(Afar depression, Ethiopia) Magmatic cycles pace tectonic and morphological expression of rifting Earth and Planetary Science Letters

Earth and Planetary Science Letters 446 (2016) 77–88

Contents lists available atScienceDirect

Earth and Planetary Science Letters

www.elsevier.com/locate/epsl

Magmatic cycles pace tectonic and morphological expression of rifting (Afar depression, Ethiopia)

S. Medynski

, R. Pik

, P. Burnard

, S. Dumont

, R. Grandin

, A. Williams

,

P.-H. Blard

, I. Schimmelpfennig

, C. Vye-Brown

, L. France

, D. Ayalew

, L. Benedetti

, G. Yirgu

, ASTER team

aCRPG,UMR7358CNRS–UniversitédeLorraine,15rueNotreDamedesPauvres,BP20,54500Vandoeuvre-lès-Nancy,France bNordicVolcanologicalCenter,InstituteofEarthSciences,UniversityofIceland,Reykjavik,Iceland

cInstitutdePhysiqueduGlobedeParis,SorbonneParisCité,UnivParisDiderot,CNRS,F-75005Paris,France

dAix-MarseilleUniversité,CNRS-IRDUM34CEREGE,Technopôledel’EnvironnementArbois-Méditerranée,BP80,13545Aix-en-Provence,France eBGS,MurchisonHouse,WestMainsRoad,Edinburgh,EH93LA,UnitedKingdom

fSchoolofEarthSciences,AddisAbabaUniversity,POBox1176,AddisAbaba,Ethiopia

a r t i c l e i n f o a b s t ra c t

Articlehistory:

Received17November2015 Receivedinrevisedform13April2016 Accepted14April2016

Availableonline9May2016 Editor:A.Yin

Keywords:

tecto-magmaticprocesses cosmogenicisotopes Afar

Theexistenceofnarrowaxialvolcaniczonesofmid-oceanicridgestestifies oftheunderlyingconcentra- tion of both melt distributionand tectonic strain. As aresult of repeateddiking and faulting, axial volcaniczones thereforerepresent aspectacular topographic expressionof platedivergence. However, thesubmarinelocationofoceanicridges makesit difficulttoconstrainthe interplaybetweentectonic and magmatic processesintime andspace. Inthisstudy,weuse theDabbahu–MandaHararo(DMH) magmatic rift segment (Afar,Ethiopia) toprovidequantitative constraintsontheresponse oftectonic processestovariationsinmagmasupplyatdivergentplateboundaries.TheDMHmagmaticriftsegment isconsideredananalogueofanoceanicridge,exhibitingafaultpattern,extensionrateandtopographic reliefcomparabletointermediate- toslow-spreadingridges.Here,wefocusonthenorthernandcentral parts ofDMHrift, wherewe presentquantitative sliprates forthe past40 kyr formajor and minor normalfaultscarpsinthevicinityofarecent(September2005)dikeintrusion.Thedataobtainedshow thattheaxialvalleytopographyhasbeen createdbyenhancedslipratesthatoccurredduringperiods oflimitedvolcanism,suggestiveofreducedmagmaticactivity,probablyinassociationwithchangesin straindistributioninthecrust.Ourresultsindicatethatthedevelopmentoftheaxialvalleytopography hasbeenregulatedbythelifetimesofthemagmareservoirsandtheirspatialdistributionalongtheseg- ment,andthustothemagmaticcyclesofreplenishment/differentiation(<100kyr).Ourfindingsarealso consistentwithmagma-induced deformationinmagma-richrift segments.The recordoftwotectonic eventsofmetricverticalamplitudeonthefaultthataccommodatedthemostpartofsurfacedisplace- mentduringthe2005dikeintrusion suggeststhatthelattertypeofintrusionoccursroughlyevery10 kyrinthenorthernpartoftheDMHsegment.

©^{2016ElsevierB.V.}All rights reserved.

1. Introduction

Thevariabilityofmagmaproductioninthemantleandsubse- quenttransfersofmagmatothecrust,andpotentiallythesurface, are fundamental, first-order controls on the style and morphol-

*

E-mailaddress:medynski@ucdavis.edu(S. Medynski).

1 NowatUniversityofCaliforniaDavis.

2 Deceased.

3 ASTERTeam:MauriceArnold, GeorgesAumaître, DidierBourlès, KarimKed- dadouche.

ogy of mid-ocean ridges (MOR) (MacDonald and Atwater, 1978;

Carbotte et al., 2001; Macdonald, 2001; Macdonald etal., 2005).

