JOURNALOFGEOPHYSICALRESEARCH:SOLIDEARTH,VOL.118,3899–3916,doi:10.1002/jgrb.50318,2013 Multiphased tectonic evolution of the Central Algerian margin from combined wide-angle and reflection seismic data off Tipaza, Algeria A.Leprêtre,1 F.Klingelhoefer,2 D.Graindorge,1 P.Schnurle,2 M.O.Beslier,3 K.Yelles,4 J.Déverchère,1 andR.Bracene5 Received23October2012;revised4July2013;accepted30July2013;published26August2013. [1] TheoriginoftheAlgerianmarginremainsoneofthekeyquestionsstilldiscussedin theWesternMediterraneansea,duetotheimprecisenatureandkinematicsofthe associatedbasinduringtheNeogene.Forthefirsttime,thedeepstructureofthe MaghrebianmarginwasexploredduringtheSPIRALseismicsurvey.Inthiswork,we presentaN-StransectoffTipaza(westofAlgiers),aplacewherethemarginbroadens duetoatopographichigh(Khayr-al-DinBank).Newdeeppenetrationseismicprofiles allowustoimagethesedimentarysequenceintheAlgerianbasinandthecrustalstructure atthecontinent-oceanboundary.Modelingofthewide-angledatashowsthinningofthe basement,frommorethan15kminthecontinentaluppermargintoonly5–6kmof oceanic-typebasementintheAlgerianbasin,andrevealsaverynarroworabsent transitionalzone.Analysisofthedeepstructureofthemarginindicatesfeaturesinherited fromitscomplexevolution:(1)anoceanic-typecrustinthedeepbasin,(2)similarities withmarginsformedinatransform-typesetting,(3)aprogressivedeepeningofthewhole sedimentarycover,andthethickeningofthePlio-Quaternarysedimentsatthemarginfoot, coevalwith(4)adownwardflexureofthebasementinthebasin.Thesefeaturesarguefor amultiphasedevolutionofthemargin,including(1)anearlystageofriftingand/or spreading,(2)alatetranscurrentepisoderelatedtothewestwardmigrationoftheAlboran domain,and(3)adiffusePlio-Quaternarycompressionalreactivationofthemargin. Citation: Leprêtre,A., F.Klingelhoefer, D. Graindorge,P. Schnurle,M. O.Beslier, K.Yelles, J.Déverchère, and R.Bracene (2013),MultiphasedtectonicevolutionoftheCentralAlgerianmarginfromcombinedwide-angleandreflectionseismicdataoff Tipaza,Algeria,J.Geophys.Res.SolidEarth,118,3899–3916,doi:10.1002/jgrb.50318. 1. Introduction [e.g., Bown and White, 1994; Louden and Chian, 1999; Geoffroy, 2005; Ziegler and Cloetingh, 2004]. Neverthe- [2] Research on continental passive margins is largely less, during rifting and/or subsequent oceanic spreading focused on the understanding of rifting processes and stages, the above factors, together with the geodynamical mechanisms of lithosphere thinning leading to continen- setting, may evolve, conferring to the margins a com- tal breakup and spreading. The way in which continen- plex structure. In addition, rifting in back-arc basins might tal margins exhibit various structural styles according to be different in some points from cratonic rifting, though the setting of rifting, inheritance, magmatic supply, and/or the mechanics of actual fracturing of the continental crust mantle conditions during their formation is well studied remainssimilar.Themaindifferencebetweenback-arcand cratonic rifting is the presence of a subducting slab in the 1Domainesocéaniques,UMR6538,InstitutUniversitaireEuropéende mantle beneath the back-arc basin [Currie and Hyndman, laMer,UniversitédeBretagneOccidentale,Plouzané,France. 2006;DunnandMartinez,2011]. 2DepartmentofMarineGeosciences,Ifremer,Plouzané,France. [3] The North African Algerian passive continentalmar- 3UniversitédeNiceSophia-Antipolis,CNRS(UMR7329),Observa- gin results from back-arc opening of the Western Mediter- toiredelaCôtedAzur,Valbonne,France. ranean basin during Oligo-Miocene times [Schettino and 4CentredeRechercheenAstronomieAstrophysiqueetGéophysique, Algiers,Algeria. Turco,2006].Whiletheopeninghistoriesoftheneighboring 5SONATRACHExploration,Boumerdès,Algeria. basins (Liguro-Provençal and Tyrrhenian basins, Figure 1) are fairly well understood today [Gueguen et al., 1998; Corresponding author: A. Leprêtre, Domaines océaniques, UMR Jolivet and Faccenna, 2000; Rosenbaum et al., 2002], the 6538, Institut Universitaire Européen de la Mer, Université de Bretagne Occidentale, place Nicolas Copernic, 29280 Plouzané, France. mechanisms for the opening of the Algerian basin are still ([email protected]) controversial regarding (1) the rifting processes (asymmet- ric, symmetric), (2) the rate and direction of opening, and ©2013.AmericanGeophysicalUnion.AllRightsReserved. 