple et al. youngages in the age spectra. [1981a].A hornblendeflux monitor of knownage (Mmhb-1), Dalryrnpleetal. [1981b]suggested that, in caseswhere 39Ar [Samsonand Alexander, 1987] was irradiated with the basalts in recoil is evident, the best age estimate•:br a samplecomes from orderto determinethe neutron flux density and the capturecross summing the gas composition of all the heating steps into a sectionof 39At for neutronsduring the irradiation,i.e., the "recalculated"40Ar-39Ar total fusion age. Thiscalculation is efficiencyofconversion of39K to 39Ar expressed asthe J-factor basedon the assumption that phases which lose 39Ar during (Table2). irradiationhave also lost their 40Ar* overgeological time. Samplesand standardswere individually placed in an However,since 39Ar recoil may be an incomplete process, a total outgassedMo crucible,which was then heated in a high-vacuum fusion age which is not supportedby both a clear plateau extractionline. The standardswere fusedin a singleheating step, (particularly for the middle temperature steps) and a well whilethe basaltswere incrementallyheated in a seriesof five constrainedisochron age, mustbe interpretedcautiously. Other steps(each of 30 minduration) and then melted during a sixth criteria,such as the consistencyof agesfor multiplesamples from andfinal step. The isotopic composition ofargon (40At, 39At, the same locality or independentgeological constraints, can 37Arand 36At) released from each individual heating or fusion increaseconfidence in suchtotal fusionages. stepwas measured immediately by meansof an AEI MS-10S Weightedmean plateau ages have been calculated for samples massspectrometer, after activegases had beenremoved by yielding an age-temperature plateau, primarily for middle coolingTi-TiO2 getters. The apparentages of the individual temperatureheating steps. Thesemiddle temperaturesites are heatingsteps were calculated from the measured 40At* to 39At consideredtobe least disturbed by the loss or additionof 39Ar ratios,after correctionsfor all interferingnuclear reactions had resulting from recoil effects, as discussedabove. The slope beenapplied using the equation of Brereton[1970]. Hencean formedby the correlation ofthe 40Ar/36Ar and 39Ar/36Ar ratios age-temperaturespectrum was obtained for eachsample, based for selectedheating steps yielded an 40Ar*/39Ar ratio, from onthe 40Ar*/39Ar compositions of the gas fractions released which an isochronage was determinedfor mostsamples. Ideally, incrementally,from low to high temperaturesites. Argon the40Ar/36Ar intercept of suchisochrons should reflect the isotopicdata and apparent ages are given in Table2. compositionof the rock at the time of crystallization,i.e., 295.5, consistingof only atmosphericargon without any contribution RESULTS from potassium decay. Isochron slopes and intercepts were Themost common age-temperature spectrum observed in this calculatedusing the leastsquares fitting techniqueof York [ 1969], which allows for correlatederrors in both 40Ar/36Ar and studyconsists of a middletemperature plateau bounded by higher agesat thelow temperaturesteps, and conversely, by lowerages 39Ar/36Arisotopic ratios. Errors involved inmeasuring Ar ratios at thehigh temperaturesteps (e.g., Figure 3e). Turner and and J-factors(a typical error of 0.5% has been assigned)and in Cadogan[1974] first proposed that such an "inversestaircase" making corrections for interfering nuclear reactions were age-temperatureprofile might result from neutron-capture recoil combinedto yield a standarddeviation for eachheating age. The of39Ar from K-rich to K-poor sites within fine-grained basalts, most reliable age-temperaturespectrum and isochronplot for duringirradiation. Because the low-temperature sites are K-rich, eachsample site are shown in Figure3. Plotsof 36Ar/40Ar recoileffects may lead to the transfer of39At (but not 40At*) versus39Ar/40Ar (inverse diagram) produced the same results as fromlow temperature (i.e., alteration minerals and groundmass) those estimated from isochron plots, as demonstrated by tohigh-temperature sites (e.g., feldspar and pyroxene), resulting Dah•ympleet al. [ 1988]. ina descendingage-temperature release pattern. Dalrymple and Samples AII-93-3-1 and AII-93-3-25 are from the same Clague[ 1976],Seidemann [ 1978], and McDougall and Duncan dredgehaul and have very similarcompositions [Humphris and [1988]have shown that significant amounts of39At may be lost Thompson,1982]. Sample3-1 produceda plateauage of 29.2 + 17,480 O'CONNORAND DUNCAN: WALVIS RIDGE AND RIO GRANDE RISE HOT S POT SY•EM

TABLE2a. ArgonIsotopic Data for Walvis Ridge and Rio Grande Rise Whole Rock Samples

Increment40Ar/36Ar 40Ar/39Ar 37Ar*/40Ar% Radiogenic %39Ar Age__1 c• 40Ar ofTotal x 10-6 years

, ,

Sample AII-93-3-1 J = 0.00368 1 339.8 65.520 0.0159 13.2 9.9 56.4 + 2.2 2 404.0 19.140 0.0816 27.5 17.0 34.6 + 0.5 3 952.8 6.278 0.4944 72.7 26.9 30.1 + 0.2 4 1492.6 4.959 0.7498 85.9 25.2 28.4 + 0.2 5 1076.2 4.120 0.9291 79.6 10.1 21.7 + 0.3 6 340.0 2.948 7.3437 70.2 1 1.0 13.9 + 0.5

.S..ampleAII-93-3-25 J = 0.00368 I 322.7 61.240 0.0121 8.5 10.9 34.4 + 3.4 2 356.7 29.520 0.0640 17.6 15.2 34.3 + 0.8 3 669.0 7.763 0.4805 59.5 26.7 30.5 + 0.3 4 1443.3 5.590 0.6147 84.2 26.6 31.3 + 0.2 5 974.0 4.515 0.5951 74.2 9.5 22.1 q-0.3 6 466.7 4.133 4.0620 68.2 1 1.2 18.8+ 0.2

Sample AII-93-5-3, J = 0.00368 1 359.5 46.900 0.0230 18.0 2.4 55.1 q- 1.0 2 371.6 31.250 0.0250 20.7 7.9 42.4 + 0.6 3 8211.2 6.060 0.2164 98.0 28.2 39.0 + 0.6 4 3474.4 6.985 0.1503 92.6 26.7 42.5 q-0.3 5 3397.8 6.259 0.1196 92.2 16.9 37.9 q-0.3 6 17197.0 6.377 0.8137 89.1 17.9 37.5 q-0.2

Sample AII-93-7-1 J ---0.00368 I 363.1 36.600 0.0200 18.7 13.0 45.0 + 2.3 2 392.2 26.900 0.0449 25.0 15.2 44.0 q-0.9 3 654.9 10.500 0.2036 56.4 18.2 39.0 + 0.2 4 1926.0 6.501 0.6132 89.3 26.4 38.3 q-0.2 5 I 132.7 5.820 0.7350 79.5 18.1 30.5 q- 0.2 6 481.4 4.762 3.6127 66.7 9.1 21.2 q- 0.4

Sample AII-93-8-11 ! = 0.00368 1 389.1 34.769 0.0348 24.3 19.1 55.3 q- 17.9 2 373.6 36.660 0.0522 21.3 12.0 51.2 + 1.4 3 784.8 8.517 0.5394 66.5 24.0 37.3 q- 0.2 4 1148.2 6.722 0.8279 80.6 16.1 35.8 + 0.4 5 732.9 5.313 1.0019 67.4 16.1 23.7 q- 0.1 6 537.7 4.565 2.5495 64.8 14.7 19.7 q- 0.3

Sample AII-93-10-11, J = 0.00341 1 509.9 24.890 0.0319 42.3 25.0 63.6 q- 0.4 2 571.4 18.420 0.0872 48.9 9.7 54.7 q- 0.3 3 561.2 18.000 0.1533 48.5 10.3 53.0 + 0.4 4 529.2 16.830 0.2173 45.8 1 1.7 46.5 q- 0.4 5 513.0 16.800 0.2529 45.4 8.6 45.4 q- 0.6 6 475.0 18.390 0.4143 41.0 34.7 46.0 q- 0.3

Sample AII-93-11-8, J = 0.00320 1 1485.0 14.900 0.0158 80.2 13.4 67.5 + 0.4 2 30782.6 11.290 0.0102 99.1 24.6 63.5 q-0.8 3 58224.5 11.570 0.0108 99.6 28.3 65.3 q-0.5 4 26671.7 11.400 0.0187 99.0 17.1 64.0 + 0.4 5 13009.2 11.250 0.0579 98.1 1 1.7 62.6 + 0.3 6 5467.1 11.260 0.0815 95.2 4.9 60.9 + 0.3

Sample V29-9-1, J = 0.00262 1 340.3 105.300 0.0073 13.2 16.1 64.6 q- 1.4 2 337.9 81.740 0.0157 12.7 14.0 48.3 + 2.0 3 371.5 53.580 0.0493 20.8 25.7 52.1 + 0.5 4 380.3 45.100 0.0984 23.1 24.5 48.6 q-0.5 5 302.1 181.400 0.0292 2.4 8.7 20.6 +_1.3 6 292.4 151.870 0.1846 0.4 1 1.0 2.8 q- 1.7 O'CONNOR AND DUNCAN: WALVIS RIDGE AND RIO GRANDE RISE HOT S POT SYSTEM 17,481

TABLE 2a. (continued) ..... Increment40Ar/36Ar 40Ar/39Ar 37Ar,/40Ar % Radiogenic%39Ar Age+ 1c• 40Ar ofTotal x 10'6 years

Sample AII-93-i4-1 J = 0.00320 1 602.8 23.460 0.0394 51.3 12.4 68.2 + 0.4 2 10398.0 11.350 0.9980 97.9 14.6 63.1 + 0.4 3 22614.3 11.030 0.0679 99.2 27.9 62.1 + 0.3 4 28535.1 10.850 0.0037 99.2 20.9 61.1 + 0.3 5 8278.9 10.490 0.0691 96.9 12.5 57.8 + 0.3 6 3229.6 10.900 0.1022 91.6 11.8 56.8 + 0.4

Sample AII-93-14-19. J = 0.00368 1 439.4 38.200 0.0135 32.8 10.3 81.5 + 4.0 2 1183.0 13.910 0.0507 75.4 7.2 68.3 + 0.4 3 7882.0 10.070 0.0891 96.9 27.7 63.6 + 0.3 4 18683.4 9.495 0.0802 99.0 39.4 61.4 + 0.3 5 3838.2 9.109 0.1050 93.1 8.4 55.4 + 0.3 6 2549.5 8.255 0.6993 93.8 6.9 50.9 + 0.5

SampleDSDP 525A-57-5. J = 0.00341 1 1795.1 15.790 0.1190 84.4 16.6 80.3 + 0.5 2 5600.9 14.320 0.0900 95.4 14.5 82.2 + 0.4 3 6807.2 13.606 0.1862 97.1 23.5 79.6 + 0.4 4 4512.4 13.790 0.2129 95.1 21.0 79.0 + 0.4 5 5178.7 13.300 0.1198 95.2 17.7 76.3 + 0.4 6 1866.4 13.700 0.8566 90.8 6.6 75.3 + 0.5

SampleDSDP 528-40-5 (73-75), .J = 0.00341 1 419.4 45.380 0.0887 30.3 7.2 82.7 + 0.5 2 869.8 19.970 0.2555 68.0 11.1 81.9 + 0.4 3 883.5 18.130 0.4984 70.4 13.5 77.3 + 0.4 4 641.3 22.590 0.3894 56.9 20.2 77.9 + 0.4 5 1010.2 16.350 0.2826 72.9 26.4 72.1 + 0.4 6 1365.9 13.320 1.3602 88.9 21.7 72.3 + 0.4

SampleDSDP 528-41-2(40-42),. J = 0.00341 1 3212.4 10.953 0.0538 91.2 17.6 60.5 + 0.3 2 9613.1 10.881 0.0785 97.5 13.6 64.1 + 0.8 3 5510.4 11.645 0.0566 95.3 17.5 66.9 + 0.5 4 915.4 17.902 0.0173 67.8 14.8 73.2 + 0.4 5 6957.0 10.000 0.0373 95.9 21.5 58.1 + 0.3 6 3308.4 9.555 0.2873 93.2 I4.9 54.1 + 0.3

SampleAII-93-21-1 J = 0.002620 1 440.7 55.310 0.0106 33.0 2.8 84.3 + 0.7 2 480.4 52.740 0.0085 38.5 7.9 93.6 + 0.8 3 467.9 59.250 0.0083 36.9 21.5 100.5 + 1.1 4 484.7 40.550 0.0303 39.3 19.1 73.9 + 3.5 5 848.2 19.800 0.2334 67.0 22.8 61.8 + 0.3 6 445.6 23.583 0.2478 35.9 26.1 39.4 + 0.3

SampleDSDP 516F-128-1(22-24), J = 0.002785 1 509.5 51.590 0.0188 42.1 5.4 106.1 + 1.8 2 581.9 40.100 0.0163 49.4 10.8 96.9 + 0.6 3 761.0 32.500 0.0969 61.9 24.4 98.6 + 0.6 4 1185.8 23.450 0.2542 77.0 22.8 88.9 + 0.5 5 678.4 26.840 0.3936 59.5 17.9 79.1 + 0.5 6 419.9 36.090 1.4133 40.6 18.6 74.7 + 0.6

SampleDSDP 516F-128-1 (104-107), J = 0.002785 1 534.2 67.210 0.0198 44.8 15.1 145.5 + 1.! 2 756.0 35.510 0.0760 61.5 22.8 106.7 + 0.6 3 1377.4 22.130 0.2410 87.6 22.5 87.6 + 0.4 4 1358.4 20.050 0.3920 81.3 14.7 80.5 + 0.4 5 881.5 22.100 0.4537 70.0 10.8 76.5 + 0.4 6 476.6 32.100 1.0919 40.8 14.1 66.1 + 0.5

Here,)• = 5.53x 10'10/year.Data in table have not been corrected forneutron interferences. Correctionfactors: (36Ar/37Ar) Ca= 0.000264; (39Ar/37Ar) Ca= 0.000673; (40Ar/39Ar) K= 0.0006. * Correctedfordecay since neutron irradiation ()•37Ar = 1.975x 10'2/day). 17,482 O'CONNORANDDUNCAN: WALVIS RIDGE AND RIO GRANDE RISE HOT SPOT SYSTEM

TABLE 2b. Age CalculationsFrom Argon Isotopic Data

PlateauC'aiculation .... IsochronCalculatio__•n--'