Few quantitative constraintsexiston howmagmatic andtectonic processes are coupled via dyke injection and fault slip in such a way as to maintain crustal accretion and produce typical ax- ial morphologies along a magmatic rift at the scale of a few to tens of thousands of years (White et al., 2006; Standish and Sims, 2010; Grandin et al., 2012). The building of ridge topog- raphy results from competition between tectonic activity, which creates the topography via normal faulting, and magmatic activ- ity, which tends to erase the topography by filling the growing depressionwithvolcanicproducts (Behnetal.,2006).Manystud- http://dx.doi.org/10.1016/j.epsl.2016.04.014

0012-821X/©^{2016ElsevierB.V.}All rights reserved.

ies have documented the contributions of diking and faulting to the extension process (most notably in Iceland), on both long- term(MastinandPollard,1988; ForslundandGudmundsson,1991) and short-term timescales (Rubin, 1992; Gudmundsson, 2003;

Doubre and Peltzer, 2007; Calais et al., 2008; Biggs et al., 2009;

Dumontetal.,2016).However,animportantyetunaddressedissue isthequantificationofthistectonicactivityintermsofvariations inmagmaticactivityinthelongterm.Inparticular,isthetectonic activityconstantthrough timeor,on thecontrary,isitrelatedto themagmaticprocesses?

To address thisquestion, we examine the subaerial Dabbahu/

Manda–Hararo(DMH)Afaractivemagmaticriftsegment(Ethiopia, Fig. 1A), which represents a natural laboratory for investigating topographicevolution inresponse tocomplex magmaticandtec- tonicinteractionsatdivergentplateboundaries.AlthoughtheDMH rift segment is currentlyat the ocean–continent transitionstage, its morphology and extension rates are comparable to those of intermediate to slow-spreading ridges, suggesting that the same processesareatworkinthetwosettings.

The ∼^{55 km long DMH rift segment in Central Afar is char-} acterised by anarrowaxial graben(∼^{3km)flankedby} <100 m- high fault scarps. The total relief of the axial valley is <300 m atthe DMH rift,typical of intermediate- to slow-spreadingMOR morphologies and similar to that observed in Iceland. (Fig. 1, Gudmundsson, 2005.) Four magmatic reservoirs have been iden- tified along the DMH segment (Fig. 1) (Grandin et al., 2009;

Wrightetal.,2006,2012;Ebingeretal.,2008; Barisinetal.,2009;

Fieldetal.,2012):twoaxialmagmaticcentresatthenorthernend of the segment; one below Dabbahu volcano; and one midway along thelength ofthe segment,referred to asthemid-segment magma chamber (MSMC). Dabbahu andthe MSMC both possess a shallow reservoir that is connected to a deeper reservoir be- low 15 km depth (Grandinet al., 2010b; Field etal., 2012). The shallowDabbahu magma storage area may consist of a seriesof stacked sills at 1 to 5 km depth (Field et al., 2012). In addi- tion,twomagmaticreservoirsarelocatedoff-axis:Gabhovolcano (in the north-east) and the Durrievolcanic centre (in the west) (Fig. 1). In September 2005, a major intrusion ruptured the en- tire length of the DMH rift,initiating a 5-yr-long rifting episode that involved 13 further smaller dike intrusions and that high- lightedcomplexmagmaticinteractionsbetweenthreeofthemag- maticcentres:Dabbahu,GabhoandtheMSMC(Wrightetal.,2012;

Grandinetal.,2010a,2010a;Hamlingetal.,2009).Themajordike that initiated the crisis was modelled to be 60 to 70-km-long,

∼^{1–2 km3 intrusion and produced 4 m ofaverage regional hor-} izontalopening (Wright etal., 2012, 2006; Grandinetal., 2009).

Thirteensubsequentandsmaller dikes(∼^{0.1km3 each)were in-} truded between2005and2010, all emittedfrom theMSMC and producing a lesser degree of deformation (Buck, 2006; Hamling etal., 2009; Grandinetal., 2010a, 2010b; Belachew etal., 2011;

Wrightetal.,2012).Thetransferofmagmatoshallowdepthsacti- vatednumeroussurfacefaultsandfissuresalongthelengthofthe dikeintrusion(Fig. 2A,Rowland etal.,2007; Wright etal., 2006;

Grandinet al., 2009; Dumont etal., 2016). The grounddisplace- mentassociatedwiththisfirstdikeintrusion wastoolargeinthe near-fielddiscriminate theroleofindividual faults usinggeodetic approaches (Wright et al., 2006; Grandin etal., 2009). However, the modelling of Grandin et al. (2009) and, more recently, the analysis of surface fault displacements during inter-diking peri- ods between 2005 and 2010 atthe DMH rift segment (Dumont et al., 2016) suggest that only faults dipping toward the dike were able to release the dike-induced stresses. In the DMH rift segment, a substantial opening component also prevails, but no evidenceforreversefaultinghasbeenidentified.Similarphenom- ena have also been observed in Iceland during the Krafla crisis

Fig. 1.Regionalgeologicalsetting.A:RegionalsettingoftheAfarRiftandlocationof theDabbahu/MandaHararosegment–modifiedafter(Ebingeretal.,2008).B: Dab- bahu/MandaHararoriftmagmaticcomplexesandfaultpattern.Theblackdotted lineshowsthe principalriftaxis,definedasthelowestpointofthedepression.