2169-9313/13/10.1002/jgrb.50318 (3) the postrift evolution of its southern margin. This is 3899 LEPRÊTREETAL.:CENTRALALGERIANMARGINEVOLUTION Figure 1. Present-day tectonic map of the Western Mediterranean area. Bathymetry and topography are from ETOPO1 1min Global relief (www.ngdc.noaa.gov). Major tectonics feature are from Frizon de Lamotte et al. [2000] and Billi et al. [2011]. The white arrow shows velocity of the African plate relativetothestableEuropeanplate((cid:2)5mm/yr)fromGPSmeasurements[NocquetandCalais,2004]. TheredrectanglemarksthelocationoftheregiondisplayedinFigure2.ThetwoinsetsshowtheWestern Mediterraneansetting(a)at35Maand(b)at18Ma,simplifiedandmodifiedfromLonerganandWhite [1997],withthemigrationoftheInternalZonesbehindtheTethyansubductionfrontandtheassociated back-arcopeningoftheAlgerianbasin.Theredcrossshowstheapproximativepositionofourstudyarea. C:Corsica,S:Sardinia,GK:GreatKabylia,LK:LesserKabylia,Pe:Pelorian,Ca:Calabria,Al:Alboran, B:Betic,R:Rif.,M.E:MazarronEscarpment. especially due to the lack of knowledgeon the deep geom- physique; and Directorate-General for Scientific Research etry of the basin and surrounding margins. The Algerian andTechnologicalDevelopment)andFrenchResearchorga- margin is now one of the few examples of a margin that nizations (Centre National de la Recherche Scientifique, experiencedatectonicinversion,resultingintherecentand InstitutFrançaisdeRecherchepourl’ExploitationdelaMer actual compressional field [Serpelloni et al., 2007] attested (Ifremer), Institut de Recherche pour le Développement, bytheseismicityaswellastectonicandkinematicevidences anduniversities).Specifically,newwide-angleseismictran- [Yellesetal.,2009]. sects,togetherwithcoincidentmultichannelseismic(MCS) [4] Because of its setting, the Central Algerian margin dataprovidethefirstconstraintsonthemargin’sdeepstruc- (Tipazaregion,westofAlgiers)isakeyareaforattempting ture,onthenatureoftheocean-continenttransition(OCT), thereconstructionoftectonicevolutioninthissouthernpart and the associated Algerian basin, as well as on the recent oftheWesternMediterraneanseaandforunderstandingthe compressive reactivation at crustal scale. In this study we modification of passive margins by reactivation processes. present first results from a deep seismic transect across Furthermore, it is the only place where a large-scale tilted the Central Algerian margin based on forward modeling of block,calledtheKhayr-al-DinBankandinheritedfromthe wide-angleseismicdataandacoincidentmultichannelseis- rifting stage, is proposed to be present [El Robrini, 1986; micprofileandcompareitwithothermarginsoftheWestern Domzig et al., 2006; Yelles et al., 2009; Strzerzynski et al., MediterraneanseaandtheAtlanticocean. 2010]. [5] Among the major unsolved questions, we would like to address the following points: (1) What is the nature 2. GeologicalSetting and thickness of the crust underlying the Algerian basin? (2) Where is the ocean-continent transition, and what is its 2.1. TheAlgerianMargin origin? (3) What is the nature and deep geometry of the [7] TheAlgerianmargincorrespondstotheplatebound- Khayr-al-Din Bank? (4) Is there evidence for deep mark- arybetweentheEuropeanandAfricanplates.Itisbounded ers of margin reactivation? (5) What are the implications to the north by the Algerian basin and to the south by an oftheseresultsonmodelsfortheevolutionoftheAlgerian Alpine-type belt called Maghrebides (known as the “Tell” basinanditssouthernmargin? north Algeria, Figure 1), resulting from the subduction and [6] Inordertounravelthedeepgeometryandstructuresof closure of the Tethyan ocean under the European plate theMaghrebianmarginbothonshoreandoffshore,theSPI- inMiocenetimes[Auzendeetal.,1973;FrizondeLamotte RAL (Sismique Profonde et Investigations Régionales en etal.,2000]. Algérie)projectwaslaunchedinSeptember2009incollab- [8] TheevolutionoftheAlgerianmarginiscloselyrelated oration between Algerian scientific institutions (Sonatrach; to the rollback of the Tethyan slab and the related back- CentredeRechercheenAstronomie,AstrophysiqueetGeo- arcopeningoftheWesternMediterraneanbasins(Figure1). 3900 LEPRÊTREETAL.:CENTRALALGERIANMARGINEVOLUTION Figure 2. Location of the wide-angle seismic profile in the Bou Ismail Bay, sector of Tipaza. Ocean bottom seismometer (OBS) positions are marked by red circles and the land stations by blue triangles. The numbers of the stations shown in Figure 6 are shaded. The multichannel seismic profile Spi06 is indicatedwithawhiteline.ThegeologicalandtectonicframeworkonlandisextractedfromYellesetal. [2009].KADF:Khayr-al-Dinfault,CF:Chenouafault,SF:Sahelfault,SA:Sahelanticline,Al:Algiers massif,Ch:Chenouamassif. ThereisageneralconsensusthattheAlgero-Provençalbasin etal.,1998;Gelabertetal.,2002;LonerganandWhite,1997 openedduringlateOligocene-earlyMiocenetimesinaback- andSchettinoandTurco,2006].Whatevermodelchosen,the arc position behind the Tethyan subduction zone [Jolivet westernmostAlgerianmargincanbeassumedtorepresenta andFaccenna,2000;Gelabertetal.,2002;Speranzaetal., purelystrike-sliptypemargin[Domzigetal.,2006],having 2002]. In the early Miocene, the stretching of the Euro- formedasaSTEP-faultsystem(subduction-transformedge peanplateisassumedtohavecausedthedrifting,spreading, propagator,GoversandWortel[2005]). and finally the collision of parts of a continental block, the [10] TheAlgerianmarginandbasinarethenmarkedbya Internal Zones, called AlKaPeCa (for Alboran, Kabylies, majorsalinitycrisisduringMessiniantimes,whichaffected Peloritan, and Calabria; Bouillin [1986]), with the African the whole Western Mediterranean domain and surrounding continent (Figure 1a). Currently, those Internal blocks are margins ((cid:2)5.96–5.32Ma, Hsu