Recalculated %of TotalFusion Sample Age(m.y.) + 1(5 StepsUsed Total39Ar Age.(m.¾.) +lcy...... Steps Used Intercept Age(m.y.) _+ 1• AII-93-3-1 29.2+ 0.2 2,3, 4 69.1 28.4+ 0.6 1,2, 3, 4 317.4+5.2 30.3+0.3 AII-93-3-25 31.0+ 0.2 1,2, 3, 4 79.2 30.8+ 0.4 1,2, 3, 4 299.9+ 3.6 29.5+0.4 AII-93-5-3 38.7+ 0.2 2,3, 4, 5, 6 97.6 38.2+ 1.1 1,2,3, 4, 5, 6 309.4+ 9.2 40.1+0.3 AII-93-7-1 38.5+ 0.2 3,4 44.6 38.0+ 0.2 1,2,3, 4 307.6+ 1.5 37.2+0.4 AII-93-8-11 37.2+ 0.2 1,2, 3, 4 69.0 35.6+ 0.3 1,2, 3, 4 321.8+ 4 37.5+3.5 AII-93-10-11 46.2+ 0.3 4,5, 6 55.0 49.2+ 5.5 4,5, 6 280.9+ 26 52.0+0.3 AII-93-11-8 64.1+ 0.3 2,3, 4, 5 81.7 62.0+ 1.3 1,2, 3, 4, 5, 6 384.0+ 44 64.4+0.4 V29-9-1 50.3+ 0.4 2,3, 4 64.0 50.0+ 7.1 2,3, 4 295.6+ 10.6 44.6+0.5 AII-93-14-1 61.4+0.3 2,3,4 63.4 60.2+1.0 2,3,4 949.0+556 61.6+0.3 AII-93-14-19 61.5_+ 0.3 3,4 67.0 59.9+ 2.0 2,3, 4, 5, 6 389.0+ 90 63.4+ 0.5 DSDP 525A-57-5, 104-106 DSDP525A-57-5, 41-43 79.4+ 0.4 3,4 62.3 82.0+ 1.9 1,2,3, 4 288.0+ 128 79.1+0.4 DSDP528-40-5 (73-75) 77.6+0.5 3,4 33.6 79.1+6.3 1,2,3,4 296.8+45 75.9+0.5 DSDP528-41-2 (40-42) 62.5+ 0.3 AII-93-21- 1 69.8+ 0.8 DSDP516F-128-1 (22-24) 87.2+ 8.9 1,2, 3,4, 5,6 293.2+ 41.9 88.7+ 0.5 DSDP5.16F-128-1 (104-107) 95.5+0.5

0.2Ma (69.1% of the total 39Ar released), anisochron age of 28.4 Ma (interceptof 281 + 26) hasbeen calculated using the heating + 0.6 Ma (interceptof 317 + 5.2), and a recalculatedtotal fusion stepswhich were included in theplateau age calculation (Figure age of 30.3 + 0.3 Ma. Sample3-25 yielded a distinctmiddle 3i). Theproportion of radiogenic 40Ar released during this temperatureage plateau of 31.0+ 0.2Ma (79.2%of the total 39Ar experimentwas low, notexceeding 50% for anyof thesix heating released) (Figure 3a) and an isochronage of 30.8 + 0.4 Ma steps.The recalculated total fusion age of 52.0 + 0.3 Ma provides (intercept of 299.9 + 3.6) (Figure3b). The total fusionage is a minimumestimate of the ageof thissample site. 29.5 + 0.4 Ma. The bestestimate of the apparentage of this site SampleAII-93-11-8 yieldeda goodplateau age of 64.1+ 0.3 is considered to be between 30 and 31 Ma. Ma (81.7%of thetotal 39Ar released) (Figure 3j), a well- SampleAII-93-5-3 produceda plateauage of 38.7 + 0.2 Ma constrainedisochron age of 62.0+ 1.3 Ma (interceptof 384+ 44) (97.6%of the total 39Ar released) (Figure 3c), an isochron ageof (Figure3k), anda recalculatedtotal fusion age of 64.4 + 0.4Ma. 38.2 + 1.1 Ma (intercept of 309 + 9.2) (Figure 3d), and a The apparentage of thissample site is 64 Ma. recalculatedtotal fusionage of 40.1 + 0.3 Ma. The crystallization SampleV29-9-1 resultsshow a middletemperature plateau age at this site appearsto be well constrainedat between39 and ageof 50.3 + 0.4Ma (64%of the total 39Ar released) (Figure 3/) 40 Ma. anda somewhatpoorly constrained isochron age of 50.0+ 7.1Ma SamplesAI!-93-8-11 andAII-93-7-1 arefrom separate dredge (interceptof 296 + 11) (Figure3m). The recalculatedtotal fusion hauls from the same volcano. Sample 8-11 yielded a middle ageof 44.6 + 0.5 Ma is significantlylower than the plateauage. temperatureplateau age of 37.2+ 0.2 (69%the total 39Ar Thepercentage radiogenic 40Ar released during incremental released),an isochronage of 35.6 + 0.3 Ma (interceptof 322 + heatingwas very low, reachinga maximumof 21%. Thebest 4.0), anda recalculatedtotal fusion age of 37.5 + 3.5 Ma. Sample estimateof theage of thissample is takento be the50.0 + 0.4Ma 7-1 resultsshow a middle temperatureplateau age of 38.5 + 0.2 plateauage. Ma (45%of the total 39Ar) (Figure 3e), an isochron age of 38.0 + Samples AII-93-14-19 and AII-93-14-1 are from the same 0.2 Ma (interceptof 308 + 1.5) (Figure3g), and a recalculated dredgehaul. Sample14-19 yielded a plateauage of 61.5+ 0.3 totalfusion age of 37.2 + 0.4 Ma. The bestestimate of theage of Ma(67% of the total 39Ar released), anisochron age of 59.9 _-,2 this volcano is between 38 and 39 Ma. 2.0 Ma (interceptof 389 + 90), anda totalfusion age of 63.4-+ The crystallizationages of dredgedsamples AII-93-5, 7 and 8 0.5 Ma. Sample14-1 produced a reasonable plateau age of 61.4 (38 to 40 Ma) are supportedby resultsfrom the nearbyDeep Sea + 0.3 Ma (63.4%of thetotal 39Ar released) (Figure 3n), an Drilling Project (DSDP) site 359, which encounteredvolcanic isochronage of 60.2 + 1.0 Ma (interceptof 949 + 556) (Figure tuff dated radiometrically (K-Ar) at 40.3 + 1.0 Ma [Fodor et al., 3o), anda totalfusion age of 61.6 + 0.3 Ma. The crystallization 1977]. age of this site is between61 and 62 Ma. SampleAII-93-10~1 l showedevidence in its argonrelease patternof 39Arrecoil; however, we tentativelyidentify a DSDPTransect of theWalvis Ridge questionableplateau age of 46.2 + 0.3 Ma (the final 55% of the DSDP sites525A, 528, and 527 constitutea SE-NW transect total39Ar released) (Figure 3h). Anisochron age of 49.2+ 5.5 acrossthe Walvis Ridge (Figure 2). Theages of drilledbasalts O'CONNORAND DUNCAN: WALVIS RIDGE AND RIO GRANDERISE HOT SPOTSYSTEM 17,483

o i i i i I ß ß ß ß

!

.-- -

(elai)oSe ).uoJeddv

I i I , i i i i i _1 I ß ' I I I I ß

o

ß

-H

.,4

..

Q -- -' FI .o , C::)0 C::)C:) 0 C::)C:) C:) (e•) eSe lueJeddv

I I I ß , ! I I I I

i i i i i i i ß i

OJ -

o • • •0 o • o • o •0 o o • o • o • o • o • o o o o o o o o (elN)eõe •,ueJeddv (elN)eõe •,ueJeddv J•9S/J•ot• 17,484 O'CONNORAND DUNCAN: WALVIS RIDGE AND RIO GRANDE RISE HOT SPOT SYSTEM

o o ,I i , i , ,l i i i . • i i i,

11

o

v

ß ! '" (eft) eõe •,ueJeddv (eft)

, i i i i ,

• ..... C) 0 C) (• 0 C; C; 0 0 0 0 0 0 0 0 0 0 0 0 0

,,-- .,..- O'CONNORAND DUNCAN: WALVIS RIDGE AND RIO GRANDE RISE HOT SPOT SYSTEM 17,485 recoveredfrom sites 525A and 528 are broadly constrained by maximumfor a middletemperature step (73.2 + 0.4 Ma) andthen magneto-biostratigraphicinformation from the overlying decreaseagain to approximatelythe age of thefirst heating step, sediments[Manivit, 1984; Chave, 1984]. Thesedata provide soforming an arched release pattern (Figure 30. Thisage-release minimumestimates on the formationages of the underlying spectrummay reflect significant loss of 40Ar* from this sample. basementandso act as an independent control on 40Ar-39Ar The recalculatedtotal fusionage is 62.5 + 0.3 Ma, which is incrementalheating age determinations. substantiallylower thanthe radiometricage of the overlying Rockchips from core intervals DSDP 74-525A-57-5, 104-106 sample (DSDP 528-40-5). It was therefore concluded that the andDSDP 74-525A-57-6, 41-43 were combined, yielding a apparenttotal fusion age of DSDP 528-41-2 is erroneously convincingmiddle temperature plateau age of 79.4 + 0.4 Ma young;this result clearly demonstrates the pitfallsof conducting (62.3%ofthe total 39Ar released) (Figure 3p), an isochron ageof singlestep total fusion rather than incremental heating 40Ar-39Ar 82.05 1.9 Ma (interceptof 288 +_128) (Figure3q), and a experiments on seawater altered whole rock basalts. Such an recalculatedtotal fusion age of 79.1+ 0.4 Ma. The apparentage archedrelease spectrum (Figure 3t) mayp•ove to be characteristic ofDSDP 525A basement is79 Ma. This 40Ar-39Ar incremental of sucherroneously young total fusion ages. heatingage is compatiblewith the biostratigraphically SampleAII-93-21-1, recovered from the most northeasterly of constrainedearly Maestrichtian to late Campanian(between-73 the samplesites investigated in this study,produced an extreme and78 Ma) sedimentswhich immediatelyoverlie site 525A exampleof an inversestaircase age-temperature spectrum, and basement[Manivit, 1984]. Magnetostratigraphiccorrelation did not producea recognizableage plateau. The percentageof indicatesthat 525A basementformed within magneticC32 (---72 radiogenic40Ar released during incremental heating was low, to74 Ma) [Chave,1984]. 60% being the maximumvalue attained. The recalculatedtotal SampleDSDP 74-528-40-5,73-75 (locatedto the northwest fusionage of 69.8 + 0.8 Ma is the bestage estimatefor this of DSDP site 525A) produced an indistinct age-temperature samplesite; this apparentage probablyrepresents a minimum pl'•eauof77.6 + 0.5Ma (33.6% of the total 39At released) estimate, rather than the truesample age. (Figure3r), an isochron age of 79.1+ 6.3 Ma (interceptof 297+ The oldestage determinedfor the Walvis Ridge is provided 45)(Figure 3s), and a recalculatedtotal fusion age of 75.9 + 0.5 by DSDP 363, where drilling encounteredlower Aptian (about Ma. A bestestimate of the apparentcrystallization age at this site 113 to 119 Ma) sediments, a short distance above the inferred is between 78 and 79 Ma. This radiometric age range is basalticbasement [Bolli et al., 1978]. To a first approximation somewhatolder than the age of overlyingsediments, which have the age of Walvis Ridge basement increases linearly with beenbiostratigraphically constrained at middleMaestrichtian age increasingdistance from Tristanda Cunha(Figure 4), thuslinking ibetween~72 and 69 Ma) [Manivit, 1984], with a corresponding recent hot spot volcanism on Tristan da Cunha, via the Walvis magnetostratigraphicassignment of C31 to C32 times(- 69 to 74 Ridge,to its 120-130Ma expressionon the Namibiancoast, the Ma) [Chave,1984]. Etendeka continental flood basalt field. The age-temperaturerelease spectrumof sample DSDP Rio Grande Rise 528-41-2, 40-42 contrasts with the generally plateaulike or inversestaircase release spectra which are characteristicof the SamplesDSDP 516F-128-1,22-24 andDSDP 516F-128-1, samplesexamined in thisstudy. Heatingstep ages increase to a 104-107both failed to produceconclusive plateau ages. The first

140

130 - J_• X Fossilage range 120 - • • ß K-Artuff age •/• - • X • Y' • - 110 -

100 - X -

90-

80-

70-

60-

50 -

40-

- 30-

20 -

10-

, , 0 5000 4000 3000 2000 1000 Tristan 1000 2000 3000 4000 5000

Distance from Tristan da Cunha along volcanic trails (kin) Fig. 4. All availablebasement age constraints for the Walvis Ridgeand Rio GrandeRise are plottedagainst their respective distancesalong volcanic trail, from Tristan da Cunha.A straightline has been fitted to thesedata, with an ageof 0 Ma assigned to Tristanda Cunha.Migration rates for theAfrican and South American plates over the Tristan plume have been estimated at 3.1 and2.9 cm/yr, respectively. 17,486 O'CONNORAND DUNCAN: WALVIS RIDGE AND RIO GRANDE RISE HOT SPOT SYSTEM yieldeda recalculatedtotal fusion age of 88.7+ 0.5 Ma andan spreading-axiswasin close proximity tothe hot spot (Figure 9b): isochronage of 87.2 + 8.9 Ma (interceptof 293 + 42). The thecontribution from renewed volcanism on the Rio Grande Rise secondproduced an evenmore extreme example of an inverse plateauduring the Eocene should bekept in mind when making staircaserelease pattern, with a recalculatedtotal fusion age of suchan assumption. An obviousexplanation is that the Tristan 95.5 + 0.5 Ma. An apparentage of 88 Ma waschosen as the best plumedecreased in strength through time. Changingeruption availableconstraint on the time of crystallizationof this site. ratesmay alsoreflect the changingtectonic character of the DSDP 516F basaltswere sampledbeneath poorly lithified and overridingplate, into which the Tristan magmas were eraplaced recrystallizedsediments which have been biostratigraphically (i.e.,from rifting continental crust, to spreading-axis,tofinally. datedat betweenConiacian to Santonianage (84 Ma to 88.5Ma) an intraplateenvironment). Changes in platethickness and/or [Weiss,1983]. Sincethis time waswithin the longCretaceous velocitymay have controlled,in part, the volumesof Tristan normal magnetizationperiod, magnetic stratigraphyis not magmawhich erupted. Variations in thedynamics of theplume. informative.Musset and Barker[1983]also investigated DSDP andthe surrounding mantle, could also have played a contributor). 516Fbasalts using the 40Ar-39Ar incremental heating method and role. concludedthat 86.0 + 4.0 Ma wasthe bestestimate of the ageof crystallization. ABS()LU'I'E Mt)TI()N M()DEI.IN(J DSDP site 21, located on the northeasternshoulder of the Rio t[()T SP()T REFERENf'E FRAME) GrandeRise (Figure 6), providesa minimumbiostratigraphic age of Campaniantime (74.5 to 84 Ma) for the underlyingbasement. RotationParameters for Afi'icanPlate Motion Over Hot Spots This drill hole did not actuallyreach basement but terminatedin Previouslyestimated sets of reconstructionparameters for shallow water sediments consisting mainly of megafossil Africanplate motion over fixed hot spots(i.e., absolutemoti0nt fragments;the actual thicknessof this rock layer is unknown [Morgan, 1981, 1983; Duncan, 1981] have been adjustedsuch [Bukry and Bramlette, 1970; Gartner, 1970; Blow, 1970]. The that the predictedtrail of the African plate over the Tristanhot distribution of available Rio Grande Rise basement and fossil spotwould reconstruct the distributionof Walvis Ridgebasement ages, in conjunction with present estimatesof the narrow ages (Figures 5 and 6). The geographicalcoordinates of the age-range of the Parana flood basalts, is compatible with adjustedrotation poles were located geometrically by requiring age-progressivevolcanism linking the northeasternshoulder of that small circles, when fitted about each pole, should the Rio Grande Rise (-75 to 84 Ma) to 120-130 Ma continental simultaneouslyfit the geometryof threeAfrican plate hot spot flood basaltvolcanism (Figure 4). lineaments: the Walvis Ridge and the volcanic trails of the However, volcanismmay not have migratedalong the Rio Reunionand Marion hot spots. This is illustratedby the factthat GrandeRise in a simple,linear age-progressive fashion. Portions the geometryof the predictedtracks for the Tristan, Reunionand of the Rio GrandeRise-Walvis Ridge systemappear to have Marion hot spots,coincides with that of their postulatedvolcanic developedin a mannersimilar to thatof presentday Iceland [Vogt trails (Figure 7). The Africa/hot spotsreconstruction parameters and Johnson, 1975; Saemundsson,1979; Helgason, 1985] and given in Table 3 differ from thoseof earlierversions, primarily in Figure 9b. Barker [ 1984] hasobserved that a sequenceof planar that the rates of plate motion are now constrainedto matchthe dipping reflectorsexists beneath the reflectorassociated with the distributionof new basementages along the Walvis Ridge. The basalt at the base of DSDP 516F, located on the Rio Grande Rise geometryof the predictedtrails of the St. Helena and Comores plateau. When Icelandic plateau basalts eventually become hot spotsis also in broad agreementwith that of their postulated submerged, they will show similar characteristics. Such a volcanictrails (Figure7). dipping reflector sequenceis similar to thosefound along many The principalassumptions made in this particularmodel are: continental margins and is usually consideredto represent a (1) the Walvis Ridge basementages and geometryperfectly subaerialseafloor spreading facies, composed of sheetlava flows record the motion of the African plate over hot spots,(2) the and shallowmarine or alluvial sediments.The centralplateau of ocean island of Tristan da Cunha representsthe centerof a the Rio Grande Rise may therefore have formed along a discrete,small diameter, vertically rising plume, as do theislands spreading-axis,which in tum was astride, or in closeproximity, of Reunion, and Marion, (3) the geometry of the Reunion, to a hot spot,as is the casefor Icelandtoday. Numerouseastward Marion/PrinceEdward hot spottrails is accuratelyknown, and spreading-axisjumps were, in all probability,associated with the therehas been no significantwander of thesethree hot spots with formation of the central plateau of the Rise. The eastern,N-S respectto one another. ridge of the Rio GrandeRise haspreviously been attributed to the eruptionof hot material along a segmentof the SouthAtlantic The PredictedTrail of the TristanPlume Comparedto WaMs spreading-axis[e.g., Vogt and Johnson,1975]. The history of Ridge Bathymetryand Age Data interactionbetween the spreading-axisand the Tristan hot spotis reconstructed in a later section. Figure5 illustratesthe relationship between the modeled trail An additionalcomplication to understandingthe evolutionof of the African plate over the Tristan plume and the bathymetry the Rio Grande Rise arises from an undefined volume of Eocene anddistribution of all availableWalvis Ridge basement ages. ^ volcanismidentified by Bryan and Duncan [1983], who have 30 Ma seamountcluster, locateddue east of Tristan da Cunha, datedvolcanoclastic, turbiditic horizons at 46 Ma, usingthe K-Ar tightly constrainsthe rotationangle aboutthe 30 Ma finite datingtechnique. This enigmaticsecond phase of volcanismhas reconstructionpole. As novolcanic trail is evidentbetween this presentlyno satisfactoryexplanation within the context of thehot 30 Ma siteand the islandof Tristanda Cunha,the 10 and20 Ma spotmodel proposed for the formationof theWalvis Ridge-Rio reconstructionpoles were calculatedby interpolation.The Grande Rise. rotationangie about the 40 Ma is very well constrainedby age Eruptionrates through time of WalvisRidge-Rio Grande Rise datafrom threesample sites, and a supportingK-Ar age. volcanismhave yet to be quantitativelyestimated. However, The 50 to 70 Ma portionof theWalvis Ridge consists ofa significantly more material was emplaced when the complexseries of subparallelvolcanic lineaments (Figure 5). It is O'CONNORAND DUNCAN: WALVIS RIDGE AND RIO GRANDE RISE HOT S POT SYSTEM 17,487