Whitelineslabelled1,2&3correspondtothetopographicsectionsinFig. 4.The redlineindicatesthelocationofthe2005dykeintrusionandfaultzonereactiva- tion.NotethattheSeptember2005dike(redline)didnotintrudeonthealignment oftheriftaxis,butdeviatedslightlytotheeast,belowtheriftshoulder.Redcircles indicatethe differentmagmabodies,locatedbelowtheDabbahuandthe Gabho volcanoes,andatthemid-axis:themid-segmentmagmachamber(MSMC)andthe slightlyoff-axisDurrievolcano.Themainstudyareasarethefollowing:onecross- sectionlocatedatthecontactoftheaxialdepressionwiththesegment-tipvolcano ofDabbahu(profile 1),andtwocross-sectionslocatedatthemid-lengthoftheseg- ment,rangingfromthemid-axistotherecent(15 ka;Medynskietal.,2015)slightly off-axisDurrievolcaniccentre(profiles2and3).(Forinterpretationoftherefer- encestocolourinthisfigurelegend,thereaderisreferredtothewebversionof thisarticle.)

S. Medynski et al. / Earth and Planetary Science Letters 446 (2016) 77–88 79

Fig. 2.Geologicalsetting,samplingandgeomorphologyofthenorthernendoftheDabbahuMRS.A:GeologicalmapoftheaxialzoneofthenorthernDabbahuMagmaticRift Segment,reconstructedfromASTERandLandSatsatelliteimageryandfieldobservations(2008,2010and2011fieldseasonsMedynskietal.,2013; Vye-Brownetal.,2012).

Thefaultsreactivatedduringthe2005eventareindicatedinred;faultsthatdidnotmovein2005areinblack.Thethicknessofthelinecorrespondstothefault’scurrent height:thicklinesareforscarpsgreaterthan15metreshigh.B:DeformationassociatedwiththeSeptember2005intrusion,andpost-dikedeformation.Theprofilesshow theverticaldeformationduringthedikeintrusionsandtheperiodDec.–Jun.2006.Thedeformationwasinducedbymagmaflowatshallowdepthsandwasobtainedfrom InSARimages(seeGrandinetal.,2009formoredetailsontheretrievaloftheverticalcomponentfortheco-dikedeformationandDumontetal.,2016forthepost-dike).

Thetopographyisshowninthebackground.Cross-sections(markedinFig. 2A),highlightthelocationsoftherift-axisandfaultreactivationcorridor.(Forinterpretationof thereferencestocolourinthisfigurelegend,thereaderisreferredtothewebversionofthisarticle.)

(1974–1985) (Brandsdottir andEinarsson, 1979; Einarsson, 1991;

Sigmundssonetal.,2015).

Detailed studies of the tectonic and magmatic processes in- volvedintheriftingepisodewereconductedoverthedecadethat followed (see for example, Wright et al., 2006; Rowland et al., 2007; Ebingeretal., 2008; Ayeleetal., 2009; Barisinetal.,2009;

Hamlingetal.,2009;Grandinetal.,2009,2010a, 2010b, 2012;Keir etal., 2011; Wrightetal., 2012). ThemajorSeptember 2005dike was remarkable in that it didnot intrude the main rift axisbut wasinsteademplacedundertheeasternriftshoulderatitsnorth- ern end(Fig. 1B). Thisoff-axis positionappears tohave resulted fromthe primary involvementof Dabbahuand Gabhoreservoirs, withtheMSMCcomingintoplayonlylater(Fig. 1B;Wrightetal., 2006;Ayeleetal.,2009;Grandinetal.,2009).Inthenorthernpart ofthesegment,theSeptember2005intrusionreactivatedfaultsto the east of the central graben Fig. 1B andFig. 2 (Ebinger et al., 2010).Inacross-sectionalongtheriftinthenorthernpartofthe segment,theassociateddeformationappearstocorrelateinversely withtheriftsegmenttopography(Fig. 2B);thehighestamountof

scarp activation occurredon therift shoulders whileno (or only limited) faulting was recorded atthe rift axiswhere the depres- sionisgreatest.

Althoughnotalldikeintrusionswillprovokeatopographicre- sponse at the surface (Gudmundsson and Philipp, 2006; Froger et al., 2004), those that can cause fault reactivation must ac- tively participate in the buildingof topography asdiking is now considered the principal mode of accommodating extension in magma-richriftsegmentssuchasthoseinAfar(RubinandPollard, 1988;Behnetal.,2006;Buck,2006).Inthisstudy,weuseterres- trial cosmogenic nuclide (TCN) dating (Gosse and Phillips, 2001;

Medynski etal., 2013) to integrate thefirst, majordikeintrusion of the 2005–2010 rifting crisis into the long-term topographical development of theDMH rift segmentand toinvestigate the re- lationships between faulting events andmagmatic differentiation anderuptiontimescales.Theadvantageofthisdatingtechnique is that itallows thetiming ofbothvolcanicandfault-slipeventsto be constrained (Gosse andPhillips, 2001; Palumbo et al., 2004).