et al. [1973]; Gautier et al. scattered around the Western Mediterranean basin, part of [1994]; Krijgsman et al. [1999]). This event resulted in the themhavingbeenaccretedalongtheAlgerianmargin,such progressive closure of the connection between the Atlantic astheKabylianblocks(Figure1). Ocean and the Mediterranean Sea, responsible for a total [9] Models of the opening of the Algerian basin remain sea level fall of more than 1500m [Ryan and Cita, 1978]. controversial regarding the kinematics and nature of the It led to an intense erosion of the basin margins, the accu- margins:(1)Someauthorspromoteanopeningofthisbasin mulation of erosional products in the downslope domain attherearofadoublesubductiontowardthewest(Alboran) [SavoyeandPiper,1991;Sageetal.,2005],andthedeposi- and the east (Calabria) [Malinverno and Ryan, 1986; tionofthethickevaporiticMessiniansequencesinthedeep Lonergan and White, 1997], after the Kabylian collision Mediterranean basin [Montadert et al., 1970; Hsu et al., withtheAfricanplate(18Ma),resultinginadominantE-W 1973;Lofietal.,2011]responsibleformarkedsalttectonics. openingbetween16and8MabehindtheGibraltararcroll- The Messinian units form a good temporal seismic marker, backandAlboranblockmigrationtowardthewest[Mauffret easilyrecognizableintheMediterraneanarea. etal.,2004].ThewestwardmigrationoftheAlboranblock [11] The Plio-Quaternary period was then characterized [Mauffret et al., 2004] would have induced a left-lateral bythetectonicinversionoftheAlgerianmargin.Thismajor deformationalongtheWesternandCentralAlgerianmargins tectonic episode is still in progress and contributes to the and right-lateral deformation along the Balearic Promon- general structure of the Algerian margin. Recent kinematic tory [Camerlenghi et al., 2009] (Figure 1b). (2) Other studiesindicateapresent-dayshorteningassociatedwiththe authors propose an older NW-SE opening of the Algerian NW-SE[Stichetal.,2006]convergencebetweentheAfrican basinbehindtheretreatingofasubductionzonetowardthe andEuropeanplatesofabout5–6mm/yratthelongitudeof S-SE, whereas no significant displacement (i.e., less Algiers[NocquetandCalais,2004](Figure1).Asignificant than(cid:2)200km)oftheAlboranblockisconsidered[Gueguen part (between 1.6 and 2.7mm/yr) of the deformation may 3901 LEPRÊTREETAL.:CENTRALALGERIANMARGINEVOLUTION Figure3. High-resolutionseismicprofilesfromtheMARADJAcruise(2003)acrosstheKhayr-al-Din Bank(modifiedafterYellesetal.[2009](verticalexaggeration:9).Thepositionofeachlineisindicated by red lines on the map, and the deep seismic profile SPIRAL Spi06 is indicated by the black line. (a) Seismicsectionshowingthesteepnorthernslopeofthebanctowardthedeepbasin,andthesedimentary sequenceatthetopofthebancandatitsfoot.(b)Seismicsectionacrossthecentralpartofthebanc.The inset shows the compressive bulge at the foot of the margin identified by Yelles et al. [2009], which is assumedtoberelatedtothepresenceofasouthdippingblindthrustbeneaththeKADB. currently occur offshore Algeria (Serpelloni et al. [2007]; AlgiersInternalmassifstothewestandtheeast,respectively Meghraoui et al. [1996], respectively), and even further (ChandAl,Figure2),andtothenorthbythedeepAlgerian north in the SE Iberian margin [Maillard and Mauffret, basin. 2013]. [14] The KADB was recently investigated using mor- [12] Contractional deformation is supported by the exis- phological and high-resolution seismic data (MARADJA tence of dominantly reverse-type fault plane solutions in 2003 and 2005 cruises [Déverchère et al., 2005; Domzig the present seismicity, as exemplified by earthquakes of et al., 2006, Yelles et al., 2009; Strzerzynski et al., 2010]) Chenoua(M =6.0,1989[Bounifetal.,2003]),AinBenian and is interpreted as a tilted block inherited from the rift- w (M =5.7, 1996), and Boumerdès (M =6.8, 2003 [Delouis ing of the Algero-Provençal basin, as first suggested by El w w et al., 2004]) (Figure 2). In recent papers, authors describe Robrini [1986]. It is also assumed to represent relics of the active offshore structures as folds and south dipping blind Kabylian basement, originally part of the Internal Zones, thrusts both east and west of Algiers [Domzig et al., 2006; and the offshore extension of the Chenoua and Algiers Déverchère et al., 2005; Yelles et al., 2009; Strzerzynski internalmassifswhichoutcroponshore[Domzigetal.,2006; et al., 2010], one of them being tentatively related to the Strzerzynskietal.,2010](Figure2). destructive Boumerdès earthquake in May 2003 and attest- [15] Evidence for recent compression is indicated by ing to the recent compressional reactivation of the margin the presence of an asymmetric (cid:2)100m high bulge at the (Figure2).TheinversionoftheAlgerianmarginremainsan foot of the slope visible in the morphology and imaged active process, and the North African margin could repre- by seismic reflection profiles from the MARADJA cruise sentanearlystageofincipientsubduction,asfirstsuggested (Figure 3b). The asymmetric steeper northern flank of the by Auzende et