15S

113 1-1 9(363) r•]?._¾:._._e120 I •nn ';.•-" \ Etendeka 100*"• r" • 120-130 20 S

25 S

• • 78-79(528

•,• .... •• - 30 S :o•61-62 (14) • X X •

• • ••• *, (35•) - ( ) - 35 S Tristan•' 20••• • Group-, _••••

40 S 15W 10W 5 W 0 5 E 10E 15E 20 E Fig.5. Thedirection ofmodeled African/hot spots motion isshown by the heavy line connecting Tristan da Cunha (0 Ma volcanism)to the 120to 130 Ma Etendekacontinental flood basalts. The smallsolid circles distributed along this line represent the modeledprogress of the Africanplate over the Tristanhot spotat intervalsof 10 Ma. Samplesites are shownas largesolid circles,along with ages,and site labels in parentheses.Two reconstructionsfrom the period80 Ma to 120Ma are presented;the dashedhot spottrack follows the geometryof the Walvis Ridge (modelB), whereasthe solidtrack (modelA) is a linear, bestfit to the bathymetryof the northernportion of the Walvis Ridge. The 130 Ma intervalpoint on the hot spottracks reconstructs the Etendekabasalts to the southeasternedge of a hypotheticallylarge diameter plume, which may have existed at the time of continentalrifting. Mercator projection.

somewhatof a quandaryas to whichof theseminor ridges best reactivationof the volcanic lineamentmay actually have occurred recordsthe motion of theAfrican plate over the Tristan hot spot. in this region of the Walvis Ridge. Two sets of rotation Thenorthern lineament formed along a spreading-axis,in contrast parametershave beenestimated for Africa/hot spotmotion for the to thetwo southernlineaments, which formed in an intraplate period 80 to 120 Ma (Figure 5). Model A predictsa track which situation, as will be shown in a later section. While these follows the bathymetry for this portion of the Walvis Ridge in southernlineaments may in fact accuratelyrecord the motionof basically a linear fashion. In contrast, the track predicted by theAfrican plate over the Tristan plume, the rotation parameters model B closely follows the curvatureof the Walvis Ridge. If, givenin Table3 havebeen constrained soas to predicta hotspot indeed,this sectionof the Walvis Ridge-Rio GrandeRise system trailwhich follows the northern limb. Theserotation parameters representsa twin of presentday Iceland (to be discussedin the predictsmooth, linear hot spottrails for otherAfrican hot spots following section),the changingcurvature of the Walvis Ridge whichare compatiblewith the geometryof their postulated does not, in all probability, record changesin the direction of volcanictraces. As no basementages were availablefor the African plate motion over hot spots. Following this reasoning, northernlimb, reasonableestimates were made on the basis of the model A is a preferablemodel and so will be usedin following age data available for the southerntwo limbs, in order to discussions. constrainthe anglesof rotationabout the 50, 60, and 70 Ma At ---130Ma the initial expressionof the Tristanplume on the reconstructionpoles. However, significantlymore basement surfaceof the rifting African and SouthAmerican plateswas in samplingis obviouslyessential in order to better define the the form of the Parana-Etendekacontinental flood basalt(Figure tectonicand geochemicalevolution of this, and other such 9a). These lavas may have erupted in responseto the rapid complextectonic regions of theWalvis Ridge-Rio Grande Rise partial melting of the large diapiric head of the Tristan plume system;increased sampling will revealwhich specific portions of [Richards eta!., 1989]. Subsequentvolcanism linked to the thevolcanic trails best record the motion of theAfrican plate over Tristan plume, i.e., the Walvis Ridge-Rio Grande Rise, are thehot spot. The rotationangle about the 80 Ma finite possibly related to smaller volumes of plume material rising reconstructionpole is well constrainedby the 78-79 Ma DSDP throughthe conduitor "tail", formedby the diapir. Alternatively, transectof theWalvis Ridge. the Parana-Etendeka basalt• may have been formed by the Between80 and120 Ma littlereliable age data are available; pressure-releasemelting of the uppermantle, which would have theenigmatic 70Ma age for sample AII-93-21 was interpreted as occurredif continentalrifling commencedin a regionwarmed by a minimumage and therefore was not usedin the estimationof a hot spot,previously unable to penetratethe lithosphere[White rotationparameters. However, future studies may confirm that and McKenzie, 1989; Cox, 19891. A 130 Ma Africa/hot spots 17,488 O'CONNOR AND DUNCAN: WALVIS RIDGE AND RIO GRANDE RISE HOT SPOT SYSTEM

TABLE 3. Finite ReconstructionPoles for African Plate reconstructionpole wasestimated such that the Etendekabasalt Motion Over Hot Spots fieldwould straddle, on reconstruction,the eastern edge of sucha

, largehot spot(Figure 9a). An additionalconstraint on the reliabilityof the adjusted

,,Age, Ma Latitude,øN Longitude,øE Angle* rotation parametersis that the predicted trail of the South Americanplate over the Tristanplume shouldbe in reasonable ModelA (LinearFit ThroughNE Wah,isRidge) agreementwith the evolutionary history of theRio GrandeRise, as it is presentlyunderstood. South American plate motionover 10 5 2.0 - 16.3 - 1.0 hot spots has been reconstructedby the addition of South 20 51.4 -24.3 -2.4 America/Africafinite reconstruction poles [Cande et al., 1988]to 30 51.5 -23.3 -3.5 Africanmotion over hot spots. Figure 6 showsthat the hot spot 40 47.3 -45.9 -7.1 track predictedby model A trendsin a linear, NW-SE direction 50 45.7 -50.9 -8.7 throughthe Rio GrandeRise. ModelB, whichmore closely 60 44.9 -50.7 - 10.6 follows the bathymetry of the Walvis Ridge, predicts a more 70 43.6 -54.2 -14.0 northerlytrending track until 100 Ma when it changesdirection to 80 42.4 -57.5 -16.4 the west. Africa/hot spotsrotation parameters do not predicta 90 40.0 -59.6 -20.2 trail for the SouthAmerican plate over the Tristan plume,which 100 39.7 -60.6 -23.2 followsthe N-S ridgeof theeastern Rio GrandeRise (Figure6). 110 39.4 -61.3 -26.0 However, as will be shown in a later section,this portionof the 120 41.5 -64.9 -29.2 riseprobably formed along a southwardlypropagating spreading- 130 32.8 -73.3 -26.4 axis (Figure 9b), and so did not record the true direction of South Americanplate motion over the Tristan hot spot. This point Model B (Following the Curvature of NE WalvisRidge illustratesthe critical importance of accuratelyunderstanding the Bathymetry) evolutionaryhistory of individual hot spot-relatedvolcanic trails. in the estimation of valid absolute motion reconstruction 90 41.2 -58.7 -21.0 parameters. 95 38.6 -64.2 -23.8 Supportfor a NW-SE trendinghot spottrack throughthe Rio 100 40.2 -64.5 -26.1 Grande Rise is provided by recent studies of the South Atlantic 110 40.9 -61.6 -28.5 geoid. At shortwavelengths, seafloor topography and variations 120 41.1 -58.0 -30.0 in the geoid are strongly correlated and so provide a better constrainton seafloortopography in poorly surveyedareas, such as the South Atlantic. A map of short-wavelength regional ',Negativeangles indicate counterclockwise rotation. undulationsin the amplitude of the geoid shows a good

lOS I I I I

Parana 120-130 Ma 12o-130 Ma 13 211 Martin Vas 20 S

125 121 , 1!,3

21 13 3O S 25 13/

Ma 34 34

40 S 21 13 i I

50 S

60 W 50 W 40 W 30 W 20 W 10 W 0 10 E 20 E

Fig. 6. Predictedtrail of the African and SouthAmerican platesover the Tristan hot spot. SouthAmerican/African relative motion[Cande et al., 1988]was added to Africa/hotspots motion given in Table 3. The modeledtrail of the Martin Vas plumeis also shown. Mercator projection. O'CONNORAND DUNCAN:WALVIS RIDGE AND RIO GRANDERISE HOT SPOT SYSTEM 17 489

correlationbetween the bathymetryof the Walvis Ridge and its andPiton de la Fournaise,which began to eruptearlier than 0.5 topographydetermined from the geoid [Gibert et al., 1989;Plate Ma and is presentlyone of the mostactive volcanoes in the world 1]. The geoid-inferredtopography of the Rio GrandeRise, [McDougall,1971; Gillot andNativel, 1989]. Morgan[1981] however,forms an elongated NW-SE trending lineament, as proposedthat Reunion locates a hotspot, which formed the island predictedbythe hot spot track, which does not correlate well with of Mauritius(7-8 Ma [McDougall,197 l ]) andthe volcanic ridge availablebathymetry (Figure lb). A NW-SE trendingvolcanic extending northward beneath the Mascarene Plateau. On the trailis moresymmetrical with respectto the Walvis Ridgethan is Indian plate, the trail of the Reunionhot spotconsists of the thebathymetry of the Rio GrandeRise. Fleitout et al. [1989] ChagosBank, the Maldive and Laccadiveislands, and the Deccan havedetected numerous small-wavelength elongated features on continentalflood basalts. Thus, the track of theReunion hot spot filteredgeoid and topographymaps of the South Atlantic. A stretches for ,--5,000 km from Reunion to the flood basalts of numberof these features are orientated at N50øE for the African westernIndia. The age progressivecharacter of this hot spot plate,and N65øW for theSouth American plate: this agrees with lineamenthas been confirmedby recentbiostratigraphic and thegeometries of thepredicted hot spottrails shown in Figure6. magnetostratigraphicdata [Backmanet al., 1988], radiometric Theauthors attribute these elongated features to magmatictraces dating [Duncan and Hatgraves, 1990], and isotopicstudies left over a large numberof convectingplumes, which recordthe [Whiteet ai., 1990]. Biostratigraphicand magnetostratigraphic directionsof African and SouthAmerican absolute plate motions. informationprovides an ageestimate of betweenchrons C13N-1 and C13N-2 time (-36 Ma) for OceanDrilling Program(ODP) Afi'icanand South American Plate Motions Over SouthAtlantic site 706 basalts,located at the northeasternmargin of the andIndian Ocean Hot Spots NazarethBank (Figure 7). Extendingnorthward from this plateau is the Sayade MalhaBank, dated at 47 Ma by 40Ar-39Ar The island of Reunion is composed of two volcanoes, the radiometricage determinationof basaltrecovered from industrial dormantPiton des Neiges (>2 to 0.02 Ma [McDouga!l, 1971' drill site SM-1 (locatedbetween ODP sites706 and707, Figure Chevallierand Vatin-Perignon, 1982; Gillot and Nativel, 1984]) 7). Duncan and Hatgraves [1990] have shown that this bank