A keyobjectiveofthestudyistoquantifythelong-term(100-kyr

Fig. 3.Faultscarppreservation.A:Circlesrepresentexamplesofsamplestakenonfaultscarps(thisstudy)andstarsrepresentsamplestakenonthelavaflowatthetop ofthescarp(frompreviousstudies:Medynskietal.,2013,2015).Gab-A:Photography ofarecentlyslumpedblockthatprovidedaccesstosamplesfromthetop-mostten metresofthescarp.Giventhegoodpreservationstateofthescarp,itwaspossibleto usesamplesfromtheblocktodeterminetheexposureageofthetopofthescarp.

Durrie15:Sampledonthetopofthelavaflowinordertoconstrainthedevelopmentofthefault(datingdescribedinMedynskietal.,2015).Durrie10:Locationofsample D-10(star),takenfromthefaultscarp,ontheflanksoftheDurrievolcano.Durrie16:LocationofsampleD-16,takenfromthetopofthelavaflowdissectedbya20 m highfault(scarpedgemarkedbythedottedlineintheforeground).Thedottedlineinthebackgroundshowstheextentofthelavaflowfieldsinthedepression.Thedating ofthelavaunitsisdescribedinMedynskietal. (2015).B: LiDARviewsoftheDikika-1faultillustratinghowlavamorphologycanbeusedtoconstraintherateoffault movement.Aswewereunabletodirectlydatethefaultitself(whichcutsthinp ¯ahoehoeflowsthatareinappropriateforcosmogenicdatingduetopalaeoexposureofthe lavasurfaces;seeMethods),weusedtheageofthegreenlavaflow(20–25ka),whichbuttsagainstthefaultinthenorthandflowsoverthefaultinthesouth,toconstrain amaximumupliftrateoftheDikika-1faultof1.35±⁰.3 mm/yr.LiDARdatacourtesyBarbaraHofmannandTimWright,NERCAfarRiftConsortium.Seetextfordiscussion.

(Forinterpretationofthereferencestocolourinthisfigurelegend,thereaderisreferredtothewebversionofthisarticle.)

timescale)dike-inducedtopographicsurfacedisplacementinorder tobetterunderstanditscouplingwithmagmaticactivity.

2. Methods

2.1. Terrestrialcosmogenicnuclidesandscarpdating

Terrestrial cosmogenic nuclides (TCN) provide a robust tech- niquefordetermining chronologies ina variety ofgeological set- tings(seereviewsinGosseandPhillips,2001; Niedermann,2002;

Dunai and Wijbrans, 2000). The technique requires the perfect preservationofsurfaces (thefault scarpsandthelavasurfacesin thisstudy)asremoval(evenpartial)ofthesurface wheremostof theTCNaccumulatewillinduce abiasintheageestimation.The low rainfall anderosion in the Afar present ideal conditions for surfacepreservation(Fig. 2) andthusTCNexposuredating(Gosse andPhillips,2001).

TCNconcentrations along tectonic scarpsin volcanic environ- ments havetwo origins: TCNaccumulationthrough thelava sur- face(i.e.,sincetheemplacementofthelavaflow);andaccumula- tion through thevertical face of thescarp itself (i.e.,since scarp rupture).Specialcaremust betakento evaluatetherelative con- tributionsofthesetwocomponents,eitherbydatingthelavaflow emplacement or by sampling the scarp at sufficient depth rela- tivetothelava-topinordertoavoidsurface-exposurecontamina- tion (TCN production isnegligible below depths of 4–5 m). It is thereforeeasierto datescarpsthat dissectrelatively thick (upto

20 m inthis area)massivea¯a flows thanthose thatdissect the p ¯ahoehoe flowunits (usually1to 4m thick),whereit isimpos- sibletoassesspotential cosmogenicinheritancebetweentheem- placement oftwosuccessiveflow units.Allsamplesinthisstudy were thereforetakenatleast1.5m belowthetop ofthescarpin order to ensure dating ofthe scarp exposure alone andto avoid anybias causedby lava top exposure.The maximumcontamina- tion fromtheflow topswas 3%, whichis negligiblecompared to the measurement uncertainties. Sample location and site details are givenFigs. 2 and 3,and cosmogenic details are given in Ta- ble 1.

2.2. Cosmogenic³Heproductionrate

Forthisstudy,we useda localproductionratedeterminedby cross-datingusingAr–Arandcosmogenic³Hetechniquesandpre- viously published in Medynski et al. (2013). In that study, the compilation ofGoehringetal. (2010) provideda goodagreement withAr–Ar ages,butwas unable to take intoaccount the paleo- magneticevolutionnecessarytoconserveconsistencybetweenthe two methodsofdating,andalocalproductionratewas therefore calculated.