al. [1972] and later by Yelles et al. [2009] bulgeandtheassociatedbasementupliftcouldbecontrolled and Strzerzynski et al. [2010], based on recent studies con- byanactivesouthdippingthrustsystemlocatedbeneaththe ductedontheKhayr-al-DinBank,abathymetrichighinthe Khayr-al-Din Bank, although this structure is not directly region of Tipaza (Figure 2). This area is thus assumed to imaged[Domzigetal.,2006;Yellesetal.,2009](Figure3b). haverecordedallthetectonicepisodesthathaveaffectedthe In north Algeria, structures inherited from the Miocene Algerianmargin,fromtherollbackoftheTethyanslabtothe phasearegenerallycharacterizedbyasouthwardvergence, recentcompressionalreactivationofthemargin. liketheMiocenesutureborderingtheInternalZonestothe south, whereas newly formed reverse structures from the 2.2. SectoroftheKhayr-al-DinBank(KADB) reactivation exhibit an opposite northward vergence [Yelles [13] West of Algiers, the continental shelf significantly etal.,2009;Déverchèreetal.,2005](Figure2). widensandformsthebathymetrichighoftheKhayr-al-Din Bank (KADB, Figure 2). This structure extends over (cid:2)80km in a roughly E-W direction and 45km in a N-S 3. SeismicData direction,overlookingthedeepbasinofabout2000mdepth (Figure 2). The northern KADB limit shows a steep slope 3.1. DataAcquisition with a basinward dip of about 12ı (Figure 3a). It is bor- [16] A wide-angle seismic profile close to Tipaza and dered onshore by the Sahel structure, the Chenoua and the coincident MCS cross-section Spi06 are presented in 3902 LEPRÊTREETAL.:CENTRALALGERIANMARGINEVOLUTION channels.MCSandOBSdatawererecordedwithasample rateof4ms. 3.2. MultichannelSeismic(MCS)DataandProcessing [18] The SPIRAL seismic sources were chosen to image deep targets, such as the top and the base of the crust, the OCT, and the deep rooting of the structures. There- fore, the deep penetrating and low-frequency MCS data set is complementary to the high-resolution and superfi- cialdataacquiredduringtheMARADJAcruises(2003and 2005) and was used to image deep structures underneath the salt layer (Figures 4 and 5). A first quality control was undertaken on groups of traces using the SISPEED soft- ware, and further processing of the MCS data was then performedusingtheGEOCLUSTERsoftware.Theprocess- ingsequenceincludedexternalandinternalmutes,spherical divergence correction, bandpass filtering (3–5–95–105Hz), anddynamiccorrections.Twoconsecutivevelocityanalyses wereconductedevery200CMP(commonmidpoint)leading to the final stack. The last processing step was the appli- cation of a frequency-wavenumber migration on the data using a constant 1550km/s water velocity. The Spi06 pro- file exhibits a higher resolution than the Spi25 profile, due tothehigherfrequenciesofthesource.Therefore,theMCS interpretation presented in this paper is based on the Spi06 profile, whereas interfaces from MCS data integrated dur- Figure 4. (a) Stratigraphic units identified on the line ingtheforwardmodeling(seethenextsection)werepicked Spi06inthedeepAlgerianbasin,(b)theircorrelationswith from the Spi25 profile to avoid even minimal differences velocities from the velocity forward modeling (this study), in time and/or space between MCS and wide-angle arrival andwith(c)high-resolutionseismicreflectiondatafromthe times.Thewide-angleseismicdataaidsgeologicalinterpre- MARADJAcruise(profilea,Figure3). tation of the crust by providing deeper and complementary informationsuchasP-wavevelocities(Vp)onthestructure ofthemargin. this work (Figure 2). Deep seismic data were acquired duringtheSPIRALcruiseconductedontheR/VL’Atalante 3.3. SeismicVelocityModelingoftheWide-Angle (IFREMER) in October–November 2009. Thirty-nine SeismicData four-component ocean bottom seismometers (OBS) spaced [19] The refraction data were modeled using forward at3kmintervalweredeployedalongthe120kmNNW-SSE modeling technique, taking into account first as well as profileacrossthewesternAlgerianmargin(BouIsmailbay) secondary arrivals from OBS and land stations, and reflec- offshore,and23landstationswerealsodeployed,extending torspickedfromthecoincidentmultichannelseismicsection the marine transect by 110km on land. The seismic profile (Figure7). crosses the deep basin, the Khayr-al-Din Bank, the Sahel 3.3.1. DataQualityandPreprocessing structure,andtheMitidjabasin(Figure2). oftheWide-AngleSeismicData [17] Two different seismic sources were used during the [20] OBSdatawerecorrectedforclock-drift,andseafloor cruiseinordertoachievetwoobjectives:(1)TheMCSpro- positionswerecalculatedusingthedirectwaterwave.Apre- file Spi06 coincident with the wide-angle line off Tipaza processing sequence was applied to all data (land stations presentedinthisstudywasacquiredusingaseismicairgun andOBS)inordertoincreasethesignal-to-noiseratioandto arrayof13airgunsofvariousvolumes(synchronizedonthe better image far-offset arrivals. This sequence is composed firstbubble)togenerate2299shotsoflowfrequency(max- ofadeconvolutionwhitening,a3–17HzButterworthfilter, imum frequency of (cid:2)70Hz) to allow for