• Aucun résultat trouvé

Stability of rift axis magma reservoirs: Spatial and temporal evolution of magma supply in the Dabbahu rift segment (Afar, Ethiopia) over the past 30 kyr

N/A
N/A
Protected

Academic year: 2021

Partager "Stability of rift axis magma reservoirs: Spatial and temporal evolution of magma supply in the Dabbahu rift segment (Afar, Ethiopia) over the past 30 kyr"

Copied!
13
0
0

Texte intégral

(1)

HAL Id: hal-01469936

https://hal-amu.archives-ouvertes.fr/hal-01469936

Submitted on 22 Feb 2017

HAL is a multi-disciplinary open access

archive for the deposit and dissemination of

sci-entific research documents, whether they are

pub-lished or not. The documents may come from

teaching and research institutions in France or

abroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, est

destinée au dépôt et à la diffusion de documents

scientifiques de niveau recherche, publiés ou non,

émanant des établissements d’enseignement et de

recherche français ou étrangers, des laboratoires

publics ou privés.

Stability of rift axis magma reservoirs: Spatial and

temporal evolution of magma supply in the Dabbahu rift

segment (Afar, Ethiopia) over the past 30 kyr

S. Medynski, Raphaël Pik, Pete Burnard, C. Vye-Brown, Lyderic France, Irene

Schimmelpfennig, K. Whaler, N Johnson, Lucilla Benedetti, D Ayelew, et al.

To cite this version:

S. Medynski, Raphaël Pik, Pete Burnard, C. Vye-Brown, Lyderic France, et al.. Stability of rift axis

magma reservoirs: Spatial and temporal evolution of magma supply in the Dabbahu rift segment

(Afar, Ethiopia) over the past 30 kyr . Earth and Planetary Science Letters, Elsevier, 2015, 409,

pp.278-289. �10.1016/j.epsl.2014.11.002�. �hal-01469936�

(2)

Stability

of

rift

axis

magma

reservoirs:

Spatial

and

temporal

evolution

of

magma

supply

in

the

Dabbahu

rift

segment

(Afar,

Ethiopia)

over

the

past

30 kyr

S. Medynski

a

,

,

R. Pik

a

,

P. Burnard

a

,

C. Vye-Brown

b

,

L. France

a

,

I. Schimmelpfennig

c

,

K. Whaler

d

,

N. Johnson

d

,

L. Benedetti

c

,

D. Ayelew

e

,

G. Yirgu

e

aCRPGUMR7358CNRS,UniversitédeLorraine,15rueNotreDamedesPauvres,54500Vandoeuvre-lès-Nancy,France bBritishGeologicalSurvey,MurchisonHouse,WestMainsRoad,Edinburgh,EH93LA,UnitedKingdom

cAix-MarseilleUniversité,CNRS-IRD-CollègedeFrance,UM34CEREGE,Aix-en-Provence,France dUniversityofEdinburgh,TheKing’sBuildingsWestMainsRoad,Edinburgh,EH93JW,UnitedKingdom eSchoolofEarthSciences,AddisAbabaUniversity,Ethiopia

Keywords:

continent–oceantransition focussed/unfocussedmagmatism diking

cosmogenic36Cland3He

Unravelling the volcanic history of the Dabbahu/Manda Hararo rift segment in the Afar depression (Ethiopia)usingacombinationofcosmogenic(36Cland 3He)surfaceexposuredatingofbasaltic

lava-flows,field observations,geologicalmapping andgeochemistry, weshowinthispaper thatmagmatic activity inthisrift segment alternates betweentwo distinct magma chambers.Recent activityin the Dabbahurift(notablythe2005–2010dykingcrises)hasbeenfedbyaseismicallywell-identifiedmagma reservoir withintherift axis,and weshow herethatthismagma bodyhas beenactive overthe last 30 kyr.However,inadditionto thisaxialmagmareservoir,wehighlightinthispaperthe importance of asecond, distinct magmareservoir, located 15 km west of the currentaxis, which hasbeen the principalfocusofmagmaaccumulationfrom15katothesubrecent.Magmasupplytotheaxialreservoir substantially decreasedbetween20 kaandthe presentday,whiletheflankreservoirappearstohave beenregularlysuppliedwithmagmasince15kaago,resultinginlessvariablydifferentiatedlavas.The trace elementcharacteristics of magmas fromboth reservoirs weregenerated byvariable degrees of partialmeltingofasinglehomogeneousmantlesource,buttheirrespectivemagmasevolvedseparately indistinctcrustalplumbingsystems.

Magmatism in the Dabbahu/Manda Hararo rift segment is not focussed within the current axial depressionbutinsteadisspreadoutoveratleast15km on thewesternflank. Thisisconsistentwith magneto-telluricobservationswhichshowthattwomagmabodiesarepresentbelowthesegment,with themainaccumulationofmagmacurrentlylocatedbelowthewesternflank,preciselywherethemost voluminousrecent(<15ka)flankvolcanismisobservedatthesurface.

Applyingtheseobservationstoslowspreadingmid-oceanridgesindicatesthatmagmabodieslikelyhave alifetimeofaleast20 ka,andthatthe continuityofmagmaticactivityismaintainedbyasystemof separate relaying reservoirs,which could in return control the location ofspreading. Thislong term (>105yr) alternationbetweendistinct crustal reservoirslocated broadlyatthesamelocation relative

tothesegmentappearstobeakeyfeaturefororganizing andmaintainingactivespreadingcentresover stablesoftpointsinthemantle.

Abbreviations: DMH, Dabbahu/MandaHararo; MRS, Magmatic RiftSegment; TCN, Terrestrial TCN; COT, Continent–Ocean Transition; MSMC, Mid-Segment MagmaChamber.

*

Correspondingauthor.

E-mailaddress:smsorcha@gmail.com(S. Medynski).

1. Introduction

Extensionconstitutesamajorfeatureofplatetectonics,mainly expressed at mid-oceanic ridges (MOR) and continental rifts. In both marine and subaerial rifting environments, tectonic and magmatic processes (e.g. faulting and dyking)interact in various proportions to accommodate extension, depending on the ma-turity of the rifting system (e.g. early continental rifting stage,

(3)

Fig. 1. A: RegionaltopographyoftheAfarregion(afterHaywardandEbinger,1996).Activemagmaticsegmentsareinred. B: RegionaltopographyoftheDMHsegmentand itsrelationwiththeAlaytasegmentandtheTransformVolcanicZone. C: DetailoftheDMHsegmentandtheextentofvolcanicproductsissuedfromthevolcaniccomplexes oftherift.AVCistheAdo-AleVolcanicComplexandMSMCisforthemid-segmentmagmachamberwhichfeedsthecurrentriftaxis(stripedarea).NotethesmallDurrie volcano,onthewesternflankoftherift,betweentheAVC,theBadivolcanoandtheriftaxis.

ocean–continent transition stage or mature oceanic ridge stage) and the along-axis distribution of magma at the segment scale

(Ebinger and Hayward, 1996; Standish and Sims, 2010; Colman

etal., 2012).The developmentofa magmaplumbing systemand riftarchitecture thatare stablethrough timeremains poorly doc-umentedatridgesettingsdueto theinaccessibilityofmid-ocean ridges(MOR)andtheresultinglackofchronologicalconstraintson themagmaticprocesses.

The Afartriple junction, Ethiopia,hasoften been takenasan analogue of a mature oceanic spreading centre as, being sub-aerial,itismoreaccessiblethanMOR(EbingerandHayward,1996;

HaywardandEbinger, 1996). Evenifthe process of formationof

oceanic crust in the Afar is not entirely complete with respect to the nature ofthe crust (Bastow andKeir, 2011; Hammond et al., 2011), this area allows the morphological evolution of indi-vidualrift segments tobe studieddirectly. Additionally, the Dab-bahu/MandaHararo(DMH)segment,inthewesternAfar(Fig. 1A), hasbeenintensivelystudiedfollowingamajorriftingcrisiswhich began in 2005 and affected the northern half (Dabbahu

sec-tion) of the DMH rift (Wright et al., 2006; Ayele et al., 2009; Ebinger etal., 2010; Ferguson etal., 2010; Grandinetal., 2010a, 2010b).Thiscrisisallowedthemagmaticreservoirsresponsiblefor successiveshallowintrusionstobeidentified,andthetopographic response induced by successive dike intrusions over the period 2005–2010tobequantified(Wrightetal.,2006;Ayeleetal.,2007, 2009;Grandinetal.,2009; Keiretal.,2009; Fergusonetal.,2010; Belachewetal.,2011;Keiretal.,2011;Desissaetal.,2013;Fig. 1B and detail in Fig. 1C). However, several unsolved questions re-main concerning how rift topographydevelops. Forexample, the roleexertedbyindividualmagmareservoirsremainsdebated,due to a lack of constraints on parameters such as their replenish-ment/recurrence time,orthe persistenceoftheir spatial distribu-tionparticularlyovertimescalesrangingfrom103to105yr.These questions are fundamental for understanding how magmatic ac-cretioncanbesustainedbyeitherephemeralorlong-livedmagma chambers and on which timescale MOR morphology is acquired

(Macdonald,2001).According to Ferguson etal. (2013)the DMH

(4)

axial depression) as is the case in mature oceanic ridges. How-ever, recent magneto-telluric measurements have shown that a massive magma body is present in the crust and upper mantle inaslightlyoff-axisposition(Westoftheaxialmagmachamber), representing at least 500 km3 of magma encompassing a depth rangeofabout15–30 km(Desissaetal., 2013). Thismagma vol-umeislargeenoughtofeedabout100episodesofthemagnitude ofthe2005event(Buck,2013).Thelocationofthismagmabody, ifactive, isinconsistent with“focussed”magmatic activityatthe DMHriftsegment.