20 N

10N % - x % AFRICA

• •••*' •_ Ascension• (707) '- lOS I •o•-,, ß e l, St.Helena 20 s I /

30 s / 40 s

i•&P. Edward Cr•Bank 50 s

I I I / I I • BouvetI ConradI Rise1 60 S 60W 50 W 40 W 30 W 20 W 10 W 0 10 E 20 E 30 E 40 E 50 E 60 E 70 E Fig. 7. The trailsof all SouthAtlantic and Indian Ocean hot spotslocated on theAfrican and South American plates are shown. Thinportions of modeledhot spot tracks illustrate the connection between volcanic trails which are linked to a commonhot spot, anddo not correspond to actualvolcanic trails. Biostratigraphic and magnetostratigraphic ages shown for theMascarene Plateau arefrom Backman et al. [ 1988];drill sitenumbers are shown in parentheses.K-At agerange for Mauritiusis t¾omMcDougall [ 1971]. Mercatorprojection. 17,490 O'CONNORAND DUNCAN: WALVIS RIDGE AND RIO GRANDE RISE HOT SPOT SYSTEM

formedsimultaneously with the ChagosBank (48 Ma) on the et al., 1988]. The oldestin-situ or reworkedmicrofossils at ODP Indian plate, and that these featureswere separatedby the site703, locatedon the MeteorRise areEocene, in contrastto the initiationof seafloorspreading at about36 Ma alongthe Central Late Cretaceousmicrofossils found on the Islas OrcadasRise IndianRidge. The volcanicridge connecting the Saya de Malha [Ciesielskiet al., 1988]. This age differencehas been attributed Bank with the SeychellesBank has been biostratigraphically and to theMeteor Rise having experienced more recent volcanism, magnetostratigraphicallydated at ODP site 707 as C26R, between whichwould account for itsshallower basement and more rugged 62 and63 Ma, andat-64 Ma by radiometricdating. This ridge is relief. An additional constrainton the age of these featuresis consideredto have formed along the southwesternedge of the givenby thefact thatthe oldestanomaly bounding the eastside of Deccan Traps, and to have been rifted away from India as the the Islas OrcadasRise, and the correspondingwest side of the Arabian Seabegan to open(Duncan and Hatgraves,1990]. Meteor Rise, is anomaly 24 (55 Ma) [Cande et al., 1989]. The Africa/hotspots reconstruction parameters given in Table The Trinadade-Columbiaseamount chain representsthe trail 3 havebeen estimated in sucha manneras to predicta trail for the of the SouthAmerican plate over the Martin Vas hot spot[Herz, Reunionplume that would follow the geometryof the Mascarene 1977] (Figure 6). Ages of 0.7 Ma and 0.2 to 3 Ma havebeen Plateau (for the period between0 and about 36 Ma). However, determinedfor Martin Vas and its adjoiningisland to the west, the predicted rate of migration of the African plate over the Trinadade,respectively [Herz, 1977]. The predictedtrail of this Reunionplume (Figure 7) is somewhatslower than that indicated hot spotfollows the Trinadade-Columbiaseamount chain and by availablebasement age data for its volcanictrail (Figure 7). broadlyagrees with the 42-52 K-Ar age rangedetermined for the Relative motion betweenthe Reunionand Tristanhot spotcould Abrolhos archipelago,located somewhatto the north of the explain this misfit. An alternativeexplanation might involve Trinadade-Columbia lineament [Herz, 1977, Figure 1]. In relative motion across the east African rift zone, i.e., the western additionto the Abrolhos archipelago,the Hotspur, Rodgers,and boundary of the Sorealia plate, which could have added an Minerve Seamountsare locatedalong the northernfringe of the easterlycomponent of motionto Somaliaplate migration over the Trinadade-Columbia seamount chain, while to the south, is Reunion plume [Emerick and Duncan, 1982]. locatedthe Almirante Saldanhaseamount [Cherkis et al., 1989]. The geometryof the predictedtrail of the Comoreshot spotis This broad zone of seamounts may indicate that, as the in generalagreement with that of its postulatedvolcanic trail, the continentalmargin of the SouthAmerican plate migrated over the Comores,Cosmoledo, Farquhar and Amirante islandgroups, and Martin Vas plume, a wide diameterhot spot was in existence. the Tertiary igneousactivity locatedat the northeasternregion of Sucha suggestionis supportedby the fact that thishot spottrack the Seychellescontinental block [Emerick and Duncan, 1982]. coincidesin agewith a groupof alkalicigneous intrusions on the This intense episode of widespread igneous activity in the GoianiaArch rangingin age from 70 to 95 Ma [Neill, 1973],as SeychellesIslands, first noted by Baker and Miller [1963], has notedby Hartnady and le Roex [I985]. About 200-300 km been dated at 63 Ma by Maclntyre et al. [1985]; this age is in southeastward,a series of alkalic intrusions between Pocosde good agreementwith that predictedby the Comoreshot spottrail Caldas(63-80 Ma) andCabo Frio (51 Ma) [Neill, 1973] havealso (Figure 7). The zero age location of the Comores plume is beenattributed to the Martin Vas plume [Hartnady and le Roex, assumed to be beneath the island of Grande Comore. As is the 1985]; thesecontinental intrusives are in broad agreementwith casewith the Reunionhot spot,extension across the East African the agespredicted for the Martin Vas hot spot. A seriesof three rift zone could alsohave augmentedthe rate at which the Somalia datedkimberlite deposits (80, 86 and 122 Ma) [Crough et al., plate migratedover this hot spot. 1980,Figure 6] alsocoincide with the predictedtrail agesfor the Africa/hot spots reconstruction parameters were also hot spot. In addition,Crough et al. [1980] pointedout thatthe constrainedto follow closely as possible,the trail of the African trail of the Martin Vas hot spotcoincides with mostof the alluvial plate over the Marion/PrinceEdward plume. However,there is a diamonddeposits in Brazil. The agesof thesealluvial deposits small degree of misfit between the predicted trail and the are still uncertain,however, some geological evidence suggests a somewhat indistinct volcanic trail of the Marion/Prince Edward Cretaceousage [Sviseroet al., 1979], but other depositsare plume, the Walters Shoals and the MadagascarRidge. The considered to be Precambrian. postulated zero age location of this hot spot trail is the The modeled trail of the St. Helena hot spot follows the volcanicallyactive Marion Island,located on the Antarcticaplate southernboundary of itspostulated volcanic trail, a broadNE-SW (Figure 7). swathof seamountsand ridges. K-Ar agesof between14 and8 Hartnady and le Roex [ 1985] haveproposed the existenceof a Ma have been determinedfor St. Helena [Baker et al., 1967], hot spot,known as Shona,located near the southernmosttip of althoughnew dataindicate an age rangeof between9-7 Ma the South Atlantic spreading-axis,in close proximity to the [Chaffeyet al., 1989]. The misfitbetween the predicted trail of Bouvet hot spot(Figure 7). The authorssupport the existenceof the St. Helenaplume and its volcanictrail maybe explainedby sucha hot spotwith geochemicaldata, which distinguishbetween uncertaintyin accuratelylocating the zero age centerof the the compositionof Bouvetand Shona magmas. The zigzagshape upwellingplume. An alternativeexplanation could be that of the Cape Rise-Meteor Rise lineamenthas beenattributed to Helenaand Tristan hot spotsare of significantlywider diameters the crossingof the Shona hot spot by the active transform (-500 km) than has been assumedfor the purposesof this segmentof theFalkland-Agulhas Fracture Zone in LateMesozoic discussion[O'Connor, 1989]. times [Hartnadyand le Roex, 1985]. The relationshipsbetween An agerange of 1.81to 12.3Ma hasbeen determined for the the predictedtrail of the Shonahot spot, on theAfrican and South islandof Fernandode NoronhaiCordani, 1970]. The modeled American plates, and the geometriesof the Meteor and Islas trail of the Fernandode Noronhahot spoton the Africanplate OrcadasRises, respectively, suggest that both of thesefeatures coincideswith the Sierra Leone Rise, although the predicted trail were formed simultaneously,at about 60 Ma (as suggestedby passessomewhat to the southof thisfeature (Figure 7). On the LaBrecque [1986]) by the Shonahot spot (Figure 7). Basal basis of both ocean drilling, and its distancefrom the sedimentson the Islas OrcadasRise (ODP site 702) show that this spreading-axis,Kumar and Embley[1977] suggestedthat the featureis >61 Ma and so is possiblyLate Cretaceous[Ciesielski CearaRise originated in conjunctionwith the Sierra Leone Rise, O'CONNORAND DUNCAN: WALVIS RIDGE AND RIO GRANDE RISE HOT SPOT SYSTEM 17,491 atapproximately 80Ma, as shown by Sibuet and Mascle [1978]. technique,to a maximumof 1.9 Ma (unpublisheddata). Two Thepredicted trail of theAfrican plate over the Fernando plume undatedseamounts located at 9 ø 04øS, 19ø 40øW and 8ø29øS, 17ø supportsthesuggestion thatthe Ceara and the Sierra Leone rises 07.6øW, some 570 and 300 km west of Ascension,respectively wereformed simultaneously by the Fernando hot spot, at about [e.g.,Cherkis et al., 1989],are located along the predicted trail of 80Ma (Figure 8): theNorth Brazilian Ridge may also have been the Ascensionplume on the SouthAmerican plate (Figure 7). formedby this particularplume. At the intersectionof the Thepredicted ages of thesetwo seamounts are -20 and,--10 Ma, Fernandode Norhonalineament and the SouthAmerican respectively. continentnear Fortaleza, Tertiary volcanic centers have been identifiedand attributed to theFernando hot spot [Alineida, 1958; North American Plate Motion Over Hot Spots I andorosand Oliveira, 1968]. Rb/Sr ages yielded an age of 34 + 2 Ma[Guimaraes et al., 1982],and five K-Ar analysesyielded an North American plate motion over hot spotshas been ageof 28.7 + 2.5Ma. [Vandoros and Oliveira, 1968]. Recent reconstructed by the additionof North America/Africafinite K-Ardating of thesevolcanics indicates an age range of 26 to36 reconstructionpoles [Klitgord and Schouten, 1986] to Africa/hot Ma(unpublished data), in broad agreement with the age predicted spots motion (Table 3). A numberof interestingobservations can bythe trail of the Fernando hot spot (Figure 7). be maderegarding various central Atlantic hot spotsystems, on AscensionIsland is locatedon the SouthAmerican plate, in thebasis of thisadjusted absolute motion reconstruction (Figure closeproximity to themid-Atlantic spreading-axis (Figure 7). 8). Thepresence of unweathered basalt flows suggests that there has Themost prominent volcanic trail in thecentral Atlantic basin beenvolcanic activity on this island within the past few hundred is theNew England Seamount chain, which probably formed in years.Nine K-Ar ages from widely distributed sample sites, association with the Muir and Corner seamount complexes rangefrom ages too young to beaccurately dated by the K-Ar (Figure8). Radiometricages for thisseamount chain confirm

60 N

Greenland Knoll o o ß

- [ ... p- : /

• ModelA 50 N Model B ', ,,," ;'• . •000ro

O • '- -• - North America • • •--'x. •';'v

40 N • •' ,,, -•-• ß

,.' o,?; Bermuda o , O• / .' • Madeira 30 N '/' '•_•. •/' ?__• ' P Canaries. earCherSrnt• eo,. '"'= "" ] Hotspot Anomaly/ NewEngland Africa 20 N

' '• [ •t Verdesape •. ReSearcherRingeI • .•x•xe.• 10N South America •razilianRidge 0 0 lOO W 90W 80W 70W 60W 50W 40W 30W 20W 10W Fig.8. Themodeled trails of the African, North American andEarasian plates over postulated hotspots. Tick marks along hot spottrails represent theprogress ofplates in10 m.y. intervals. Relative motion parameters forNorth American/Africa motion [KlitgordandSchouten, 1986], and for Eurasia/North America [Srivastava andTapscott, 19861 were added toAfrican/hot spots motion(Table 3) Theintermediate wavelength geoid anomaly shown isfrom Vogt et al. [1984]. Solid circles represent seamountswitharelief greater that50 fathoms (300 feet) (91 m) [after Epp and Smoot, 19891. Radiometric agesshown 1orthe NewEngland Seamounts arefrom Duncan [1984]. Mercator projection. 17,492 O'CONNORAND DUNCAN:WALVIS RIDGE AND RIO GRANDERISE HOT SPOTSYSTEM thatit is ageprogressive [Duncan, 1984] (Figure 8). Previous + 1.5 Ma has been determinedfor trachytedredged from the authorshave postulated that the zero age locationof the New Newfoundlandseamount [Sullivan and Keen, 1977], compatible Englandhot spot is eitherthe Great Meteor seamount, or a nearby with the agepredicted by the reconstructionof NorthAmerican location.However, this study indicates that a zeroage position of platemotion over a GreatMeteor plume (Figure 8). Earlier theNew England hot spot needs to be locatedat, or in theregion activity causedby this hot spot may be marked by 115-145 of 27øN,43.0øE, on the African plate, for thepredicted trail of the alkalineintrusions along the coastof Newfoundland[Clark, NorthAmerican plate over the New Englandhot spotto agree 1977]. broadlywith the geometryand distribution of radiometricages Usingthe empiricalage-depth curve for the AtlanticOcean alongthe New EnglandSeamount lineament. The agreement crust, Tucholkeand Smoot [1990] have shown that the Comer betweenthe geometriesof the hot spottrack and the volcanic seamountsformed at about80 Ma to 76 Ma, while on theAfrican lineamentis nonethelessstill poor. The Corner Seamounts plate,the conjugate Cruiser plateau, formed at 76 Ma. Volcanism formedwhen the spreading-axiswas in closeproximity to a hot appearsto havemigrated northward from GreatMeteor Seam0unt spot between80 Ma and 76 Ma [Tucholkeand Smoot, 1990] and in early Cenozoic time, and then southwardto Great Meteor may also have been formed by the New England hot spot. Seamount in the late Cenozoic, with renewed volcanism Interestingly,this postulated location of theNew Englandplume occurringon some of theseseamounts 20 to 30 Ma aftertheir adjoinsa largeintermediate wavelength anomaly, which straddles initial formation over a hot spot. The existence of the centralAtlantic spreading-axis [Vogt eta!., 1984] (Figure8). "intralithosphericconduits" of several hundred kilometersin Suchintermediate wavelength geoid anomalies generally coincide lengthmay have maintaineda link betweenthese seamounts and with major topographicstructures such as subductionzones, the hot spot[Tucholke and Smoot,1990]. If plume materialfrom mid-plateswells, and mid-oceanicridges, and are consideredto the Great Meteor and New England plumes were supplied reflect sublithosphericprocesses. There is, however, no simultaneouslyto the spreading-axisbetween 80 Ma and76 convincingevidence to supportthe existenceof thisplume from the combinedflow of magmafrom thesetwo hot spotsmight currently available studiesof the isotopic and trace element account for the construction of the N-S elongated, Comer composition of lavas erupting along this section of the Seamounts-CruiserPlateau (Figure 8). mid-Atlanticspreading-axis [Schilling, 1986]. The motionof the Eurasianplume over the Azoreshot spot The Mesozoic plutonic complexesof northeasternNorth was calculatedby the additionof Eurasian/NorthAmerica relative America, the White Mountain Intrusivesof New Englandand the motion [Srivastava and Tapscott, 1986] to that of African/hot MonteregianHills of Quebec,probably represent a continuation spots. The predicted trail of the Azores plume intersectsthe of the trail of the New Englandhot spotbeneath the North southernopening of the Rockall Trough (Figure 8) at the 105Ma American continent,as first proposedby Morgan [1972]. locationon hot spottrack. This resultsuggests a link betweenthe Regionaluplift of southeasternCanada and New Englandin Azoresplume and the initiationof rifting in the RockallTrough. Cretaceous-earlyTertiary time has been attributed to thepassage Srivastava and Tapscott [1986] concluded that the initiationof of this region of the North Americanplate over a hot spot rifting betweenthe PorcupineBank and the OrphanKnoll andin [Crough,1981]. Foland et al. [1988] havereported that the Nd the Rockall Trough began at anomaly M0 (118 Ma). From and Sr isotopicsignatures of the parentalmagmas for the New between anomalies M0 and 34 active seafloor spreadingwas England-Quebeccontinental plutons are similar to one another, initiated in the North Atlantic and Rockall Trough (all regions regardless of whether the magma intruded Precambrian or southof the Charlie Gibbs FractureZone). By anomaly34 time, Paleozoiccrust, and also similar to thosefor the New England seafloor spreading had ceased in the Rockall Trough and seamounts,described by Taras and Hart [1987]. Foland and commenced to the west in the Labrador Sea. A more recent Fau! [1977] reportedthat K-Ar ages,supported by Rb-Sr and expressionof the Azores plume would appear to be the Azores U-Pb data, indicate that the White Mountain Intrusives constitute BiscayRise (Figure 8), as first suggestedby Morgan [1981]. On distinctage groupsof 230, 200-156, and 124-100 Ma. Morgan the North American plate, the Milne seamounts, locatedon [1983] andCrough [1981] accountedfor thisgrouping of agesby betweenapproximately anomaly 32 to anomaly34 seafloor,may suggestingthat a series of different hot spotshave been active recordthe trail of the Azores plume on the North Americanplate. beneaththe White Mountain region, with the most recent age period of volcanismassociated with the New Englandhot spot. Hot Spot Fixity Foland et al. [1986] and Gilbert and Foland [1986] determined, onthe basis of 40Ar-39Ardating, that the Monteregian Hill The cornerstone of all hot spot or absolute motion centersformed within a shortinterval of timeat 124Ma. The age reconstructionsis the assumptionthat hot spots occur above distributionalong the New England-Quebecplutonic lineament is plumeswhich rise from, as yet undefined,depth(s) in the mantle in very broadagreement with the hot spottrack shown in Figure to the baseof the lithosphere,and that suchplumes remain fixed 8. However, it shouldbe kept in mind that the 120 and 130 Ma globally with respectto one anotherover geologicallysignificant rotationpoles for Africa/hotspots, on which thisreconstruction is periodsof time (e.g., 100 m.y.). However,a globalextension of based, are not well constrained. the finite reconstructionparameters presented here in orderto The Great Meteor Seamount(30øN, 29øW) (Figure8) may furthertest this hypothesis, is beyondthe scopeof thispaper. The reflect the earlierlocation of a plume,now extinct,which formed researchreported here has focused on estimatingthe mostvalid the Newfoundlandseamounts and/or Fogo Seamounts (Figure 8); setof reconstructionparameters for Africanplate motion over hot both of these seamountcomplexes are locatedon Cretaceous spots, on the basis of presently available data for hot spot Quiet Zone to olderseafloor [e.g., Klitgord and Schouten,1986]. generatedvolcanic lineamentson the African plate. A better The modeledtrack of thisplume assumes a zeroage for the Great understandingof the historyof interactionbetween the mid- Meteor Seamount;however, K-Ar agesof 11 Ma and 17 Ma have Atlanticspreading-axis and the Tristan hot spothas contributed beendetermined [Wendt etal., 1976]. An 40Ar-39Ar age of 97.7 informationas to how well particularportions of the Walvis O'CONNORAND DUNCAN: WALVIS RIDGE AND RIO GRANDE RISE HOT S POT SYSTEM 17.493