The calculation of the reference ³He production rate (P_3local) scaledtosea-levelhigh-latitudeisgivenby:

P_3local

= [

]

(

) ∗

∗

∗

S.Medynskietal./EarthandPlanetaryScienceLetters446(2016)77–8881 Table 1

Backgroundsampleinformation.Self-shieldingfactors,scalingfactor(Stone,2000) andP³HecosrateswerecalculatedusingtheCosmocalccalculatorofVermeesch(2007).A0.5topographic-shieldingcorrectionisrequiredfora verticalsurface.

Sample Altitude Latitude

(^◦N)

Longitude (^◦E)

Lava morphology

Petrology Scarp

description

Sample thickness (cm)

Thickness correction

Topo shielding

Stone scaling factor

Local P³Hecat g⁻¹yr^−1a

±

NORTH of study area:³He dating Gabbole main scarp (65 m)

Gab A2 444 12.49192 40.53872 Massiveaa

lavaflow

CumulativeOl, Cpx&Pl phenocrysts

Preservedslumped blockrepresentativeof thefirst10 mofthe scarp(fromthetop)

3 0.98 0.5 0.86 45.5 5

Gab A3 444 12.49192 40.53872 4 0.97 0.5 0.86 45.2 5

Gab A4 444 12.49192 40.53872 4 0.97 0.5 0.86 45.2 5

Gab A7 448 12.49192 40.53872 Pahoehoe

flow

Ol&Pl phenocryst rich

Baseofthescarp(above the2005reactivation mark)

5 0.96 0.5 0.86 44.8 5

Gab-B old scarp (11 m)

Gab B 305 & 370 386 12.483331 40.5339 Upper

pahoehoe flows

Ol&Cpx phenocryst rich

Upperscarpzone partiallyrefreshed

3 0.98 0.5 0.83 43.6 5

GabB540to990 GabB614

386 12.483331 40.5339 Lower

pahoehoe flows

Ol&Cpx phenocryst rich

Scarpwellpreserved eolianpolishfeature(?)

3 0.98 0.5 0.83 43.6 5

Dikika 3 scarp (8 m)

Dik 3 samples 413 12.49107 40.543027 Massiveaa

lavaflow

CumulativeOl, Cpx&Pl phenocrysts

Coreoftheflow The 2005eventclearly appears

3 0.98 0.5 0.84 44.2 5

Two subsequent base of minor faults (<10 m)

Gab G3 406 12.490611 40.53105 Massiveaa

lavaflow

CumulativeOl, Cpx&Pl phenocrysts

Coreoftheflow Well preservedscarp

8 0.94 0.5 0.84 42.5 5

Yem 7 436 12.51344 40.52374 Massiveaa

lavaflow

CumulativeOl, Cpx&Pl phenocrysts

Coreoftheflow Well preservedscarp

8 0.94 0.5 0.86 43.5 5

Altitude Latitude (^◦N)

Longitude (^◦E)

Lava morphology

Petrology Scarp

description

Sample thickness (cm)

Thickness correction

Topo shielding

Stone scaling factor (neutrons)

Scalingfactor formuonic production

SOUTH of study area

Two minor faults dated with 36Cl (4 m)

D-10 465 12.392917 40.505583 Pahoehoe

flows

Aphyric Coreoftheflow 3 0.98 0.5 0.87 0.78

D-12 454 12.402033 40.515450 Pahoehoe

flows

Aphyric Coreoftheflow 3 0.98 0.5 0.86 0.77

a Theproductionratesappliedtoeachsamplewerecalculatedusingalocal³Heproductionrateof108 at/g/yr.

where [³He_{Gab D}] is the concentration in at/g of cosmogenic ³He determinedinsampleGab-D,AgeAr–AristheAr–Aragedetermined for the Gab-D sample, f is the Stone correction factor (Stone, 2000), Correc_prof isthecorrection factorforthesample thickness, and Correc_magnetic is the correction factor due to paleomagnetic variations(Stone,2000). P_3local=¹⁰⁸±^{12 at}/g/yr.

Thislocalproductionrateislowerthan theratespublishedin (Blard etal.,2013) butisconsistent withrecentlocal production globalmeancalculatedattheFogoIslands(86–109 at/g/yr;Foeken etal.,2009)atsimilarlatitude(14.9^◦NatFogo;12^◦Ninthisstudy).

ThecosmogenicagesofallsamplespresentedinSupplementary MaterialNo. 3&4werecalculatedusing:

Age

= (

3local∗f∗Corr_prof

)

Cosmogenic³Hewasmeasuredonseparatedolivineandpyroxene grainsintheNobleGasLaboratoryatCRPG,byheatingto1300^◦C undervacuumfollowingthemethodsdescribedinMedynskietal.

(2013).FormoredetailsseeSupplementaryMaterials5.