In this study, we combine geological mapping of surface to-pography,structureandlavaarchitecture,majorandtraceelement analyses and cosmogenic 36Cl and 3He exposure dating of lavas

erupted along an East/West transect of the DMH rift segment (Figs. 1 and 2). Theaim ofthiswork is toassess thestability of themagmaticreservoirsinariftingenvironment.

2. Geologicalsetting

The Afarregion formsthe junctionbetweenthree extensional systems:theGulfofAdenRidge, theRedSeaRidgeandtheMain EthiopianRift(Fig. 1A).TheseparationoftheNubianandArabian platesledtothecreationofthetriangularAfardepression,cutting into a massive pile of continental flood basalts (CFB), emplaced around 30 Ma ago (Hofmann et al., 1997) and linked to the ac-tivityofanunderlyingplume(Martyetal.,1996; Piketal., 2006; Bastowetal.,2008).TheriftingstageoftheRedSeaRidgestarted 29–25 Ma ago (Wolfenden et al., 2005), and since 2–1 Ma the riftsegmentation hasbeenorganized alongfouren-échelon prin-cipal magmatic rift segments (MRS): Erta’ Ale, Tat’ Ale, Alayta andDabbahu/MandaHararo(DMH) (Figs. 1A and 1B).Those MRS are typically 60–100 km long and 20–40 km wide, associated withhighly faulteddifferentiated volcanoes (Lahitte et al., 2003; Barberietal.,1972; Fieldetal.,2012; Rowlandetal.,2007).

ThecurrentspreadingrateforAfarobtainedbygeodeticdatais

15 mm/yr (Calaisetal., 2006; McCluskyetal., 2010), compara-bletothatofIcelandandotherslowspreadingridges(Macdonald,

2001; Carbotte, 2005). However, complete continental break-up

has not yet occurred in the Afar, resulting in a stretched and heavily intruded crust (Tiberi et al., 2005; Bastow et al., 2010;

Hammond etal., 2012) more akinto thecontinent–ocean

transi-tion(COT)stagethantoamatureoceanicspreadingcentre. Morphologically,the DMHspreadingcentre canbe subdivided intotwo sub-segments:the MandaHararosegment inthesouth, andtheactiveDabbahusegmentintheNorth(Fig. 1Aand B).The currentmorphologyoftheMandaHararosegmenthaslikelybeen inplacesince 220kaandwas activeuntil atleast

31–39 ka at the axis(Lahitte etal., 2003). It seems likely, however, that vol-canic activity more recent than this has occurred in the Manda Hararosegmentasunweatheredlavaflowtopsandalackofcover by aeoliansediments(similarto surfacesofyoungdated lavasin theDabbahusegment)havebeenobserved(Fergusonetal.,2010; Medynskietal.,2013),whichimplies

<

5 kyr activity.The tran-sitionbetweenthesetwosub-segmentsischaracterized byashift oftheaxial depressiontothe west(Fig. 1Aand B). TheDabbahu sub-segment–whichstepstothewestrelativetheMandaHararo segment – is about 60 km long, and presents an axial rift val-leyof

40to 100 m deep.Its lastrecordedvolcanic activitywas linkedtothe2005riftingevent,withtheemissionoffissurelavas in2007, 2009(Fergusonetal., 2010) and 2010. Therift axiscuts therhyolitic andatpresentundatedAdo-Ale Gommoytavolcanic complex(AVC)in themiddleofthesegment (Fig. 1C).There isa small caldera in an elevated section of the rift (about 1 km di-ameter)presentattheintersectionoftherift(Dabbahusegment) withtheAVC.Thisportionoftheriftsegmentischaracterized by a complexfault pattern(Rowland etal., 2007),associated witha

Fig. 2. A: Detailedmapoftheareastudiedwithsamplelocations.Smallcircles de-notesamplesanalysedforchemistryalone.Largecirclesanddiamondswerealso datedbycosmogenicnuclides.Lavasurfacesarecolored asafunctionoftheir erup-tivelocation(riftflanks/riftaxis):bluerepresentsflankvolcanism,whereasgreen represents volcanismfromthe currentneo-volcanic zone;volcanicspattercones areinpurple.Chemicalanalysesshowthatsomeflanklavaspresentthechemical characteristicsoftheriftaxislavas(discussedinthetext).Asaresult,the corre-spondingflowfieldsaremarkedingreen(=riftaffinity)despitethefactthatthey aregeographicallylocatedontheadjoiningflank.DF-1andDF-2 (squares)arefrom Fergusonetal. (2013)andweredatedbytheAr–Artechnique.Theboundariesof theselavaflowfields(aflowfieldcanencompassmultipleindistinguishablelava flows)aredefinedbyacombinationofmapping(includingremotesensingdata), fieldobservations, agedeterminations andchemical data. B: Topographicprofile (fromthe LidarDEMoftheNERC-fundedAfarRiftConsortium,courtesyof Bar-baraHoffman)alongthetransect,showingthetopographicinfluenceoftheDurrie flankvolcanoontheriftmorphology.NotethatthelimitbetweenDurrievolcanism and theriftaxisiscontrolledbypre-existingtopography(seethenormal West-dippingfaultbetweensamplesD-15andD-30).Alsonotethatthecontactwiththe axial/Durrievolcanismandthe basement(mostlikelythedissectedAdo-Ale vol-caniccomplex)ishypothetical,becauseitdoesnotoutcropinthefield.Formore detailsonthemappingoftheNorthofthesegment–e.g.thecontactzonewith theDabbahuvolcanics–pleaseseeMedynskietal. (2013).Formoredetailsonthe mappinginunsampled areas,pleaseseeVye-Brownetal. (2012).

(5)

re-orientationoftheaxialdepression.Southofthecaldera,the ax-ialriftvalleyisorientedNW–SE,whereasfurthernorththefaults reorient toward the Dabbahu volcano in an NNW–SSE direction (Fig. 1C).

Thenorthernextremityofthesegmentismarkedbythe pres-enceof Dabbahu volcano, a strato-volcano supplied by series of stackedsill-like magmareservoirs (Fieldetal., 2012), which pro-ducedlavasfrom72 ka(Medynskietal.,2013) to5 kaago(Field et al., 2013). This composite volcanoforms the northern end of an NE–SWalignment of numerous volcanic centres anderuptive fissureventsthat define atransformvolcanic zone (Fig. 1B). This volcanictransformzoneisPleistoceneinage(Lahitteetal.,2003;

Fergusonetal.,2013) andextendsSWtotheEthiopianescarpment

(Fig. 1A).

Onthe western flank (betweenthe rift axis,thewestern part oftheAVCandBadivolcano–Fig. 1)standsthesmallDurrie vol-caniccomplexwhichoverlapstheprevioustopography.Thecentral portionoftherift(includingtheDurrievolcaniccomplex)is char-acterized by the emission of p ¯ahoehoe lavas (Vye-Brown et al.,

2012,submittedforpublication;Vye-Brown,2012).

Recent magneto-telluric studies show that two low resistiv-ity zones are present at depth below the DMH segment, which mostlikelycorrespondto magmareservoirs,oneatabout10 km, theother between15and atleast 30km depth,suggestingthat magma storage beneath the rift axis is composite and not re-strictedtoa singlereservoir (Desissaetal., 2013).The first,axial reservoirmatchesthepositionofthemid-segmentmagma cham-berasrecordedbytheseismicactivityduringdikeinjections(Keir etal.,2009; Grandinetal.,2009; Belachewetal.,2011; Ebingeret al.,2008).The second,larger reservoir,is locatedslightlyoff-axis, betweenthecurrentriftaxisandtheBadivolcano(Desissaetal., 2013) directlybelowtheDurrievolcaniccomplex.

Inthisstudywefocusonatransectextendingacrossthe mid-segmentpartoftheDabbahusegment fromtheaxistothewest. ThissectionoftheDMHriftlackstemporalconstraintsontectonic and/ormagmaticactivitywiththeexceptionofsparse Ar–Ar dat-ing oflavas onthe easternmost shoulder ofthe rift(Fergusonet al.,2013).

3. Mappingdetails

The area studied covers

270 km2 betweenthe rift mid-axis

andtheBadivolcano(Figs. 1 and 2)onthewesternflank.This re-gion is beyondthe influence of Dabbahu volcano (15 km to the north),whichcontrolstopography acquisitioninthenorthern ex-tremityoftheMRS(Medynskietal.,2013).