deep penetration andanautomaticgaincontrol. oftheseismicsignal[Avediketal.,1993].Thissourcepro- [21] The OBS data acquired along the Tipaza profile are videdatotalvolumeof50L,withanintershotof20sleading of good quality, with a better signal-to-noise ratio on the to 50m spacing. Our objective was to increase the seismic vertical geophone component than on the hydrophone. The coverage to allow for better processing results. (2) For the OBS sections show clear sedimentary (Ps1, Ps2, Ps3) and wide-angleacquisition,aseismicairgunarraycomposedof crustalarrivals(Pg1,Pg2),anddeeparrivalsfromtheupper eight airguns of 16L and two airguns of 9L was used to mantle (Pn) are identifiable (Vp (cid:3) 7.6km/s) up to 50km generate751low-frequencyshots,synchronizedonthefirst offsetawayfromsomeOBS(Figures6aand6b).Sedimen- peak.Thissourceprovidedatotalvolumeof146L,withan taryreflections(PsP1,PsP2)aswellasreflectionsfromthe intershot of 60s leading to 150m spacing. Simultaneously top of the basement (PgP) are clearly observed in the deep withthewide-angleacquisition,thecoincidentMCSprofile basin. Moho reflections (PmP) are not always easily dis- Spi25wasacquired.AllMCSprofileswererecorded,using cernibleinthedeepAlgerianbasin,evenafterapplyingthe the 4.5km streamer of Ifremer, composed of 360 12.5m preprocessingsequence.Fortheforwardmodeling,picking 3903 LEPRÊTREETAL.:CENTRALALGERIANMARGINEVOLUTION Figure 5. (a) Time migrated multichannel seismic profile Spi06. (b) A corresponding line drawing showingthedeepeningofboththetopandthebaseoftheMessinianunitsandseveraldeepreflections. The inset shows deep reflection of low amplitude probably corresponding to the Moho discontinuity beneaththeKADBat7.2–7.5s,and(c)sectionSpi06withvelocitiesfromforwardmodeling(Figure7) convertedintimeunderlain.KADB:Khayr-al-DinBank.Plio-Quaternarysedimentthicknessespresented insection4.3areestimatedusingtheforwardmodelin“e1”and“e2”(seedetailsinthetext). uncertaintiesweredefinedforeachphaseusingthemethod phases and multichannel data into the model and, on the of Zelt [1999], based on the ratio of the amplitude 250ms other hand, to verify that all structures from the forward before and after the picked arrival. A mean error depend- model are required to fit the data. For the modeling, a ing on the signal-to-noise ratio was calculated from all the minimum structure for the continental crust was used to picks,foreachphaseofeachstation,andthenconvertedto successfullyexplainarrivalsatthelandstations. atraveltimepickingerrors,forarangeinvaluesbetween20 [24] Seismic velocities were modeled using the 2-D ray and125ms.Phaseswithahighratioarecharacterizedbya tracing software XRAYINVR developed by Zelt and Smith low uncertainty, whereas phases with a low ratio are char- [1992].Thismodelingused alayer-strippingstrategy,from acterized by a higher uncertainty. Phases names and picks thetopofthemodeldownward.Thevelocitymodeliscon- aredetailedinTable1.Wide-angledataacquiredinthedeep structed layer after layer and composed of velocity and basinareratherhomogeneous,whereasdatarecordedclose interface nodes. Depth and velocities were modeled such to the coastline show significant lateral variations, proba- as to minimize the difference between the observed arrival blyinducedbystrongchangesinbathymetryandbylateral timesandthearrivaltimescomputedinthemodel(Figures6 structural variations, especially in the crustal part of the and7). profile(Figures6aand6b). [25] The set of observed traveltimes, including refracted [22] Amongthe23landstationsdeployed,only11exhib- and reflected phases, were picked from the 39 OBS and ited a sufficient quality to allow picking identification of 11 land stations recorded sections. Geometries of the sed- arrivals. Most of them were located close to the coastline imentary layers were determined from interfaces picked (Figure 7). Land station sections did not show sedimentary fromtheMCScoincidentline.Theseinterfacesincludethe arrivalsduetothelargedistancebetweenthestationandthe Messinian erosion surface on the upper margin, as well as closest shot but only deep arrivals (PmP, Pg) (Figure 6c). the top of the Messinian units, and, where visible, the base Pnarrivalsfromtheuppermantlewerenotrecordedbythe of the Messinian salt layer in the deep basin (see geologi- landstations. cal units, Figures 4 and 5). Arrival times picked from the 3.3.2. ForwardModeling MCS data were converted to depth using velocities from [23] Construction of a forward ray tracing model allows the forward modeling. For these layers, only the veloci- us on the one hand to include information from reflected ties were adjusted to reduce the misfit between observed 3904 LEPRÊTREETAL.:CENTRALALGERIANMARGINEVOLUTION andcalculatedtraveltimes.Theforwardmodelingintegrated 28,586picks. 