Ridge-RioGrande Rise volcanic system actually record African axisaway from the plume. The absence of a volcanictrail linking andSouth American plate motionsover the Tristan hot spot. the southeastern terminus of the Rio Grande Rise to Tristan da Additionalconstraints on Africanplate motion over hot spotsare, Cunha may best be explainedin terms of sucha westward however,critical to further adjustmentsof African absolute migrationof the spreading-axis.which eventually shut off the motion. Ongoing radiometricand isotopic studiesof basalts supplyof hot materialfrom the upwellingplume. The changein recoveredfrom the St. Helena seamountlineament, the seamount Walvis Ridge morphologyfrom that of aseismicridge to an chainextending eastward from Gough Island, and additional indistinctline of discreteseamounts and guyors (Figure 1 a) also samplesfrom the Walvis Ridge-Rio Grande Rise, should provide reflects,no doubt,this transitionfrom on-axisto intraplate suchadditional constraints on Africanplate motionover hot spots volcanism. [J.M. O'Connoret al., manuscriptsin preparation]. Africa/hot In orderto betterillustrate the changingspatial relationship spotsreconstruction poles given in Table 3 (adjusted,if betweenthe spreading-axis, the hot spot. and the evolving Walvis necessary,in the light of ongoing studiesof South Atlantic Ridge-Rio Grande Rise, a seriesof reconstructionsfor selected volcanictrails) will be extendedglobally [J. M. O'Connoret al., spreading anomaly times is shown in Figure 9. If a hot manuscriptin preparation].Such a studywill evaluatediffering spot-generatedseamount or ridge formed at a spreading-axis,it adjustmentsto Africa/hotspots reconstruction parameters and will be surroundedby spreadinganomalies of the same age, leadto a further test as to whether or not hot spotsare fixed with whichmay on occasionbe mappedonto the hot spottrail (Figure respectto oneanother. While uncertainty ellipses, following the 2). Conversely, if hot spot magma erupts through older, methodof Stock and Molnar [ 1983], have not been calculated for previouslyformed seafloor,an age differencewill be apparent the reconstructionsdiscussed above, typical errors involved in between the age of the hot spot feature and that of the extendingAfrican/hot spotsmotion globally throughrelative surroundingseafloor. The age data determinedin this study motionplate circuits are +_ 100 km [Molnarand Stock,1987]. thereforepresent an excellentopportunity to modelthe changing spatialrelationship between the spreading-axis,the hot spot,and SPATIAL REI•AT!()NSHIP BE'I'WI:t-N I'!iE SPREADING the evolving Walvis Ridge-Rio Grande Rise volcanic system, AXIS AND 'I'!tE TRIS'!'AN H{)T SPI)T sinceanomaly 34 (84 Ma) time. A suggestedprerifting-rifting configurationof Africa, South For the period of time when the Walvis Ridge-Rio Grande America, the Parana-Etendekaflood basalts,and a postulated, Rise system was forming in a synchronous,age-progressive largediameter hot spot,is shownin Figure9a. From anomaly34 manner (apart from possible Eocene reactivation), the time, spreadinganomalies, which were forming at the time of mid-Atlanticspreading-axis must have been locatedastride, or in each particular reconstruction,were rotated, along with the closeproximity to the Tristan hot spot. Tristan da Cunha, which Walvis Ridge bathymetry (and African continent), to their is situatedon anomaly C6 (19 Ma) African plate seafloor, is respectivepredrift positions. An appropriaterotation parameter locatedsome 550 km east of the present day spreading-axis was calculatedfor each reconstructionfrom the Africa/hot spots (Figure2). This---550 km separationbetween the spreading-axis finite polesgiven in Table 3. The correspondinganomaly picks andthe hot spotpoints to a westwardmigration of the spreading- on the South American plate, the Rio Grande Rise bathymetry

10S

20 S

SOUTH AMERICA AFRICA I 30 S

40 S

50 S

1500 m African contour

...... 1500' m South AmericanI contour "I,..,.,"•' ~13• MarecOnstructiOn 6O S 60 w 50 W 40 W 30 w 20 w 10 w 0 10 E 20 E 30 E Fig. 9a. The shadedregion represents a nypuum•,/•cd, large-diameter hot spot, which formed above the upwelling Tristan plume (opencircle) at -130 Ma. A 130 Ma reconstructionparameter for Africa/hotspots motion was calculated such that the Etendeka floodbasalts lie alongthe southeastern edge of thislarge diameter plume. Theclosure parameter of Martin et al. [1981],when combinedwith this 130 Ma Africa/hotspot rotation parameter, rejoins South America to Africa. Projectionis centeredon Tristan da Cunha. 17,494 O'CONNORAND DUNCAN:WALVIS RIDGE AND RIO GRANDE RISE HOT SPOT SYSTEM

0 0 0 0 0 •- C•I CO • LOlli 0 0 0 0 • Ill - 0

LU 0

Ill 0

0

50 (0 •o o .•o •/_.'•-• 6452(I)

0 0 0 0 0 0 0 0 0 O'CONNOR AND DUNCAN: WALVIS RIDGE AND RIO GRANDE RISE HOT S POT SYSTEM 17,495

0 0 0 0 0 0 0 0 0 0 ,-- •X• C'r) •!' tt• I•1 •- • CO • lad 0 0

0 0 ED 0 .-- {Xl CO • CO ili

•-' • I•Iøo

o

o

o 17,496 O'CONNORANDDUNCAN: WALVIS RIDGEAND RIO GRANDERISE HOT SPOTSYSTEM

(and the SouthAmerican continent) were, in turn,rotated to their GrandeRise. The migration over the plume of thefracture zone, conjugate,predrift positions, by the additionof the appropriate which representsthe southernboundary of spreading-axis SouthAmerica/Africa relative motion [Cande et al., 1988]. The asymmetry(located to the north of Tristanda Cunha in Figure2), extentto whichthe WalvisRidge-Rio Grande Rise bathymetry may have switchedthe flow of plume materialfrom theRio had evolved at the time of each reconstruction is constrained Grande Rise to the Zapiola seamount region of the primarilyby thebasement ages reported in thisstudy. spreading-axis.The northwardmigration of thisfracture zone, The C34 (84 Ma) reconstructionplaces the spreading-axisover the hot spot, could explain why untilabout C21 (48.6Ma) astridethe Tristan plume. Largevolumes of hotmaterial erupted theseafloor isochrons identified to the south of theWalvis Ridge alongthis divergent boundary and formed the conjugate western wereolder than those located on andto thenorth of theridge and eastern plateausof the Rio Grande Rise and the Walvis (Figure2). The subsequentrapid disappearance of thisage offset Ridge, respectively(Figure 9b). The significantage-offset betweenthe seafloor located north and south of theWalvis Ridge between seafloor located to the north versus that to the south of is thereforebest explained by the combinedeffects of a westward theWalvis Ridge (Figure 2) canbe attributed to a priorhistory of migrationof the spreading-axisaway from the hot spotand an easterlyand southeasterly spreading-axis migrations beginning in associated northward migration over the hot spot of the Albian to late Aptiantime, asdiscussed in an earliersection. The previouslyformed African lithosphere. configurationof the spreading-axis,volcanic structure, and the At C13 (35.3 Ma) the Zapiola Seamountcomplex ceasedto hot spotin anomaly34 time is strikinglysimilar to thatwhich evolve. A mirror imageof theZapiola Complexcould have been existsin presentday Iceland. formed on the African plate by the westwardflow of plume At C33r (80 Ma) the northeasternshoulder of the Rio Grande materialto the spreading-axis,beneath the lithosphere. If sucha Rise andthe westernedge of the WalvisRidge plateau were feature existed it would eventually have been carried over the formed. The narrowridge which extendssouthward from the Tristanplume and have been incorporated into the Walvis Ridge. Wa!vis Ridge plateau,was constructedalmost as far southas the There is geochemical evidence to suggest that basaltic 78-79 Ma DSDP transect,in conjunctionwith thewestern side of magmas,which have eruptedat the presentday spreading-axis, the Rio GrandeRise Plateau. BetweenC33r (80 Ma) and C32 locateddue west of Tristanda Cunha,are slightly "enriched" by (71.4 Ma) time, this narrowridge continuedto developin the continuationof westwardflow of plume materialbeneath the conjunctionwith the N-S easternridge of the Rio GrandeRise. African plate [Humphris et al., 1985]. Such a continuous This N-S ridgeapparently formed along a spreading-axis,which enrichmentof the magmasforming African lithosphereat the was continuouslypropagating southward and southeastward latitudeof Tristanda Cunhawould have varied with increasing toward the hot spot,apparently in orderto retainits position separationof the plume and the spreading-axis,which began above the upwelling plume. Massive amountsof hot material about70 Ma. This enrichedgeochemical signal, trapped in the musthave been supplied to the spreading-axisby the plumein Africanplate, might have been mixed into Walvis Ridge magmas, order for this massive lineament to have formed within such a as this variable enrichedAfrican lithospheremigrated overthe relativelyshort time interval. As the spreading-axispropagated hot spot. Additionalinformation about the history of hot spotand southwardit invadedthe previouslyformed seafloor which was in spreading-axisinteractions is provided by a number of other theprocess of migratingnorthward over the hot spot. Thus, while approaches(e.g., geochemistry and isostasy) discussedin the the SouthAmerican plate may in fact havemigrated in a NW-SE following section. directionover the hot spot, the N-S ridge of the easternRio Grande Rise formed along a southwardlypropagating, hot GEOCHEMICAL COMPOSITIONS spot-fedspreading-axis, and apparentlydid not recordthe true directionof SouthAmerican plate motion over the plume. As a Magmasderived from hot spots(ocean island basalts or OIB) resultof thiscontinued southward propagation, the asymmetry of and from seafloorspreading-axes (mid-ocean ridge basaltsor the spreading-axisbecame most pronounced between C33r and MORB) are compositionallydistinct, especially in certaintrace C32 times. element and isotopic ratios. The compositionsof Walvis By C31 (68.5 Ma), or earlier, the Walvis Ridge had Ridge-Rio GrandeRise basaltsvary in incompatibleelement commencedto bifurcateinto three subparallellineaments. The abundancesand ratios, rare earth elements(REE), and Sr, Nd, Pb northernridge formed on thespreading-axis, in conjunctionwith isotopic ratios [Humphris and Thompson, 1982, 1983; the southeasternterminus of the easternRio Grande Rise, while Richardsonet al., 1982, 1984;Thompson and Humphris,1984; to the southeasta subparallelridge was forming in an intraplate Thompsonet al., 1983;Humphris et al., 1985]. situation. Between C32 (71.4 Ma) and C30 (66.7 Ma) times a A decreasein the Zr/Nb ratio of the hot spothas occurred majorspreading-axis jump occurred,which may have represented sincethe formationof the Etendeka-Paranaflood basalts(Figure the final effortby thespreading-axis to regainits position over, or 10), which may be related to changesin the plate tectonic close to, the hot spot. Thus, by about C30 (66.7 Ma) time, or environmentassociated with the westwardmigration of the earlier, the northern side of the Walvis Ridge (and the spreading-axis(Figure 9b). The elementsZr and Nb (and alsoY southeasternportion of theRio GrandeRise) was evolving as a andthe REE elements)are of particularuse in the investigationof plume-fed spreading-axis,while the southernportion of the magmasource regions, as theseelements are highlyincompatible Walvis Ridgewas forming in an intraplateenvironment, as older and their ratio will be influencedonly slightlyby variationsin seafloorbegan to migrateover the hot spot. partial melting and crystal fractionation [Erlank and Kable, By aboutC21 (48.6 Ma) time,the hot spotcould no longer 1976; Sunet al., 1979;Pearce and Norry, 1979]. They are also sufficientlyaugment the volume of hotmaterial upwelling along among the trace elementswhich are most highly resistantto the westwardly migrating spreading-axis for continued mobilizationby seawateralteration [Thompson, 1973]. constructionof the Rio GrandeRise. The Zapiola seamount High Zr/Nb ratios are evident in Etendeka-Parana,DSDP complex beganto evolve aboutthis time to the southof the Rio 516F (Rio GrandeRise plateau),CH 18 and CH 19, AII-93-21, O'CONNOR AND DUNCAN: W ALVIS RIDGE AND RIO GRANDE RISE HOT S POT SYSTEM 17,497

'Low Zr/Nb' hasalta lO0 X DSDP528 * ill DSDP 527 90 ß DSDP 530A ß Aii-93-3-1D * M -3 -9 -3-25 * 80 -5-3 * -7-! * -8-1! * 70 -12-4 -14-IA" -14-19 * -14- (3 samples) 60 AK-I699-3 II• RC16-10-!l(Eocene volcanism?) 50 'Intermediate Zr/Nb' basalts ß AII93-10-11 * V29-9-1" 40 AII-93-10 (3 samples) 'High Zr/Nb' basalts _

30 [] DSDP525* + DSDP 516F" O AII-93-21-1 * 2O AII-21 (N = 2) V29-11(N = 4) a CH18&CH19 (N = 7) 10 COlatinental Flood Basalts

-•- Etendeka(N = 39,15 & 5 ) A Parana(N = 14. 8, & 2) 0 5O 100 150 200 250 300 350 400 450 500

Zr (ppm) Fig. 10. Zr/Nb ratiosfor theWalvis Ridge-Rio Grande Rise hot spot system can be dividedinto "high", "intermediate", and "low"ZrfNb basalts. "High" Zr/Nb basalts (apart from present-day spreading-axis basalts) are associated with continental flood andon-spreading-axis, hotspot volcanism. In contrast,"low" Zr/Nb ratios are associated with intraplate hot spot volcanism and hotspot-fed spreading-axis (i.e., DSDP 527,528, and 530). Figure9b illustratesthis correlation between decreasing Zr/Nb ratios andthe transition from on-spreading-axis to intraplate volcanism, within the Walvis Ridge-Rio Grande Rise system. Gough Islandis a recentlyactive volcano located to thesoutheast of Tristanda Cunha.Sources of data:Pararia [Fodor et al., 1985], Etendeka[Marsh, 1987], Walvis Ridge dredges [Humphris and Thompson, 1982, 1983], DSDP 530 and CH18 and 19 [Humphris andThompson, 1983], DSDP 525A, 527 and 528 [Thompson and Humphris, 1984], DSDP 516F [Thompson etal., 1983;Weaver etal., 1983],Tristan da Cunha [Weaver et al., 1987],Gough Island [le Roex,1985; Weaver et al., 1987],and MORB [Humphris etal., 1985].Asterisks denote samples for which 2*0Ar-39Ar ages have been determined. N = numberofanalyses forwhich a mean value has been determined.