Onthewesternflank,10kmfromthepresent-dayaxis,stands asmallflank volcano, Durrie, whichis characterized bya central spatter cone surrounded by numerous (

>

20) smaller cones dis-tributedovertherift flank.The lavaflow fieldseruptedfromthe flankcones spreadover

160 km2 (Fig. 2).Thisvolcanism resur-faced the western rift margin (the term “resurfacing” is used to indicatea periodof volcanic activitysufficiently intense to erase theunderlyingtopography,forexample,bycompletelyinfillingthe axial valley): thetopographic profile(Fig. 2B)and the geological

map (Fig. 2 and Vye-Brown et al., 2012, submitted for

publica-tion)show thatthe densityoffaults diminishesinthe vicinityof the Durrie volcano, which was also confirmed by field observa-tions.

In order to focus sampling on relevant morphological objects relative to rift topography acquisition and the different volcanic complexes, the methods, software and manipulation of spectral imagesusedtoproducethenewgeological mapofthe DMHRift (Vye-Brown etal., 2012) wereapplied. Inthisportion oftherift, only lobate p ¯ahoehoe lava flows outcrop, making lava unit con-tactsdifficult todistinguish in thefield. Moreover, the petrologic

textures of thelavas are similar, withmicrolithic assemblages of clinopyroxenes and plagioclases (and rare olivine), further com-plicating identification of individual units. In this situation, re-mote sensingtechniques(Landsat,ASTER,andLiDAR;seeSOM1) can be used in order to distinguish the different lavaflow units (Vye-Brownetal.,submittedforpublication).

4. Samplingdetails

Between 12 to 15 different eruptive units were identified on Durriebasedonremotesensingandfield-baseddata,whereasonly 3unitsaredistinguishableintheaxialvalley,likelyduetostacking oflavasinthedepression:thelimitedsurfaceavailableforlava ex-pansionimpliesmoreefficientresurfacingintheaxialvalleythan ontheriftflankswherelavaflowscanspreadradially.Sample lo-cations,carefullyselectedinordertoberepresentativeoftheflow unitsidentifiedfromthedetailedmapping,arereportedinFig. 2, and sample details are summarized in Table 1. All 24 lava flow samples described here were analysed for chemical composition (majorandtraceelements).Inaddition,15lavaflowsampleswere dated using cosmogenic 36Cl (see SOM 2) and two others were dated with cosmogenic 3He, following the protocol described in

Medynskietal. (2013)(D-2 and D-29).

4.1. Flankvolcanismsamples

The Durrie volcanic cones and lavas are composed of piled p ¯ahoehoe flows, mainly focused around the 40 m high central spatter cone (Fig. 2A, Table 1 for sample details). Samples D-9 and D-14 are partofthe sameeruptiveunit, which flows down-slopeontheeasternflankoftheDurriemainspattercone,clearly identified byremote sensingasasingle flowfield withadistinct contact with adjacent flow fields visible on the high resolution SPOTDEM(seeSOM1).Thisunitwas sampledatitstwo extrem-ities (Fig. 2) in order to test the homogeneity of exposure ages onthe sameunit, andalso tovalidate themappingby geochem-istry. Sample D-4, takenfromthe topof themain Durriespatter cone, and sample D-5, part of a pre-existing, partly dismantled cone (Fig. 2), were not suitable for datingbut were analysed for chemistry.Sample D-3wastakenfromthenorthernflank-unit(see

Fig. 2) was also unsuitablefor dating,andis thereforeonlyused hereforcomparativechemistry.SamplesD-32andD-31,fromthe upper unitsofthe Durriecone, were only intended forchemical analyses(notdated).

4.2. Riftaxisvolcanism

Thestudied portionoftheriftaxisis characterized byahorst withawell-preservedvolcaniccone(Fig. 2B).Eastofthishorstlies the mainrift axisdepression,partially in-filledby extremelylow albedo(i.e.recent)lavas.Onthewesternsideofthehorst,adeep, narrowdepression(30–40 mdeep,1–2 kmwide)extends(withan NW–SEorientation)uptoanaxialcaldera(

10 kmSEofthe ax-ialhorst, attheintersectionwiththeAVC)(Fig. 2).Severalfissure ventscanbeobservedinthedepressionbetweenthehorstandthe caldera.Lavaflowsarepiledupinamonotonoussequence, charac-terized bythepresenceofamassiveandthick(

>

4 m)doleritelava layer (Fig. 3). Thisdoleritelayer outcropsinthe field atthebase ofthemainfaults,spreadingoverseveralhundredmeters,and re-curs severaltimes along therift depression. Thedolerite layer is thick andlaterallyextensive;flows emplacedafterthisare gener-allythinner,andtheheightofthesurroundingp ¯ahoehoelavapile gradually decreasesaway fromthehorstcone (Fig. 3), suggesting thatthedoleritelayerrepresentstheinitialstageofanintense vol-canicperiod.Itwassampledattheaxis(sampleD-20,seeFig. 2)in ordertocompareitschemicalcharacteristicswithotherlavaunits.

(6)

Table 1

Samplelabelsanddetailsforallsamplesanalysedhere.Thosegiveninitalicswereanalysedforchemistryalone.Samplesinregularfontweredatedbythecosmogenic36Cl

technique,andthoseinboldusingcosmogenic3He.

Latitude (◦N) Longitude (◦E) Altitude (m)

Geomorphology Distancetoriftaxis (km)a

Petrology Exposureage (ka)

± (1σ) Durrie volcano lavas

D1 12◦22.148 40◦26.732 574 pahoehoe lava top 13 Basalt 8.9 1.1 D2 12◦22.844 40◦26.359 554 pahoehoe lava top 13.2 Basalt 14.5 3.6

D6 12◦22.780 40◦29.382 560 pahoehoe lava top 8.2 Basalt 14.9 1.5

D4 1222.482 4028.650 599 spatter cone 10 Basalt D5 1222.425 4028.700 591 spatter cone 10 Basalt

D9 12◦23.483 40◦30.255 466 pahoehoe lava top 6.5 Basalt 11.6 1.2

D14 12◦24.385 40◦32.013 434 pahoehoe lava top 3 Basalt 11.9 1.2

D11 12◦23.903 40◦30.777 449 pahoehoe lava top 5 Basalt 10.1 1.2

D13 12◦24.588 40◦31.120 442 pahoehoe lava top 4.3 Basalt 5.4 0.6

D30 12◦22.620 40◦31.848 473 pahoehoe lava top 5 Basalt 9.6 1.0

D31 12◦21.707 40◦31.108 490 pahoehoe lava top 7 Basalt 9.5 1.0

D32 1221.530 4029.840 548 pahoehoe lava top 9 Basalt D33 1221.547 4027.810 585 pahoehoe lava top 12 Basalt

D36 12◦28.575 40◦28.957 367 pahoehoe lava top 5.9 Basalt 10.9 1.1

Rift Axis lavas

D15 12◦23.903 40◦32.742 436 pahoehoe lava top 2.3 Basalt 21.8 2.2

D17 12◦23.783 40◦33.367 440 pahoehoe lava top 1.4 Basalt 24.4 2.2

D19 1223.870 4033.055 405 pahoehoe lava top 1.9 Basalt D20 1223.652 4033.275 429 pahoehoe lava core 1.7 Dolerite

D23 12◦23.932 40◦32.795 396 pahoehoe lava top 2 Basalt 19.6 2.3

D28 1222.890 4033.575 463 spatter cone 2.4 Basalt

Cald-1 12◦21.157 40◦34.957 576 pahoehoe lava top 0 Basalt 24.3 2.6 D29 12◦22.553 40◦33.448 495 pahoehoe lava top 0 Basalt 25.0 1.7

D16 12◦23.860 40◦33.532 430 pahoehoe lava top 0 Basalt 7.0 1.1

Gab-C3 12.4834 40.5351 383 pahoehoe lava top 0 Basalt 6.4 1.1 a The“riftaxisrefersheretothemiddleoftheaxialvalley(dottedlineonFig. 2).

Fig. 3. Topfigureshowstheorientationofthephotomontage,orientedtowardthe Dabbahuvolcano(coordinates:lat.12◦2251.84N,long. 40◦3326.39E),notethe eruptivespatterconeontheaxialhorst.Weidentifiedtwofaultlocations(circles) wherethesamemassivedoleriticlavaunitoutcrops,suggestingthatits emplace-mentextendsovermorethan2kmfromtheriftaxis.Thislayerisrecognizableas shownonthelowerphotos(thephotoontherightcorrespondstothesampling siteofsampleD-20–coordinates:lat.12◦2332.02N,long:40◦3322.96E).Itis notablethatthe heightofthe lavapileoverlyingthisrecurrentmassivedolerite lavaflowdiminishesfromtheriftaxistowarditsflanks,suggestinganaxial emis-sionpoint(seeleftphoto–coordinates:lat.12◦2251.30N,long.40◦336.74E). Therecurrenceofthispatternseemstoberepresentativeofacompletemagmatic phase,startingwiththedoleritelayer,andfollowedbytheemplacementofpiled basaltflowfields.