2.3. Cosmogenic³⁶Cl

The TCN ³⁶Cl is applicable to timescalesof ∼^{103 to} >10⁶ yr (DunaiandWijbrans,2000).It hasbeenextensivelyused fordat- ing surfaces with carbonate lithologies (including fault scarps;

Palumboetal.,2004)ascalciumisoneofthemaintargetelements fortheinsituproductionof³⁶Cl(besidespotassium).Afewstud- ieshave alsoused³⁶Clmeasurements involcanicwhole rocksto datetheemplacementoflavaflows(Zredaetal.,1993) andtocal- ibrateTCNproductionrates(Licciardi andPierce, 2008). Although pure Ca- and K-bearingminerals,such asfeldspars,are preferred for³⁶Cldatingbecauseoftheirsimplerchemicalcompositioncom- paredtowholerocks(Zredaetal.,1993;LicciardiandPierce,2008;

Schimmelpfennig et al., 2009, 2011), the lack of sufficient phe- nocrystsinthelavasstudiedoftenimposestheuseofwhole-rock samples.In this study,the sampled lavas presentaphyric or mi- crolithic textures,preventingtheuseofseparatedmineralphases for exposure dating. We therefore used whole-rock ³⁶Cl analysis inordertoestimate theexposure agesofthescarps. Cosmogenic 36Clwas measuredattheASTERfacilityatCEREGE,followingthe methods described inMedynskiet al. (2015). SeeSupplementary Material4and 5fordetails.

3. Long-termfaultslipratesandthebuildingoftopography

Building of axial depression topography over the long term (∼^{100 kyr timescale) can be} constrained using (i) the sliprates ofthemajorfaults(determinedherefromcosmogenic³Heor³⁶Cl exposure ages on the escarpments), and (ii) dating oflava flows emplaced alongthe riftsince 60ka,inorderto estimateaverage sliprates(Medynskietal.,2013,2015).

ThetwomajortectonicfeaturesofthenorthernDMHrift(sec- tion 1 on Fig. 1 andFig. 4) are the Gabbole and Dikika-1 faults (Fig. 3).Thesearesyntheticwest-dippingfaultsthatoffsettherift floorby55and30m,respectively,atthesamplingsites(Figs.3A and4),andconstitutetheeasternboundaryofthemainaxialde- pression.Intheaxialvalleyandonthewesternflank,topography isdistributed along asuccession of smallerscarps(<20 m) such asthe Gab-B site(Fig. 3and SOM 1). The slip rateson the two boundaryfaults (Fig. 4) wereconstrainedasfollows.TheGabbole fault cuts a 51.1 ka lava flow which implies a minimum aver- age vertical slip rate of 1.1±0.5 mm/yr (SOM 1). Additionally, the scarp itself was dated tenmetres below the top ofthe fault withages ranging from19 to 29.5ka (Fig. 4A andSOM 1–3 for

details),withtheoldestageprovidingtheminimumexposuredu- ration of this upper portion of the fault. This corresponds to a maximumverticalsliprateof1.5±0.1 mm/yr (Fig. 4A),henceat 29.5kaatleast10 mofthe present-dayaxialdepressionhadal- readybeenbuilt.Similarly,thesouthernendoftheDikika-1fault cuts a flow unit dated at 20–25ka (Medynski etal., 2013) at a point where the scarp is 30 m high, and therefore has an aver- ageverticalsliprateof1.35±⁰.15 mm/yr.Fieldrelationsindicate that the fault propagated southwards (Fig. 3) andinitiated north of our samplinglocation. Assuming a mean constant sliprate of 1.5±⁰.1 mm/yr,then thissystemoffaultscarpsfirstinitiatedat 37.6±².7 ka (Fig. 4A).Toconstraintheslipratesthataffectedthe westernmarginoftheaxialdepression,threesmall(<15m)faults (Gab-G3,Yem-7 andGab-B) were dated.The Gab-G3 andYem-7 scarps(sampled 1 and1.5 mabove the presentgroundlevel,re- spectively)displayexposureagesof8.4±⁰.9 ka and21.7±².5 ka, corresponding to significantly lower average vertical sliprates of 0.18±⁰.02 and0.07±⁰.01 mm/yr,respectively,comparedtothe majorscarps.TheGab-Bfaultcutsanoldervolcanicunitemplaced at 72.1±⁴.3 ka (Medynski et al., 2013) (SOM 1). Thus, small (<15 m),rarelyreactivatedfaultswithlowaverageslipratesexist inthisnorthernpartoftheaxialgraben.Thedifferenceintheslip ratesofthemajorandminorfaultscombinedwiththeasymmetric rift axistopography (Fig. 4A) suggest that thedeformation isnot uniformlydistributedacrossthenorthernpartofthesegment.