V29-I 1 (easternWalvis Ridge) and DSDP 525A (centralWalvis of binary mixing between"enriched" (Tristan) and "depleted" Ridge)basalts (Figures 5 and 6). Apart from the flood basalts, (MORB) end-membersources is evidentin spreading-axisbasalts thesehigh Zr/Nb ratios may be characteristicof an on-axis recovered due west of Tristan da Cunha. This indicates a situation(Figure 9b), as proposedby Humphriset al. [1985]. In continuedsub-lithospheric connection between the Tristan hot sucha tectonicenvironment large volumesof magmafrom the spotand the spreading-axis[Humphris et al., 1985];however, uppermantle MORB sourceregion might have diluted the signal earlier enrichmentor fertilization of the upper mantle by the of the OIB source,as occursat the main central rifting zone in plumemay alsobe of importance. Iceland[e.g., Schilling et al., 1982, 1983]. The "enriched"and The remainingbasalts recovered from the WalvisRidge all "depleted"MORB type recoveredfrom the spreading-axisabout exhibit low to intermediate Zr/Nb ratios, characteristic of their -550 km to the west of Tristanda Cunha [Humphriset al., 1985] havingerupted in an intraplatetectonic environment (Figure 9b). are the only known volcanicrocks in this regionwith higher Basaltsdredged from the Rio GrandeRise plateau(.probably of Zr/Nb ratios. Eocene age) exhibit low Zr/Nb ratios, in contrastwith the A transition to low Zr/Nb ratios is first recorded within the underlying,on-axis, high Zr/Nb basementbasalts (DSDP 516F) WalvisRidge-Rio Grande Rise systemat DSDP sites528 and (Figure9b). 527 (79-80 Ma). However, low Zr/Nb basalts have been Low Zr/Nb ratios for the island of Tristan da Cunha may recoveredfrom DSDP site 530, locatedabout 20 km to the north resultfrom thefact that,at very low degreesof partialmelting, Zr of the Walvis Ridge (Figure 7), close to the African coast. doesnot behaveas incompatiblyas Nb; this couldexplain why Variationsin ratios of Zr/Y or Y/Nb, with Zr/Nb, displayedby the trend of the Tristan da Cunha Zr and Nb data intersectsthe Zr DSDP530, 527, and528 basalts,correspond well with predicted axis,rather than the origin[Weaver et al., 1987]. GoughIsland mixingcurves calculated for depletedMORB and an enriched, samplesshow a slight,but distinct decrease in Zr/Nbbetween the Tristan-typeend-member source, in contrastto DSDP 525A and lower basalt series(about 1.0 Ma) and the upper basalt series dredgedbasalts [Humphris and Thompson, 1983]. Thispoints to (about 0.3 Ma) [le Roex,19851, due possibly to the influenceof 527and 528 basaltshaving formed at a hot spot-fedspreading decreasingdegrees of partialmelting. ridge,as was the casefor 530 basalts,which had formed at an Humphrisand Thompson [1983] and Humphris et al. [1985], earliertime [Humphrisand Thompson, 1983]. A similarpattern afterChen and Frey [1983],have suggested that as a transitionto 17,498 O'CONNORAND DUNCAN: WALVIS RIDGE AND RIO GRANDE RISE HOT SPOT SYSTEM an intraplateenvironment occurs, low degreesand low volumes an eastern(African coast to about3øE), a central(about 3øE to of partial melting of the surroundingdepleted lithosphere (i.e., about0ø), and a westernsection (about 0 ø to Tristanda Cunha). previously formed African plate) will become important in An Airy modelcalculated with a crustalthickness of 20 krnis in determiningthe composition of thevolcanic trail. Spreading-axis goodagreement with the experimental results obtained for both and hot spot interactions may therefore have controlled the the easternWalvis Ridge and the Rio GrandeRise. Experimental mixing, and relative importance,of depletedMORB, enriched admittancesusing the platemodel with an elasticthickness of 8 OIB, and low degrees and amounts of partial melts of the km and 6 km best fit the data from the central and western depletedlithosphere, thus indirectly determining the composition sectionsof the Walvis Ridge, respectively. This gradationin of Walvis Ridge andRio GrandeRise basalts. modeledplate thicknesses between the central and western study As a plume rises,however, it warms the surroundingmantle areasprobably records the transitionfrom on-axisto intraplate material by thermal diffusion, thus lowering its viscosity hotspot volcanism. Diament and Goslin [1987], however, have sufficiently to cause large-scale entrainment and mixing of shown that the contributions of mechanical isostasy (off-axis asthenosphericsource material [Griffiths, 1986]. The transitionto formation)and densitycontrasts at depth(near-axis formation) lessdepleted magmas within the Reunionhot spotsystem (i.e., are very differentover adjoiningdomains along the Walvis decreasingMORB component)is bestexplained by reductionin Ridge.This can best be explainedfor theWalvis Ridge in terms the entrainmentof uppermantle material [White et al., 1990]. of a rather complex history of hot spot spreading-axis Sucha changein theextent of uppermantle entrainment may also interactions,as broadly documented in this study. haveplayed a role in controllingthe degreeof enrichmentof WalvisRidge-Rio Grande Rise basalts. Sr, Nd, andPb isotopic CHANGESIN SOUTH ATLANTIC OCEAN CIRCULATION analysesare presently being conducted on available Walvis Ridge and Rio Grande Rise basalts,in order to determinethe factors The transition from on-axis to intraplate hot spot volcanism responsiblefor controllingtemporal and spatialvariations in within the Walvis Ridge-Rio Grande Rise may be indirectly composition[J. M. O'Connoret al., manuscriptin preparation]. recordedby an abruptimprovement in the oxidationstate of In conclusion,the changefrom high to low Zr/Nb ratioswithin South Atlantic bottom water at 55 to 58 Ma. Liu [1982], Liu and the Walvis Ridge-Rio GrandeRise systembroadly supports a Schmitt[1984], Wanget al. [1986], Liu et al. [1988], andHu et model of transitionfrom on-axisto intraplatevolcanism. al. [1988] have proposedthat under reducingconditions, Ce existsin seawateras a trivalent ion and is thereforeincorporated ISOSTASYOF THE WALVIS RIDGEAND RIO GRANDERISE along with the other trivalent rare earth elements (i.e., lanthanides) into seafloor sediments, by adsorption onto The transitionfrom on-axisto intraplatehot spotvolcanism carbonateminerals. In contrast,under oxidizing conditions, Ce3+ withinthe Walvis Ridge-Rio Grande Rise system is supportedby isoxidized toCe 4+ and precipitates asvery insoluble hydroxides. numerousgravity and geoid studies [Derrick and Watts, 1979; This resultsin depletionof Ce relativeto the othertrivalent rare Bulotet al., 1984;Kogan et a/.,1985; Gibert and Court#lot, 1987; earth elements in both seawater and sediments. Such Ce Freedman, 1987;Diament and Goslin, 1987]. The gravity variationswithin marine carbonates represent a recordof changes signaturesof aseismicridges and seamounts are sensitive to the in palaeo-oceanbottom water redox conditions. age(i.e., strength) of thelithosphere at the time of construction. Carbonatesediment samples from DSDP 525A [Liu, 1982; Thosefeatures which formed in an intraplatesetting (i.e., old, Liu and Schmitt,1984], DSDP 530A, B, locatedin oceanfloor thickseafloor) are supported by theelastic stresses produced by 20 km northof the easternend of the Walvis Ridge [Wang et al., thebending of theplate; as the plate distributes the compensating 1986], and DSDP 516, 516F [Hu et al., 1988]have been analyzed massesover a widearea, the gravitysignature measured directly by the INAA techniquein orderto determinedowncore Ce overthe load will belarge. In contrast,volcanoes which form on variations. All three sites show gradual Ce depletions(i.e. younglithosphere (i.e., near a spreading-axis)will be locally reducingseawater conditions) from the late Campanian (-80 Ma) supportedso that the compensating masses are located directly until about 55 to 58 Ma, at which time an abrupt transitionto beneaththe load;the resulting gravity anomaly measured above presentday Ce values(i.e., oxidizingseawater conditions) the feature will be small. occurred.Hu et al. [1988] andLiu et al. [1988] haveattributed Gravityand geoid data measured over the Walvis Ridge and suchchanges in oxidationstate to improvements in the circulation the Rio Grande Rise have been correlatedwith bathymetry in of South Atlantic bottom water. Breachingof the Walvis orderto determinethe mechanism by whichthey are isostatically Ridge-RioGrande Rise barrier to bottomwater circulation, as a compensated[Detrick and Watts, 1979; Bulot et al., 1984;Kogan resultof thedivergence of thespreading-axis and the hot spot, is et al., 1985;Gibert and Courtillot,1987; Diament and Goslin, a probableexplanation for this improvementin circulation. 1987]. Derrickand Watts[1979] have shown that from the Wideningof theSouth Atlantic ocean basin and subsidence of the Africancoast to about3øE the Walvis Ridge is locallysupported RomancheFracture Zone (which actedas a barrierto deepbottom by an Airy-typethickening of thecrust with a modeledcrustal waterflowing between the Northand SouthAtlantic basins), thicknessin therange 15 km to 25 km. Westof longitude3øE, may,however, have also played significant roles. however,the WalvisRidge is bestexplained by a platemodel (i.e.,a thinelastic plate overlying a weak fluid) with a modeled CONCLUSIONS elastic thickness of between 5 and 8 km. These results are consistentwith a transitionfrom on-axis to intraplatehot spot Volcanism within the Walvis Ridge-Rio Grande Rise volcanismon theWalvis Ridge at about70 Ma. volcanic systemhas been shown to be fundamentally The mechanicalisostasy of the Walvis Ridgeand the Rio age-progressive.The originof therecently active island of GrandeRise hasalso been investigated by Bulotet al. [1984] Tristanda Cunha, the Walvis Ridge and Rio Grande Rise, and the usingconstraints imposed by SEASAT geoid data. In theirstudy, Etendekaand Paranacontinental flood basalts,can thereforebe theWalvis Ridge was divided into three overlapping subsections: attributed to a commonsource, the Tristan plume. O'CONNORAND DUNCAN: WALVIS RIDGE AND RIO GRANDE RISE HOT SPOT SYSTEM 17,499

The Walvis Ridge and Rio Grande Rise formed in a (J.O'C)wishes to acknowledgethe adviceand guidance of Lew Hogan synchronous,age-progressive fashion along a sectionof the South duringthe analyticalstages of thisstudy. Suggestions by reviewerSteve Atlanticspreading-axis, asthe African and South American plates Cande,an anonymousreviewer, associate editor Richard Gordon, and divergedapart, astride (or in closeproximity to) the Tristanhot unofficialreviewer Gordon Ness, were both insightful and extremely helpful.Pierrick Roperch gave helpful advice on techniques in computer spot.Reconstructions of the spatialrelationship between the graphics.A mapof WalvisRidge bathymetry was provided by Jean spreading-axis,thehot spot, and the evolving Walvis Ridge-Rio ClaudeSibuet. This work was supported primarily by the Office of Naval Grande Rise system have shown that at about 70 Ma the Research(grant N00014-87-K-0420). spreading-axismigrated to thewest, away from the hot spot. This westwardmigration of thespreading-axis was accompanied by a REFERENCES northwardmigration over the hot spot of previouslyformed Almeida,F. F. M., Geologiae petrologiado argquipelago de Fernandodo seafloor. As a consequence, a transition from on-axis to Noronha,Monogr., 13, Div. Geol.Min. Dep.Nac. Prod. Mineral., Rio de Janeiro, 1958. intraplatehot spot volcanismoccurred, which resultedin the Amaral, G., U. G. Cordani, K. Kawashita,and J. H. Reynolds, terminationof Rio GrandeRise formation. Supportfor this Potassium-argondates of basaltic rocks from Southern Brazil, conclusionisprovided by thefact that on-axis basalts have higher Geochim.Cosmochim. Acta, 30, 159-189, 1966. Zr/Nbratios in comparisonto those which erupted in anintraplate Anderson,D. L., Hot spots,polar wander, Mesozoic convection and the situation. Changesin the spatialrelationship between the geoid, Nature, 297, 391-393, 1982. Aumento,F., Vesicularityof mid oceanpillow lavas,Can. J. Earth Sci., spreading-axisandthe hot spot may have controlled the mixing, 8, 1315-1319, 1971. andrelative importance of, depletedMORB, enrichedOIB, and Austin,J. A., Jr., andE. Uchupi,Continental-oceanic crustal transition off partialmelts of the depletedocean lithosphere. Isostasy studies SouthwestAfrica, Am. Assoc.Pet. Geol. Bull., 66, 1328-1347, 1982. alsosupport a transitionto intraplatevolcanism on the Walvis Backman,J., R. A. Duncanet al., Proc. Ocean Drill. Program, Initial Ridgebeginning at about70 Ma. Rep.,115, 1085pp., 1988. Baker, B. H., and J. A. Miller, Geology and geochronologyof the In the light of an improved understandingof the Seychelles Islands and structureof the floor of the Arabian Sea, temporal-spatialevolution of thesurface expression of theTristan Nature, 199, 346, 1963. plume,i.e., the Walvis Ridge-Rio Grande Rise and associated Baker, I., N.H. Gale, and J. Simons,Geochronology of the St. Helena continentalflood basalt fields, rotation parameters for the motion volcanoes,Nature, 215, 1451-1454, 1967. Barker, P. E., Tectonicevolution and subsidencehistory of the Rio of the African plate over the Tristan hot spot have been GrandeRise, Initial Rep.Deep Sea Drill. Proj., 72,953-976, 1984. recalculated.These reconstruction poles predict a trail for the Barker, P. E., I. G. Gass, P. G. Harris, and R. W. LeMaitre, The Tristanhot spot on the African plate which agreeswith the volcanologicalreport of the Royal Societyexpedition to Tristan da distributionof agesalong the Walvis Ridge. The N-S trending Cunha, Philos. Trans. R. Soc.London., Ser. A, 256, 439-578, 1964. Blow, W. H., Deep Sea Drilling Project, Leg 3, Foraminiferafrom eastemridge of the Rio GrandeRise formed, in all probability, selectedsamples, Initial Rep.Deep Sea Drill. Proj., 3, 629-661, 1970. alonga hot spot-fed,southwardly propagating spreading-axis. Bolli, H. M., W. B. F. Ryanet al., WalvisRidge-sites 362 and363, Initial Thisportion of the Rio GrandeRise apparentlydoes not record Rep. Deep SeaDrill. Proj., 40, 183-356, 1978. theactual motion of the SouthAmerican plate over the hot spot. Brereton,N. R., Correctionsforinterfering isotopes in the 40Ar-39Ar Africa/hot spots rotation parameters were therefore not dating method,Earth Planet. Sci. Lett., 8, 427-433, 1970. constrainedto predict a trail which, on addition of South Bryan,W. B, andR. A. Duncan.,Age and provenanceof clastichorizons from hole 516F, Initial Rep. Deep Sea Drill. Proj., 72, 475-477, America/Africarelative motion poles, would follow the geometry 1983. of theN-S ridgeof the Rio GrandeRise. Rather,the hot spot Bukry, D., and M. N. Bramlette,Coccolith age determinationsleg 3, trackcuts a NW-SE pathacross the Rio GrandeRise, following Deep SeaDrilling Project,Initial Rep.Deep Sea Drill. Proj., 3,589- theNW-SE trending, elongated topography suggested for therise 612, 1970. Bulot, A., M. Diament,M. G. Kogan,and J. Dubois, Isostasyof aseismic fromstudies of theSouth Atlantic geoid. This resultillustrates tectonic units in the South Atlantic Ocean and geodynamic the critical importanceof a detailed understandingof the implications, Earth Planet. Sci. Lett., 70, 346-354, 1984. evolutionof individualhot spot systemssuch as the Walvis Cande, S.C., and P. D. Rabinowitz, Mesozoic sea floor spreading Ridge-RioGrande Rise, in estimatingreliable reconstruction borderingconjugate continental margins of Angolaand Brazil, Proc. Annu. OffshoreTechnol. Conf., 3, 1769-1776, 1978. poles. Cande, S.C., and P. D. Rabinowitz, Magnetic anomalies of the Thereconstructed motion of theAfrican plate over hot spots continentalmargin of Brazil, Offshore Brazil Map Ser, American hasbeen extended to theNorth and South American plates by the Associationof PetroleumGeologists, Tulsa, Okla., 1979. additionof appropriaterelative motion poles. The evolution of Cande, S.C., J. L. LaBrecque,and W.B. Haxby, Plate kinematicsof the South Atlantic: Chron 34 to present, J. Geophys. Res., 93, allsignificant hot spot systems on both of theseplates has been 13,479-13,492, 1988. discussedin detail. In orderfor thepredicted track of theNew Cande, S.C., J. L. LaBrecque, R. L. Larson, W. C. Pitmann III, X. Englandhot spot to passthrough the region of theNew England Golovchenko,and W. F. Haxby, Magnetic Linearionsof the WorM's seamounts,the zeroage locationof this plumemust be locatedin Ocean Basins (map). Scale: approximately 1:27,400,000 at the equator,American Association of PetroleumGeologists, Tulsa, Okla., thevicinity of thecentral Atlantic spreading-axis at about 27øN. 1989. Theexistence of a prominentintermediate-wavelength geoid Chaffey, D. J., R. A. Cliff, and B. M. Wilson, Characterizationof the St. anomalyin this locality has been interpretedas supporting Helena magmasource, edited by A.D. Saunders,and M. J. Norry, evidencefor this conclusion.The Azoreshot spothas been Magmatism in the Ocean Basins, Spec. Publ. 42, Geol. Soc. of linkedto the initiationof rifting, and consequentseafloor London,pp. 257-276, 1989. Chave, A.D., Lower Paleocene-upperCretaceous magnetostratigraphy, spreading,in the Rockall Trough. sites525, 527, 528, and 529, Initial Rep. Deep Sea Drill. Proj., 74, 525-532, 1984. Acknowledgments.The authorswish to thankGeoff Thompson who Chen, C.-Y., and F. A. Frey, Origin of Hawaiian tholelite and alkalic providedthe dredge samples used in thisstudy. Steve Cande generously basalt, Nature, 302,785-789, 1983. providedmagnetic anomaly data, and poles of relativemotion in advance Cherkis, N. Z., H. S. Fleming, and J. M. Brozena, Bathymetry of the of publication.Brent Dalrymple and colleagues kindly provided the South Atlantic Ocean-3øSto 40øS, Map and Chart Ser. MCH-069, computerprogram used for the reductionof Ar-Ar data. One of us Geol. Soc.of Am., Boulder,Colo., 1989. 17,500 O'CONNOR AND DUNCAN: WALVIS RIDGE AND RIO GRANDE RISE HOT S POT SYSTEM