The youngest samples are Gab-C3 (which was reanalysed for cosmogenic 36Cl after being dated with cosmogenic 3He by

Medynskietal. (2013)and D-16,whichbelongstothesame

vol-canicunitbutatitssouthernmostextremity(Fig. 2).Thiseruptive unitlookssimilartothelavaseruptedin2007,2009and2010, fol-lowing shallowdikeintrusions(Ferguson etal., 2010;Grandinet

al., 2010a, 2010b). Although thevolumeserupted since 2005 are

muchsmallerthantheGab-C3/D-16unit,both ofthesep ¯ahoehoe flow fieldsseemto haveinvolvedasimilar eruptionstyle,issuing fromsmallalignederuptiveventssuggestingtheinvolvementofa shallowdike.

The oldest samples in the stratigraphic lava pile to be dated were samples D-15, D-17 and D-23 (seeFig. 2).SampleD-19was usedforchemistryalone.

Itshouldbe notedthatwhilemostoftheflowscanberelated to a spatter cone in the area, there is no obvious volcanic cone associatedwiththeflowfromwhichD-29wastaken,buttheslope variationismoreconsistentwithanoriginfromthecalderarather thanfromafissurewithintheaxialdepression.

5. Chronologicalconstraints

While dating lava emissions using cosmogenic nuclide accu-mulation in lava surfaces is well adapted to this geological and climatological context,onlythe uppermostlava inagivenpileof lavascanbedated.Theprinciplesbehindcosmogenicnuclide dat-ing(by36Cland3He)andthetechniquesusedaredescribedinthe

Supplementary Online materials (SOM 2). The lava surface expo-sureagescalculatedforthemid-segmentDabbahuMRSlava-flows rangefrom5

.

4

±

0

.

6 ka (D-13)to24

.

7

±

1

.

6 ka (D-29)(Table 1and SOM2Tables3and4).Thevolcanicstratigraphyestablishedonthe basis oftheseresultsissummarizedinFigs. 2 and 6.A lavaflow emplacedontheedgeofthedepressionandcurrentlydissectedby theEasternfaultsofthedepressionwas datedbytheAr–Ar tech-nique at30

.

0

±

5

.

4 ka byFerguson etal. (2013) (sampleDF-1 in

(7)

Twomainresurfacingeventsintheaxialdepressionandonthe riftwesternflank canbe identifiedfromtheagesofthedifferent flows.

5.1.Rift-axisvolcanism:amajorresurfacingeventofthedepressionat

about25–20 ka

Thefirstmajorresurfacingeventspreadsnorthfromthe vicin-ity of the caldera, and took place at about 25–20 ka. Samples Cald-1 and D-29 (thatrepresent thesamples closest to the axial caldera – Fig. 2) yield indistinguishableages (respectively 24

.

3

±

2

.

6 and 25

.

0

±

1

.

7 ka) suggesting a rapid succession of eruptive units.About 5 km north of the caldera stands an eruptive cone (Fig. 2),preservedduetoitslocationonthehorstdescribedabove. This cone was active coevally with the caldera units, and pro-ducedlavas between24

.

4

±

2

.

2 ka (sampleD-17)and19

.

6

±

2

.

3 (D-23).It islikely that thisepisodeended withthe formationof thecaldera,witharapidemptyingofashallowreservoir.Thelavas datedaround24–20 kainthisstudysharesimilargeomorphologic characteristics with the 30 ka lavas of the Eastern rift shoulder (Fergusonetal.,2013).Thebroadspatialdistributionoftheir asso-ciatederuptiveventswithintheriftandtheformationofanaxial caldera,suggest thatthiswas amajorresurfacingeventthatmay havespreadover thewhole Dabbahu riftsegment. This resurfac-ing eventrepresents a considerable volume oflava forthe DMH itself;atleast5 km3oflavawasemitted.Thisisaminimum

esti-matebecausethebaseofthiseruptiveepisodeisnotconstrained –weestimatedanaveragelavapileheightofclosely-spaced (tem-porally and spatially) eruptions of about 20 m for a surface of 280 km2.

The youngest activity recorded in the depression consists of lavas that were emplaced at the rift axis at about 6 ka. These younger flows filled the graben on the eastern side (Fig. 2) in the vicinity of eruptive fissure vents, and may represent much smaller lava volumes (less than 100 km2 covered with an esti-mated lava pilethickness of less than 10 m from field observa-tions).Theseflowfieldsareclearly identifiableonsatelliteimages aslow albedo lava flows, andwhich, incontrast to the previous resurfacingevent(20–25 ka),wereclearlyflowedupagainst exist-ingfaultscarps.

5.2.Flankvolcanism:arecent(

<

15 ka)resurfacingofthewestern

shoulderoftherift

The second major resurfacing event occurred on the western flank ofthe rift,at theDurrie volcaniccomplex. The ages ofthe variousunitsoftheDurrieflankvolcanorangefrom16

.

3

±

2

.

8 ka (D-2) to 5

.

4

±

0

.

6 ka (D-13), coeval with the last resurfacing episode in the rift axis (samples D-16 and Gab-C). Emission of individual lava units was distributed along the different mapped eruptivecentres. Forinstance,the oldest recordedlava flow field (samplesD-2andD-6,datedat15

.

0

±

1

.

5 ka and14

.

9

±

1

.

5 ka re-spectively)spreadconcentricallyawayfromthemainspattercone, whichweinterpretasthesourceofthisflowfield.Thesubsequent unit(D-9 andD-14,datedat 11

.

7

±

1

.

2 ka and 11

.

9

±

1

.

2 ka re-spectively) erupted from an eastern cone that is located

1 km away from that of the D-2/D-6 flow field. Samples D-30, D-31 and D-1 display similar ages (at 9 ka) and were emitted from smallcones to the south-east of themain spatter cone. The last (youngest)lavawas eruptedat5

.

4

±

0

.

6 ka (D-13)further north, closerto theriftaxis(Fig. 2)andischaracterized bythinner and lesslaterallyextensiveflowsthan thoserelatedtothemain spat-tercone(Fig. 2).The widelydispersederuptivecentres(inpurple inFig. 2)associatedwithvolumesoflava

<

2 km3stronglysuggest that thisflank volcanism corresponds to a significant resurfacing eventat15–10 ka.The volumeestimation was madebasedon a

160 km2 area covered andan average lava pileheight of 10 m.

However,thelavavolumescouldpossiblybehigher,dependingon howthebasementtopography isestimated:amaximumvalue of 4 km3isobtainedifweinsteaduseanaveragepileheightof25 m,

a plausiblepossibilitygiven thatsome relics ofthe AVC complex outcrop in the vicinity ofthe Durrie main spattercone – Fig. 2, suggestingashallowbasement.

6. Twogeochemicallydistinctmagmaspresentwithinthesame

riftsegment

Major and trace element concentrations were determined by ICP-OES and ICP-MS respectively, at the Service d’Analyse des RochesetdesMinéraux(SARM,CRPG–Nancy,France)followingthe protocolestablishedbyCarignanetal. (2001),eitheronwholerock materialoronseparatedmatrixforphenocryst-bearinglavas.The samplesareallsub-alkalinebasaltsandthevariationsofsome se-lected major andminorelements are presentedin Fig. 4(for the completelavacompositionsseeSOM 3).

The axial and flank (Durrie) lavas are chemically distinct, no-tably withaxially erupted lavas beingricher inFe2O3T andTiO2,

in incompatible elements (except Sr), and depleted in Al2O3 for

a given MgO content (Fig. 4). The use of compatible elements is particularly appropriate to assess the crystallization sequence occurringwithin magma chambers;especially Ni that is compat-ible with olivine and pyroxene, whereas Cr is compatible only withpyroxene. The decreasein Ni correlates perfectlywithMgO depletion in both series, while Cr decreases only in axial lavas (Fig. 4); these variations show that only olivine crystallizes in the flank magma chamber, whereas fractionation of both olivine and clinopyroxene occur in the axial magma (Fig. 4). Neverthe-less, this crystallization sequence cannot account for the differ-ences in Fe2O3T, TiO2, and Al2O3 between the two (axial and

flank) lava series. This major element variability is also associ-ated with variations in trace element concentrations, with ax-ial lavas being slightly enriched in incompatible elements (ex-cept Sr) in comparison to flank lavas (e.g. Durrie – Figs. 4, 5). These compositional differences can be most easily attributedto eithercrustal contaminationorto primary liquiddifferences (dif-ferentdegreesofmeltingordifferentmantledomains).An imma-ture, more reactive plumbing system below the flank volcanoes might be expected to result in a greater proportion of crustal contamination at Durrie relative to axial lavas. However, crustal contamination alone cannot account for the observed chemical variations, notably the difference in Fe2O3T (at a given MgO)

is unlikely to result from assimilation of a predominantly fel-siccrust. Also, contamination via assimilationof previously crys-tallized and possibly hydrothermally altered basaltic rocks (the mostlikelylithologies constitutingthe magmachamber margins) would result in the contaminated melts displaying negative Eu and Sr anomalies (France et al., 2014), which are not observed