In the central partof the DMHsegment, closerto the MSMC (section 2 on Fig. 1andSOM 1), theages oflavaflows offsetby faults(30to6.4ka,Medynskietal.,2013,2015)canalsobeused toestimateslipratesonthemainfaults.Here,theaxialgrabenis symmetrical, withantithetic faults thatexhibit mean verticalslip ratesof1.3to1.6mm/yr(Fig. 4B). Thesevaluesareclosetothose of themajor faults inthe northernDMHsegment overthe same periodofactivity,suggestingcommonreactivationmechanismsin thenorthernandcentralpartsofthesegment.However,unlikein thenorthern partofthe segment,thetopography(andsliprates) across the central part is/(are) symmetrical (Fig. 4B), suggesting that themechanisms areuniformlydistributedhere. Additionally, one ofthesefaults displacesa youngerflow(6.4 ka,Medynskiet al.,2015)byover20 m,resultinginahighermeansliprateof2.8 mm/yrthatsuggeststhatfaultingmayhaveincreasedrecently(i.e.

after6.4ka).

The most recent active volcanism in the central part of the DMHriftsegmentoccursoutsidetheaxialdepression,centredon the15–5kaDurrieVolcanicComplex(SWoftheDabbahuvolcano, spreadingfromtheaxisupto15kmtothewest)(Medynskietal., 2015).Thisriftshouldervolcaniccomplexischaracterisedbyopen fissures and small (<5 m) recent fault scarps (Medynski et al., 2015).Acombinationoflava-flowsurfaceandfault-scarpexposure datinggivesconsistent slipratesof0.34–0.35mm/yr(Fig. 4C)for these faults,which are associatedwiththe currentoff-axis high- intensitymagmaticphase(Medynskietal.,2015).Thesesliprates aremuchlowerthanthosedeterminedfortheequivalentperiodin theaxial graben(wheretheslipratesashighas2.8mm/yrwere measured).

4. RecurrencetimescaleofSeptember2005-typedikeintrusion

Our field observations and scarp dating have allowed us to place constraints on the development of the two specific faults in the northern portion of the segment: Dikika-3 and Dikika-1 (Fig. 2A and Fig. 4). These west-dipping faults are separated by a50-mdeepgrabenthatisboundedbytheeast-dippingDikika-3 fault on its western border(Fig. 2andFig. 5). Thissmall graben is notobserved inanyother location along therift,andits posi- tion along theeastern rift shoulder suggeststhat itmight be re- latedtooff-axisintrusionofthetypeobservedinSeptember2005

S. Medynski et al. / Earth and Planetary Science Letters 446 (2016) 77–88 83

Fig. 4.DenudationratesalongtheDMHRift.Leftcolumn:A:theorthernsegmentextremity.Theratesoffaultmovementarediscussedinthetext.Twodifferentfault displacementregimesareevident:themainscarps,GabboleandDikika-1,presentaverageslippingrateswhicharetentimeshigherthanontheminorscarps(Dik-3,Gab-G3 andYem-7).ThisisconsistentwithsubduedtectonicmovementduringrapidemplacementoflavasduringDabbahumagmaticcycle1(between70and58kyr;orangebar) followingMedynskietal. (2015),whereasthelessvoluminousDabbahumagmaticcycle2(50–20kyr;blueandgreen)ischaracterisedbyrapiddenudation.B:Centralpart ofthesegment.Thefaultslippingratesarehomogeneousalongthedepression,andcomparabletothehighslipratesrecordedinthenorthernextremityofthesegment, leadingtothesymmetricalriftmorphology.C:Westernriftshoulder:deformationmarkersarelimitedtoopenfracturesandsmallscarps.Theslowslipratesrecordedalong thefaultsindicatesubduedtectonicactivityduringphasesofhighmagmainput,whichiscurrentlythecaseattheDurrievolcaniccomplex(Medynskietal.,2015).The correspondingsamplingsitesareshownoncross-sectionsontheright(foralargerviewofthecross-sections,refertoFig. 1).Notetheclearasymmetryofthenorthernpart oftheriftdepression,comparedtothesymmetriccentralpartoftherift.AlsonotetheinfluenceoftheDurrieyoungvolcaniccentreontheriftshouldermorphology:the intenseandrecent(<15 ka)volcanicactivityerasedtheprevioustopographyandonlysmallfaultactivescarpsarepresent.Thecosmogenicdatingofthesescarps–andthe deducedsliprates–showsthattheirlimitedheightresultsfromsubduedtectonicactivityandnotjustrapidlavainfilling.(Forinterpretationofthereferencestocolourin thisfigurelegend,thereaderisreferredtothewebversionofthisarticle.)