Chevallier,L., andN. Vatin-Perignon,Volcanostructural evolution of Province,DSDP site 359, leg 39, Initial Rep. Deep Sea Drill., Proj., Pitondes Neiges, Reunion Island, Indian Ocean, Bull. Volcanol.,45, 39, 525-536, 1977. 287-298, 1982. Fodor,R. V., C. Corwin,and A. Rosinenberg,Petrology of SerraGeral Ciesielski,P. F., Y. Kristoffersenet al., Proc. OceanDrill. Program, (Parana)continental flood basalts, southernBrazil: Crustal Initial Rep.,114, 801 pp., 1988. contamination,source material, and South Atlantic magmatism, Clark,D. B., TheTertiary volcanic province of BaffinBay, Geol. Assoc. Contrib. Mineral. Petrol., 91, 54-65, 1985. Can. Spec'.Pap., 16, 445-460, 1977. Foland,K. A., andH. Faul,Ages of the White MountainIntrusives-New Cordani,U. G., Idadedo vulcanismono OceanoAtlantico Sul, Bol. Inst. Hampshire,Vermont and Maine, U.S.A., Am. J. Sci.,277, 888-904, Geocienc.Astron., Univ. Sao Paulo, 1, 9-75, 1970. 1977. Cordani,U. G., P. L. Sartori,and K. Kawashita,Geoquimica dos Foland,K. A., L. A. Gilbert,C. A. Sebring,and J.-F. Chen, 40Ar/39Ar isotropesde estroncioe a evolucaoda atividadevulcanica na Baciado agesfor plutonsof the MonteregianHills, Quebec:Evidence for a Parana,An. Acad.Bras. Cienc.,52, 811-818, 1980. singleepisode of Cretaceousmagmatism, Geol. Soc. Amer. Bull., 97, Cox, K. G., Therole of mantleplumes in thedevelopment of continental 966-974, 1986. drainagepatterns, Nature, 342, 873-877, 1989. Foland,K. A., J.-F. Chen,L. A. Gilbert,and A. W. Hofmann,Nd andSr Crough,S. T., Mesozoichot spotepeirogeny in easternNorth America, isotopic signaturesof Mesozoic plutons in northeasternNorth Geology, 9, 2-6, 1981. America, Geology,!6, 684-687, 1988. Crough,S. T., W. J. Morgan,and R. B. Hargraves,Kimberlites: Their Freedman,A. P., Marinegeophysical applications of Seasataltimetry and relationto mantlehot spots,Earth Planet. Sci. Lett., 50, 260-274, the lithosphericstructure of theSouth Atlantic Ocean, Ph.D. thesis, 1980. Mass.Inst. Technol., Cambridge, 1987. Dalrymple,G. B., andD. A. Clague,Age of theHawaiian-Emperor Bend, Gartner,S., Coccolithage determinationsleg 3, DeepSea Drilling Earth Planet. Sci. Lett., 31, 317-321, 1976. Project,Initial Rep. Deep Sea Drill. Project, 3, 613-627, 1970. Dalrymple, G. B., and M. A. Lanphere,Potassium Argon dating: Gibert, D., and V. Courtillot, Seasataltimetry and the South Atlantic Principles,Techniques, and Applicationto Geochronology,W. H. geoid, 1, Spectralanalysis, J. Geophys.Res., 92, 6235-6248, 1987. Freeman,New York, 258 pp., 1969. Gibert, D., V. Courtillot,and J.-L. Olivet, Seasataltimetry and theSouth Dalrymple,G. B.,and M. A. Lanphere,40Ar/39Ar technique of K-Ar Atlanticgeoid, 2, Short-wavelengthundulations, J. Geophys.Res., 94, dating:A comparisonwith theconventional technique, Earth Planet. 5545-5559, 1989. Sci. Lett., 12,300-308, 1971. Gilbert,L. A., andK. A. Foland,The Mont St. Hilaireplutonic complex: Dalrymple,G. B., andJ. G. Moore,Argon 40: Excessin submarinepillow Occurrenceofexcess 40Ar and short intrusion history, Can. J. Earth basalts from Kilaueau Volcano, Hawaii, Science, 161, 1132-1135, Sci., 23, 948-958, 1986. 1968. Gillot, P.-Y., andP. Nativel, K-Ar chronologyof the ultimateactivity of Dalrymple,G. B., E. C. Alexander,Jr., M. A. Lanphere,and G. P. Kraker, Piton des Neiges, Reunion Island, Indian Ocean, J'. Volcanol. Irradiationofsamples for 40Ar/39Ar dating using the Geological Geotherm. Res., 13, 131-146, 1984. SurveyTRIGA reactor,U.S. Geol.Surv. Prof. Pap., 1176, 1981a. Gillot, P.-Y., and P. Nativel, Eruptive history of Piton de la Fournaise Dalrymple,G. B., M. A. Lanphere,and D. A. Clague,Conventional and volcano,Reunion Island, Indian Ocean, J. Volcanol. Geotherm.Res., 40Ar/39Arages of volcanic rocks from Ojin (site 430), Nintoku (site 36, 53-65, 1989. 432), and Suiko (site 433) seamountsand chronologyof volcanic Griffiths, R. W., The differing effects of compositionaland thermal propagationalong the Hawaiian-EmperorChain, Initial Rep. Deep buoyancieson the evolution of mantle diapirs, Phys. Earth Planet. Sea Drill. Proj., 55, 659-676, 1981b. Inter., 43, 261-273, 1986. Dalrymple, G. B., M. A. Lanphere, and M. S. Pringle, Correlation Guimaraes, I. P., A. N. Sial, and A. F. Silva Filho, Petrologiae diagramsin40Ar/39Ar dating: Is therea correctchoice?, Geophys. geochimicada provincia alcalina Terciaria Fortaleza, Cera, An. Res. Lett., 15, 589-591, 1988. Congr.Bras. Geol., Salvador2, XXXII, 577-588, 1982. Detrick, R. S, and A. B. Watts., An analysisof isostasyin the world's Hartnady,C. J. H., and A. P. le Roex,Southern Ocean hot spottracks and oceans:Three aseismicridges, J. Geophys.Res., 84, 3637-3655, the Cenozoic absolute motion of the African, Antarctic, and South 1979. Americanplates, Earth Planet.Sci. Lett., 75, 245-257, 1985. Diament, M., and J. Goslin, Mechanical and thermal isostasy of Hawkesworth, C. J., M. S. M.P. N. Mantovani, P. N. Taylor, and Z. submarineplateaus, (abstract), Eos Trans. AGU, 68, 1462, 1987. Palacz, Evidence from the Parana of south Brazil for a continental Duncan, R. A., Hot-spotsin the southernoceans-An absolute frame of contributionto Dupal basalts,Nature, 322,356-359, 1986. referencefor the motionof the Gondwanacontinents, Tectonophysics, Helgason,J., Shifts of the plate boundaryin Iceland: Some aspectsof 74, 29-42, 1981. Tertiary volcanism,J. Geophys.Res., 90, 10,084-10,092, 1985. Duncan, R. A., Age progressive volcanism in the New England Herz, N., Timing of spreadingin the South Atlantic: Information from Seamountsand the openingof the centralAtlantic Ocean, ./. Geophys. Brazilian alkalic rocks, Geol. Soc. Am. Bull., 88, 101-112, 1977. Res., 89, 9980-9990, 1984. Hu, X., Y. L. Wang, and R. A. Schmitt, Geochemistryof sedimentson Duncan,R. A., andR. B. Hargraves,40Ar-37Ar geochronology of the Rio Grande Rise and the redox evolution of the South Atlantic basementages from the MascarenePlateau, ChagosBank, and the Ocean, Geochim. Cosmochim. Acta., 52, 201-207, 1988. MaldivesRidges, Proc. OceanDrill. ProgramSci. Results, in press, Humphris,S. E, and G. Thompson,A geochemicalstudy of rocksfrom 1990. the Walvis Ridge,South Atlantic, Chem. Geol., 36, 253-274, 1982. Dymond, J., ExcessAr in submarinepillow basalts,Bull. Geol. Soc.Am., Humphris,S. E, and G. Thompson,Petrology and geochemistry of rocks 181, 1229-1232, 1970. from the Angola Basin adjacentto the Walvis Ridge: Deep Sea Emerick, C. M., and R. A. Duncan, Age progressivevolcanism in the Drilling Projectleg 75, site530, Initial Rep.Deep Sea Drill. Proj., 75, Comores Archipelago,western Indian Ocean and implicationsfor 1099-1105, 1983. Somali platetectonics, Earth Planet.Sci. Lett., 60, 415-428, 1982. Humphris,S. E., G. Thompson,J.-G. Schilling,and R. H. Kingsley, Epp, D., and N. C. Smoot, Distribution of seamountsin the North Petrologicaland geochemical variations along the mid-Atlantic Ridge Atlantic, Nature, 337, 254-257, 1989. between46 S and 32 S; Influenceof the Tristan da Cunhamantle Erlank, A. J., and E. J. D. Kable, The significanceof incompatible plume,Geochim. Cosmochim. Acta, 49, 1445-1464,1985. elementsin Mid-Atlantic Ridge basaltsfrom 45øN, with particular Kent,D. V., andF. M. Gradstein,A Jurassicto recentchronology, in: The reference to Zr/Nb, Contrib. Mineral. Petrol., 54, 281-291, 1976. Geologyof North America: The WesternNorth Atlantic Region, Erlank, A. J., J. S. Marsh, A. R. Duncan, R. M. Miller, C. J. DNAG Ser., vol. M, edited by B. E. Tucholke and P. R. Vogt, pp. Hawkesworth, P. J. Betton, and D. C. Rex, Geochemistry and 45-50, GeologicalSociety of America,Boulder, Colo., 1986. petrogenisisof the Etendekavolcanic rocks from SWA/Namibia, Klitgord,K. D., andH. Schouten,Plate kinematics of thecentral Atlantic, Spec.Publ. Geol.Soc. S. AriA.,13, 195-245, 1984. in: The Geologyof North America: The WesternNorth Atlantic' Fleitout, L., C. Dalloubeix,and C. Moriceau,Small-wavelength geoid and Region,DNAG Ser., vol. M, editedby B. E. Tucholkeand P. R. Vogt, topographyanomalies in the SouthAtlantic Ocean: A clueto newhot pp.351-378, Geological Society of America,Boulder, Colo., 1986. spot tracks and lithosphericdeformation, Geophy. Res. Lett., Kogan,M. G., M. Diament,A. Bulot,and G. Balmino,Thermal isostasy 16, 637-640, 1989. in theSouth Atlantic Ocean from geoid anomalies, Earth Planet. Sci. Fodor, R. V., K. Keil, J. W. Husler, and E. H. McKee, Petrology and Lett., 74, 280-290, 1985. K-Ar age of volcanic tuff and ash from the Walvis Seamount Kowsmann,R., R. Leyden, and O. Francisconi,Marine seismic O'CONNOR AND DUNCAN: WALVIS RIDGE AND RIO GRANDE RISE HOT SPOT SYSTEM 17,501