(Fig. 5). Importantly, the geochemical markers that discriminate

axialfromflank lavas(Fe, Ti,Al,incompatible elements)also cor-relate with trace element ratios that are sensitive to the frac-tion of partial melt in the mantle source region, such as Sm/Yb (Fig. 4).Sm/Ybfractionatesinthepresenceofgarnet-bearing man-tle, whileLa/Smvariations can indicate variabledegreesof melt-ing of a spinel-bearing mantle. Additionally, a pronounced pos-itive Sr anomaly would mark a contribution from low-pressure plagioclase-bearing mantle. In the present case, flank lavas dis-play strong positive Sr anomalies (while axial lavas display no anomaly), similar La/Sm to axial lavas, and lower Sm/Yb than axial lavas. These results are consistent with higher degrees of melting of a garnet-bearing mantle in the flank lavas, a similar degree of partial meltingof spinel-bearing mantle,and an influ-ence of low-pressure plagioclase-bearing mantle present only in

(8)

Fig. 4. Chemistryoflavasamples.Heavyandlightrareearthelement(REE)compositionsarenormalizedtoE-MORB(Galeetal.,2013).TheSranomaly(notedSr∗)was cal-culatedafternormalizingthetraceelementratiotoN-MORBS(Galeetal.,2013).Sr∗=Sr/((Pr+Nd)/2).Diamonds:samplesoutcroppingontheflankandwithflank-affinity chemistry;filledcircles:axialsamples;emptycircles:samplesoutcroppingontheflankbutwithaxialchemicalaffinity(seetextfordescriptionofaxialandflankchemical affinities).

Fig. 5. Spiderdiagram,normalizedtoE-MORB(Galeetal.,2013).Theyoungestlavas(eruptedat6–7 kaandin2007)areinblackandareenrichedinincompatibleelements comparedtotheotherlavas(fromtheaxisandtheflanks)duetotheirhigherdegreeofdifferentiation.Asexpected,theaxislavas(green)displayhighervaluesthanthe flanklavas(blue),duetothehigherpartialmeltfractionoccurringbeneaththeflanks.However,itseemsthatthereisacompositionalcontinuumbetweenthetwogroups oflavas,illustratedbytheblueemptysquaresoftheflanklavas.

theflanklavas(Chalot-Pratetal.,2010).Mantle-derivedmeltsthat haveequilibrated, atleastpartially,withplagioclasebearing man-tle havebeenshownto be Al-richer,andpoorerin Fe

+

Tithan mantlemeltsoriginatedindeeper(spinel- orgarnet-bearing) man-tle domains (Chalot-Prat et al., 2010), consistent withthe differ-encesobservedbetweentheflank(influenceofplagioclase-bearing mantle),andaxiallavas(no influenceofplagioclasebearing

man-tle).Higherdegreesofpartialmeltingforflanklavasarealso con-sistent with their lower concentration in incompatible elements (Figs. 4, 5).

Giventhat thereis botha higherpartial meltfractionderived fromdeepgarnet-bearingmantleandaninfluencefromashallow plagioclase-bearing mantlein the flank lavas, we expect a larger meltingcolumnintheflankareathanattheaxis.Thisalsoimplies

(9)

that the thermal anomaly is centred slightly to the west of the presentdaymorphologicalaxis(

15 kmtothewest).These con-clusionsareconsistentwithrecentmagnetotelluricdatathatimage alargermagmabody

15 kmtothewestofthepresentday mor-phologicalaxis(Desissaetal.,2013;Fig. 7).

Thus we conclude that two distinct parental magmas are presentinthisportionoftheDMH,withtheflanklavas character-ized by slightlyhigherpartial melt fractions of thesame mantle source than that implicated in the axial magmatism. The geo-morphologicaland geochronologicalidentification oftwo distinct volcanic eruptive centres (e.g. the Durrie volcano and the mid-axis magma chamber, Fig. 2), well-separated in space and time, isthereforealsosupportedbytheir geochemistry.However, afew samples which present a “rift-axis chemical affinity” were actu-ally erupted on the flank (for example, flow field D-9/D-14 and flowfield D-36; Fig. 2).Based on thisobservation, inthe follow-ing discussionwe distinguish flank andrift-axisvolcanics on the basisoftheir composition(Fig. 2),bearinginmind thatlavas ge-netically linked to the rift-axis reservoir can also erupt up to 6 km west of the axis. In addition, some infiltration of axial-type magma intothe flank volcanismmay have occurred, particularly when looking at the spread of Sr/Sr∗ in Durrie (flank) volcanic products(Fig. 4).

7.Discussion

7.1.DistributionandlongevityofmagmareservoirsalongtheDMHrift

Thevolcanismencasedintheriftaxisdepressionandthe west-ern flank volcanism linked with the Durrievolcanic cones were suppliedbyatleasttwodistinctreservoirs,whosepeaksofactivity areasynchronous, e.g.the axialreservoir hadits maximuminput ratearound30–20 kawhiletheDurrievolcanohasbeenthemain sourceofmagmaticactivitysince15 ka.Thischangeinthefocus ofmagmatic activityconstrains thestability ofmagma dynamics in spaceand time, demonstrating that unfocussed magmatic ac-tivityisa featureoncertaintime andlength scalesinthecentral DMHrift.Inthissectionweexaminethedistinctstagesof volcan-ism on the flank andin the rift and show that variable magma supplytothesurface (influxandinspatialdistribution) islinked tothedifferentiationandlifetimeofdiscretemagmareservoirsin thecrust.

7.1.1. A“dying”axialreservoir

Thepresenceofamagmachamber(the “mid-segmentmagma chamber”,MSMC)locatedbelowtheriftaxisapproximately1 km south of the caldera (Fig. 1) has been identified from geome-chanical modellingof crustal movement following the 2005dike injection(Grandinetal.,2009; Wrightetal.,2006) andfrom seis-micity (e.g. Ebinger et al., 2008). Our rift-axis lavas were likely eruptedfromthismid-segmentmagma reservoir. Theselavas ex-hibitadrasticdecreaseinMgOcontentsince10 ka(Fig. 6),which probablyresultfromdifferentiationprocesses.Itappearsfromthe age/composition correlation inFig. 6 that thisreservoir was pre-viously at steady state (i.e. extrusion

=

supply) due to periodic replenishment balancing lava production. The subsequent period ofintensevolcanicactivityandresurfacingfrom25to10 kalikely startedwiththeeruption ofthemore primitivemassivedoleritic lavas (D20: MgO

=

9.2%) that recurrently outcrop at the base of the 20 m thick lava pile throughout the rift axis. This major eruption episodecould have triggered the formation ofthe axial caldera,present directly above the MSMC. However, after 10 ka, the MSMC evolvedtowards distinctly more differentiated basalts

(Fig. 6), most likely associated with a decrease or a permanent

breakinmagmasupplyfromdepth.

Fig. 6. Magmareplenishmentanddifferentiationthroughtimeattheriftaxisand onthe flank(Durrie).Symbolsas forFig. 3.AlthoughtheMSMChasbeen con-tinuouslyactivesinceatleast∼30 kaago,thisdiagramillustratesthattherewas adramaticchangeinbehaviourataround10ka,afterwhichmoredifferentiated lavaswereeruptedattheaxis,indicativeofareducedmagmasupplytotheMSMC. Flankvolcanismwasfirstrecordedatabout15 ka,bycontrastalllavaproductson theflanksarerelativelyundifferentiated,indicatingasustainedmagmasupply. Co-evallavaemissionoccurredintheaxialdepressionandontheflank(∼15 kmto thewest)overaperiodofatleast15 ka.

7.1.2. A“fullyactive”riftflankreservoir

In contrast, chemical variations in lavas erupted on the flank from 14.5to 5.4 ka are less pronounced. During thistime span, MgO content ismaintained within a restrictedrange from7.9to 8.6 wt%, equivalent to the composition of axial lavas olderthan 10 ka.These limitedcompositional variations over a long period oftimereflectthefact thatshallowreservoirshavebeen periodi-cally refilledbymore primitivemagmas fromdeeperinthecrust oratthecrust/mantleboundary.Fromtheperspectiveofthe mag-matic cycles described above, this would put theDurrie volcano inaphaseofhighmagmainput,withrapid,andpossiblyfrequent magmareplenishments(Figs. 6 and 7).

7.1.3. Comparisonwithpresent-daymagmarepartitionwithinthecrust

Desissaetal. (2013) collectedmagnetotelluric data(MT)along

the same transect as our samples (Fig. 2). Thesedata indicate a 35 km-wide zone of highelectrical conductivity atcrustal/upper mantledepths.Usingcompositionalconstraintsfromgeochemistry oflavasamplesandtwo-phasemixinglaws,theydeducedthatthe highconductivityzonecontainsatleast500 km3 ofmagma(with up to

13% ofmelt).Two magmabodies can be identified from the MT, one located beneath themid-segment axis andthe sec-ond locatedbeneaththe Durrievolcaniccomplexon thewestern

riftflank(Fig. 7).AlthoughMTimagingcannotdeterminewhether

these two magma bodiesare connected,it nevertheless provides strongconstraintsontherelativevolumesandlocationsofmagma thatmightbeavailable.