3.3.3. ErrorAnalyses [26] The quality of the forward model can be quantified using the fit between predicted arrival times and traveltime picks.Thecorrespondingmisfitis121msusing93%ofthe picks. The number of picks, RMS errors, and (cid:2)2 for each phase obtained for the final forward model are detailed in Table1. [27] Two-point ray tracing between source and receiver (Figure 8) shows the well-resolved and the unconstrained areas. Ray coverage for both diving and reflected waves is generally very good due to the excellent data quality and closeinstrumentspacing(Figures8aand8b).Allsedimen- tarylayersarewellsampledbyreflectedandturningraysin themarinepartofthemodel.Thecrustallayers,theoceanic Moho,andtheuppermantlearewellsampled. [28] Resolution is a measure of the number of rays passing through a region of the model constrained by a particular velocity node and is therefore dependent on the node spacing [Zelt, 1999]. If a layer can be modeled with one single velocity gradient, then the resolution parameter will be high even in areas which have lower ray cover- age,as theareaisrelated to onlyonevelocitynode.Nodes with values greater than 0.5 are considered well resolved (Figure9).Thevelocitiesthroughoutthemodelshowares- olution higher than 0.5 except at the southern end of the model. The resolution decreases at the ends of the model wherenorayspassthroughthelayersandalsodecreasesat the very shallow onshore sedimentary layer due to missing reverseshotsonland.Uppermantlevelocitiesarewellcon- strainedathigherlevels,howeverlesssoatincreasingdepth due to fewer rays penetrating into this deeper portion of themodel. [29] In order to estimate the velocity and depth uncer- tainty of the final velocity model, a perturbation analysis was performed. The depths of key interfaces were var- ied, and an Ftest was applied to determine if a significant change between models could be detected. The 95% con- fidence limit gives an estimate of the depth uncertainty of theinterface(Figures9and10).Inordertobetterconstrain uncertaintyattheMoho,boththedepthofthisinterfaceand velocitiesinthelowercrustallayerwerechangedsystemati- cally(Figure10).Weobtainonourfinalmodeluncertainties of+0.3/–0.4kmand˙0.1km/sfortheMohodepthandfor Figure6. Threeexamplesofrepresentativerecordsections velocitiesinthelowercrust,respectively.Resultsfromthis (wide-angle data). (a) Seismic section of OBS 06 on the analysis show that our preferred model allows a maximum upper margin (top), the corresponding ray paths in the for- of picks to be explained, with a minimum resulting misfit ward model (middle), and the observed traveltime picks between the picked traveltimes and arrivals predicted from (thick grey lines) and calculated traveltimes (thin black themodeling.Solutionsleadingtobetterfitsexplainalower lines) in the forward model (bottom). (b) Seismic section numberofpicksandarethuslessreliable. of OBS 26 in the deep basin (top), the corresponding ray [30] In order to additionally test the validity of the for- pathsintheforwardmodel(middle),andtheobservedtrav- ward velocity model, we may also convert velocity to eltime picks (thick grey lines) and calculated traveltimes density using an empirical law. This density model is then (thinblacklines)intheforwardmodel(bottom).(c)Seismic used to generate a predicted gravity anomaly which can be sectionoflandstation43(top),thecorrespondingraypaths compared with the measured gravity anomaly. The gravity intheforwardmodel(middle),andtheobservedtraveltime anomalywasmodeledusingthesoftwareGRAVMOD[Zelt picks(thickgreylines)andcalculatedtraveltimes(thinblack and Smith, 1992] and free-air gravity anomaly data col- lines)intheforwardmodel(bottom).Thesethreestationsare lected during the SPIRAL cruise. This modeling approach shadedinFigure2.Alltheexamplescorrespondtotheverti- isbasedontheempiricalrelationshipexistingbetweenseis- calcomponentrecording,representedwitha6km/svelocity mic velocities and densities proposed by Ludwig et al. reduction. [1970].Themisfitbetweencalculatedandpredictedgravity 3905 LEPRÊTREETAL.:CENTRALALGERIANMARGINEVOLUTION Table1. ResidualTraveltimesandChi-SquareErrorsforAllthePhasesfortheTipazaTransect,UsingForwardModeling Phase Name NumberofPicks RMS[s] (cid:2)2Error Uncertainties Water 1893 0.014 0.460 0.020–0.035 SedimentrefractioninthePlio-Quaternaryunit Ps1 748 0.086 1.261 0.020–0.100 SedimentrefractionintheMessinianunits Ps2 1695 0.104 0.761 0.100–0.125 SedimentrefractioninthePresaltunit Ps3 1494 0.132 1.895 0.020–0.125 ReflectionatthetopoftheMessinianunit PsP1 1710 0.072 6.330 0.025–0.050 ReflectionatthebaseoftheMessinianSaltunit PsP2 1263 0.097 2.399 0.100–0.125 Reflectionatthetopofthebasement PgP 1465 0.128 3.256 0.020–0.125 Refractionintheuppercrust Pg1 4562 0.137 1.232 0.125 Refractioninthelowercrust Pg2 6037 0.121 1.456 0.100–0.125 ReflectionattheMoho PmP 5866 0.142 1.403 0.100–0.125 Refractionintheuppermantle Pn 1853 0.127 1.073 0.100–0.125 AllPhases 28586 0.121 1.720 anomaliesisabout15.5mGal,whichrepresentsagoodval- the seismic profile in order to (1) constrain the structure of idation of our velocity forward model (Figures 7a and 7b). thesedimentarysequenceandthebasementoftheAlgerian The largest misfit observed, between 15 and 30km model margin and basin off Tipaza and (2) better understand distance,mightbeduetothe3-DtopographyoftheKADB the kinematic and tectonic history of the Algerian margin. (Figure7). The main sedimentary and crustal features identified are describedbelow. 