(Figs. 2 and6A).The8 m-highDikika-3scarp(DIK-3,Figs.3Aand 4) played a major role during the September 2005intrusion, ac- commodatingmostofthedeformationthatoccurredinthenorth- ernportionofthe rift(Fig. 2,Wrightetal., 2006; Rowlandetal., 2007; Ebinger et al., 2008; Grandin et al., 2009). Geodetic mea- surementshaveshownthatthe faults abovethe September2005

dike, particularly DIK-3 andto a lesserextent the Dikika-1 fault, continuedtoslipinthemonthsfollowingthisfirstmainintrusion asaresultofmagmadrainingoutofDabbahudeepreservoirand into the path ofthe September 2005 dike (Ebinger et al., 2008;

Grandinetal.,2009; Dumontetal.,2016).Incontrast,muchlarger scarpsclosertotheaxis(e.g.the65mGABBOLEscarp,350 mW

Fig. 5.CaptureofmetrictectoniceventalongtheDik-3scarp.A:³HeprofileoftheDIK-3scarp.³Heconcentrationsincreaseverticallyupthescarpwithinflectionsthatare consistentwithexhumationofpalaeosurfaces.Asascarpisexhumed,itupliftstheportionofthescarppreviouslyburiedimmediatelybelowthesurface;thispalaeosurface thereforepresentsaTCNconcentrationprofilewhichdecreasesexponentiallywithdepth.Onceexhumed,thewholesurfaceofthescarpstartsaccumulatingTCNwitha constantandhomogeneous productionrate.Theresulting³Heconcentrationisthereforethesumofthe³Heproducedbelowthescarppriortotheseismicevent,andthe 3Heaccumulatedabovethesurfaceafterexhumation.Afterseveraltectonicevents,thescarpdisplaysaconcentrationprofilewithaseriesofexponentials,separatedby discontinuitiesmarkingthedifferenttectonicevents(Palumboetal.,2004).B: PhotographoftheDik-3scarpcorrespondingtothe³HeprofileshowninFig. 3A.Theabsence offallenblocksatthebaseofthescarpindicatesgoodpreservationofthefaultsurface(therubbleatthebaseisfromthetopofanaaflow,notduetofallenblocks).Note theobviousupliftresultingfromthe2005event(paleportionofthescarpvisibleinthephoto).Thetopofthe2005scarissetas0 m.C:Simulationsof³Heaccumulation ontheDik-3scarpfordifferentscarpemplacementscenarios:Model1(blue)considers3eventsthataffectedthefaultat16–20ka,5–6kaandin2005.Thisfitsthe datarelativelywellbutfailstoadequatelyreproducethedatabetween0and1 m(between2005andthe5–6kaevent).Model2(red)usesthesameupliftscenarioas model 1butaddsslighterosionofthescarpbetween80–25 cm,whichcanexplainthesteep³Heprofile(steeperthanexponential)observedonthescarpatthislevel.This erosioncouldcorrespond,forexample,toarefreshmentofthescarpsurfaceofabout15 cm,whichmaywellhaveoccurredduringthetectoniceventitself.Model3(black) illustratestheexpectedprofileforregularuplift,assumingatectoniceventevery2kyr,with50cmofscarpdenudationeachtime(inordertomatchthelong-termuplift rate).Ourdatastronglysuggestthatatleastthreelarge(>1 m)displacementscreatedtheDik-3scarpwithlongperiodsofinactivity(upto12kyr)betweendisplacements.

Irrespectiveofthedisplacementmodelused,themeanupliftrateontheDIK-3scarpis0.3±⁰.04 mm/yr overthepast20kyr,and0.16 mm/yrsinceemplacementofthe dissectedlava.(Forinterpretationofthereferencestocolourinthisfigurelegend,thereaderisreferredtothewebversionofthisarticle.)

ofDIK-3; Fig. 3A) show littleorno evidence fordisplacementin 2005(Figs.3Aand4 andSOM 1). Thisisalmost certainlydueto theirdistancefromthedikeand,inthecaseofGABBOLE,itswest- dippingorientation,i.e.awayfromthedike,whichisunfavourable toaccommodationofthedike-inducedstresses.

To investigate the past recurrence of events similar to the September 2005 intrusion, we have reconstructed the displace- menthistory oftheDik-3 fault usingcosmogenic³He concentra- tions measured on a continuous vertical profile along the well- preservedDik-3 scarp(Fig. 5).The concentrationsincreaseupthe faceofthe scarp,withwell-definedinflectionsinthe profilethat

areconsistentwithmultipleepisodesoffaultmovement(Fig. 5).In the scarpsection upliftedduring the2005event(recognisableby a whitehorizontalstripepresenton thescarp-face; Fig. 5), there isanexponentialincreasein³Heconcentrationover∼1 m;thisis the fossilattenuationprofile whichwasburied priorto 2005and wasonlyexposedbythe2005faultmovement.Twosimilarexpo- nentialprofiles are visiblefurther up thescarp, corresponding to twoearlierslipevents.Thesetwoeventsbothalsoproduced>1 m verticalmovementonthescarp(Fig. 5),similarinamplitudetothe 2005event. Thesethree tectoniceventscan bemodelledby: i) a 2 mslipin2005, ii)a1mslipat5–6kyrbpandiii)an eventof