OurconclusionsareremarkablyconsistentwiththeMTimagery

(Fig. 7). Indeed, it appears that the volume of magma currently

available belowthe riftaxis (i.e.the mid-segmentmagma cham-ber)isconsiderablylessthanthatavailablebelowtheflank(Fig. 7), consistentwiththerecentvigorousactivityatDurriethatwehave documentedhere.Moreover,thisrestrictedamountofmagma be-low the current neo-volcanic zone is concentrated in the upper crust,whereasitextendsdowntoatleastthecrust/mantle bound-arybelowtheDurrieflankvolcanism.Therefore,themost straight-forwardexplanation of thechemical evolution observed inFig. 6

isthatthere wasarecentdeficitinmagmasupplyfromdepthto theMSMCresultinginmoreextensivecrystallization andincreased differentiationofthemagmaticproducts. Accordingtoourdating results,thislikelyoccurredaround10 ka.Asaresultofthis crys-tallization phase, the magma body imaged by MT is presumably

(10)

Fig. 7. Correspondencebetweensurfacetopographyandsub-surfaceresistivity(magneto-telluricimageandMohodepthfromDesissaetal.,2013).Thevoluminousmagma storagezonecanbedividedinto(atleast)tworeservoirs:ashallowaxialmagmachamber(whichwasalsoidentifiedfromdatacollectedduringthe2005crisis)andaslightly deeperreservoirbelowthewesternflankvolcanicsystem.Remnantinter-connectivityisstillpossiblebetweenthetworeservoirs,consistentwithsomeofthesampledlava compositionswhichshowevidenceofmixingbetweentheaxialandflankaffinities(magneto-telluriccrosssectionandmagmavolumeestimationsfromDesissaetal.,2013). Thesill-shapeillustratedfortheDurriereservoirisinspiredfromHammond (2014),whoshowedthattheseismicanisotropyobservedhereisbestexplainedbythepresence ofmagmastoredinsills.TheseMTobservationperfectlyfitthegeochemicalobservationswhichpredictthepresenceofagreatermeltcolumn(extendingtogarnet-bearing mantle)beneaththeDurrievolcaniccomplexthanbeneaththeriftaxis(spinel-bearingmantle).

smallerthanit wasat

25kawhenitwas connectedwithfresh magmastoreddeeperinthecrust.Thepresentdaylargermagma body,locatedunderthewesternflank, firstappearedto beactive at

15 ka.Thisdemonstratesthatsuchlargemagmabodiesstored atthebaseofthecrust arestableatleastforperiodsof10–15 ka overwhichtheycansustainandbufferthecompositionofshallow reservoirsanderuptedlavasbyfrequentreplenishment withfresh magma.

A recent seismic study (Hammond, 2014) showed that the reservoirs below the Durrie volcanic centre are most likely sill-shaped. This reservoir geometry is compatible with our model (Fig. 6) and could account for the larger compositional variabil-ityobservedatDurrie.However,oneofthemaininterpretationsof theHammondetalstudywasthatthepresent-dayaxialvolcanism is fed fromdeep off-axis reservoirs whereas, from thechemistry ofthe different lavas,we show that theaxial and flank magmas evolvedinseparate reservoirs,in agreement withFerguson etal.

(2013). Minor mixing between rift-axis and flank magmas may

nevertheless occur, whichisconsistent withsomeinterconnected plumbingbetweenDurrieandtheaxis.

7.1.4. Relationwiththerecentmagmato-tectonicactivityintheDMH

andfutureevolution

The two magma bodieswith vastly different volumesimaged by MT strengthens the idea of a mid-segment reservoir that is magma-starved andthat the magma supplyhas relocated to be-low the western margin (Fig. 7). Our dating results suggest that thisstartedaround 15 ka.Thisingoodagreementwiththe2005 rifting event which involved the participation of three magma reservoirs, including the mid-segment magma chamber, begin-ning with those beneath the volcanic centres of the northern end of the segment (Wright et al., 2006; Grandin et al., 2009;

Ayele et al., 2009). This magma injection disrupted the

stabil-ity of the mid-segment magma chamber, leading to the intru-sion of the “mega-dike” (a 60 km long, 6–8 m wide dike that opened in 2005; Wright et al., 2006) from the MSMC. The ab-senceofMSMCreplenishment(whichwouldhavebeenseenwith InSAR and/or seismic techniques; Hamling et al., 2009) explains why the latest lavas derived from the MSMC (2007 and 2009,

Ferguson etal., 2010) are positioned at the end of the

(11)

beenrejuvenatedtomoreprimitivecompositions.Ifthe mid-axis magma chamber continues to evolve as a closed system (with-out further replenishment of primary magma), the next logical step in the evolution would be eruption of even more differen-tiated products. This has been frequently observed in the older volcanic activityof Afar(Lahitte etal., 2003). This long term re-currenceofalternatingbasicandacidicproductsatthesame loca-tion appears to be a key feature of the organization and main-tenance of such active spreading centres on top of stable soft points in the mantle (e.g. areas of the mantle extending over 10s ofkilometres,softenedbylocalizedmeltingprocess;Geoffroy, 2005).

7.2.Morphologicalevolutionalongtheaxisoverthepast30 kadueto

unfocussedmagmaticactivity

There have been at least two major resurfacing events dur-ingthe past30 ka overthestudied transect,spatially distributed between theaxialdepressionandtheflanklocation oftheDurrie volcaniccones.

Thefirstmajorresurfacingeventissourcedfromthemid-axial magmareservoir. Ourcosmogenic exposure ages, combined with theAr–AragefromFergusonetal. (2013)ontheeastern margin, showthata voluminous,monotonouslavapilewasemplaced be-fore construction of the modern fault-bounded axial valley. This intense magmatic phase took place around 20–25 ka,and likely erasedallthepre-existingtopography(Fig. 8).Becauselavasinthe depression are stacked, and cosmogenic dating can only be per-formed onthe latest, currentlyoutcropping lava, it is impossible tosayiflavaemplacementintheaxialdepressionwascontinuous between19kaand6ka, oriftherewas a hiatusinvolcanic ac-tivityatthecurrentriftaxis.However,basedonfieldobservations, the6 kaeventseems tobe significantly smallerin termsof vol-umeerupted(although lavathicknesses are notavailable forthis unit).

The second major resurfacingevent is dueto the flank activ-itywhicherased thepre-existing topography, andwas also suffi-cientlyintensetobuilda significantvolcaniccone(Figs. 2 and 7). Currently,theflanksoftheDurrievolcanoarenottectonically dis-sected,althoughopenfractureshavestartedtodevelop.

Amajor implicationisthat,forthepast15kyrsatleast, mag-matism is not limited to the axial topographic depression, and asynchronousvolcanic activitycanbe distributedover morethan 15 kmfromit.Thiscontrasts withmostmid-oceanridge models where focussed (

<

5 km) magma supply is inferred (Macdonald, 2001). This could be due to the fact that the Dabbahu rift is stillimmatureandisnotrepresentativeofa trueoceanic spread-ing centre. However Standish and Sims (2010) have shown that off-axis magmatism (up to 10 km from the spreading axis) oc-curredconcomitantlywithon-axismagmatismattheSouthWest IndianRidge(SWIR).Ourdatathereforesupporttheideathatthe DMHriftsegmentisrepresentativeofslowoceanicridgesystems, wheremagmachambersexist fora fewtens ofkyrs, andcan be distributedin a 15–20 km wide zone. The steady state buffered compositionofvolcanismoccurringby theDMHactivemagmatic reservoirs (7.9–8.6 wt%MgO)is alsoingood agreementwiththe compositionofmagmasevolvinginasystemcontrolledbyaslow spreadingratewithlowmeltsupplieswhichareuniformlyless dif-ferentiatedthanothertypesofridge(e.g.fastspreadingridges)but aremorelikely toretain variationsinheritedfromtheunderlying mantle(RubinandSinton,2007).

Giventhattherehasbeenan establishedmagmasupplyunder the western flank of the rift for at least 15 ka, it seems proba-blethatthecentreofmagmaticaccretionisshifting.Wespeculate that future intrusion of dikes will be focussed where magma is currentlymostabundant,i.e.15 kmtothewestofthepresentday

axialdepression:thecentralDMHsegmentisundergoingaminor “ridge-jump”. Whilethetimescalesinvolved arehighly debatable, itseemslikelythatwithinthenexttensofkyrsanewaccretionary axiswillbeinplacewestwardofthepresent-dayrecognized “neo-volcaniczone”.