4. Results 4.1. StructureoftheSedimentaryUnits [31] Forwardmodelingwascarriedoutandcoupledwith [32] While the MARADJA data were limited by their the interpretation of the MCS data on the marine part of penetration (Figures 3 and 4), the MCS profile SPIRAL Figure7. ResultsofforwardvelocityandgravitymodelingalongtheTipazaprofile.(a)Resultsofthe gravitymodeling.RedlinerepresentsgravityanomaliesfromtheSPIRALcruisemeasurement,andthe dashed black line represents gravity anomalies calculated from conversion of the seismic velocity pre- dictedbytheforwardmodelingtodensities.(b)Resultsoftheforwardvelocitymodeling,including39 OBSand11landstations.OBSlocationsareindicatedbyredcircles.Locationsofthelandstationsused inthemodelingareindicatedbyredtriangles.Seismicrecordsfromthelandstationsindicatedbyyellow trianglesareoftoolowaqualitytobeintegrated.Isovelocitycontoursarerepresentedevery0.25km/s. Areasunconstrainedbyraytracingareshaded.Redlinesmarkvelocity-depthprofilesshowninFigure11. (c)Isostaticgravityanomalycalculatedalongthetransectassumingalocalisostaticequilibriumandcon- stantvaluesof2700and3300kg/m3forcrustandmantledensities,respectively[e.g.,BehnandLin,2000; Balmino et al., 2012], and a level of compensation at a depth of 15km. Density for the crust is chosen to be typical of continental crust. The Bouguer gravity anomaly is from the International Gravimetric Bureau(BGI)(http://bgi.omp.obs-mip.fr/)(seetextfordetails). 3906 LEPRÊTREETAL.:CENTRALALGERIANMARGINEVOLUTION strong undulations of the refracted arrivals on the OBS sections(Figures6aand6b). [36] Inthedeepbasin,variationsinsedimentthicknessare observed at two scales: (i) At short wavelengths, both the Plio-Quaternarysediments(1.9km/s(cid:4)Vp(cid:4)2.7km/s)and theMessiniansequence(3.9km/s(cid:4)Vp(cid:4)4.20km/s)exhibit strong variations in thickness associated with diapirism induced by the Messinian salt. Below these levels, the deepest presalt sedimentary layer (Figure 4) shows a rela- tivelyconstantthicknessofabout1.3–1.4kmalongthebasin (Figure7),withvelocitiesrangingfrom4.50km/satthetop to 5.0km/s at the base. (ii) At larger wavelengths, the total sedimentary cover depicts a regular 3.7km thickness cor- responding to the sedimentary infilling of the distal basin (Figure 7). However, the whole sedimentary cover shows a progressive thickening toward the margin foot, where it reachesmorethan4kminthickness. 4.2. StructureoftheCrustandUpperMantleVelocities [37] Beneath the upper margin (KADB), the Moho evolvesatadepthgreaterthan15kmbelowthesouthernpart oftheKADB(distance0onmodel,Figure7)andbecomes progressivelyshallowertowardthedeep basin.Thisresults inacrustalthicknessofabout15kmwheretheperchedsed- imentary basin is observed (between 5 and 20km in the model,Figure7).IntheMCSdatasection,theMohoproba- blycorrespondstosomediscontinuousreflectionsobserved at 7.2–7.5 seconds two-way travel time (stwtt) between 0 Figure 8. (a) (top) Ray coverage of diving waves with and12kmalongtheprofile(Figure5b),comparablewiththe every twentieth ray from two-point ray tracing plotted. time-convertedforwardvelocitymodel(Figure5c).Crustal (bottom) Observed traveltime picks and calculated travel- P velocities change from 5.2km/s in the upper part of the times (line) for the same phases for all receivers along the crust to 6.3km/s in the lower part, resulting in a very low model.(b)SameasFigure8abutforreflectedphases. verticalvelocitygradientof0.065˙0.015km/s/km. [38] Thetransition toward thedeep basin is marked by a thinningofthecrustfrommorethan15kmthickintheupper Spi06 (Figure 5) and the coincident wide-angle data define margin to only (cid:2)6km at the margin foot (Figure 7), over the overall geometry of the margin and locally allow to adistanceof50km.Underneaththesedimentarycover,the imagebelowthesaltlayerandfarthertowardthedeepbasin basement is characterized by a two-layered velocity struc- (untilabout120kmawayfromtheAlgeriancoast).Fromthe ture and depicts an average total thickness of (cid:2)5.5km in uppermargintowardthedeepAlgerianbasin,wecandiscern the deep Algerian basin (Figure 7). Velocities evolve from threestructuralregions: 5.4 to 6.2km/s in the upper layer and from 6.6 to 7.2– [33] 1. The top of the KADB is marked by a perched 7.3km/sacrossthelowerlayer(Figure4).Thecrustinthis sedimentary basin filled with several kilometers of Plio- region can be modeled using only one layer, as no strong Quaternary((cid:2)1.2kmthick)andMiocenesediments((cid:2)1km thick). This basin is imaged on the Spi06 profile at the top ofthebank(Figure5),wheretheMessinianErosionSurface (MES)formsadepression.Itisvisibleintheforwardmodel byanareaoflowvelocitieswhereisovelocitycontourdeep- ensatthetopofthebankbetweendistancesof5and25km (Figure7). [34] 2. Asharp12ıslopeformsthenorthernborderofthe