investigations,southern Brazil margin,Bull. Am.Assoc. Pet. Geol., Morgan,W. J., Hot-spottracks and the openingof the Atlanticand Indian 546-557, 1977. Oceans,in The Sea, vol. 7, pp. 443-488, Wiley Interscience,New Kumar,N., Origin of "paired"aseismic rises: Ceara and SierraLeone York, 1981. Risesin theequatorial, and the Rio GrandeRise andWalvis Ridgein Morgan, W. J., Hot-spot tracks and the early rifting of the Atlantic, the South Atlantic, Mar. Geol., 30, 175-191, 1979. Tectonophysics,94, 123-139, 1983. Kumar,N., and R. W. Embley, Evolutionand origin of Cera Rise: An Musset,A. E. andP. F. Barker,40Ar/39Ar age spectra ofbasalts, Deep aseismicrise in the westernequatorial Atlantic, Geol. Soc.Am. Bull., Sea Drilling Projectsite 516, Initial Rep.Deep SeaDrill. Proj., 72, 88, 683-694, 1977. 467-470, 1983. Kumar,N., and L. A. P. Gamboa,Evolution of the Sao Paulo Plateau National GeophysicalData Center, Worldwide gridded bathymetry, (southeasternBrazilian margin) and implications for theearly history DBDB 10, Announc. 85-MGG-01, Boulder, Colo., 1985. of the South Atlantic, Geol. Soc. Am. Bull, 90, 281-293, 1979. Needham, H. D., D. Carte and J. C. Sibuet, Bathymetriede la Ride de LaBrecque,J. L.(Ed.) SouthAtlantic Oceanand adjacentcontinental Walvis et du Bassin du Cap, Ocean Atlantique Sud, report, Dep. marginReg. Atlas Set'., Atlas 13, OceanDrill. Program,College Geosci. Mar., Inst. Franc. de Rech. pour l'Exploitationde la Mer, Station,Tex., 1986. Orleans, France, 1986. LaBrecque,J. L., J. Phillips, and J. A. Austin, The crustalage and NeilI, W. M., Possiblecontinental rifting in Brazil and Angolarelated to tectonicfabric at the leg 73 sites,Initial Rep. Deep Sea Drill. Proj., the openingof the SouthAtlantic, Nature Phys. Sci., 245, 104-107, 73, 791-798, 1984. 1973. Ladd,J. W., SouthAtlantic sea floor spreadingand Caribbean tectonics, O'Connor,J. M., On the trail of SouthAtlantic hot spots:Implications for Ph.D.thesis, 251 pp., ColumbiaUniv., New York, 1974. reconstructingAfrican andSouth American plate motions (abstract), Ladd,J. W., Relative motion of South America with respectto North Eos Trans. AGU, 70, 1135, 1989. America and Caribbean tectonics,Geol. Soc. Am. Bull., 87, 969-976, Pearce,J. A., andM. J. Norry, Petrogeneticimplications of Ti, Zr, Y and 1976. Nb variationsin volcanic rocks, Contrib. Mineral. Petrol., 69, 33-47, le Roex,A. P., Geochemistry,mineralogy and magmaticevolution of the 1979. basalticand trachytic lavas from Gough Island, South Atlantic, J. Ponte,F. C., and H. E. Asmus,The Brazilian marginalbasins: Current Petrol., 26, 149-186, 1985. stateof knowledge,An. Acad.Bras. Cienc,48 (suppl.),215-239, Leyden,R., H. Asmus,S. Zembruscki,and G. Bryan,South Atlantic 1976. diapirstructures, Am. Assoc. Pet. Geol. Bull., 60, 196-212,1976. Rabinowitz,P. D., Geophysicalstudy of the continentalmargin of Liu, Y.-G., Chemicalelement profiles by instrumentalneutron activation southernAfrica, Geol. Soc.Am. Bull., 87, 1643-1653, 1976. analysis2, Representativesediment and basaltsamples taken from Rabinowitz,P. D., Geophysicalstudy of the continentalmargin of DSDP 678 m core, site 525A, leg 74, Walvis Ridge, M.S. thesis, southernAfrica, Reply, Geol. Soc. Am. Bull., 89, 793-796,1978. OregonState Univ., Corvallis,1982. Rabinowitz, P. D., and J. L. LaBrecque,The Mesozoic South Atlantic Liu, Y.-G., and R. A. Schmitt,Chemical profiles in sedimentand basalt Oceanand evolution of its continentalmargins, J. Geophys.Res., 84, samplesfrom deepsea drilling project leg 74, hole 525A, Walvis 5973-6002, 1979. Ridge,Initial Rep.Deep Sea Drill. Proj., 74, 713-730,1984. Richards, M. A., R. A. Duncan, and V. E. Courtillot, Flood basalts and Liu, Y.-G., M. R. U. Miah, and R. A. Schmitt,Cerium: A chemicaltracer hot-spottracks: Plume heads and tails, Science,246, 103-107, 1989. for paleo-oceanicredox conditions, Geochim Cosmochim Acta, 52, Richardson, S. H., A. J. Erlank, A. R. Duncan, and D. L. Reid, Correlated 1361-1371, 1988. Nd, Sr and Pb isotope variations in Walvis Ridge basalts and MacIntyre,R. M., A. P. Dickin, A. E. Fallick, and A. N. Halliday,An implicationsfor the evolution of their mantle source,Earth Planet. isotopicand geochronological study of theyounger igneous rocks of Sci. Lett., 59, 327-342, 1982. theSeychelles, Eos Trans. AGU, 66, 1137, 1985. Richardson,S. H., A. J. Erlank, D. L. Reid and A. R. Duncan,Major and Manivit, M., Paleogeneand upper Cretaceous nannofossils from Deep traceelement and Nd and Sr isotopegeochemistry of basaltsfrom the SeaDrilling Project, leg 74, Initial Rep.Deep Sea Drill. Proj., 74, DSDP leg 74 Walvis Ridge transect,Initial Rep. Deep Sea Drill. 475-499, 1984. Proj., 74, 739-754, 1984. Marsh,J. S., Basaltgeochemistry and tectonicdiscrimination within Saemundsson,K., Outline of the geologyof Iceland,Joekull, 29, 7-28, continentalflood basalt provinces, J. Volcanol.Geotherm. Res., 32, 1979. 35-49, 1987. Samson,S. D., and E. C. Alexander,Jr., Calibrationof the interlaboratory Martin.,A. K., C. J. H. Hartnady,and S. Goodlad,A revisedfit of South 40Ar-39Ardating standard, Mmhb-1, Chem. Geol., 66, 27-34, 1987. America and South Central Africa, Earth Planet. Sci. Lett., Schilling, J.-G., Geochemicaland isotopic variation along the Mid- 54, 293-305, 1981. Atlantic Ridge axis from 79øN to 0øN, in The Geology of North Mascle,J., and J. D. Phillips,Magnetic quiet zones in theSouth Atlantic, America, The Western North Atlantic Region, DNAG Ser., vol. M, Nature, 240, 80-84, 1972. editedby B. E. Tucholkeand P. R. Vogt, pp. 137-156,Geological McDougall,I., Thegeochronology and evolution of theyoung volcanic Societyof America,Boulder, Colo., 137-156, 1986. islandof Reunion(Indian Ocean), Geochim. Cosmochim. Acta, 35, 261-288, 1971. Schilling,J.-G., R. H. Kingsley,and J. G. DeVine, Galapagoshot McDougall,I., andR.A. Duncan, Age progressivevolcanism in the spreadingcenter system: Spatial, petrological and geochemical TasmantidSeamounts, Earth Planet.Sci. Lett.,89, 207-220, 1988 variations83øW-101øW, J. Geophys.Res., 87, 5593-5610, 1982. Schilling,J.-G., M. Zajac,R. Evans,T. Johnston,W. White,J. D. Devine, McDougall, I., and T. M. Harrison, Geochronology and andR. Kingsley,Petrologic and geochemical variations along the Thermochronologyby the 40Ar/39Ar Method, 212 pp., Oxford Mid-AtlanticRidge from 29øN to 73øN,Am. J. Sci.,283, 510-586, UniversityPress, New York, 1988. 1983. McDougalI,I., and C. D. Ollier, ApparentK-Ar agesfrom Tristanda Schlanger,S. O., M. O. Garcia,B. H. Keating,J. J. Naughton,W. W. Cunha,South Atlantic, Geol. Mag., 119, 87-93, 1982. Sager,J. A. Haggerty,J.A. Philpotts,and R. A. Duncan,Geology and McDougall,I., andN. R. Ruegg,Potassium-argon dates on the Serra geochronologyof the Line Islands,J. Geophys.Res., 89, Geral formation of South America, Geochim. Cosmochim.Acta, 30, I 1, 261-11,272, 1984. 191-195, 1966. Sclater,J. G., and D. P. McKenzie,Palaeobathymetry of the South Melfi,A. J., Potassium-argonages for coresamples from Southern Brazil, Atlantic,Geol. Soc. Ant. Bull.,84, 3203-3216,1973. Geochim. Cosmochim.Acta, 31, 1079, 1967. Scrutton,R. A., Geophysicalstudy of thecontinental margin of southern Melson,W. G., and G. Thompson,Glassy abyssal basalts, Atlantic sea Africa-Discussion,Geol. Soc.Am. Bull., 89, 791-793, 1978. floornear St. Pauls Rocks: Petrography and composition of secondary Seidemann,D., Effectsof submarinealteration on K-Ar datingof deep- clayminerals, Bull. Geol.Soc. Am., 84, 703-716, 1973. Molnar,P., and J. Stock,Relative motions of hot spotsin the Pacific, seaigneous rocks, Geol. Soc. Am. Bull., 88, 1660-1666,1977. Atlantic and Indian Oceanssince late Cretaceoustime, Nature, 327, Seidemann,D.,40Ar/39Ar studies ofdeep-sea igneous rocks, Geochim. 587-591, 1987. Cosntochim.Acta, 42. I721-1734, t978. Morgan,W. J., Convectionplumes in the lowermantle, Nature, 230, Sibuet,J.-C., and J. Mascle,Plate kinematicimplications of Atlantic 42-43, 1971. equatorialfracture zone trends, J. Geophys.Res., 83, 3401-3421, Morgan,W. J., Deepmantle convection: Plumes and plate motions, Am. 1978. Assoc.Pet. Geol. Bull., 56, 203-213, 1972. Sibuet,J.-C., W. W. Hay, A. Prunier,L. Montadert,K. Hinz, andJ. 17,502 O'CONNORAND DUNCAN: WALVIS RIDGE AND RIO GRANDE RISE HOT SPOT SYSTEM

Fritsch, Early evolutionof SouthAtlantic: Role of the rifting phase, Vogt,P. R., andG. L. Johnson,Transform faults and longitudinal flow Initial Rep. Deep Sea Drill. Proj., 75,469-481, 1984. beneaththe mid-oceanridge, J. Geophys.Res., 80, 1399-1428,1975. Siedner,G., and J. A. Miller, K-Ar agedeterminations on basalticrocks Vogt,P. R., B. Zondek,P. W. Fell,N. Z. Cherkis,and R. K. Perry,Seasat from south-westAfrica andtheir bearing on continentaldrift, Earth altimetry,the North Atlantic geoid, and evaluationby shipborne Planet. Sci. Lett., 4,451-458, 1968. subsatelliteprofiles, J. Geophys.Res., 89, 9885-9903,1984. Siedner,G., andJ. G. Mitchell,Episodic mesozoic volcanism in Namibia Walker,D. A.,and I. McDougall,40Ar/39Ar and K-Ar dating of altered and Brazil: A K-Ar isochronstudy bearing on the openingof the glassyvolcanic rocks: The Dabi Volcanics,P.N.G., Geochem. South Atlantic, Earth Planet. Sci. Lett., 30, 292-302, 1976. Cosmochim.Acta., 46, 2181-2190, 1982. Srivastava,S. P., and C. R. Tapscott,Plate kinematicsof the North Wang, Y. L., Y.-G. Liu, and R. A. Schmitt,Rare earth element Atlantic, in The Geology of North America: The westernNorth geochemistryof southAtlantic deep sea sediments: Ce anomaly AtlanticRegion, DNAG Ser.,vol. M, editedby B. E. Tucholkeand P. changeat • 54 My, Geochim.Cosmochim. Acta, 50, 1337-1355, R. Vogt, 379-404,Geological Society of America,Boulder, Colo., 1986. 1986. Weaver,B. L., N. G. Marsh,and J. Tamey, Traceelement geochemistry Stock,J. M., andP. Molnar,Some geometrical aspects of uncertaintiesin of basalticrocks recovered at site516, Rio Granderise, Initial Rep. combinedplate reconstructions, Geology, I 1,697-701, 1983. DeepSea Drill. Proj., 72, 451-455, 1983. Sullivan, K. D., and C. E. Keen, NewfoundlandSeamounts: Petrology Weaver,B. L., D. A. Wood, J. Tarney,and J. L. Joron,Geochemistry of andgeochemistry, Geol. Assoc. Can. Spec. Pap., 16,461-476, 1977. ocean island basaltsfrom the South Atlantic: Ascension, Bouvet, St. Sun, S.S., R. W. Nesbitt, and A. Y. Sharaskin, Geochemical Helena, Gough, and Tristan da Cunha, Alkaline IgneousRocks, characteristicsof mid-ocean ridge basalts, Earth Planet. Sci. Lett., 44, edited by J. G. Fitton, and B. G. J. Upton, Spec. Publ. 30, pp. 119-138, 1979. 253-267, Geol. Soc. of London, 1987. Svisero,D. P., H. O. A. Meyer, andH. Tsai, Kimberlitesin Brazil:An Weiss,W., UpperCretaceous foraminiferal biostratigraphy from the Rio initial report,in KimberlitesDiatremes, and Diamonds:Their GrandeRise: Site 516 of leg 72, Initial Rep. Deep Sea Drill. Proj., Geology,Pen'ology and Geochemistry, edited by H. O. A. Meyer,and 72, 715-721, 1983. F. R. Boyd,pp. 92-100,AGU, Washington,D.C., 1979. Wendt,I., H. Kreuser,P. Muller, U. von Rad, andH. Raschke,K-Ar ages Taras,B. D., andS. R. Hart, Geochemicalevolution of theNew England of basaltfrom Great Meteor and Josephineseamounts (eastern North Seamountchain: Isotopic and trace element constraints, Chem. Geol., Atlantic),Deep SeaRes., 23,849-862, 1976. 64, 35-54, 1987. White, R., and D. McKenzie, Magmatismat rift zones:The generationof Thompson,G., A geochemicalstudy of thelow temperature interaction of volcaniccontinental margins and flood basalts, J. Geophys.Res., 94, seawaterand oceanic igneous rocks, Eos Trans. AGU, 54, I015-1019, 7685-7729, 1989. 1973. White, W. M., M. M. Cheatham,and R. A. Duncan,Isotope geochemistry Thompson,G., andS. E. Humphris,Petrology and geochemistry of rocks of leg 115basalts and inferences on the history of theReunion mantle fromthe Walvis Ridge: Deep Sea Drilling Project leg 74, sites525, plume,in Proc.Ocean Drill. ProgramSci. Results, in press,1990. 527, and528, Initial Rep. Deep Sea Drill. Proj., 74,755-764, 1984. Wilson,J. T., A possibleorigin of the HawaiianIslands, Can. J. Phys., Thompson,G., S. Humphris,and J.-G. Schilling,Petrology and 41,863-870, 1963. geochemistryof basalticrocks from the Rio GrandeRise, South York, D., Leastsquares fitting of a straightline with correlatederrors, Atlantic:Deep Sea Drilling Project,leg 72, hole516, Initial Rep. Earth Planet. Sci. Lett., 5,320-324, 1969. Deep SeaDrill. Proj., 72,457-466, 1983. Tucholke,B. E., and N. C. Smoot,Evidence for age andevolution of Corner Seamounts and Great Meteor seamount chain from multibeam J. M. O'Connorand R. A. Duncan,College of Oceanography,Oregon bathymetry,J. Geophys.Res., in press,1990. StateUniversity, Corvallis, OR 97331. Turner,G., andP. H. Cadogan,Possible effects of 39Arrecoil in 40Ar-39Ardating of lunar,Proc. Lunar Sci. Conf., 5, 1601-1615, 1974. (Received March 14, 1989; Vandoros,P., and M. A. F. Oliveira, Sobre o fonolito de Messejana, revised March 14, 1990; Ceara,An. Acad.Bras. Cienco.,40, 203-206, 1968. acceptedMarch 22, 1990.)