8. Conclusions

In this study,cosmogenic 36Cland 3He lava surface exposure dating,combinedwithfieldobservations,geological mappingand geochemistry, show that the magmatic activity in a 15 km sec-tion across the Dabbahu–Manda Hararosegment is sustained by two distinct reservoirs: one beneaththe axis anda second lying 15km tothewestbeneaththeDurrievolcaniccomplex.Thetrace elementcharacteristicsofthesemagmasshowthattheywere gen-erated by variable degrees of partial melting of a homogeneous mantlesource.Themagmasevolvedseparatelyindistinct plumb-ing systems.Theaxialmagma chamberdifferentiated slowlyover time,consistentwithadecreaseinmagmasupplybetween20 ka andthepresentday.Conversely,theslightlyoff-axis(“flank”) reser-voir appears to have been regularly supplied with magma since 15 ka, resulting in lessvariably differentiated lavas. Interconnec-tions betweenthesetwo reservoirs haveoccurred, aswell asthe eruption of lavas displaying an “axial” signature in an off-axis position (Fig. 8). The steady state buffered composition of vol-canism emitted on top of the DMH active magmatic reservoirs (7.9–8.6 wt%MgO)isingoodagreementwiththecomposition of magmas evolving in a system controlled by slow spreading rate withlowmeltsupplies.

OurdatashowthatmagmatismintheDMHsegmentisnot fo-cussed within the currentaxial depression but instead is spread out over at least 15 km of the western flank. Coeval lava pro-duction occurredfrom volcanoes that were separatedby at least 15 km.Magma supplyfromtwo differentreservoirs isconsistent withmagneto-telluricobservationsby Desissaetal. (2013) which showthattwomagmabodiesarepresentbelowthesegment,with themainmagmabodycurrentlylocatedbelowthewesternflank, preciselywherethemostvoluminousflankvolcanismoccurs.The axial reservoir only represents 25 km3 of melt, i.e. about 5% of

the total(but still 10timesthe amountinjected during the dyk-ingeventsinthecurrentcrisis e.g.Desissaetal.,2013).However, giventheprincipalmagmalocation(i.e.inaseparatereservoir be-low thewesternflank, inactiveduring the2005event)it islikely that future dikeintrusions(once the currentepisodehas ceased, i.e.all thestress hasbeenrelieved)willoriginate fromandclose to theDurrie(flank)reservoir, ratherthan fromtheaxial magma chamber.

We infer that the Durrie reservoir is currentlytaking over as theprincipal magmachamber inthemid-segment,whilethe ax-ial magma chamber (the source ofthe 2005dikingevent) isnot being replenished and is therefore dying. As a consequence,the expressionofmagmatism overawider areathanthepresent-day depression alone could be an indicator for a future minor ridge jump.Atthescaleofslowspreadingmid-oceanridges,thiscould indicate that magmabodies havea lifetimeofatleast 10–15 ka, andthatthecontinuityofthemagmaticactivityismaintainedby asystemofdistinctreservoirsbroadlydistributedbetweenthe cur-rentaxisandtheflanks.Inreturn,thelongtermrecurrenceofthis changeinthefocusofmagmaticactivitycouldcontrolthelocation ofspreading.Therefore,magma distributionandmagmachamber longevityappeartobekeyfeaturesoftheorganization and main-tenance of active spreading centres on top of stable soft points inthemantle.Theseobservationsofariftclosetothecontinent– oceantransitionprovidevaluableinformationformodelsofmature oceanicridgedevelopment.

(12)

Fig. 8. SchematicevolutionoftheDMHriftoverthepast30 kaillustratingthetwomainresurfacingeventsat30–25 ka(fromthemid-axisreservoir)andat15ka (from theDurriereservoir).Theevolutionofthesetwomagmareservoirsthroughtimesuggestsrelayingmagmaticactivityfromtheaxistothewesternflank,andmayprefigure arift-jump.

Acknowledgements

Fortheir helpinthe field,we thankthe membersoftheAfar regionalgovernmentatSemera.WearegratefultoB. Tibari(CRPG) forassistance withcosmogenic3He andUandThmeasurements

andL. Zimmerman(CRPG)fortechnicalassistancewithmass

spec-trometry. WethankBarbaraHoffman forprovidingtheLidarDEM oftheNERC-fundedAfarRiftConsortium.CVBpublisheswith per-mission oftheExecutiveDirectoroftheBritish GeologicalSurvey (NaturalEnvironmentResearchCouncil).WeareverygratefultoR. Buck andanonymousreviewer fortheir constructivereview.This isCRPGcontributionno. 2345.

(13)

Appendix A. Supplementarymaterial

Supplementarymaterialrelatedtothisarticlecanbefound on-lineathttp://dx.doi.org/10.1016/j.epsl.2014.11.002.

References

Ayele,A.,Keir, D., et al., 2009.September 2005mega-dikeemplacement inthe Manda-Harraronascentoceanicrift(Afardepression). Geophys.Res. Lett. 36 (20).

Ayele,A.,Stuart,G.,et al.,2007.TheAugust2002earthquake sequenceinnorth Afar:insightsintotheneotectonicsoftheDanakilmicroplate.J.Afr.EarthSci. 48 (2–3),70–79.

Barberi,F.,Tazieff,H.,etal.,1972.VolcanismintheAfardepression:itstectonicand magmaticsignificance.Tectonophysics 15(1–2),19–29.

Bastow,I.D.,Keir,D.,2011.Theprotracteddevelopmentofthecontinent–ocean tran-sitioninAfar.Nat.Geosci. 4,248–250.

Bastow,I.D., Nyblade, A.A.,etal., 2008.Uppermantle seismicstructure beneath theEthiopianhotspot:riftingattheedgeoftheAfricanlow-velocityanomaly. Geochem.Geophys.Geosyst. 9(12).

Bastow,I.D.,Pilidou,S.,etal.,2010.Melt-inducedseismicanisotropyandmagma assistedriftinginEthiopia:evidencefromsurfacewaves.Geochem.Geophys. Geosyst. 11(6).

Belachew,M.,Ebinger,C.,Coté,D.,Keir,D.,Rowland,J.V.,Hammond,J.O.S.,Ayele, A.,2011.Comparisonofdikeintrusionsinanincipientseafloor-spreading seg-mentinAfar,Ethiopia:seismicityperspectives.J.Geophys.Res. 116,B06405. http://dx.doi.org/10.1029/2010JB007908.

Buck,W.R.,2013. Plate tectonics:magmafor 50,000 years. Nat.Geosci. 6 (10), 811–812.

Calais,E.E.C.,Hartnady,C.,Nocquet,J.M.,2006.KinematicsoftheEastAfricanRift fromGPSandEarthquakeSlipVectorData.

Carbotte,S.M.,2005.Seismicstructureatmid-oceanridges.In:Richard,C.S.,etal. (Eds.),EncyclopediaofGeology.Elsevier,Oxford,pp. 405–417.

Carignan,J.,Hild,P.,etal.,2001. Routineanalysesoftraceelementingeological samplesusingflowinjectionandlowpressure on-lineliquidchromatography coupledtoICP-MS:astudyofgeochemicalreferencematerialsBR,DR-N,UB-N, AN-GandGH.Geostand.Newsl. 25(2–3),187–198.

Chalot-Prat,F.,Falloon,T.J., Green,D.H.,Hibberson,W.O.,2010.An experimental studyofliquidcompositionsinequilibriumwithplagioclase+spinel lherzo-liteatlowpressures(0.75 GPa).J.Petrol. 51(11),2349–2376.http://dx.doi.org/ 10.1093/petrology/egq060.

Colman,A.,Sinton,J.M.,etal.,2012.Effectsofvariablemagmasupplyonmid-ocean ridgeeruptions:constraintsfrommappedlavaflowfieldsalongtheGalápagos SpreadingCenter.Geochem.Geophys.Geosyst. 13,Q08014.

Desissa,M.,Johnson,N.E.,etal.,2013.Amantlemagmareservoirbeneathan incip-ientmid-oceanridgeinAfar,Ethiopia.Nat.Geosci. 6(10),861–865.

Ebinger,C.,Ayele,A.,etal.,2010.Lengthandtimescalesofriftfaultingandmagma intrusion:theAfarriftingcyclefrom2005topresent.Annu.Rev.EarthPlanet. Sci. 38(1),439–466.

Ebinger,C.,Keir,D.,Ayele,A.,Belachew,M.,Calais,E.,Wright,T.,Campbell,E.,Buck, R.,2008.Magmaintrusionandfaultingprocessesinazoneofcontinental rup-ture:seismicity ofthe Dabbahu (Afar)rift. Geophys.J.Int.http://dx.doi.org/ 10.1111/j.1365-246X.2008.03877.x.

Ebinger,C.J.,Hayward,N.J.,1996.Softplatesandhotspots:viewsfromAfar.J. Geo-phys.Res. 101(B10),21859–21876.

Ferguson, D.J., Barnie, T.D., et al., 2010. Recent rift-related volcanism in Afar, Ethiopia.EarthPlanet.Sci.Lett. 292(3–4),409–418.

Ferguson,D.J.,Calvert,A.T.,et al.,2013.Constrainingtimescalesoffocused mag-maticaccretionandextensionintheAfarcrustusinglavageochronology.Nat. Commun. 4,1416.