KADBwhichmarksthetransitionfromtheuppermarginto theAlgerianbasin.OntheslopebetweenOBS8andOBS11 (seelocation,Figures2and7),onlyafewPs1phaseswere observed, and no PsP and no PgP phases were recorded. Across this second region, the Plio-Quaternary unit is very thinandtheslopeparticularlysteep,renderingthemodeling difficult. Figure 9. Resolution parameter for depth nodes of the [35] 3. In the deep basin, evidence for intensive salt velocity model. The depth uncertainties of the most tectonics,includinglocaldiapirsthatoutcropattheseafloor, important boundaries calculated from the 95% confidence is imaged by multichannel and wide-angle seismic data limitofthef-testaregivenintheframedboxes(Figure10). (Figures5and6).Tallsaltdiapirsatthemarginfootinduce Velocitynodesareindicatedbybluecircles. 3907 LEPRÊTREETAL.:CENTRALALGERIANMARGINEVOLUTION Figure10. Erroranalysisbymodelperturbation.(a)Resultsfromsimultaneousvariationofthedepthof theMohoandvelocitiesinthelowercrustallayer.Contoursindicatethenumberofpicksexplainedbythe forwardmodel.Theuncertaintiesofthemostimportantboundariescalculatedfromthe95%confidence limit of the f-test are given in the grey boxes. (b) Results from variation of the lower crustal velocities only.(c) ResultsfromvariationoftheMohodepth.Theuncertaintiesofthemostimportantboundaries calculatedfromthe95%confidencelimitofthef-testaregiveninthegreybox. reflections from intracrustal boundary are clear in the data. 4.3. NatureoftheCrust However,asafirstarrivaltomographicmodelingperformed [41] According to 1-D velocity-depth profiles from for- onthemarinepartclearlyimagestwolayers,anupperlayer ward modeling (Figure 11), three different domains can be characterized by a high velocity gradient and a lower layer distinguished along the transect (Figure 7). These profiles with a weak gradient, we use a two-layered velocity model were compared with preexisting compilations of velocity- forthisregion.There,inthedistaldeepbasin,thetopofthe depth profiles extracted from below the top of the base- crustislocatedataconstantdepthof(cid:2)6.5km((cid:2)5.5stwtt) ment,fortypicalthinnedcontinentalcrust[Christensenand andtheMohodiscontinuityat(cid:2)12km((cid:2)7stwtt)(Figures5 Mooney,1995]andAtlantic-typeoceaniccrust[Whiteetal., and 7). Both the top and the base of the crust, as 1992] in order to provide information on the nature of the well as the isovelocity contours, slightly deepen toward basementacrossthedifferentdomains. the margin foot where the sedimentary cover is thicker [42] 1. The first domain corresponds to the upper margin (Figure7). marked by theKhayr-al-DinBank (Domain 1,Figure7).It [39] The southern end of the model, between –80 and is located between 0 and 30km from the coastline. In this 0km(Figure7),correspondstotheonshorepart.There,the domain, the crust shows velocities and a velocity gradient sedimentary layers cannot be imaged by the seismic data consistentwithtypicalcontinentalcrust(curve1,Figure11). because of the large offset between land stations and off- The vertical velocity gradient is low, and the velocities shoreshots.Landstationsdonotprovideagoodresolution are lower than those of oceanic-type crust. The velocity- on land but rather help us to constrain the deep structure depth profile falls into the range of velocities compiled by of the margin with the contribution of Pg and PmP arrivals Christensen and Mooney [1995] of velocities for an (Figure 6c). At the southern end of the model, the deep extendedcontinentalcrusttype(Figure11).Thecontinental arrivalsenableustomodeltheMohodepthbetween–35and natureandgeometryoftheKADBsupportthehypothesisof 0km in the profile where it reaches (cid:2)20km depth at about itsoriginasablockinheritedfromtheriftingstage,aspro- 35kmfromthecoastline(Figure7). posedinearlierwork[ElRobrini,1986;Domzigetal.,2006; [40] Upper mantle velocities are constrained by Pn Yellesetal.,2009]. arrivals between distances of 20 and 115km along the for- [43] 2. The second domain is located at the foot of the ward model (Figure 8a). The velocities range from 7.9 margin,between30and40kmalongthesection(Domain2, to 8.0km/s just below the crust, when using velocities of Figure7).Inthisarea,themodeldepictsintermediateveloci- 8.2–8.3km/sat30kmdepthduringmodeling.PmParrivals tiesfasterthanintypicalcontinentalcrustandslowerthanin reflected on the Moho beneath the margin foot and the typicaloceaniccrust(Figure11),inaverynarrowtransition deep basin are of lower amplitude when compared with zone((cid:2)10kmwideorless).However,theresolutionofour thosereflectedbeneaththeKADB(Figure6).Thisobserva- velocitymodeldoesnotallowustodiscriminatebetweena tion supports a lower velocity contrast between crustal and narrowtransitionzoneordirectcontactbetweencontinental mantle velocities at the transition between lower crust and andoceaniccrust. uppermantlealongDomains2and3relativetoDomain1, [44] 3. The third domain is located beneath the deep wherevelocitiesareloweratthebaseofthecrust. basin at (cid:2) 40km from the coastline, toward the north 3908