Field,L.,Blundy,J.,etal.,2012.MagmastorageconditionsbeneathDabbahu Vol-cano(Ethiopia)constrainedbypetrology,seismicityandsatellitegeodesy.Bull. Volcanol.,1–24.http://dx.doi.org/10.1007/s00445-012-0580-6.

Field,L.,Blundy,J.,etal.,2013.MagmatichistoryofDabbahu,acompositevolcano intheAfarRift,Ethiopia.Geol.Soc.Am.Bull. 125(1–2),128–147.

France,L.,Koepke,J.,MacLeod,C.J.,Ildefonse,B.,Godard,M.,Deloule,E.,2014. Con-taminationofMORB byanatexis ofmagmachamberroof rocks:constraints from a geochemical study of experimental melts and associated residues. Lithos 202–203,120–137.http://dx.doi.org/10.1016/j.lithos.2014.05.018. Gale,etal.,2013.Geochem.Geophys.Geosyst. 14(3),489–518.

Geoffroy,L.,2005.Volcanicpassivemargins.C.R.Géosci. 337(16),1395–1408. Grandin,R.,Socquet,A.,Binet,R., Klinger,Y.,Jacques,E.,deChabalier,J.-B.,King,

G.C.P.,Lasserre,C.,Tait,S.,Tapponnier,P.,Delorme,A.,Pinzuti,P.,2009. Septem-ber2005 Manda Hararo-Dabbahu rifting event, Afar (Ethiopia): constraints

provided by geodetic data. J. Geophys. Res. 114, B08404. http://dx.doi.org/ 10.1029/2008JB005843.

Grandin,R.,Jacques, E.,etal., 2010a.Seismicityduringlateraldikepropagation: insightsfrom newdatainthe recentMandaHararo/Dabbahurifting episode (Afar,Ethiopia).Geochem.Geophys.Geosyst. 12,Q0AB08.

Grandin,R.,Socquet,A.,etal.,2010b.Transientriftopeninginresponseto multi-pledikeinjectionsintheMandaHararorift(Afar,Ethiopia)imagedby time-dependent elasticinversion ofinterferometricsynthetic aperture radardata. J. Geophys.Res. 115(B9),B09403.

Hamling,I.J.,Ayele,A.,Bennati,L.,Calais,E.,Ebinger,C.J.,Keir,D.,Yirgu,G.,2009. GeodeticobservationsoftheongoingDabbahuriftingepisode:newdyke intru-sionsin2006and2007.Geophys.J.Int. 178(2),989–1003.

Hammond,J.O.S.,2014.ConstrainingmeltgeometriesbeneaththeAfarDepression, Ethiopiafromteleseismicreceiverfunctions:theanisotropicH-κstacking tech-nique.Geochem.Geophys.Geosyst. 15(4),1316–1332.

Hammond, J.O.S., Kendall, J.M., et al., 2011. The nature of the crust beneath theAfartriplejunction:evidencefromreceiverfunctions.Geochem.Geophys. Geosyst. 12(12),Q12004.

Hammond, J.O.S., Kendall, J.M., et al., 2012. The nature of the crust beneath theAfartriplejunction:evidencefromreceiverfunctions.Geochem.Geophys. Geosyst. 12(12).

Hayward,N.J.,Ebinger,C.J.,1996.Variationsinthealong-axissegmentationofthe AfarRiftsystem.Tectonics 15,244–257.

Hofmann,C.,Courtillot,V.,etal.,1997.TimingoftheEthiopianfloodbasaltevent andimplicationsforplumebirthandglobalchange.Nature 389,838–841. Keir,D.H.,Hamling,I.J.,Ayele,A.,Calais,E.,Ebinger,C.,Wright,T.,Jacques,E.,

Mo-hamed,K.,Hammond,J.O.S.,Belachew,M.,Baker,E.,Rowland,J.,Lewi,E., Ben-nati,L.,2009.Evidenceforfocusedmagmaticaccretionatsegmentcentersfrom lateraldikeinjectionscapturedbeneaththeRedSeariftinAfar.Geology 37, 59–62.http://dx.doi.org/10.1130/G25147A.1.

Keir,Derek,Pagli,Carolina,Bastow,IanD.,Ayele,Atalay,2011.Themagma-assisted removalofArabiainAfar:evidencefrom dikeinjectionintheEthiopianrift capturedusingInSAR andseismicity. Tectonics 30,TC2008. http://dx.doi.org/ 10.1029/2010TC002785.

Lahitte,P.,Gillot,P.-Y.,et al.,2003. Siliciccentral volcanoesas precursorstorift propagation:theAfarcase.EarthPlanet.Sci.Lett. 207(1–4),103–116. Macdonald,K.C.,2001.Mid-Oceanridgetectonics,volcanism,andgeomorphology.

In:John,A.,Steele,H.,Karl,K.T.,Steve,A.T.(Eds.),EncyclopediaofOcean Sci-ences.secondedition.AcademicPress,Oxford,pp. 852–866.

Marty,B.,Pik,R.l.,etal.,1996.HeliumisotopicvariationsinEthiopianplumelavas: nature ofmagmatic sources and limit on lower mantle contribution. Earth Planet.Sci.Lett. 144(1–2),223–237.

McClusky,S., Reilinger,R.,et al.,2010.Kinematicsofthesouthern RedSea–Afar TripleJunctionandimplicationsforplatedynamics.Geophys.Res.Lett. 37(5), L05301.

Medynski,S., Pik,R., etal.,2013.Controlsonmagmaticcyclesanddevelopment ofrifttopographyoftheMandaHararosegment(Afar,Ethiopia):insightsfrom cosmogenic3Heinvestigationoflandscapeevolution.EarthPlanet.Sci.Lett. 367,

133–145.

Pik,R.,Marty,B.,etal.,2006.HowmanymantleplumesinAfrica?Thegeochemical pointofview.Chem.Geol. 226(3–4),100–114.

Rowland, J., Baker, E., et al., 2007. Fault growth at a nascent slow-spreading ridge:2005Dabbahuriftingepisode,Afar.Geophys.J.Int. 171.http://dx.doi.org/ 10.1111/j.1365-246X.2007.03584.x.

Rubin,K.H.,Sinton,J.M.,2007.Inferencesonmid-oceanridgethermalandmagmatic structurefromMORBcompositions.EarthPlanet.Sci.Lett. 260(1),257–276. Standish,J.J.,Sims, K.W.W.,2010. Young off-axisvolcanism along the

ultraslow-spreadingSouthwestIndianRidge.Nat.Geosci. 3(4),286–292.

Tiberi,C.,Ebinger,C.,Ballu,V.,Stuart,G.,Oluma,B.,2005.Inversemodelsofgravity datafromtheRedSea-GulfofAden-Ethiopiantriplejunctionzone.Geophys.J. Int. 163(2),775–787.http://dx.doi.org/10.1111/j.1365-246X.2005.02736.x. Vye-Brown,C.,2012.GeologicalmappinginthenorthernpartoftheDabbahuRift

Segment,Afar,Ethiopia.BGSInternalReport.

Vye-Brown,C.,Medynski,S.,etal.,2012.GeologicalmapoftheDabbahu (Manda-Hararo)Rift,North,1:100,000scale.BritishGeologicalSurvey.

Vye-Brown,C.,Smith,K.,Napier,B.,Hofmann,B.,Wright,T.,submittedfor publi-cation.Three-dimensionalremote mappingofthe AfarRift,Ethiopia.J.Appl. Volcanol.

Wolfenden,E.,Ebinger,C.,etal.,2005.Evolutionofavolcanicriftedmargin: South-ernRedSea,Ethiopia.Geol.Soc.Am.Bull. 117(7–8),846–864.

Wright,T.J.,Ebinger,C.,etal.,2006.Magma-maintainedriftsegmentationat conti-nentalruptureinthe2005Afardykingepisode.Nature 442(7100),291–294.

Références

Documents relatifs

In an orthogonal opening model, it accommodates pure dip-slip normal faulting with extension orthogonal to the rift segments and a general E-W extension for the entire EARS..

The Kivu rift region forms the central segment of the western branch of the East African rift system, between the northern termination of the Tanganyika rift and the

The Cenozoic volcanism in the Kivu rift: Assessment of the tectonic setting, geochemistry, and geochronology of the volcanic activity in the South-Kivu and Virunga regions.. As

7 shows, stations from regions I-A and II and station ARMI have high STDV during summer (panel c) and low STDV during fall (panel d), while stations in regions I-B and I-C exhibit

A metal sculpture mirroring the image was installed on the wall, in order to create a public monument resembling a lightning bolt that lasted until the school was

Moreover, the sediment fluxes measured on the different nonmesquite plant/soil complexes are very similar: minimum 62 g cm −1 for the 3‐month sampling time for creosote and maximum

We note that in the data of both satellites, the VTEC and also the Ne increases are observed when the satellites cross the Eastern Paci fic region (90°W ÷ 120°W), which is close to

Cependant, comme le signalent Bernadette Charlier et Marie Lambert dans leur contribution à cette revue, les décideurs, concepteurs et formateurs manquent d’outils